Citation
Comparison of males and females on math item performance

Material Information

Title:
Comparison of males and females on math item performance analysis of response patterns
Creator:
Feliciano, Sonia, 1938-
Publication Date:
Language:
English
Physical Description:
x, 117 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Arithmetic ( jstor )
College mathematics ( jstor )
Educational research ( jstor )
Learning ( jstor )
Mathematical variables ( jstor )
Mathematics ( jstor )
Mathematics achievement ( jstor )
Mathematics education ( jstor )
Problem solving ( jstor )
Sex linked differences ( jstor )
Dissertations, Academic -- Foundations of Education -- UF
Foundations of Education thesis Ph. D
Mathematical ability ( lcsh )
Sex differences ( lcsh )
City of Madison ( local )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1986.
Bibliography:
Bibliography: leaves 107-115.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Sonia Feliciano.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
029704760 ( ALEPH )
15167202 ( OCLC )
AEJ5569 ( NOTIS )

Downloads

This item has the following downloads:

E7GIIQLHZ_4RJU7A.xml

AA00011802_00001.pdf

AA00011802_00001_0094.txt

AA00011802_00001_0026.txt

AA00011802_00001_0235.txt

comparisonofmale00feli_0055.txt

comparisonofmale00feli_0071.txt

AA00011802_00001_0201.txt

AA00011802_00001_0067.txt

comparisonofmale00feli_0079.txt

comparisonofmale00feli_0101.txt

comparisonofmale00feli_0028.txt

comparisonofmale00feli_0104.txt

AA00011802_00001_0215.txt

AA00011802_00001_0070.txt

AA00011802_00001_0029.txt

AA00011802_00001_0095.txt

AA00011802_00001_0017.txt

comparisonofmale00feli_0037.txt

comparisonofmale00feli_0123.txt

AA00011802_00001_0021.txt

AA00011802_00001_0031.txt

AA00011802_00001_0065.txt

comparisonofmale00feli_0041.txt

AA00011802_00001_0124.txt

comparisonofmale00feli_0115.txt

AA00011802_00001_0086.txt

AA00011802_00001_0009.txt

AA00011802_00001_0160.txt

AA00011802_00001_0176.txt

comparisonofmale00feli_0058.txt

AA00011802_00001_0119.txt

comparisonofmale00feli_0002.txt

comparisonofmale00feli_0082.txt

comparisonofmale00feli_0051.txt

comparisonofmale00feli_0044.txt

comparisonofmale00feli_0090.txt

AA00011802_00001_0110.txt

comparisonofmale00feli_0106.txt

comparisonofmale00feli_0019.txt

comparisonofmale00feli_0054.txt

comparisonofmale00feli_0021.txt

AA00011802_00001_0001.txt

comparisonofmale00feli_0015.txt

comparisonofmale00feli_0085.txt

AA00011802_00001_0004.txt

comparisonofmale00feli_0031.txt

AA00011802_00001_0082.txt

comparisonofmale00feli_0094.txt

AA00011802_00001_0025.txt

AA00011802_00001_0192.txt

comparisonofmale00feli_0118.txt

AA00011802_00001_0016.txt

AA00011802_00001_0044.txt

AA00011802_00001_0143.txt

AA00011802_00001_0219.txt

AA00011802_00001_0040.txt

comparisonofmale00feli_0017.txt

AA00011802_00001_0147.txt

comparisonofmale00feli_0110.txt

comparisonofmale00feli_0046.txt

AA00011802_00001_0136.txt

comparisonofmale00feli_0049.txt

comparisonofmale00feli_0072.txt

comparisonofmale00feli_0020.txt

AA00011802_00001_0039.txt

AA00011802_00001_0210.txt

AA00011802_00001_0208.txt

AA00011802_00001_0248.txt

comparisonofmale00feli_0117.txt

AA00011802_00001_0232.txt

comparisonofmale00feli_0077.txt

comparisonofmale00feli_0048.txt

comparisonofmale00feli_0006.txt

comparisonofmale00feli_0099.txt

AA00011802_00001_0171.txt

AA00011802_00001_0060.txt

AA00011802_00001_0184.txt

comparisonofmale00feli_0119.txt

AA00011802_00001_0059.txt

AA00011802_00001_0130.txt

AA00011802_00001_0175.txt

AA00011802_00001_0178.txt

AA00011802_00001_0172.txt

AA00011802_00001_0083.txt

AA00011802_00001_0209.txt

AA00011802_00001_0011.txt

AA00011802_00001_pdf.txt

comparisonofmale00feli_0024.txt

AA00011802_00001_0223.txt

AA00011802_00001_0108.txt

AA00011802_00001_0157.txt

AA00011802_00001_0179.txt

comparisonofmale00feli_0066.txt

AA00011802_00001_0084.txt

AA00011802_00001_0211.txt

AA00011802_00001_0071.txt

comparisonofmale00feli_0063.txt

AA00011802_00001_0154.txt

AA00011802_00001_0194.txt

comparisonofmale00feli_0014.txt

AA00011802_00001_0227.txt

AA00011802_00001_0015.txt

AA00011802_00001_0080.txt

comparisonofmale00feli_0061.txt

AA00011802_00001_0239.txt

AA00011802_00001_0141.txt

AA00011802_00001_0023.txt

AA00011802_00001_0207.txt

AA00011802_00001_0096.txt

AA00011802_00001_0164.txt

AA00011802_00001_0234.txt

comparisonofmale00feli_0053.txt

AA00011802_00001_0035.txt

AA00011802_00001_0010.txt

AA00011802_00001_0032.txt

AA00011802_00001_0002.txt

comparisonofmale00feli_0089.txt

AA00011802_00001_0052.txt

AA00011802_00001_0203.txt

AA00011802_00001_0206.txt

comparisonofmale00feli_0000.txt

AA00011802_00001_0243.txt

AA00011802_00001_0006.txt

AA00011802_00001_0100.txt

comparisonofmale00feli_0039.txt

comparisonofmale00feli_0010.txt

AA00011802_00001_0198.txt

AA00011802_00001_0214.txt

AA00011802_00001_0180.txt

AA00011802_00001_0079.txt

comparisonofmale00feli_0105.txt

AA00011802_00001_0107.txt

AA00011802_00001_0205.txt

comparisonofmale00feli_0120.txt

AA00011802_00001_0224.txt

AA00011802_00001_0074.txt

comparisonofmale00feli_0036.txt

AA00011802_00001_0241.txt

AA00011802_00001_0151.txt

comparisonofmale00feli_0033.txt

AA00011802_00001_0181.txt

comparisonofmale00feli_0025.txt

AA00011802_00001_0173.txt

AA00011802_00001_0077.txt

comparisonofmale00feli_0009.txt

comparisonofmale00feli_0097.txt

comparisonofmale00feli_0109.txt

comparisonofmale00feli_0030.txt

AA00011802_00001_0113.txt

AA00011802_00001_0185.txt

AA00011802_00001_0245.txt

AA00011802_00001_0042.txt

AA00011802_00001_0118.txt

comparisonofmale00feli_0005.txt

comparisonofmale00feli_0043.txt

AA00011802_00001_0078.txt

AA00011802_00001_0104.txt

AA00011802_00001_0246.txt

AA00011802_00001_0222.txt

comparisonofmale00feli_0075.txt

AA00011802_00001_0048.txt

AA00011802_00001_0068.txt

AA00011802_00001_0144.txt

AA00011802_00001_0097.txt

AA00011802_00001_0088.txt

AA00011802_00001_0076.txt

AA00011802_00001_0255.txt

comparisonofmale00feli_0040.txt

comparisonofmale00feli_0073.txt

AA00011802_00001_0166.txt

AA00011802_00001_0046.txt

AA00011802_00001_0132.txt

AA00011802_00001_0140.txt

AA00011802_00001_0091.txt

AA00011802_00001_0225.txt

comparisonofmale00feli_0121.txt

AA00011802_00001_0024.txt

AA00011802_00001_0202.txt

AA00011802_00001_0155.txt

AA00011802_00001_0055.txt

AA00011802_00001_0122.txt

AA00011802_00001_0226.txt

AA00011802_00001_0034.txt

AA00011802_00001_0218.txt

comparisonofmale00feli_0128.txt

comparisonofmale00feli_0070.txt

AA00011802_00001_0090.txt

AA00011802_00001_0251.txt

AA00011802_00001_0186.txt

AA00011802_00001_0145.txt

comparisonofmale00feli_0084.txt

AA00011802_00001_0075.txt

AA00011802_00001_0216.txt

AA00011802_00001_0102.txt

AA00011802_00001_0258.txt

AA00011802_00001_0087.txt

AA00011802_00001_0148.txt

comparisonofmale00feli_0102.txt

AA00011802_00001_0187.txt

comparisonofmale00feli_0124.txt

comparisonofmale00feli_0026.txt

AA00011802_00001_0114.txt

comparisonofmale00feli_0095.txt

AA00011802_00001_0162.txt

AA00011802_00001_0169.txt

comparisonofmale00feli_0107.txt

AA00011802_00001_0053.txt

AA00011802_00001_0238.txt

AA00011802_00001_0242.txt

AA00011802_00001_0182.txt

comparisonofmale00feli_0050.txt

AA00011802_00001_0204.txt

AA00011802_00001_0195.txt

comparisonofmale00feli_0088.txt

AA00011802_00001_0064.txt

AA00011802_00001_0117.txt

AA00011802_00001_0135.txt

AA00011802_00001_0045.txt

comparisonofmale00feli_0069.txt

AA00011802_00001_0231.txt

AA00011802_00001_0253.txt

AA00011802_00001_0250.txt

comparisonofmale00feli_0067.txt

AA00011802_00001_0163.txt

comparisonofmale00feli_0076.txt

AA00011802_00001_0098.txt

AA00011802_00001_0105.txt

AA00011802_00001_0043.txt

comparisonofmale00feli_0004.txt

AA00011802_00001_0240.txt

AA00011802_00001_0054.txt

comparisonofmale00feli_0042.txt

AA00011802_00001_0128.txt

AA00011802_00001_0212.txt

AA00011802_00001_0193.txt

AA00011802_00001_0146.txt

AA00011802_00001_0188.txt

comparisonofmale00feli_0086.txt

comparisonofmale00feli_0023.txt

comparisonofmale00feli_0022.txt

AA00011802_00001_0049.txt

comparisonofmale00feli_0065.txt

AA00011802_00001_0170.txt

comparisonofmale00feli_0122.txt

AA00011802_00001_0103.txt

AA00011802_00001_0237.txt

AA00011802_00001_0099.txt

AA00011802_00001_0190.txt

comparisonofmale00feli_0056.txt

AA00011802_00001_0012.txt

AA00011802_00001_0038.txt

AA00011802_00001_0093.txt

comparisonofmale00feli_0057.txt

comparisonofmale00feli_0093.txt

AA00011802_00001_0022.txt

AA00011802_00001_0116.txt

AA00011802_00001_0092.txt

comparisonofmale00feli_0127.txt

AA00011802_00001_0159.txt

AA00011802_00001_0061.txt

AA00011802_00001_0057.txt

AA00011802_00001_0072.txt

AA00011802_00001_0126.txt

AA00011802_00001_0062.txt

AA00011802_00001_0030.txt

AA00011802_00001_0230.txt

comparisonofmale00feli_0112.txt

AA00011802_00001_0134.txt

AA00011802_00001_0005.txt

AA00011802_00001_0161.txt

AA00011802_00001_0233.txt

comparisonofmale00feli_0113.txt

AA00011802_00001_0156.txt

AA00011802_00001_0158.txt

comparisonofmale00feli_0038.txt

AA00011802_00001_0213.txt

AA00011802_00001_0137.txt

AA00011802_00001_0153.txt

comparisonofmale00feli_0013.txt

AA00011802_00001_0196.txt

comparisonofmale00feli_0027.txt

AA00011802_00001_0221.txt

AA00011802_00001_0111.txt

AA00011802_00001_0041.txt

AA00011802_00001_0236.txt

AA00011802_00001_0174.txt

AA00011802_00001_0191.txt

comparisonofmale00feli_0100.txt

AA00011802_00001_0257.txt

AA00011802_00001_0020.txt

AA00011802_00001_0199.txt

AA00011802_00001_0127.txt

AA00011802_00001_0120.txt

comparisonofmale00feli_0078.txt

comparisonofmale00feli_0016.txt

comparisonofmale00feli_0116.txt

comparisonofmale00feli_0018.txt

AA00011802_00001_0028.txt

comparisonofmale00feli_0080.txt

AA00011802_00001_0003.txt

comparisonofmale00feli_0092.txt

AA00011802_00001_0066.txt

E7GIIQLHZ_4RJU7A_xml.txt

AA00011802_00001_0167.txt

comparisonofmale00feli_0114.txt

AA00011802_00001_0013.txt

AA00011802_00001_0123.txt

comparisonofmale00feli_0008.txt

comparisonofmale00feli_0103.txt

AA00011802_00001_0152.txt

AA00011802_00001_0244.txt

AA00011802_00001_0027.txt

AA00011802_00001_0036.txt

comparisonofmale00feli_0074.txt

AA00011802_00001_0019.txt

AA00011802_00001_0142.txt

AA00011802_00001_0089.txt

AA00011802_00001_0051.txt

comparisonofmale00feli_0001.txt

AA00011802_00001_0069.txt

AA00011802_00001_0168.txt

comparisonofmale00feli_0125.txt

comparisonofmale00feli_0034.txt

AA00011802_00001_0228.txt

AA00011802_00001_0183.txt

comparisonofmale00feli_0111.txt

AA00011802_00001_0177.txt

AA00011802_00001_0165.txt

AA00011802_00001_0247.txt

AA00011802_00001_0047.txt

AA00011802_00001_0007.txt

AA00011802_00001_0018.txt

comparisonofmale00feli_0012.txt

AA00011802_00001_0033.txt

comparisonofmale00feli_0052.txt

comparisonofmale00feli_0035.txt

comparisonofmale00feli_0064.txt

AA00011802_00001_0254.txt

AA00011802_00001_0101.txt

AA00011802_00001_0139.txt

AA00011802_00001_0249.txt

comparisonofmale00feli_0081.txt

AA00011802_00001_0200.txt

AA00011802_00001_0121.txt

comparisonofmale00feli_0068.txt

AA00011802_00001_0189.txt

comparisonofmale00feli_0060.txt

comparisonofmale00feli_0011.txt

comparisonofmale00feli_0087.txt

AA00011802_00001_0106.txt

AA00011802_00001_0138.txt

AA00011802_00001_0150.txt

comparisonofmale00feli_0029.txt

AA00011802_00001_0073.txt

comparisonofmale00feli_0091.txt

AA00011802_00001_0125.txt

AA00011802_00001_0037.txt

AA00011802_00001_0217.txt

comparisonofmale00feli_0032.txt

AA00011802_00001_0085.txt

AA00011802_00001_0252.txt

comparisonofmale00feli_0045.txt

comparisonofmale00feli_0126.txt

AA00011802_00001_0058.txt

AA00011802_00001_0112.txt

AA00011802_00001_0256.txt

AA00011802_00001_0056.txt

comparisonofmale00feli_0059.txt

comparisonofmale00feli_0047.txt

AA00011802_00001_0081.txt

AA00011802_00001_0197.txt

AA00011802_00001_0229.txt

AA00011802_00001_0131.txt

AA00011802_00001_0149.txt

AA00011802_00001_0129.txt

comparisonofmale00feli_0083.txt

AA00011802_00001_0008.txt

comparisonofmale00feli_0062.txt

AA00011802_00001_0063.txt

AA00011802_00001_0133.txt

AA00011802_00001_0220.txt

AA00011802_00001_0014.txt

comparisonofmale00feli_0096.txt

AA00011802_00001_0115.txt

comparisonofmale00feli_0007.txt

AA00011802_00001_0050.txt

comparisonofmale00feli_0003.txt

comparisonofmale00feli_0098.txt

AA00011802_00001_0109.txt

comparisonofmale00feli_0108.txt


Full Text











COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS





By

SONIA FELICIANO


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1986











ACKNOWLEDGMENTS


I would like to acknowledge my gratitude to several

people who have influenced my formal education and/or made

this study possible. My special thanks to Dr. James

Algina, Chairman of the doctoral committee who contributed

to the development of my love for research and statistics.

He has been insuperable as professor and valued friend; his

help and guidance in the preparation and completion of this

study were invaluable. I extend my thanks to Dr. Linda

Crocker for her advice and help during my doctoral studies

at University of Florida. Thanks also go to Dr. Michael

Nunnery, member of the doctoral committee. To Dr. Wilson

Guertin, who was a friend for me and my family, I extend my

special thanks. Thanks are also extended to Dr. Amalia

Charneco, past Undersecretary of Education of the Puerto

Rico Department of Education, for her continuous support.

To my sister Nilda Santaellar who typed the thesis, I give

my sincere thanks. Special thanks go to my family and to

those friends who provided encouragement throughout this

critical period of my life.










TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS.... .......................... ii

LIST OF TABLES .. ................ ................... v

LIST OF FIGURES...................... ............... vii

ABSTRACT....... ...................................... viii

CHAPTER

I INTRODUCTION.................................. 1

Purpose of the Study ............... ........ 6
Significance of the Study................... 6
Organization of the Study................... 8


II REVIEW OF THE LITERATURE ...................... 9

Sex-related Differences in Incorrect
Response Patterns........................ 10
Sex-related Differences in Problem
Solving............... ..... ........... .... 14
Cognitive and Affective Variables that
Influence Sex Differences in Mathematics
Learning and Achievement................ 23
Differences in Formal Mathematics
Education....................... 24
Differences in Spatial Ability.......... 26
Differentiated Effect of Affective
Variables................... ....... ... 30

Problem Solving Performance and Related
Variables.... ..................... ......... 37
Computational Skills and Problem Solving
Performance. .......................... 38
Reading and Problem Solving Perfor-
mance........................... .... .. 44
Attitudes Toward Problem Solving and
Problem Solving Performance........... 50


iii










III METHOD..... ... ...... ...... ..... ............... 54

The Sample ................ ................... 54
The Instrument............. ................ 55
Analysis of the Data....................... 57
Analysis of Sex by Option by Year
Cross Classifications................ 57
Comparison of Males and Females in
Problem Solving Performance........... 66


IV RESULTS........... ..... ....... ...... .. ........ 68

Introduction ............................. 68
Sex-related Differences in the
Selection of Incorrect Responses.......... 68
Sex-related Differences in Problem
Solving Performance.................... 81
Summary.................................... 93


V DISCUSSION..................................... 101

Summary and Interpretation of the
Results ................................... 101
Implications of the Findings and
Suggestions for Further Research.......... 103
Sex-related Differences in
Incorrect Responses................ 104
Sex-related Differences in Problem
Solving Performance................... 105

REFERENCES.......................................... 107

BIOGRAPHICAL SKETCH................................ 116









LIST OF TABLES


Table Page

3.1 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex.......... 58

3.2 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex and
Arranged by Sex......... ...... ........... 58

3.3 Hypothetical Joint Probabilities of Yearr
Option, and Sex............................. 64

3.4 Hypothetical Probabilities of Option
Choice Conditional on Sex and Year.......... 66

4.1 Summary of Significant Tests of Model 3.2..... 71

4.2 Chi-square Values for the Comparison of 75
Models (3.2) and (3.3) with Actual Sample
Sizes and with the Corresponding Number
of Subjects Needed for Significant Results 75

4.3 Means, Standard Deviations, and t-Test for
the Eight Mathematical Variables............ 82

4.4 ANCOVA First Year: Multiplication Covariate 85

4.5 ANCOVA First Year: Division Covariate....... 85

4.6 ANCOVA Second Year: Subtraction Covariate... 86

4.7 Reliability of the Covariates for Each
of the Three Years of Test Administra-
tion........................................ 91

4.8 ANCOVA First Year: Other Covariates......... 92

4.9 Adjusted Means on Problem Solving, by
Covariate and Sex.......................... 94

4.10 ANCOVA Second Year: Other Covariates........ 95









Table


Page


4.11 Adjusted Means on Problem Solving,
by Covariate and Sex........................ 96

4.12 ANCOVA Third Year........................ 97

4.13 Adjusted Means on Problem Solving,
by Covariate and Sex....................... 98










LIST OF FIGURES


Figure Page


3.1 Number of Questions by Skill Area in a
Basic Skills Test in Mathematics-6.......... 56

4.1 Sex by Multiplication Interaction............. 87

4.2 Sex by Division Interaction ................ 88

4.3 Sex by Subtraction Interaction................ 89


vii














Abstract of Dissertation Presented to the Graduate
School of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy

COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS

By

Sonia Feliciano

August, 1986


Chairman: James Algina
Major Department: Foundations of Education

The first objective of this study was to investigate

sex differences in the selection of incorrect responses on

a mathematics multiple-choice test, and to determine

whether these differences were consistent over three

consecutive administrations of the test. A second

objective was to compare male and female performance in

problem solving after controlling for computational skills.

The responses of all 6th grade students from the

public schools in Puerto Rico who took the Basic Skills

Test in Mathematics-6 ("Prueba de Destrezas B&sicas en

Matem&ticas-6") during three academic years were used in

the analyses relevant to the first objective.


viii









Log-linear models were used in the analysis of

incorrect responses. The results of the analyses showed

that for 100 of the 111 items of the test, males and

females selected different incorrect options, and this

pattern of responses was consistently found during the

three years of test administration. However, for the vast

majority of the 100 items the male-female differences were

relatively small, considering the fact that the number of

subjects needed to obtain statistical significance was very

large.

The responses of approximately 1,000 randomly selected

students per academic year were analyzed in the comparison

of male and female performance in problem solving. Females

outperformed males in problem solving and in six of the

seven computational variables. Males showed superiority in

equivalence in all the three years, but statistical

significance was obtained in only one of the years.

Analysis of covariance (ANCOVA) was used in the

comparison of male and female performance in problem

solving after controlling for computational skills. Seven

analyses of covariance tests were conducted, one for

each of the covariates. Estimated true scores for observed

scores were used in the analyses. The results tend to show

that for examinees with similar levels of computational









skills, sex-related differences in problem solving

performance do not exist. Females retained their

superiority in problem solving when equivalence (in all

three years) and subtraction (in one year) were the

controlling variables.


The question of whether male-female differences in

problem solving depend on computational skills was

answered, partially, in the affirmative.













CHAPTER I
INTRODUCTION



Sex differences in mathematics learning and

achievement have been the subject of intensive research.

Research done before 1974 has shown that male performance

on mathematical achievement tests is superior to female

performance by the time they reach upper elementary or

junior high school (Fennema, 1976, p. 2). The literature

strongly suggests that at the elementary level females

outperform males in computation and males excel in

mathematical reasoning (Glennon & Callahan, 1968; Jarvis,

1964; Maccoby, 1966).

Since 1974, research findings have been less

consistent. Fennema (1974), after reviewing 36 studies,

found that during secondary school or earlier, sex-related

differences in mathematics achievement are not so evident,

but that when differences are founds they favor males in

high level cognitive tasks (problem solving) and females in

low level cognitive tasks (computation). As a result of a

further review of the literature, Fennema (1977) concluded

that at the elementary level, sex-related differences do

not exist at all cognitive levels, from computation to

problem solving.











Many variables, cognitive affective, and educational,

have been investigated since 1974 in relation to sex

differences in mathematics learning and achievement.

Fennema and Sherman (1977) investigated the effect of

differential formal mathematics education. After

controlling for the number of years of exposure to the

study of mathematics, they found sex differences in only

two of the four schools under study. However, in those

schools where boys scored higher than girls, differences

were also found in their attitudes toward mathematics.

Hilton and Berglund (1974) found significant sex

differences after controlling for the number of mathematics

courses taken, and attributed them to sex differences in

interests. "As the boys' interests in science increase

related to the girls', their achievement in mathematics

increases relative to that of the girls" (p. 234).

Wise, Steel, and McDonald (1979) reanalyzed test data

collected in a longitudinal study of 400,000 high school

students (Project Talent). They found that when the effect

of the number of high school mathematics courses was not

controlled, no sex differences emerged for 9th graders, but

that gains made by boys during the next three years were

more than twice that of the girls. These differences

between the sexes disappeared when the number of

mathematics courses taken was controlled. Results of the

1978 Women and Mathematics National Survey, Survey I,










indicated no significant sex differences for 8th grade

students on measures of problem solving or algebra.

However, females outperformed males in computation and

spatial visualization. For the 12th grade students,

statistically significant sex differences favoring males

were found in problem solving, but not in algebra,

computation, or spatial visualization. For males and

females who had enrolled in courses beyond general

mathematics and who had taken or were enrolled in courses

such as pre-calculus, calculus, or geometry, differences in

problem solving or spatial visualization did not exist.

Sex differences favoring males were found on a total score

obtained summing across the computation, problem solving,

and algebra subtests (Armstrong, 1979). The mathematics

data collected in the second survey by the 1978 National

Assessment of Educational Progress showed significant sex

differences for both 13- and 17-year-old students. The

13-year-old females outperformed males in the computational

subtest and males outscored females by 1 1/2 percentage

points in problem solving (statistically significant). No

statistically significant differences were found in

algebra. No sex differences were found for the 17-year-old

group either in the computation subtest or in the algebra

subtest. Males surpassed females in problem solving. A

reanalysis of the data from the 17-year-old group confirmed

male superiority in problem solving after controlling for











mathematics preparation. Males who were enrolled or had

completed algebra II outperformed the females in

computation and problem solving but not in algebra. Males

who studied beyond algebra II outscored females on all

three subtests: computation, algebra, and problem solving

(Armstrong, 1979).

Carpenter, Lindquist, Mathews, and Silver (1984)

analyzed the results of the Third National Assessment of

Educational Progress (NAEP), and compared them with the

First and Second Surveys. Between 1978 and 1982, the

differences between the average performance of males and

females remained stable at each age level. At ages 9 and

13, the overall performance of males and females was not

significantly different. At age 17r males scored higher

than females by about 3 percentage points. When course

background was held constant, achievement differences still

existed at age 17. For each category of course background,

male achievement exceeded female achievement. Consistent

with previous assessments, sex differences in problem

solving in favor of males were found for the 17-year-old

sample. At ages 9 and 13, no large differences were found

between the sexes within any level of course background.

Marshall (1981, 1984) investigated sex differences in

mathematics performance. She found that males and females

excel each other in solving different types of problems.











Females were better on items of computation and males were

more successful on word-story problem items (problem

solving). She also found that females successful

performance in the problem solving items was more dependent

on their successful performance in the computation items.

Males did not need, as much as females, to succeed in the

computation items in order to answer correctly the problem

solving items.

Although the general findings seem to support sex

differences in mathematics learning and achievement, the

research done does not consistently support superiority for

either sex. Most of the research has been concerned with

how the sexes differ on subtests or total test scores in

mathematics. Moreover, the great majority of the studies

deal with correct responses. Sex differences in incorrect

responses at the item level have not been fully researched.

Only two studies dealing with sex differences in incorrect

responses at the item level were found in the research

literature (Marshall, 1981, 1983). Marshall investigated

whether boys and girls made similar errors in computation

and story problems. She analyzed boys' and girls' answers

to six mathematics items and found that the sexes made

different errors, possibly reflecting different problem

solving strategies. Her original findings were supported

when she studied the same problem using a large number of

items three years later.











Purpose of the Study

The purpose of this study was twofold. The first was

to investigate sex-related differences in the selection of

incorrect responses in a multiple-choice achievement test

in mathematics. For each test item, the research questions

were as follows:

1. Is there a difference in the proportion of males

and females choosing each incorrect option?

2. Is the same pattern of differences found in data

obtained in three different administrations of the test?

The second objective was to investigate sex-related

differences in test scores in mathematics problem solving.

The following questions were studied:

1. Do males and females differ in problem solving

performance?

2. Do sex differences in problem solving persist

after controlling for computational skills, and does the

differential success of males and females on problem

solving items depend on their success on the computation

items?

Significance of the Study

Item response patterns are very useful techniques in

the assessment of mathematics learning and achievement.

Total test scores can be very misleading in the assessment

of student performance and provide no diagnostic

information about the nature and seriousness of student











errors (Harnisch, 1983). Item response patterns are

valuable for the identification of large group differences,

including district-to-district, school-to-school, and

classroom-to-classroom variations on different subsets of

items. The response patterns can provide diagnostic

information about the type of understanding the student has

on various mathematics topics (e.g.r problem solving).

Marshall (1981, 1984) has used the item response

pattern technique and her findings indicate sex differences

in mathematics performance at the item level. Females

outperformed males in computation and males outscored

females in problem solving. Also the success of girls in

the problem solving items was dependent upon their success

in the computation items; for boys, success in the problem

solving items did not depend as much on their computational

performance. Marshall (1981, 1983) has also reported that

males and females differ in the selection of incorrect

responses, reflecting differences in reasoning abilities.

In Puerto Rico, a high percentage of children promoted

to the 7th grade in the public schools does not master the

basic skills in mathematics. If 6th grade male and female

children can be diagnosed as having different problem

solving abilities, as Marshall found with California

children, teachers may need to provide tailor-made

mathematics instruction for each sex, in order to ensure

equal access to formal education and enhance mathematics











achievement. Since there are no investigations reported in

sex differences in item response patterns in Puerto Rico,

research is needed.


Organization of the Study

A review of the literature on sex differences in

mathematics performance is reported in Chapter II. The

research methodology is presented in Chapter III. Research

questions, sample, instrument, and data analysis are

discussed in that chapter. Chapter IV is an exposition of

the results of the study. Chapter V contains a summary and

interpretation of the results of the study and the

implications of the findings together with suggestions for

further research.













CHAPTER II
REVIEW OF THE LITERATURE



Sex differences in mathematics learning and

achievement have been a subject of concern for educators

and psychologists. Many studies found in the literature

support the existence of these differences. Boys show

superiority in higher level cognitive tasks (problem

solving or mathematical reasoning) in the upper elementary

years and in the early high school years (Fennemar 1974;

Maccoby & Jacklin, 1974).

Almost all the research carried out has dealt with

analysis of total correct scores in mathematics aptitude

and achievement tests or scores in subtests. The

literature related to sex differences in incorrect

responses, the main subject of the present study, is

surprisingly sparse. For the most part, the studies have

investigated the differences between the sexes in

mathematics learning and achievement and the underlying

variables causing the differences. Cognitive and affective

variables have been the matter at issue in the

establishment of sex differences.

Although research in mathematics problem solving, the

secondary subject of this investigation, is extensive, most











of the studies consider sex differences incidental to the

major study findings. The available literature offers very

little research directly related to the problem of sex

differences in this area.

The review of the literature has been divided in four

sections. The first section consists of a detailed summary

of the available research on sex differences in incorrect

responses. The second section deals with sex-related

differences in problem solving performance. These sections

are directly related to the objectives of the study. The

third section is more peripheral, and contains a discussion

of the more prevalent issues about the influence of

cognitive and affective variables on sex differences in

mathematics learning and achievement. The fourth is a

summary of the research dealing with variables considered

as influential to mathematics problem solving performance.



Sex-related Differences in Incorrect Response Patterns



Research findings tend to suggest that boys and girls

may be approaching problem solving differently (e.g.,

Fennema and Sherman, 1978; Marshall, 1981, 1983; Meyer,

1978, among others). Marshall (1981) investigated whether

6th grade boys and girls approach mathematical problem

solving with different strategies. Her specific interest

was whether the sexes made the same errors.











She analyzed the responses of 9,000 boys and 9,000

girls to 6 selected items, 2 computation items, and 1 story

problem item from each of 2 of the 16 forms of the Survey

of Basic Skills test administered during the academic year

1978-79. The Survey is a 30-item achievement test

administered every year to all 6th grade children in

California through the California Assessment Program.

There are 16 forms of the test, to which approximately

9,000 boys and 9,000 girls respond each year. Of the 160

mathematics items contained in the 16 forms of the test, 32

are on measurement and graphing, 28 on number concepts, 28

on whole number arithmetics, 20 on fraction arithmetics, 20

on decimal arithmetics, 20 on geometry, and 12 on

probability and statistics.

The item analysis performed on the 1978-79 data showed

that boys and girls tended to select different incorrect

responses. In the first computation item (Form 1 of the

test) both sexes reflected similar mistakes in carrying,

but in different columns. In the second computation item,

both sexes ignored the decimal points and selected the same

incorrect response. However, more girls than boys chose

this response.

In the first computation item (Form 2 of the test) the

incorrect choice of both sexes was option cr but the

second most frequently selected option was a for boys

and b for girls. In the second computation item of this











test form, no sex differences were found in response

patterns. Approximately 45% of each sex selected option

c. The next popular choice for both sexes was option

d, selected by approximately 35% of both boys and girls.

On the story problem of Form 1, males and females

responded alike. Their most popular incorrect response

choice was option a for both males and females. The

second most popular incorrect choice was option c for

both sexes.

Response to the story problem in Form 2 showed sex

differences in response choice. Including the correct

option, 33% of the girls selected option a, 20% chose

option c, and 20% option d. For males approximately

25% selected option a and the same percent selected

option d.

Marshall concluded that although the analysis of

incorrect responses does not explain why boys and girls

differ in their responses, the analysis shows that boys and

girls approach problems in different ways and these varying

strategies can be useful in identifying how the sexes

differ in reasoning abilities.

Two years later, Marshall (1983) analyzed the

responses of approximately 300,000 boys and girls to

mathematics items contained in the 16 test forms of the

Survey of Basic Skills during the years 1977, 1978, and

1979. She used log-linear models (explained in Chapter












III) to investigate sex related differences in the

selection of incorrect responses, and the consistency of

such differences over three years of administration of the

test.

Based on her findings that sex differences were found

in 80% of the items, Marshall classified the students'

errors according to Radatz' (1979) five-category error

classification. The categories are language (errors in

semantics), spatial visualization, mastery, association,

and use of irrelevant rules.

It was found that girls' errors are more likely to be

due to the misuse of spatial information, the use of

irrelevant rules, or the choice of an incorrect operation.

Girls also make relatively more errors of negative transfer

and keyword association. Boys seem more likely than girls

to make errors of perseverance and formula interference.

Both sexes make language-related errors, but the errors are

not the same.

Available research is not extensive enough to make

definite judgments about the sex-related differences

observed in incorrect responses. Clearly more research is

needed.











Sex-related Differences in Problem Solving



It has already been acknowledged that the subject of

problem solving has been extensively researched. However,

as early as 1969, Kilpatrick criticized the fact that the

study of problem solving has not been systematic; some

researchers have studied the characteristics of the problem

while others have given their attention to the

characteristics of the problem solvers. Moreover,

differences in the tests used to measure problem solving

performance also constitute an obstacle when trying to

compare the results of the studies carried out.

In order to avoid this pitfall and provide a basis for

comparison, the studies reviewed in this section, dealing

with sex-related differences in problem solving have been

divided in two groups. The first comprises those studies

that used the Romberg-Wearne Problem Solving Test. The

second contains other relevant studies in which problem

solving performance has been measured by means of other

instruments.

The Romberg-Wearne Problem Solving Test merits special

mention because it was the first attempt "to develop a test

to overcome the inadequacies of total test scores in

explaining the reasons why some students are successful

problem solvers and others are not" (Whitaker, 1976, pp. 9,

10).











The test is composed of 23 items designed to yield 3

scores: a comprehension score, an application score, and a

problem solving score. The comprehension question

ascertains whether a child understands the information

given in the item stem. The application question assesses

the child's mastery of a prerequisite concept or skill of

the problem solving question. The problem solving question

poses a question whose solution is not immediately

available, that is, a situation which does not lend itself

to the immediate application of a rule or algorithm. The

application and problem solving parts of the test may refer

to a common unit of information (the item stem) but the

questions are independent in that the response to the

application question is not used to respond to the problem

solving question.

Meyer (1978) Whitaker (1976), Pennema and Sherman

(1978), and Sherman (1979) have used the Romberg-Wearne

test in their studies. Meyer (1978) investigated whether

males and females differ in problem solving performance and

examined their prerequisite computational skills and

mathematical concepts for the problem solving questions. A

sample of 179 students from the 4th grade were administered

19 "reference tests" for intellectual abilities and the

Romberg-Wearne test. The analysis showed that males and

females were not significantly different in the

comprehension, application, and problem solving questions











of the test. The sexes differed in only 2 of the 19

reference tests, Spatial Relations and Picture Group

Name-Selection. A factor analysis, however, showed

differences in the number and composition of the factors.

For females, a general mathematics factor was determined by

mathematics computation, comprehension, application, and

problem solving. For males, the comprehension and

application parts determined one factor; problem solving

with two other reference tests (Gestalt and Omelet)

determined another factor.

Meyer concluded that comprehension of the data and

mastery of the prerequisite mathematical concepts did not

guarantee successful problem solving either for males or

for females. Problem solving scores for both sexes were

about one third their scores in comprehension and one

fourth their scores in application.

She also concluded that the sexes may have approached

the problem solving questions differently. The methods

used by females for solving problem situations may have

paralleled their approach to the application parts.

Males may have used established rules and algorithms for

the application parts, but may have used more of a Gestalt

approach to the problem solving situation.

Whitaker (1976) investigated the relationship between

the mathematical problem performance of 4th grade children

and their attitudes toward problem solving, their teachers'











attitude toward mathematical problem solving, and related

sex and program-type differences. Although his main

objective was to construct an attitude scale to measure

attitudes toward problem solving, his study is important

because his findings support Meyer's regarding the lack of

significant sex-related differences in problem solving

performance. Performance in the problem solving questions,

for both males and females, was much lower than performance

in the application questions, and much lower than

performance in the comprehension questions. In fact, the

mean score for each part of the item, for both males and

females, was almost identical to the mean scores obtained

by males and females in Meyer's study. Whitaker noted that

each application item is more difficult than its preceding

comprehension item, and that each problem solving item is

more difficult than its preceding application item. No

significant sex-related differences were found for any of

the three parts of the item (comprehension, application, or

problem solving).

Fennema and Sherman (1978) investigated sex-related

differences in mathematics achievement and cognitive and

affective variables related to such differences. They

administered the Romberg-Wearne Problem Solving Test to a

representative sample of 1320 students (grades 6-8) from

Madison, Wisconsin, predominantly middle-class, but

including great diversity in SES. The sample consisted of











students who had taken a similar number of mathematics

courses and were in the top 85% of the class in mathematics

achievement. They were tested in 1976. Four high school

districts were included. In only one of the high school

districts were sex-related differences in application and

problem solving founds in favor of males. They concluded

that when relevant factors are controlled, sex-related

differences in favor of males do not appear often, and when

they dor they are not large.

Sherman (1980) investigated the causes of the emerging

sex-related differences in mathematics performance, in

favor of males, during adolescence (grades 8-11). She

wanted to know if these differences emerge as a function of

sex-related differences in spatial visualization and

sociocultural influences that consider math as a male

domain. In grade 8, she used the Romberg-Wearne Test and,

in grade 11, a mathematical problem solving test derived

from the French Kit of Tests.

The analysis showed that for girls, problem solving

performance remained stable across the years. Mean problem

solving performance for boys, however, was higher in grade

11 than in grade 8. No sex-related differences were found

in grade 8, but boys outperformed girls in grade 11, where

the Stafford test was used.

Sherman found that for both sexes problem solving

performance in grade 8 was the best predictor of problem











solving performance in grade 11. Spatial visualization was

a stronger predictor for girls than for boys. Mathematics

as a male domain was a good predictor for girls only; the

less a girl stereotyped mathematics as a male domain in

grade 8, the higher her problem solving score in grade 11.

Attitudes toward success in mathematics in grade 8 was a

more positive predictor of problem solving performance for

boys than for girls; the more positive the attitudes toward

success in mathematics in grade 8, the higher their

performance in problem solving in grade 11.

None of these four studies, all of which used the

Romberg-Wearne Mathematics Problem Solving Test, show

statistically significant sex-related differences in

problem solving performance. In later studies other tests

were used to measure this variable (Kaufman, 1984;

Marshall, 1981, 1984).

Kaufman (1983/1984) investigated if sex differences in

problem solving, favoring males, exist in the 5th and 6th

grades and if these differences were more pronounced in

mathematically gifted students than in students of average

mathematical ability. The Iowa Test of Basic Skills and a

mathematics problem-solving test were administered to 504

subjects. Males in the average group as well as males in

the gifted group outperformed females, but only the gifted

group showed statistically significant differences.











As a result of her investigations, Marshall (1981)

concluded that sex-related differences in mathematics

performance may be the result of comparing the sexes on

total test scores. If the test contains more computation

items than problem solving items, girls will perform better

than boys, but if the test contains more problem solving

items than computation ones, boys will outperform girls.

With this in mind, Marshall investigated sex-related

differences in computation and problem solving by analyzing

the responses of approximately 18,000 students from grade 6

who had been administered the Survey of Basic Skills Test:

Grade 6, during the academic year 1978-79.

Two of the 16 test forms of the Survey were used to

assess skills such as concepts of whole numbers, fractions,

and decimals. These skills were tested both as simple

computations and as story problems (problem solving).

Two computation items and one story problem item were

selected because they were particularly related; both

computation items required skills needed in solving the

corresponding story problem. It was assumed that correct

solution of the computation item correlates with solving

the story problem because the story problem requires a

similar computation.

Marshall found that girls were better in computation

and boys were better in problem solving. She also found

that boys were much more likely than girls to answer the











story problem item correctly after giving incorrect

responses to both computation items. Apparently, mastery

of the skills required by the computation items is more

important for girls than for boys. If girls cannot solve

the computation items, they have little chance of solving

the related story problem item. For girls, the

probability of success in the story problem item after

giving successful answers to both computation items is

almost 2 1/2 times the probability of success after giving

incorrect responses to both computation items. For boys,

the probability of success in the story problem item after

successful responses to the computation items is about

1 3/4 times the probability of success on the story problem

item after incorrect responses to the computation items.

Three years later, Marshall (1984) analyzed more in

depth these phenomena of sex-related differences. Her

interest was twofold. First, she wanted to know if there

were differences in the rate of success for boys and girls

in solving computation and story problem items. Second,

she examined additional factors that interact with sex to

influence mathematics performance, such as reading

achievement, socio-economic status (SES), primary language,

and chronological age. Two questions were raised: Do the

probabilibities of successful solving of computation and

story problem items increase with reading score? Are these

probabilities different for the two sexes?











Approximately 270,000 students from the 6th grade were

administered the Survey of Basic Skills of the California

Assessment Program, during the years 1977, 1978, and 1979.

Responses were analyzed using log-linear models.

Successful solving of computation items was positively

associated with successful solving of story problems.

Girls were more successful in computation than boys, and

boys were more successful than girls in solving story

problems. This finding supports reports from the National

Assessment of Educational Progress (NAEP) (Armstrong,

1979).

To investigate the effects of reading, SES, language

and chronological ager only those test forms containing 2

computation items and 2 story problems were considered for

analysis; 32 items from 8 test forms were included in the

analysis.

The results of these analyses showed that at every

level of reading score, 6th grade children were more

successful in computation than in story problems.

Although the differences were not larger at every reading

score boys consistently had higher probabilities of

success in story problems than did girls, and girls

consistently showed higher probabilities of success in

computation than boys. Also, as the reading score

increased, the difference between the probability of

success in story problems and the probability of success in











computation grew larger. This difference grew larger for

girls than for boys.

Although SES was a major factor in solving computation

and story problem items successfully, the effect was

similar for each sex. Sex-related differences by primary

language or chronological age were not large.

This research carried out by Marshall with elementary

grade children supports previous research findings that

males are better than females in mathematics problem

solving (a higher order skill) and females are better than

their counterpart males in computation (a lower level

skill). Marshall's research also brought out a different

aspect of this question: the notion that girls find it more

necessary than boys to succeed in the computation items in

order to successfully solve the story problem items.


Cognitive and Affective Factors That
Influence Sex Differences in Mathematics Learning
and Achievement


The research reviewed in the literature does not

provide evidence of any unique variable that could serve

as an explanation for the observed sex-related differences

in mathematics learning and achievement. However, some

issues have been discussed, among which the most prevalent

are that sex-related differences in mathematics learning

and achievement are a result of differences in formal

education; that sex-related differences in mathematics











learning and achievement arise from sex differences in

spatial visualization; and that sex-related differences

result from a differentiated effect of affective variables

on the mathematics performance of males and females.


Differences in Formal Mathematics Education (Differential
Coursework Hypothesis)


The basis for the differential coursework hypothesis

is the fact that sex-related differences in mathematical

learning and achievement show up when comparing groups

which are not equal in previous mathematics learning.

Atter the 8th grader boys tend to select mathematics

courses more otten than girls. Therefore, girls show lower

achievement scores in mathematics tests because their

mathematics experience is not as strong as the boys'

(Fennemar 1975; Fennema & Sherman, 1977; Sherman, 1979).

Fennema and Sherman's study (1977) lends additional

support to the feasibility of viewing sex differences in

mathematics learning and achievement as reflecting

something other than a difference in mathematics aptitude.

After controlling for previous study of mathematics, they

found significant sex differences in mathematics

achievement in only two of the four schools under study,

making the attribution to sex per se less likely.

Controlling for the number of space visualization-related

courses, the sex-related differences which originally












emerged in spatial visualization scores became

non-significant. In the two schools where sex differences

in mathematics achievement were found, differences between

the sexes were also found in their attitudes toward

mathematics.

Researchers like Backman (1972), who analyzed data

from Project Talent, and Allen and Chambers (1977) have

also hypothesized that sex-related differences in

mathematics achievement may be related to different

curricula followed by males and females. Allen and

Chambers attributed male superiority in mathematics problem

solving to differences in the number of mathematics courses

taken in high school.

This issue has been seriously questioned by Astin

(1974) Fox (1975ar 1975b), and Benbow and Stanley (1980),

among others. Astin and Fox have reported large

differences in favor of males among gifted students taking

the Scholastic Achievement Test. These differences occur

as early as grade 7, when there are no sex differences in

the number of courses taken. Benbow and Stanley (1980)

compared mathematically precocious boys and girls in the

7th grader with similar mathematics background, and found

sizeable sex-related differences favoring boys in

mathematical reasoning ability. Five years later, they

conducted a follow-up study which showed that boys











maintained their superiority in mathematics ability during

high school. While Fox attributed sex-related differences

in mathematical achievement to differential exposure to

mathematical games and activities outside school, Benbow

and Stanley suggested that sex-related differences in

mathematics performance stem from superior mathematical

ability in males, not from differences in mathematics

formal education.

The differential coursework hypothesis is not totally

convincing and, as reported before, it has been challenged

by researchers such as Benbow and Stanley (1980).

However, Pallas and Alexander (1983) have questioned the

generalizaoility of Benbow and Stanley's findings based on

the fact that they used highly precocious learners. The

differential coursework hypothesis can be accepted only as

a partial explanation of differences in mathematics

performance found between the sexes.

Differences in Spatial Ability

The basic premise in this issue is that males and

females differ in spatial visualization and this explains

differential mathematics learning and achievement. Until

recently, sex differences in spatial ability in favor ot

males were believed to be a fact and were thought by some

to be related to sex differences in mathematical

achievement.











Research findings in this area have been inconsistent.

In 1966, Maccoby stated that "by early school years, boys

consistently do better (than girls) on spatial tasks and

this difference continues through the high school and

college years" (p.26). In 1972, Maccoby and Jacklin said

that the differences in spatial ability between the sexes

"remain minimal and inconsistent until approximately the

ages of 10 or 11, when the superiority of boys becomes

consistent in a wide range of populations and tests"

(p.41). In 1974, after a comprehensive literature search,

Maccoby and Jacklin concluded that sex differences in

spatial visualization become more pronounced between upper

elementary years and the last year of high school, the

years when sex-related differences in mathematics

achievement favoring boys emerge.

Guay and McDaniel (1977) supported in part Maccoby and

Jacklin's 1974 findings. They found that among elementary

school children, males had greater high level spatial

ability than females, but that males and females were equal

in low level spatial ability. This finding is inconsistent

with that portion of Maccoby and Jacklin's review that

suggests that sex differences become evident only during

early adolescence. Cohen and Wilkie (1979) however, stated

that in tests measuring distinct spatial tasks, males

perform better than females in early adolescence and

throughout their life span. Most studies carried out after











1974 have failed to support these sex differences in

spatial abilities (Armstrong, 1979; Connor, Serbin, &

Schackman, 1977; Fennema & Sherman, 1978; Sherman, 1979).

Fennema and Sherman (1978) and Sherman (1979) have

explored sex-related differences in mathematical

achievement and cognitive and affective variables related

to these differences. In a study involving students from

grades 6, 7, and 8, from four school districts, Fennema and

Sherman found that spatial visualization and problem

solving were highly correlated for both sexes (.59 and

.60). Even in the school district where sex differences

were found in problem solving, no significant sex-related

differences were found in spatial visualization.

When Sherman (1980) compared groups of males and

females in two different grades, 8 and 11, she found no

sex-related differences in problem solving or in spatial

visualization in grade 8. In grade 11, however, although

the sexes differed in their problem solving performance, no

sex-related differences were found in spatial

vizualization. Even though spatial visualization in grade

8 was the second best predictor of problem solving

performance in grade 11, sex differences in grade 11 were

not a result of spatial visualization since no differences

were found in that skill.

In spite of the fact that no sex differences were

found in spatial abilities, it is evident that males and











females may use them in a different way. Meyer (1978),

with an elementary grade sample, and Fennema and Tartree

(1983) with an intermediate level sampler found that the

influence of spatial visualization on solving mathematics

problems is subtler and that males and females use their

spatial skills differently in solving word story problems

(problems that measure problem solving ability or

reasoning). Fennema and Tartree (1983) carried out a

three-year longitudinal study which showed that girls and

boys with equivalent spatial visualization skills did not

solve the same number of items, nor did they use the same

processes in solving problems. The results also suggested

that a low level of spatial visualization skills was a more

debilitating factor for girls than for boys in problem

solving performance.

Landau (1984) also investigated the relationship

between spatial visualization and mathematics achievement.

She studied the performance of middle school children in

mathematical problems of varying difficulty, and the extent

to which a diagramatic representation is likely to

facilitate solution. She found that spatial ability was

strongly correlated to mathematical problem solving and

that the effect of spatial ability was more influential for

females. Females made more use of diagrams in the solution

of problems, reducing the advantage of males over females

in problem solving performance.











The issue of sex-related differences in spatial

visualization ability as an explanation for sex differences

observed in mathematics achievement is less convincing and

the findings more contradictory than in the issue of sex

differences in formal education. Besides these cognitive

issues, other issues, mostly affective in nature have also

been studied in trying to explain the origin of these sex

differences in mathematics achievement and learning. The

studies dealing with these affective variables are reviewed

in the next section.

Differentiated Effect of Affective Variables

Researchers have attempted to explain the effect of

sex differences in internal beliefs, interests, and

attitudes (affective variables) on mathematics learning and

achievement. A brief statement of each explanation

precedes the summary of studies conducted that support the

explanation.

Confidence as lerners of mathematics. Females, more

than males, lack confidence in their ability to learn

mathematics and this affects their achievement in

mathematics and their election of more advanced mathematics

courses.

Maccoby and Jacklin (1974) reported that self-

confidence in terms of grade expectancy and success in

particular tasks was found to be consistently lower in

women than in men. In 1978, Fennema and Sherman reported











that in their study involving students from grades 6

through 12, boys showed a higher level of confidence in

mathematics at each grade level. These differences between

the sexes occurred in most instances even when no

sex-related differences in mathematics achievement were

found. The correlation between confidence in mathematics

performance and mathematics achievement in this study was

higher than for any other affective variable investigated.

Sherman reported a similar finding in 1980; in males,

the most important factor related to continuation in

theoretical mathematics courses was confidence in learning

mathematics. This variable weighed more than any of the

cognitive variables: mathematics achievement, spatial

visualization, general ability, and verbal skill. In the

case of females, among the affective variables, confidence

in learning mathematics was found to be second in

importance to perceived usefulness of mathematics. Probert

(1983) supported these findings with college students.

A variable that needs discussion within the context of

sex differences in confidence as learners of mathematics is

causal attribution. Causal attribution models attempt to

classify those factors to which one attributes success or

failure. The model proposed by Weiner (1974) categorizes

four dimensions of attribution ot success and failure:

stable and internal, unstable and internal, stable and

external, and unstable and external. For example, if one











attributes success to an internal, stable attribute, such

as ability, then one is confident of being successful in

the future and will continue to strive in that area. If

one attributes success to an external factor such as a

teacher, or to an unstable one, such as effort, then one

will not be as confident or success in the future and will

cease to strive. Failure attribution patterns work this

way: if failure is attributed to unstable causes, such as

effort, failure can be avoided in the future and the

tendency will be to persist in the task. However, if

failure is attributed to a stable cause, such as ability,

the belief that one cannot avoid failure will remain.

Studies reported by Bar-Tal and Frieze (1977) suggest

that males and females tend to exhibit different

attributional patterns of success and failure. Males tend

to attribute their success to internal causes and their

failures to external or unstable ones. Females show a

different pattern; they tend to attribute success to

external or unstable causes and failures to internal ones.

The pattern of attributions, success attributed externally

and failure attributed internally, has become hypothesized

to show a strong effect on mathematics achievement in

females. Kloosterman's (1985) study supported these

findings. According to Kloosterman, attributional

variables appear to be more important achievement mediators

for females than for males, as measured by mathematics word

problems. More research is needed in this area.












Mathematics as a male domain. Mathematics is an

activity more closely related to the male sex domain than

to the female sex domain (Eccles et al., 1983). Thus, the

mathematical achievement or boys is higher than that of

girls.

According to John Ernest (1976) in his study

Mathematics and Sex, mathematics is a sexist discipline.

He attributed sex-related differences in mathematical

achievement to the creation by society of sexual

stereotypes and attitudes, restrictions, and constraints

that promote the idea of the superiority of boys in

mathematics. Ernest reported that boys, girls, and

teachers, all believe that boys are superior in

mathematics, at least by the time students reach

adolescence. Bem and Bem (1970) agree and argue that an

American woman is trained to "know her place" in society

because or the pervasive sex-role concept which results in

differential expectations and socialization practices.

Plank and Plank (1954) were more specific. They

discussed two hypotheses related to this view: the

differential cultural reinforcement hypothesis and the

masculine identification hypothesis. The differential

cultural reinforcement hypothesis states that society in

general perceives mathematics as a male domain, giving

females less encouragement for excelling in it. The

masculine identification hypothesis establishes that











achievement and interest in mathematics result from

identification with the masculine role.

A study related to the differential cultural

reinforcement hypothesis is that of Dwyer (1974).

Dwyer examined the relationship between sex role standards

(the extent to which an individual considers certain

activities appropriate to males or females) and

achievement in reading and arithmetic. Students from

grades 2, 4, 6, 8, 10, and 12 participated in this study.

She found that sex role standards contributed significant

variance to reading and arithmetic achievement test scores

and that the effect was stronger for males than for

females. This led to her conclusion that sex-related

differences in reading and arithmetic are more a function

of the child's perception of these areas as sex-appropriate

or sex-inappropriate than of the child's biological sex,

individual preference for masculine and feminine sex roles,

or liking or disliking reading or mathematics.

In a study which agrees with the masculine

identification hypothesis, Milton (1957) found that

individuals who had received strong masculine orientation

performed better in problem solving than individuals who

received less masculine orientation. Elton and Rose (1967)

found that women with high mathematical aptitude and

average verbal aptitude scored higher on the masculinity

scale of the Omnibus Personality Inventory (OPI) than those

with average scores on both tasks.











It is not until adolescence that sex differences in

the perception of mathematics as a male domain are found

(Fennemar 1976; Stein, 1971; Stein & Smithless, 1969;

Verbeker 1983). In a study with 2nd, 6th and 12th

graders, Stein and Smithless (1969) found that students'

perceptions of spatial, mechanical, and arithmetic skills

as masculine became more defined as these students got

older. Fennema (1976) considers that the influence each

sex exerts upon the other on all aspects of behavior is

stronger during adolescence. Since during these years

males stereotype mathematics as a male domain, they send

this message to females who, in turn, tend to be influenced

in their willingness to study or not to study mathematics.

Before that stage, girls consider arithmetic feminine,

while boys consider it appropriate for both sexes (Bobber

1971).

Usefulness of mathematics. Females perceive mathema-

tics as less useful to them than males do, and this

perception occurs at a very young age. As a results

females exert less effort than males to learn or elect to

take advanced mathematics courses.

Many studies reported before 1976 found that the

perception of the usefulness of mathematics for one's

future differs for males and females, and is related to

course taking plans and behavior (Fox, 1977). If females

do not perceive mathematics as useful for their future,











they show less interest in the subject than counterpart

males. These differences in interest are what Hilton and

Berglund (1974) suggest to account for sex-related

differences in mathematics achievement.

Although the perception of the usefulness of

mathematics is still an important predictor of course

taking for girls, there is a growing similarity between

males and females regarding the usefulness of mathematics

(Armstrong & Pricer 1982; Fennema & Sherman, 1977; Moller,

1982/1983). Armstrong and Price investigated the relative

influence of selected factors in sex-related differences in

mathematics participation. Both males and females selected

usefulness of mathematics as the most important factor in

deciding whether or not to take more mathematics in high

school. Moller's study revealed that both males and

females based mathematics course-taking decisions on career

usefulness. A Fennema and Sherman (1977) study showed only

slight differences between males and females in their

feelings about the usefulness of mathematics. In her study

of this variable among college students, Probert

(1983/1984) did not find any sex-related differences

either.

These have been the main affective variables

researched in attempting to explain the underlying causes

of sex-related differences observed in mathematics learning

and achievement. In spite ot the great diversity of studies











dealing with both cognitive and affective variables, there

are no clear-cut findings to render unequivocal support to

a particular variable as accounting for these sex-related

differences. However, everything seems to point to the

fact that affective, rather than cognitive variables play a

more significant role in the sex-related differences

observed in mathematics performance and learning. In most

of the studies dealing with affective variables, findings

consistently show that these factors influence mathematics

performance in females more than in males. In at least one

area confidence as learners of mathematics, Sherman (1980)

found that this variable influenced course election more

than all the cognitive variables previously discussed.

The case for the societal influences on sex roles and

expectations to account for the differences in mathematics

learning is also supported in one way or another in the

studies reported in the literature.



Problem Solving Performance and Related Variables

Problem solving has been perhaps the most extensively

researched area in mathematics education. Published

reviews by Kilpatrick (1969) Riedesel (1969), and Suydam

and Weaver (1970-1975) attest to this. Much of the

research done has focused on identifying the

determinants of problem difficulty and the problem features

that influence the solution process.












At presents no set of variables has been clearly

established as a determinant of problem difficulty.

Several researchers have investigated the effect or reading

and computation on problem solving performance. Others

have studied the effect of student attitudes toward problem

solving in problem solving learning and achievement.

Typically, correlational methods have been used to

investigate these questions.

Computational Skills and Problem Solving Performance

One of the first researchers to study the effect of

computation and reading on problem solving performance was

Hansen (1944). He investigated the relationship of

arithmetical factors, mental factors, and reading factors

to achievement in problem solving. Sixth grade students

were administered tests in problem solving and categorized

as superior achievers (best problem solvers) and inferior

archievers (poorest problem solvers). The two groups were

compared in selected factors believed to be related to

success in arithmetic problem solving: arithmetical, mental

and reading factors. After controlling for mental and

chronological age, the superior achievers in problem

solving surpassed the inferior achievers in mental and

arithmetical factors. The superior group did better in

only two of the six items under the reading factors:

general language ability and the reading of graphs, charts,

and tables.











The findings suggest that reading factors are not as

important as arithmetic and mental factors in problem

solving performance. However, these findings should be

taken cautiously, as the content of the Gates tests (used

to measure reading) is literary and does not include

mathematical material.

Chase (1960) studied 15 variables in an effort to find

out which ones have significant influence on the ability to

solve verbal mathematics problems. Only computation,

reading to note details, and fundamental knowledge were

primarily related to problem solving. Computation

accounted for 20.4% of the 32% variance directly associated

with problem solving.

Chase concluded that a pupil's ability in the

mechanics of computation, comprehension of the principles

that underline the number systems, and the extent to which

important items of information are noticed when reading,

are good predictors of the student's ability in solving

verbal problems.

Balow (1964) investigated the importance of reading

ability and computation ability in problem solving

performance. He objected to the approaches used by other

researchers who in their analyses dichotomized research

subjects as "poor" or "good" students, and who ignored the

recognize effect of intelligence on reading and on

mathematics achievement. Balow administered the Stanford











Achievement Test (subtests of reading, arithmetic, and

reasoning) and the California Short-Form test of mental

ability to a group of 1,400 children from the 6th grade.

All levels of achievement were included in the analysis.

Analysis of variance and covariance were used and compared.

He confirmed the findings of other researchers to the

effect that there is a direct relationship between I.Q. and

reading ability and between I.Q. and computational skills.

The results of the analysis of variance revealed that

increases in computation ability were associated with

higher achievement in problem solving. A relationship

between reading ability and problem solving was also found,

but it was not as strong. Significant differences in

problem solving performance associated with computational

ability were found when intelligence was controlled.

Balow concluded that computation is a much more

important factor in problem solving than reading ability,

and that when I.Q. is taken into consideration, the degree

of the relationship between reading and problem solving

ability becomes less pronounced. Intelligence tends to

confound the relationship between these two variables.

Knifong and Holtan (1976, 1977) attempted to

investigate the types of difficulties children have in

solving word problems. They administered the word problem

section of the Metropolitan Achievement Test to 35 children

from the 6th grade. Errors were classified in two











categories. Category I included clerical and computational

errors. Category II included other types of errors, such

as average and area errors, use of wrong operation, no

response, and erred responses offering no clues. It the

student's work indicated the correct procedure and yet the

problem was missed because of a computational or clerical

errors it was assumed that the problem was read and

understood.

An analysis of frequencies showed that clerical errors

were responsible for 3% of the problems incorrectly solved,

computational errors accounted for 49%, and other errors

for 48% of the erred problems.

Knifong and Holtan concluded that "improved

computational skills could have eliminated nearly half or

the word problem errors" (p. 111). These computational

errors were made in a context where other skills such as

reading, interpretation of the problem, and integration of

these skills necessary for the solution of word problems,

might interact. However, Knifong and Holtan state that

their findings neither confirm nor deny that improvement or

reading skills will lead to improvement in problem solving.

They conclude that "it is difficult to attribute major

importance to reading as a source of failure" (p. ll).

In a later analysis, looking for evidence of poor

reading abilities affecting children's success in word

problems, Knifong and Holtan (1977) interviewed the











children whose errors fell under the category or "other

errors." Students were asked to read each problem aloud

and answer these questions: What kind of situation does

the problem describe? What does the problem ask you to

find? How would you work the problem?

Ninety five percent of the students read the problem

correctly; 98% explained the kind of situation the problem

described in a correct manner; 92% correctly answered what

the problem was asking them to find, and 36% correctly

answered the question of how to work the problem.

The fact that a large percent of the students whose

errors were classified as "other errors" (in which reading

skills might have been a factor) correctly stated how to

work the problem, is strong evidence of their ability to

read and interpret the problems correctly. The errors made

by this group of students had a distinct origin, unrelated

to reading ability.

Zalewski (1974) investigated the relative contribution

of verbal intelligence, reading comprehension, vocabulary,

interpretation of graphs and tables, mathematical concepts,

number sentence selection, and computation to successful

mathematical word problem solution, and the relationship of

the dependent variable to the eight independent variables.

She worked with a group of 4th grade children who

were administered the subtests of the Iowa Test of Basic

Skills (ITBS) and the Weschler Intelligence Scale for











Children (WISC). Multiple regression analysis was

performed. A correlation of .769 was found between word

problem solving and the eight independent variables.

Correlations between word problem solving and the

independent variables ranged from .363 (verbal

intelligence) to .674 (mathematical concepts).

Correlations between the independent variables ranged from

.369 (verbal intelligence and computation) to .749 (reading

comprehension and vocabulary). Mathematical concepts,

computation, and number sentence selection were almost as

effective as all eight independent variables in predicting

achievement in mathematical word problem solving.

Mathematical concepts, computations number sentence

selection, and reading comprehension accounted for 58% of

the variance, whereas all eight predictors accounted for

59% of the variance. The two best predictors were

mathematical concepts and computations which accounted for

54% variance. Other variables accounted for about 40% of

the variance.

The author recommends that the findings of this study

be interpreted cautiously because the correlation between

the eight independent variables was high, and, according to

Zalewski,


in a study of this nature where the interest is
primarily in the influence of several variables
on one dependent variable, a low correlation
between the independent variables is required.
(p. 2804)











In a more recent investigations Exedisis (1983)

studied the contribution of reading ability, vocabulary,

mathematical concepts, computation, sex, and race on

problem-solving performance. The Iowa Test of Basic

Skills was administered to a group of 6th, 7th, and 8th

grade anglo and black Chicago male and female adolescents.

Problem solving was highly correlated to an understanding

of basic mathematical concepts, somewhat correlated to

race, and weakly correlated to computational and

vocabulary skills, sex, and reading ability.

Although the findings of these studies show a

relationship between computational skills and problem

solving achievement, this relationship is not strong enough

to be considered the most determinant factor in problem

solving achievement, as some of the researchers have been

careful to point out. In spite ot the dismissal of reading

as a determinant factor in problem solving achievement by

some of these same researchers, more recent studies in this

area have led others to hold different views.

Reading and Problem Solving Performance

Martin (1964) studied the contribution of reading

comprehension, computation, abstract verbal reasoning, and

arithmetic concepts to arithmetic problem solving

performance. Fourth and 8th grade students were

administered the Iowa tests of Basic Skills and the

Lorge-Thorndlike intelligence test (verbal).











He found that in the 4th grade the correlations

between problem solving and abstract verbal reasoning,

reading comprehension, arithmetic concepts, and computation

were .61, .64, .66, and .60 respectively, and .56, .68,

.69, and .63 in the 8th grade. When computation was held

constant, the correlation between problem solving and

reading was .52 in grade 4 and .54 in grade 8. When

reading was held constant the correlation between problem

solving and computation was .43 in grade 4 and .42 in

grade 8.

Creswell (1982) worked with a sample of anglo and

black adolescents from Chicago. Each subject was

administered the California Achievement test. Multiple

regression was used to analyze the data. The analysis

showed that reading is more important than computation in

predicting student performance in problem solving. Reading

accounted for 49.5% of the variance; computation accounted

for 14.6% of the variance.

Ballew and Cunningham (1982) worked with 6th grade

students in an attempt to find what proportion of students

have as their main source of difficulty with word problems

each of the following factors: a) computation skills,

b) interpretation of the problem, c) reading and,

d) integrating these skills in the solution of problems.

They also wanted to know if a student can be efficiently

diagnosed as having one of the four categories as his/her

main difficulty with mathematics word problems.












Their study is important because it represents an

attempt to demonstrate that multiple factors can interact

in the correct solution of a mathematics word problem.

They constructed three graded tests from a basal

mathematics series for grades 3 through 8. For test 1i the

problems were set in pure computational form (the effects

or reading, interpretation, as well as the necessity for

integration were removed in an effort to measure the

computational skills required by the word problems).

For test 2, the effects of reading and computation

were removed by reading the problems to the students and by

giving scores based on whether or not the students set them

up properly, in an attempt to measure problem

interpretation alone. For test 3, the effect of

computation was removed. The test yielded two scores--one

by grading the students on whether or not they set up the

problems properly and another by grading on the basis of

the correct answer.

The tests were administered to all 244 students from

the 6th grade in two different schools. A diagnostic

profile was obtained for each of the 217 students for which

complete data were available: a computational score, a

problem-interpretation score, a reading score, and a

reading-problem solving score.

They assumed that if the reading-problem

interpretation score was lower (one or more levels lower)











than the problem-interpretation score, the difficulty was

due to reading ability. If the score of the lowest of the

three areas (computation, problem interpretation, and

reading-problem interpretation score) was the same as the

reading-problem solving score, the student's area of

greatest immediate need was either computation, problem

interpretation, or reading. If the reading-problem solving

score was lower than the lowest of the other three scores,

the student's area of greatest immediate need was

integration.

Analysis of the data revealed that for 19% of the

students, problem interpretation was their major

difficulty; for 26% of the students, integration (total

problem solving) was their greatest immediate need; for

another 26%, computation was the major weakness; and for

29%, reading was their greatest immediate need.

Seventy five percent of the students demonstrated

clear strength in computation, 21% in problem

interpretation, and 4% in reading-problem interpretation.

An analysis across all students (including those without

complete data) showed that 26% of the subjects could not

work word problems at a level as high as that at which they

could computer interpret problems, and read and interpret

problems, when those areas were measured separately. This

led them to conclude that knowing the skills or the

components of solving word problems is not sufficient for











success, since the components must be integrated into a

whole process (mastery learning of the components cannot

assure mastery of the process).

Their analysis also led them to conclude that, in the

case of 6th graders, inability to read problems is a major

obstacle in solving word problems. Only 12% ot the

subjects could read and set up problems correctly at a

higher level than they could computer while 60% could

compute correctly at a higher level than they could read

and set up problems; 44% could set up problems better when

they heard them read than when they read the problem

themselves. Only 13% could set up problems better when

they read them than when they heard them read.

Muth (1984) investigated the role of reading and

computational skills in the solution of word problems. A

group of 200 students from the 6th grade were administered

a test of basic skills and a mathematics word problem test.

The word problem test consisted of 15 sample items supplied

by the National Assessment of Educational Progress. The

items were adapted to include some extraneous information

and complex syntactic structure. Four versions of the test

were constructed by combining two versions of problem

information (absence vs. presence of extraneous

information) with two versions of syntactic structure

(simple vs. complex syntax). Task performance was measured

by means of the number of problems answered correctly,











number of problems set up correctly, and amount of time

spent taking the test.

Reading ability and computational ability were both

positively correlated with number of correct answers and

with number of problems correctly set up, and negatively

correlated with test-taking time. Presence of extraneous

information was negatively correlated with correct answers

and correct set ups and positively correlated with

test-taking time. Syntactic complexity was not

significantly correlated with any of the performance

measures.

Results of a multiple regression analysis showed that

reading accounted for 46% of the variance in total correct

answers and computation accounted for 8%. Reading ability

and computational ability uniquely accounted for 14% and 8%

of the variance in the number of correct answers,

respectively. Extraneous information added significantly

to the variance explained in the number of correct answers,

but syntactic structure did not. Reading ability accounted

for 5% of the variance in test-taking time, but computation

did not add significantly to the variance explained by

reading.

Muth concluded that reading and computation both

contribute significantly to success in solving arithmetic

word problems, but that reading plays a more significant

role than does computation.











The studies reviewed in this section show a positive

relationship between reading and problem solving

performance, but in the case of Ballew and Cunningham

(1982) this relationship is not viewed singly but rather

as one among the interacting factors that produce

successful problem solving.

The third variable reviewed is the effect of student

attitudes toward problem solving on problem solving

performance. Many researchers have tried to demonstrate

that this variable is a determinant factor in problem

solving achievement.


Attitudes Toward Problem Solving and Problem Solving
Performance

Research studies support the existence of positive and
rather stable relationships between student attitudes and

achievement in mathematics. Aiken (1970) has suggested

that an individual's attitude toward one aspect of the

discipline (mathematics), such as problem solving, may be

entirely different from his/her attitude toward another

phase of the discipline, such as computation.

Research, however, has been directed to the use of

single, global measures of attitudes toward mathematics

rather than to the investigation of attitudes toward a

particular phase of the discipline.

The studies described below are only part of the few

investigations which have examined the relationship between











student attitudes and performance in the area of problem

solving.

Carey (1958) constructed a scale to measure attitudes

toward problem solving. Her interest was in general

problem solving rather than in mathematical problem

solving. Her work constitutes the first attempt to

construct a measure of attitudes toward problem solving.

The scale was used with a group of college students, and

she found, among other things, that problem solving

performance is positively related to problem solving

attitudes and that, in the case of females, positive

modification of attitudes toward problem solving brings a

significant gain in problem solving performance.

Lindgren, Silvar Faraco, and DaRocha (1964) adapted

Carey's scale of attitudes toward problem solving and

applied it to a group of 4th grade Brazilian children.

Students also answered an arithmetic achievement test, a

general intelligence test, and a socioeconomic (SE) scale.

A low but significant positive correlation was found

between arithmetic achievement and attitudes toward problem

solving. A near zero correlation was found between

attitudes toward problem solving and intelligence. Since

problem solving is one aspect of the discipline of

mathematics, this correlation between attitudes and

arithmetic achievement can lead to a conclusion or a strong












correlation between attitudes toward problem solving and

problem solving performance.

Whitaker (1976) constructed a student attitude scale

to measure some aspects of 4th grade student attitudes

toward mathematic problem solving. He included statements

reflecting children's beliefs about the nature of various

types of mathematical problems, the nature of the problem

solving process, the desirability or persevering when

solving a problem, and the value of generating several

ideas for solving a problem.

He correlated student attitudes toward problem solving

with their scores in a mathematical test which yielded a

comprehension score, an application score, and a problem

solving score. He found a significant positive

relationship between problem solving performance and

student attitude scores on the subscale which measured

reactions to such things as problem solving techniques or

problem situations, or to the frustration or anxiety

experienced when confronted with problem solving

situations.

In another part of this study, Whitaker investigated

the relationship between the attitudes of 4th grade

teachers toward problem solving and their students'

performance in problem solving. A very weak and

nonsignificant negative correlation was found between the

teacher's attitudes toward problem solving and student

performance.











The studies reviewed have confirmed the relationship

between problem solving performance and attitudes toward

problem solving (Carey, 1958; Lindgren et al., 1964;

Whitaker, 1976). However, the results reported in the

studies that investigated the relationship between problem

solving performance and computation and between reading and

problem solving fail to be consistent in their conclusions.

Hansen (1944), Chase (1960) Balow (1964), Knifong and

Holtan (1976, 1977), and Zalewski (1974) concluded that

computation is more strongly related to problem solving

than is reading. Martin (1964), Creswell (1982), Ballew

and Cunningham (1982), and Muth (1984), concluded that

reading ability and mathematical problem solving show a

stronger relationship than computation and problem solving.

Exedisis's (1983) findings led to the conclusion that the

effect or reading and computation in problem solving

performance is unimportant.













CHAPTER III
METHOD



The first objective of this study was to investigate

sex differences in the selection of incorrect responses in

a mathematics multiple-choice achievement test, and to

determine whether these differences were consistent over

three consecutive administrations of the test. The second

objective was to investigate whether males and females

differ in problem solving performance, if these differences

persist after accounting for computational skills, and if

the male-female differences depend on the level of

computational skills. This chapter contains descriptions

of the sample, the test instrument, and the statistical

analysis used in achieving the above mentioned objectives.



The Sample
To achieve the first objective of the study, all the

students who took the Basic Skills Test in Mathematics-6

("Prueba de Destrezas Basicas en Matem&ticas-6") in

each or the three years were included in the study. To

achieve the second objective of the study, approximately











1,000 students were selected randomly for each year. For

the first year, 1,002 were selected (492 boys and

510 girls); for the second year, 1,013 students were

selected (504 boys and 509 girls); and, for the third years

1,013 students were selected (509 boys and 504 girls).

The student population in Puerto Rico includes

children from the urban and rural zones and comprises

children from low and middle socioeconomic levels.

Findings can be generalized only to this population.



The Instrument

The Basic Skills Test in Mathematics-6 is a

criterion-referenced test used in the Department ot

Education of Puerto Rico as part of the annual assessment

program. The test measures academic achievement in

operations, mathematical concepts, and story problems. It

has a reported split half reliability or .95. The test was

designed specifically for Puerto Rico. Its contents and

the procedures followed for its development were formulated

and reviewed by educators from the mathematics department

of the Department of Education of Puerto Rico, in

coordination with the Evaluation Center ot the Department

of Education and mathematics teachers from the school

districts. The emphasis placed on each skill area is

depicted in Figure 3.1.


























































Fig. 3.1 Number of Questions by Skill Area in the Basic
Skills Test in Mathematics 6













Analysis of the Data

Analysis of Sex by Option by Year Cross Classifications



Log-linear models were used to analyze the sex by

option by year cross classifications for each item. Two

topics are addressed in this section, the hypotheses tested

using log-linear models and a comparison of the hypotheses

tested in this study with those tested by Marshall (1981,

1983).

The object of the analysis was to test two hypotheses:

1. The proportion of males and the proportion of

females choosing each incorrect option does not vary from

year to year. Note that this hypothesis is stated in the

null form.

2. Assuming that the first hypothesis is correct, the

proportion of males who choose each incorrect option is

different from the proportion of females who choose each

incorrect option. This hypothesis is stated in the

alternate form.

In Table 3.1 a hypothetical cross classification of

sex, option, and year is presented. Hypothesis 1 is true

for this three dimensional contingency table. In Table 3.2

the three dimensional contingency table is rearranged to

show the year by option contingency table for each gender.












Table 3.1

Hypothetical Probabilities of Option Choice
Conditional on Year and Sex


Option
Year Sex 1 2 3


First M .7 .2 .1
F .5 .3 .2

Second M .7 .2 .1
F .5 .3 .2

Third M .7 .2 .1
F .5 .3 .2



Table 3.2

Hypothetical Probabilities of Option Choice
Conditional on Year and Sex and Arranged by Sex


Option
Sex Year 1 2 3


M First .7 .2 .1
Second .7 .2 .1
Third .7 .2 .1

F First .5 .3 .2
Second .5 .3 .2
Third .5 .3 .2











Inspection of the year by option contingency tables shows

that year and option are independent for each gender.

Thus, hypothesis 1 is equivalent to the hypothesis that

year and option are independent conditional upon gender.

Hypothesis 2 is also true for Table 3.1 and Table 3.2.

Therefore, when hypothesis 1 is correct, hypothesis 2

is equivalent to the hypothesis that sex and option choice

are dependent.

In his discussion of the analysis of three dimensional

contingency tables, Fienberg (1980) presents the following

saturated model for the data:



log mijk = Ui+Uj+Uk+Uij+Uik+Ujk+Uijk. (3.1)


In this model, mijk is the expected value of the

frequency in cell ijk of the three dimensional table. The

model states that all three classification factors for a

three dimensional contingency table are mutually dependent.

In the present research i is the year index, j is the

option index, and k is the sex index. Fienberg shows that

deleting the terms Uij and Uijk yields a model in which

year and option are independent conditional upon sex. This

model is


log mijk = Ui+Uj+Uk+Uik+Ujk.


(3.2)











Fienberg also shows that deleting the Ujk term from (3.2)

to obtain



log mijk= Ui+Uj+Uk+Uik. (3.3)



yields a model that specifies that option is independent of

sex.

Based on Fienberg's presentation, an appropriate

analysis for testing the hypotheses is

1. Conduct a likelihood ratio test of the adequacy of

model (3.2). If this test is nonsignificant, then the

data support the adequacy of the model, and the hypothesis

that conditional on gender, options and year are

independent. Because model (3.1) is a saturated model,

testing the adequacy of model (3.2) is the same as

comparing the adequacies of models (3.1) and (3.2).

2. Conduct a likelihood ratio test comparing .the

adequacies of models (3.2) and (3.3). If this test is

significant, then model (3.2) fits the data better than

model (3.3) and the data support the hypothesis that the

choice of incorrect option is dependent on sex.

To summarized if the first test is nonsignificant and

the second test is significant, then the choice of option

is dependent on sex, and the pattern of dependency is the

same for all three years.











A problem arises in interpreting this analysis with

the data used in this research. Over the three years,

there were responses available from 135,340 students. Even

if only 5% of the students answered an item incorrectly,

the responses of 7,767 students would be used in analyzing

this item. On the other hand, if 90% answered an item

incorrectly, the responses of 121,806 students would be

used in analyzing the data. As a result of the large

sample size the tests described above are likely to be

very powerful. In step 1 of the analysis, then, even a

very small change from year to year in the proportion of

males or females who choose an option is apt to be

detected, and the results will indicate that

hypothesis 1 is not supported. For step 2, even a very

small dependence of option choice on sex is likely to be

detected and hypothesis 2 is likely to be supported. In

brief, the problem caused by the large sample size is that

practically insignificant differences may yield statistical

significance.

Fortunately, the form of the test statistic used in

the likelihood ratio test suggests a reasonable solution to

this problem. The test statistic is


G2 = 2 ZFo loge (Fo/Fe).


(3.4)











Here the summation is over all the cells in the contingency

table Fo refers to the observed frequency in a cell, and

Fe refers to the estimated expected frequency in a cell.

Denoting the observed proportion in a cell by Po and the

estimated expected proportion in a cell by Per the test

statistic can be written


G2 = 2 N TPo loge (Po/Pe)- (3.5)



where N is the total number of subjects. This form of the

test statistic suggests the following strategy. For any

significant G2, using Po and Pe calculated from the

total data set available for an item, calculate the minimum

N required for G2 to be significant. If the minimum N is

very larger this suggests that the statistically

significant result is not practically significant since it

can only be detected in very large samples. Of course, the

question remains as to what can be considered a minimum

large N. Although there is room for argument, it seems

reasonable to claim that if an average of 1000 subjects per

year is required to show significance, then the result is

not likely to be practically significant. On the basis of

this reasoning, it was proposed to ignore all significant

results that would be nonsignificant if there were less

than 3000 subjects available. In addition, all log-linear











model tests were conducted using a .01 level of

significance.

Since this research is based on Marshall (1981, 1983),

it is important to compare the method of analysis used in

this study to the one used by Marshall. Marshall also used

a two-step analysis. In the first step of her analysis she

deleted the Uijk term from (3.1) and tested the adequacy

of the model,


log mij k = Ui+Uj+Uk+Uij+Uik+Ujk. (3.6)


Following this, she deleted the Ujk term to obtain



log mijk = Ui+Uj+Uk+Uij+Uik. (3.7)


and compared these two models using a likelihood ratio

test. If the first test was nonsignificant and the second

significant, Marshall claimed that option choice was

dependent on sex and the pattern of dependency was the same

from year to year.

This is the same claim that this study sought to

establish. However, the approach used here was to

present evidence that model (3.2) fits the data while

Marshall tried to show that model (3.6) fits the data.

The major difference between the two approaches

concerns the operationalization of the concept that the








64



gender-option dependency is the same from year to year. In

this study, a three dimensional table was considered to

exhibit the same year to year pattern of gender-option

dependency if, conditioned on gender, the same proportion

of students chose each incorrect response over the three

year period. This seems to be a straightforward and

natural way to operationalize the concept. To illustrate

how Marshall operationalized the concept in question, a

hypothetical set of probabilities was constructed

conforming to the pattern specified by Marshall. This is

displayed in Table 3.3.




Table 3.3

Hypothetical Joint Probabilities of Year, Option,
and Sex


Option

Year Sex 1 2 3



------------------------
First M .120 .045 .015
F .060 .060 .030
I------------ I
Second M .144 .022 .019
F .072 .030 .039
I----------------------~
Third M .132 .040 .016
F .066 .054 .039



Note: Probabilities reported are truncated to three
places.











One important characteristic of this table involves

two-by-two subtables of sex cross classified with two of

the options for each year. For example, consider the three

two-by-two tables obtained by cross classifying sex and

option choices one and two for each year. These tables are

indicated by dotted lines on Table 3.3. For this table,

the ratio of the odds of a male choosing option one to the

odds of a female choosing option one is the same (within

truncating error) for each year. For example, this odds

ratio for the first year is (.120/.045) / (.060/.060) =

2.67. Within the error caused by reporting truncated

probabilities the odds ratio for years two and three is

the same as that for year one. The equality of these odds

ratios is Marshall's operationalization stage of the year-

to-year gender-option dependency.

To show that the odds ratio can be constant over

years, but that the probabilities of option choice

conditional on sex and year can change from year to year,

for both males and females, the probabilities in Table 3.3

were converted to the probabilities of option choice

conditional upon sex and year. These conditional

probabilities are reported in Table 3.4. Unlike the

probabilities in Table 3.2, those in Table 3.4 change from

year to year for both males and females.









Table 3.4

Hypothetical Probabilities of Option Choice
Conditional on Sex and Year


Option
Year Sex 1 2 3

First M .666 .250 .083
F .400 .400 .200

Second M .774 .120 .104
F .510 .212 .276

Third M .698 .214 .087
F .431 .352 .215


Note: Probabilities reported are truncated to three
places.



Which procedure is more appropriate to test the claim

that the pattern of male and female option choice remains

the same from year to year? It seems more reasonable to

test for a pattern like that in Table 3.2 than to test for

a pattern like that in Table 3.4, and, consequently, this

was the strategy adopted in this study.

Comparison of Males and Females in Problem Solving
Performance

One object of the study was to compare the performance

of males and females on problem solving. Two questions

were addressed. First, do males and females differ in

problem solving performance? Second, do these differences

persist when computational skill is controlled for, and do

these differences depend on the level of computational

skill? Seven analyses of covariance (ANCOVA) tests were











conducted, one with each of addition, subtraction,

multiplication, division, addition of fractions, decimals

in subtraction, and equivalence as covariates. A well

known problem that arises in the use of ANCOVA is that an

unreliable covariate can cause spurious differences between

the sex groups. To solve this problem, Porter (1967/1968)

proposed the use of estimated true scores for observed

scores. Porter (1967/1968) conducted a simulation that

gave empirical support to the adequacy of this strategy.

Hunter and Cohen (1974) have provided theoretical support

for this strategy.












CHAPTER IV
RESULTS


Introduction

The data gathered from 6th grade students from the

public schools in Puerto Rico who took the Basic Skills

Test in Mathematics-6 during three consecutive years were

analyzed in this study. The first objective was to

investigate whether boys and girls differ in the

selection of incorrect responses, and if the pattern of

differences was consistent throughout the three years in

which the test was administered. The second objective was

to investigate whether males and females differ in problem

solving performance, if these differences persist after

accounting for computational skills, and if the male-female

differences depend on the level of computational skills.

The results are discussed in two sections. Study

findings in the area of sex differences in the selection of

incorrect responses are discussed in the first section.

The second section is devoted to an exposition of the

findings in the area of sex-related differences in problem

solving.

Sex-Related Differences in the Selection of
Incorrect Responses

As indicated in Chapter III, there are two models of

interest. The first model indicates that year and option











are independent, conditional upon sex,



log mijk = Ui+Uj+Uk+Uik+Ujk. (3.2)



Substantively, this model implies that the pattern of male

and female option choices is consistent over the three

years of test administration.

The second model indicates that option is independent

of sex,



log mijk = Ui+Uj+Uk+Uik- (3.3)


This model implies that the pattern of option choice is the

same for males and females.

In order to determine if males and females differed in

the selection of incorrect responses and if these

differences were stable across the three years of test

administration, a two-step test was performed. First, a

likelihood ratio test of the adequacy of model (3.2) was

conducted for each of the 111 items of the test. Under

this test a model fits the data when the X2 value

obtained is nonsignificant at a specified alpha level. For

50 (45%) of the items, model (3.2) fitted the data

adequately. For these items the pattern of male and female

option choices was consistent for the three years. The X2

values for the 61 items for which the likelihood ratio











tests were significant at the .01 level are reported in

Table 4.1. Also reported in this table are the actual

sample sizes and the minimum sample sizes necessary for the

likelihood ratio tests to be significant. Of the 61 items,

59 had minimum sample sizes greater than 3,000. Thus,

although for the three years both males and females samples

had inconsistent option choices on these 61 items, on 59 of

the items the inconsistency was relatively minor.

Consequently, these 59 items were included in step 2 along

with the initial 50 items.

In step 2, a likelihood ratio test was performed

comparing the adequacy of the models in (3.2) and (3.3).

The X2 values associated with models (3.2) and (3.3) for

each of the 109 items subjected to step 2 are reported in

columns b and c of Table 4.2. Also reported in Table 4.2

is the difference between the two X2 values (see

column d). This latter figure is the test statistic for

comparing the adequacies of models (3.2) and (3.3).

Significant X2 statistics are indicated by asterisks.

The X2 statistics were significant for 100 of the items

indicating a male-female difference in option choice for

these items. In column e of Table 4.2 the actual sample

sizes are reported. In column f, the minimum sample sizes

necessary for significance are reported. For those 100

items which had significant X2 statistics reported in

column d, 94 (94%) had minimum sample sizes greater than

3,000.











TABLE 4.1

Summary of Significant Tests of Model 3.2


Item Xe for Actual Sample Min. Sample
Number Model 3.2* Size Size for
Significance

(a) (b) (c) (d)


8

13

23

25

28

29

30

31

34

36

37

42

45

48

49

52

54

56

57


46.15

25.74

29.25

35.90

46.00

61.48

25.19

23.26

58.25

84.36

23.70

23.26

22.41

179.25

27.82

26.01

41.41

21.62

28.51


16,796

15,718

60,377

26,707

84,847

88,396

87,473

32,870

82,257

86,137

37,068

59,140

63,098

97,593

16,766

39,858

74,502

44,107

53,075


3,352

5,624

19,011

6,852

16,988

13,242

31,982

13,015

13,006

9,404

14,405

23,417

25,932

5,014

5,550

14,114

16,570

18,789

17,146












TABLE 4.1 Continued


Item Xe for Actual Sample
Number Model 3.2* Size


(a) (hi (


50.96

66.43

21.39

55.03

158.36

88.01

56.28

50.51

34.28

48.01

75.29

44.39

14,425.97

24.61

22.75

99.68

79.38

64.56

23.98


---


Min. Sample
Size for
Significance

(d)

7,560

8,387

15,342

8,485

3,028

7,660

6,943

14,352

13,935

12,128

7,647

14,011

54**

19,969

34,077

6,485

7,693

10,957

18,700


41,829

60,494

35,632

50,699

52,068

73,196

42,429

78,708

51,862

63,223

62,511

67,528

84,368

53,358

84,175

70,183

66,304

76,806

48,690











TABLE 4.1 Continued


Item XZ for Actual Sample Min. Sample
Number Model 3.2* Size Size for
Significance
(a) (b) (c) (d)

80 75.01 55,277 6,787

81 95.51 57,710 5,565

82 51.99 71,315 12,633

83 20.12 81,021 37,088

84 38.62 78,959 18,830

85 1240.78 41,180 306**

86 102.85 39,903 3,573

87 66.75 48,194 6,650

88 63.01 38,742 5,663

89 43.26 39,648 8,441

90 85.53 53,557 5,767

91 82.53 62,954 7,025

93 78.39 80,723 9,484

94 70.15 87,241 11,454

95 40.59 83,969 19,053

96 29.31 74,287 23,343

98 29.87 69,052 21,291

101 23.77 83,782 32,462

102 61.06 75,708 11,419

104 40.47 72,517 16,503












TABLE 4.1 Continued


Item XZ for Actual Sample Min. Sample
Number Model 3.2* Size Size for
Significance
(a) (b) (c) (d)

105 28.66 60,712 19,510

108 26.56 86,261 29,912

111 30.71 86,190 25,849


* p<.Ol

** Minimum sample size less than 3,000











TABLE 4.2


Chi-square Values for the Comparison of Models (3.2) and
(3.3) with Actual Sample Sizes and with the Corresponding
Number of Subjects Needed for Significant Results


Item Xk for X2 for X1 for Actual
Num- Model Model Difference Sample
ber 3.2 3.3 Size


lal

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17


(b)

16.11

17.54

20.06

7.54

14.75

11.86

4.60

46.15

10.13

9.54

5.89

12.58

25.74

11.19

16.63

14.51

17.90


(c)

30.94

103.22

47.51

21.87

70.40

23.13

35.73

329.98

109.99

65.29

19.06

198.46

62.37

80.50

17.33

84.29

44.20


(_d)

14.83*

85.68*

27.45*

14.33*

55.65*

11.27*

31.13*

283.83*

99.86*

55.75*

13.17*

185.88*

36.63*

69.31*

0.70

69.78*

26.30*


(e)

7,427

11,018

11,497

6,016

7,706

13,046

8,832

16,796

23,075

35,060

24,908

40,423

15,718

30,238

42,889

35,114

43,962


Minimum
Sample
Size for
Signif.

(f)

10,061

2,583**

8,733

8,414

2,782**

23,256

5,700

1,189**

4,642

12,634

37,996

4,369

8,621

8,765

1,230,914

10,109

33,582











TABLE 4.2. Continued


Item Xe for X for X2 for Actual Minimum
Num- Model Model Difference Sample Sample
ber 3.2 3.3 Size Size for
Signif.
(a) (b) (c) (d) (e) (f)

18 11.71 15.68 3.97 52,340 264,864

19 9.37 18.77 9.40* 41,363 88,402

20 12.37 13.46 1.09 55,707 1,026,746

21 11.51 19.03 7.52 49,240 131,547

22 9.63 83.74 74.11* 51,724 14,022

23 29.25 153.74 124.49* 60,377 9,744

24 10.35 42.93 32.58* 68,426 42,194

25 35.90 147.40 111.50* 26,707 4,812

26 18.84 47.30 28.46* 56,717 40,037

27 8.78 222.87 214.09* 78,353 7,353

28 46.00 242.78 196.78* 84,847 8,662

29 61.48 170.22 108.74* 88,396 16,331

30 25.19 53.84 28.65* 87,473 61,338

31 23.26 47.42 24.16* 32,870 27,333

32 15.51 41.69 26.18* 48,777 37,430

33 7.57 48.84 41.27* 39,470 19,214

34 58.25 324.67 266.42* 82,257 6,203

35 6.91 225.66 218.75* 88,069 8,088












TABLE 4.2 Continued


Item X2 for a for Xk for Actual Minimum
Num- Model Model Difference Sample Sample
ber 3.2 3.3 Size Size for
Signif.
(a) ( (c) (d) (e) (f)


36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54


84.36

23.70

13.64

13.35

6.67

12.65

23.26

16.29

3.55

22.41

6.56

7.82

179.25

27.82

14.23

16.37

26.01

6.69

41.41


134.56

76.85

84.24

757.60

701.39

603.42

189.11

21.65

115.84

55.90

16.30

76.15

452.64

59.75

240.27

224.54

96.14

27.49

95.41


50.20*

53.15*

70.60*

744.25*

694.72*

590.77*

165.85*

5.36

112.29*

33.49*

9.74*

68.33*

273.39*

31.93*

226.04*

208.17*

70.13*

20.80*

54.00*


mE


86,137

37,068

38,870

61,057

56,438

57,145

59,140

33,712

67,192

13,098

56,727

56,555

97,593

16,766

25,410

24,715

39,858

44,694

74,502


34,472

14,011

11,0618

1,648**

1,630**

1,943**

7,154

126,357

12,021

37,851

117,007

16,628

7,172

10,549

2,258**

2,385**

11,418

43,168

27,718











TABLE 4.2 Continued


Item Xz for Xa for Xl for Actual Minimum
Num- Model Model Difference Sample Sample
ber 3.2 3.3 Size Size for
Signif.
(a) (b) (c) (d) (e) (f)


55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

72

73


90.87

70.14

35.78

51.69

73.68

148.39

394.54

300.95

278.34

246.48

94.11

113.88

180.60

225.46

251.66

271.35

208.90

37.80


163.01 140.26*


71.34*

48.52*

7.27

34.20*

22.72*

81.96*

373.15*

245.92*

119.98*

158.47*

37.83*

63.37*

146.32*

177.45*

176.37*

226.96*

191.49*

13.19*


19.53

21.62

28.51

17.49

50.96

66.43

21.39

55.03

158.36

88.01

56.28

50.51

34.28

48.01

75.29

44.39

17.41

24.61


11,770

18,263

146,668

25,873

36,987

14,828

1,918**

4,142

8,719

9,279

22,532

24,953

7,121

7,158

7,121

5,977

8,831

81,271


41,796

44,107

53,075

44,044

41,829

60,494

35,632

50,699

52,068

73,196

42,429

78,708

51,862

63,223

62,511

67,528

84,175

53,358


74 22.75


84,175 12,057











TABLE 4.2 Continued


Item X1 for X1 for Xa for Actual
Num- Model Model Difference Sample
ber 3.2 3.3 Size


(aL

75

76

77

78

79

80

81

82

83

84

86

87

88

89

90

91

92

93

94


(b)

8.69

99.68

79.38

64.56

23.98

75.01

95.51

51.99

20.12

38.62

102.85

66.75

63.01

43.26

85.53

82.53

16.36

78.39

70-15


(c)

237.07

167.72

432.36

127.04

72.59

128.13

106.82

108.12

95.18

69.05

433.47

430.40

126.27

289.14

405.04

85.48

93.12

229.02

137.27


(d)

228.38*

68.04*

352.98*

62.48*

48.61*

53.12*

11.31*

56.13*

75.06*

30.43*

330.62*

363.65*

63.26*

245.88*

319.51*

2.95

76.76*

150.63*

67.12*


(e)

67,783

70,183

66,304

76,806

48,690

55,277

57,710

71,315

81,021

78,959

39,903

48,194

38,742

39,648

53,557

62,954

88,231

80,723

87,241


Minimum
Sample
Size for
Signif.

(f)

5,963

20,723

3,774

24,696

20,123

20,906

102,511

25,525

21,685

52,129

2,425**

2,662**

12,304

3,240

3,368

428,727

23,092

10,766

26,113











TABLE 4.2 Continued


Item X1 for Xa for X for Actual Minimum
Num- Model Model Difference Sample Sample
ber 3.2 3.3 Size Size for
Signif.
(a) (b) (c) (d) (e) (f)


68.41

212.34

126.96

41.77

9.94

41.48

315.62

437.09

7.97

283.94

48.59

1,682.71

99.20

79.33

18.33

49.11

477.44


27.82*

183.03*

118.36*

11.90*

0.50

30.13*

291.85*

376.03*

1.85

243.47*

19.93*

1,669.34*

86.07*

52.77*

12.60*

31.31*

446.73*


83,969

74,287

50,028

69,052

72,647

60,931

83,782

75,708

50,663

72,517

60,712

71,946

72,482

86,261

76,800

70,980

86,190


60,638

8,154

8,492

116,576

2,918,956

40,627

5,767

4,045

550,173

5,984

61,199

866**

16,918

32,840

112,453

45,544

3,876


* P<.01


** Minimum sample size less than 3,000


95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111


40.59

29.31

8.60

29.87

9.44

11.35

23.77

61.06

6.12

40.47

28.66

13.47

13.13

26.56

5.73

17.80

30.71










Sex-related Differences in Problem
Solving Performance


In this part, results for the two questions related to

the second objective of the study are presented. These

questions serve as the framework for the presentation.

Each question is stated, followed by the results pertaining

to that question.



Question 1: Do males and females differ in problem solving

performance?

The responses of 492 males and 510 females who took

the Puerto Rico Basic Skills Test-6 during the spring of

the first year were analyzed in this study. Also, data

from 504 males and 509 females tested in the second year

and from 509 males and 504 females tested in the third year

were included in the analysis.

The mean performance scores and the standard

deviations for each of the eight variables are presented in

Table 4.3. Results of t-tests are also presented in this

table. Females outperformed males in problem solving, a

finding consistently present in all the three years of test

administration. Over the three-year period the mean

differences favored females in all variables except

equivalence. The sex-related differences in problem

solving were significant (p<.01) for all three years.













TABLE 4.3

Means, Standard Deviations, and t-Tests for the Eight
Mathematical Subtests



Year/ Males Females
Subtest X SD X SD t


First

Problem Solving

Addition

Subtraction

Multi pl i ca ti on

Division

Fracadd

Decsub

Equivalence




Second

Problem Solving

Addition

Subtraction

Multiplication

Division

Fracadd


3.436 2.32

5.412 1.00

4.550 1.66

4.014 1.80

3.136 1.87

2.475 1.75

2.648 1.83

.77 .77

N = 492


3.632

5.470

4.843

4.148

3.242

2.565


2.28

.98

1.52

1.73

1.77

1.74


3.862

5.592

4.852

4.433

3.580

2.743

3.321

.73


2.45

.86

1.56

1.68

1.91

1.81

1.86

.76


2.82*

3.05*

1.22

3.81*

3.72*

2.32**

5.77*

- .83


N = 510


4.015

5.603

4.923

4.550

3.632

2.903


2.36

.86

1.44

1.67

1.86

1.91


2.65*

2.29**

.86

3.76*

3.42*

2.94*













TABLE 4.3 Continued


Year/ Males Females
Subtest X SD X SD t



Decsub 2.863 1.87 3.440 1.80 5.00*

Equivalence .76 .76 .65 .75 -2.32**

N = 504 N = 509
Third

Problem Solving 3.927 2.49 4.341 2.44 2.67*

Addition 5.536 .84 5.704 .64 3.59*

Subtraction 4.836 1.52 4.958 1.46 1.30

Multiplication 4.168 1.74 4.541 1.68 3.47*

Division 3.343 1.88 3.795 1.83 3.87*

Fracadd 2.819 1.89 3.117 1.85 2.53**

Decsub 3.021 1.93 3.448 1.85 3.60*

Equivalance .830 .82 .800 .78 .60

N = 509 N = 504


Note: The number of items in the problem solving subtest was 9.
In each computation subtest, the number of items
was 6. An item was included in the computation subtest
only if it measured a computation skill required to solve
a problem solving item.


* p <.01
** p <.05











Consistent significant differences were also found for

addition, multiplication, division, addition of fractions,

and subtraction of decimals. For subtraction the

difference was not statistically significant.



Question 2: Do sex-related differences in problem solving

persist when computational skills are

controlled for, and is the male-female

differences in problem solving dependent on

level of computational skills?

To address the question of dependence of male-female

problem solving differences in computational skills for

each year and computation subtest, the possibility of an

interaction was investigated. For the first year,

statistically significant interactions were found between

sex and multiplication, F (1,998) = 8.59, p<.01; and sex

and division, F (1,998) = 4.25, p<.01. A significant

interaction was found between sex and subtraction in the

second year, F (1,1009) = 6.39, p<.05. No significant

interactions were found in the third year. Analysis of

covariance summary tables are shown as Tables 4.4, 4.5, and

4.6. Also, the three interactions are depicted in

Figures 4.1, 4.2, and 4.3. Each figure indicates that at

lower levels of computational skills, males outperformed

females in problem solving, with the reverse happening at

higher levels of computational skills.











TABLE 4.4

ANCOVA Summary Table: Multiplication Covariate

First Year


Source df SS MS F




Multiplication (M) 1 1195.90 1195.90 265.66*

Sex (S) 1 31.95 31.95 7.07*

M x S 1 38.80 38.80 8.59*

Error 998 4509.00 4.51



p <.01


TABLE 4.5

ANCOVA Summary Table: Divison Covariate

First Year



Source df SS MS F



Division (D) 1 1195.90 1195.90 264.66*

Sex (S) 1 13.94 13.94 3.22

D x S 1 18.39 18.39 4.25*

Error 998 4317.00 4.32


* p<.01












TABLE 4.6

ANCOVA Summary Table: Subtraction Covariate

Second Year


Source df SS MS F



Subtraction (S) 1 93.14 593.14 122.80*

Sex (S) 1 18.58 18.58 3.85*

SxS 1 30.85 30.85 6.39*

Error 1009 4873.40 4.83


*p <. 5










87



9




8




7




6 .




0
I~ 5


o

,. 4





3




2





1






1 2 3 4 5 6

Multiplication

Fig. 4.1 Sex by Multiplication Interaction.










88




9




8.




7



6




5




4




3




2




1






1 2 3 4 5 6


Division
Fig. 4.2 Sex by Division Interaction










89





9




8.





7.




6





: 5
*r*
O
U)




















1
4
























1 2 3 4 5 6


Subtraction

Fig. 4.3 Sex by Subtraction Interaction











The significant interactions found between sex and

multiplications and sex and division (first year), and sex

and subtraction (second year), answered, in part, the

question of whether male-female differences in problem

solving performance depend on computational ability.

However, the evidence is quite weak. Of 21 possible

interactions, only three were significant. No variable

exhibited a significant interaction for each of the three

years.

The analysis of covariance was also used to determine

if sex-related differences exist after controlling for

computational skills. Analyses were conducted for

those variables that did not exhibit significant

interactions with sex. As discussed in Chapter III,

estimated true scores were used for observed scores to

adjust for unreliability of the covariates (the

computational subtests). Reliability coefficients

calculated for each covariate are shown in Table 4.7.

Summaries of the analyses of covariance for the first

year are reported in Table 4.8. The results show that

females retained their superiority in problem solving

performance when equivalence was the controlling variable

in the analysis of covariancer the only variable in which

males outperformed females (nonsignificant). When the

controlling variables were addition, subtraction, addition

of fractions, and subtraction with decimals, female




Full Text
84
Consistent significant differences were also found for
addition/ multiplication/ division/ addition of fractions/
and subtraction of decimals. For subtraction the
difference was not statistically significant.
Question 2: Do sex-related differences in problem solving
persist when computational skills are
controlled for/ and is the male-female
differences in problem solving dependent on
level of computational skills?
To address the question of dependence of male-female
problem solving differences in computational skills for
each year and computation subtest/ the possibility of an
interaction was investigated. For the first year/
statistically significant interactions were found between
sex and multiplication/ F (1/998) = 8.59/ p and division/ F (1/998) = 4.25 / p interaction was found between sex and subtraction in the
second year/ F (1/1009) = 6.39/ p<.05. No significant
interactions were found in the third year. Analysis of
covariance summary tables are shown as Tables 4.4/ 4.5/ and
4.6. Also/ the three interactions are depicted in
Figures 4.1/ 4.2/ and 4.3. Each figure indicates that at
lower levels of computational skills/ males outperformed
females in problem solving/ with the reverse happening at
higher levels of computational skills.


16
of the test. The sexes differed in only 2 of the 19
reference tests Spatial Relations and Picture Group
Name-Selection. A factor analysis however showed
differences in the number and composition of the factors.
For females a general mathematics factor was determined by
mathematics computation comprehension application and
problem solving. For males the comprehension and
application parts determined one factor; problem solving
with two other reference tests (Gestalt and Omelet)
determined another factor.
Meyer concluded that comprehension of the data and
mastery of the prerequisite mathematical concepts did not
guarantee successful problem solving either for males or
for females. Problem solving scores for both sexes were
about one third their scores in comprehension and one
fourth their scores in application.
She also concluded that the sexes may have approached
the problem solving questions differently. The methods
used by females for solving problem situations may have
paralleled their approach to the application parts.
Males may have used established rules and algorithms for
the application parts but may have used more of a Gestalt
approach to the problem solving situation.
Whitaker (1976) investigated the relationship between
the mathematical problem performance of 4th grade children
and their attitudes toward problem solving their teachers'


96
TABLE 4 .11
Adjusted Means on Problem Solving/
by Covariate and Sex
Second Year
Covariate
Adjusted Means
Males
Females
Addition
3 .69
3 .96
Multiplication
3 .80
3.85
Division
3.80
3.85
Fracadd
3.78
3 .87
Decsub
3.90
3 .75
Equivalence
3.46
44.18


62
Here the summation is over all the cells in the contingency
table/ F0 refers to the observed frequency in a cell/ and
Fe refers to the estimated expected frequency in a cell.
Denoting the observed proportion in a cell by PG and the
estimated expected proportion in a cell by Pe/ the test
statistic can be written
G2 = 2 N £^P0 1 ge (P0/Pe>- (3.5)
where N is the total number of subjects. This form of the
test statistic suggests the following strategy. For any
significant G2/ using P0 and Pe calculated from the
total data set available for an item/ calculate the minimum
N required for G2 to be significant. If the minimum N is
very large/ this suggests that the statistically
significant result is not practically significant since it
can only be detected in very large samples. Of course/ the
question remains as to what can be considered a minimum
large N. Although there is room for argument/ it seems
reasonable to claim that if an average of 1000 subjects per
year is required to show significance/ then the result is
not likely to be practically significant. On the basis of
this reasoning/ it was proposed to ignore all significant
results that would be nonsignificant if there were less
than 3000 subjects available. In addition/ all log-linear


57
Analysis of the Data
Analysis of Sex bv Option by Year Cross Classifications
Log-linear models were used to analyze the sex by
option by year cross classifications for each item. Two
topics are addressed in this section/ the hypotheses tested
using log-linear models and a comparison of the hypotheses
tested in this study with those tested by Marshall (1981/
1983).
The object of the analysis was to test two hypotheses:
1. The proportion of males and the proportion of
females choosing each incorrect option does not vary from
year to year. Note that this hypothesis is stated in the
null form.
2. Assuming that the first hypothesis is correct/ the
proportion of males who choose each incorrect option is
different from the proportion of females who choose each
incorrect option. This hypothesis is stated in the
alternate form.
In Table 3.1 a hypothetical cross classification of
sexz option/ and year is presented. Hypothesis 1 is true
for this three dimensional contingency table. In Table 3.2
the three dimensional contingency table is rearranged to
show the year by option contingency table for each gender.


76
TABLE 4.2. Continued
Item
Num
ber
X2 for
Model
3.2
X2 for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
18
11.71
15.68
3.97
52/3 40
264/864
19
9.37
18.77
9.40*
41/363
88/402
20
12.37
13.46
1.09
55 /707
1/026 /7 46
21
11.51
19.03
7.52
49/240
131/547
22
9.63
83 .74
74.11*
51/7 24
14/022
23
29.25
153.74
124.49*
60/377
9/744
24
10.35
42.93
32.58*
68/426
42/194
25
35.90
147.40
111.50*
26 /707
4/812
26
18.84
47 .30
28.46*
56/717
40 /037
27
8.78
222.87
214.09*
78/353
7/353
28
46.00
242.78
196.78*
84/847
8/662
29
61.48
170.22
108.74*
88/396
16/331
30
25.19
53 .84
28.65*
87/473
61/338
31
23 .26
47.42
24.16*
32/870
27/333
32
15.51
41.69
26.18*
48/777
37/430
33
7.57
48.84
41.27*
39/470
19/214
34
58.25
324.67
266.42*
82 /257
6/203
35
6.91
225 .66
218.75*
88/069
8/088


60
Fienberg also shows that deleting the Ujterm from (3.2)
to obtain
log mijk= Ui+Uj+Ufc+Uik. (3.3)
yields a model that specifies that option is independent of
sex.
Based on Fienberg's presentation/ an appropriate
analysis for testing the hypotheses is
1. Conduct a likelihood ratio test of the adequacy of
model (3.2). If this test is nonsignificant/ then the
data support the adequacy of the modelr and the hypothesis
that conditional on gender option/ and year are
independent. Because model (3.1) is a saturated model/
testing the adequacy of model (3.2) is the same as
comparing the adequacies of models (3.1) and (3.2).
2. Conduct a likelihood ratio test comparing the
adequacies of models (3.2) and (3.3). If this test is
significant/ then model (3.2) fits the data better than
model (3.3) and the data support the hypothesis that the
choice of incorrect option is dependent on sex.
To summarize/ if the first test is nonsignificant and
the second test is significant/ then the choice of option
is dependent on sex and the pattern of dependency is the
same for all three years.


19
solving performance in grade 11. Spatial visualization was
a stronger predictor for girls than for boys. Mathematics
as a male domain was a good predictor for girls only; the
less a girl stereotyped mathematics as a male domain in
grade 8# the higher her problem solving score in grade 11.
Attitudes toward success in mathematics in grade 8 was a
more positive predictor of problem solving performance for
boys than for girls; the more positive the attitudes toward
success in mathematics in grade 8/ the higher their
performance in problem solving in grade 11.
None of these four studies# all of which used the
Romberg-Wearne Mathematics Problem Solving Test# show
statistically significant sex-related differences in
problem solving performance. In later studies other tests
were used to measure this variable (Kaufman# 1984;
Marshall# 1981# 1984).
Kaufman (1983/1984) investigated if sex differences in
problem solving# favoring males# exist in the 5th and 6th
grades and if these differences were more pronounced in
mathematically gifted students than in students of average
mathematical ability. The Iowa Test of Basic Skills and a
mathematics problem-solving test were administered to 504
subjects. Males in the average group as well as males in
the gifted group outperformed females# but only the gifted
group showed statistically significant differences.


85
TABLE 4.4
ANCOVA Summary Table: Multiplication Covariate
First Year
Source
df
SS
MS
F
Multiplication
(M) 1
1195.90
1195.90
265.66*
Sex (S)
1
31.95
31.95
7.07*
M x S
1
38.80
38.80
8.59*
Error
998
4509.00
4.51
* p <;.oi
TABLE 4.5
ANCOVA Summary
Table: Divison Covariate
First Year
Source
df
SS
MS
F
Division (D)
1
1195.90
1195.90
264.66*
Sex (S)
1
13.94
13.94
3.22
D x S
1
18.39
18.39
4.25*
Error
998
4317.00
4.32
* pC.Ol


7
errors (Harnisch# 1983). Item response patterns are
valuable for the identification of large group differences#
including district-to-district # school-to-school# and
classroom-to-classroom variations on different subsets of
items. The response patterns can provide diagnostic
information about the type of understanding the student has
on various mathematics topics (e.g.r problem solving).
Marshall (1981# 1984) has used the item response
pattern technique and her findings indicate sex differences
in mathematics performance at the item level. Females
outperformed males in computation and males outscored
females in problem solving. Also# the success of girls in
the problem solving items was dependent upon their success
in the computation items; for boys# success in the problem
solving items did not depend as much on their computational
performance. Marshall (1981# 1983) has also reported that
males and females differ in the selection of incorrect
responses# reflecting differences in reasoning abilities.
In Puerto Rico# a high percentage of children promoted
to the 7th grade in the public schools does not master the
basic skills in mathematics. If 6th grade male and female
children can be diagnosed as having different problem
solving abilities# as Marshall found with California
children# teachers may need to provide tailor-made
mathematics instruction for each sex# in order to ensure
equal access to formal education and enhance mathematics


11
She analyzed the responses of 9*000 boys and 9*000
girls to 6 selected items* 2 computation items* and 1 story
problem item from each of 2 of the 16 forms of the Survey
of Basic Skills test administered during the academic year
1978-79. The Survey is a 30-item achievement test
administered every year to all 6th grade children in
California through the California Assessment Program.
There are 16 forms of the test* to which approximately
9 *000 boys and 9 *000 girls respond each year. Of the 160
mathematics items contained in the 16 forms of the test* 32
are on measurement and graphing* 28 on number concepts* 28
on whole number arithmetics* 20 on fraction arithmetics* 20
on decimal arithmetics* 20 on geometry* and 12 on
probability and statistics.
The item analysis performed on the 1978-79 data showed
that boys and girls tended to select different incorrect
responses. In the first computation item (Form 1 of the
test) both sexes reflected similar mistakes in carrying*
but in different columns. In the second computation item*
both sexes ignored the decimal points and selected the same
incorrect response. However* more girls than boys chose
this response.
In the first computation item (Form 2 of the test) the
incorrect choice of both sexes was option £* but the
second most frequently selected option was a for boys
and b for girls. In the second computation item of this


21
story problem item correctly after giving incorrect
responses to both computation items. Apparently mastery
of the skills required by the computation items is more
important for girls than for boys. If girls cannot solve
the computation items they have little chance of solving
the related story problem item. For girls the
probability of success in the story problem item after
giving successful answers to both computation items is
almost 2 1/2 times the probability of success after giving
incorrect responses to both computation items. For boys
the probability of success in the story problem item after
successful responses to the computation items is about
1 3/4 times the probability of success on the story problem
item after incorrect responses to the computation items.
Three years later Marshall (1984) analyzed more in
depth these phenomena of sex-related differences. Her
interest was twofold. First she wanted to know if there
were differences in the rate of success for boys and girls
in solving computation and story problem items. Second
she examined additional factors that interact with sex to
influence mathematics performance such as reading
achievement socio-economic status (SES) primary language
and chronological age. Two questions were raised: Do the
probabilibities of successful solving of computation and
story problem items increase with reading score? Are these
probabilities different for the two sexes?


55
1 /000 students were selected randomly for each year. For
the first year / 1 / 00 2 were selected (492 boys and
510 girls); for the second year/ 1/013 students were
selected (504 boys and 509 girls); and/ for the third yearz
1/013 students were selected (509 boys and 504 girls).
The student population in Puerto Rico includes
children from the urban and rural zones and comprises
children from low and middle socioeconomic levels.
Findings can be generalized only to this population.
The Basic Skills Test in Mathematics-6 is a
criterion-referenced test used in the Department ot
Education of Puerto Rico as part of the annual assessment
program. The test measures academic achievement in
operations/ mathematical concepts/ and story problems. It
has a reported split half reliability ot .95. The test was
designed specifically for Puerto Rico. Its contents and
the procedures followed for its development were formulated
and reviewed by educators from the mathematics department
of the Department of Education of Puerto Rico/ in
coordination with the Evaluation Center ot the Department
of Education and mathematics teachers from the school
districts. The emphasis placed on each skill area is
depicted in Figure 3.1.


114
Sherman/ J. (197 9) Predicting mathematics performance in
high school girls and boys. Journal of Educational
Psychology/ 71/ 242-249.
Sherman/ J. (1980) Mathematics/ spatial visualization/
and related factors: Changes in girls and boys/
grades 8-11. Journal of Educational Psychology/
22/ 476-482.
Stein/ A. H. (1971). The effects of sex role standards
for achievement and sex role preference on three
determinants of achievement motiviation.
Developmental Psychology/ 4.* 219-231.
Stein/ A. H./ & Smithless/ J. (1969). Age and sex
differences in children's sex role standards about
achievement/ Developmental Psychology/ 1/ 252-259.
Suydam/ M. N. (1971). Research on mathematics education
(K-12) reported in 1970. Journal for Research in
Mathematics Education/ Z, 257-298.
Suydam/ M. N. (1972). Research on mathematics education
(K-12) reported in 1971. Journal for Research in
Mathematics Education/ 2/ 196-232.
Suydam/ M. N. (197 3) Research on mathematics education
(K-12) reported in 1972. J Mathematics. Education/ / 205-242.
Suydam/ M. N. (1974). Research on mathematics education
(K-12) recorded in 1973. Journal for Research in
Mathematics Education/ 2/ 238-272.
Suydam/ M. N./ & Weaver/ J. F. (1975). Research on
mathematics education (K-12) recorded in 1974.
Journal for Research in Mathematics Education/ £/
253-282.
Verbeke/ K. A. (1983). Sex-related differences in
mathematically gifted secondary students: An
investigation of selected cognitive/ affective/ and
educational factors (Doctoral dissertation/
University of Maryland/ 1982) Dissertation Abstracts
International/ 42/ 2267A-2268A.
Weiner/ B. (1974). Achievement motivation and attribution
theory. Morristown/ N J: General Learning Press.


76
TABLE 4.2. Continued
Item
Num
ber
X2 for
Model
3.2
X2 for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
tb)
(c)
. (d).
(e)
(f)
18
11.71
15.68
3.97
52/340
264,864
19
9.37
18.77
9.40*
41/363
88,402
20
12.37
13.46
1.09
55/707
1/026,746
21
11.51
19.03
7.52
49/240
131/547
22
9.63
83.74
74.11*
51/724
14,022
23
29.25
153.74
124.49*
60/377
9/744
24
10.35
42.93
32.58*
68/426
42/194
25
35.90
147.40
111.50*
26/707
4/812
26
18.84
47.30
28.46*
56/717
40 /037
27
8.78
222.87
214.09*
78,353
7/353
28
46.00
242.78
196.78*
84/847
8,662
29
61.48
170.22
108.74*
88/396
16/331
30
25.19
53.84
28.65*
87/473
61/338
31
23.26
47.42
24.16*
32/870
27,333
32
15.51
41.69
26.18*
48,777
37/430
33
7.57
48.84
41.27*
39,470
19/214
34
58.25
324.67
266.42*
82,257
6/203
35
6.91
225.66
218.75*
88,069
8,088


Log-linear models were used in the analysis of
incorrect responses. The results of the analyses showed
that for 100 of the 111 items of the test* males and
females selected different incorrect optionsr and this
pattern of responses was consistently found during the
three years of test administration. However for the vast
majority of the 100 items the male-female differences were
relatively small considering the fact that the number of
subjects needed to obtain statistical significance was very
large.
The responses of approximately 1000 randomly selected
students per academic year were analyzed in the comparison
of male and female performance in problem solving. Females
outperformed males in problem solving and in six of the
seven computational variables. Males showed superiority in
equivalence in all the three years but statistical
significance was obtained in only one of the years.
Analysis of covariance (ANCOVA) was used in the
comparison of male and female performance in problem
solving after controlling for computational skills. Seven
analyses of covariance tests were conducted one for
each of the covariates. Estimated true scores for observed
scores were used in the analyses. The results tend to show
that for examinees with similar levels of computational
ix


21
story problem item correctly after giving incorrect
responses to both computation items. Apparently* mastery
of the skills required by the computation items is more
important for girls than for boys. If girls cannot solve
the computation items* they have little chance of solving
the related story problem item. For girls* the
probability of success in the story problem item after
giving successful answers to both computation items is
almost 2 1/2 times the probability of success after giving
incorrect responses to both computation items. For boys*
the probability of success in the story problem item after
successful responses to the computation items is about
1 3/4 times the probability of success on the story problem
item after incorrect responses to the computation items.
Three years later* Marshall (1984) analyzed more in
depth these phenomena of sex-related differences. Her
interest was twofold. First* she wanted to know if there
were differences in the rate of success for boys and girls
in solving computation and story problem items. Second*
she examined additional factors that interact with sex to
influence mathematics performance* such as reading
achievement* socio-economic status (SES) primary language*
and chronological age. Two questions were raised: Do the
probabilibities of successful solving of computation and
story problem items increase with reading score? Are these
probabilities different for the two sexes?


37
dealing with both cognitive and affective variables/ there
are no clear-cut findings to render unequivocal support to
a particular variable as accounting for these sex-related
differences. However/ everything seems to point to the
fact that affective/ rather than cognitive variables play a
more significant role in the sex-related differences
observed in mathematics performance and learning. In most
of the studies dealing with affective variables/ findings
consistently show that these factors influence mathematics
performance in females more than in males. In at least one
area/ confidence as learners of mathematics/ Sherman (1980)
found that this variable influenced course election more
than all the cognitive variables previously discussed.
The case for the societal influences on sex roles and
expectations to account for the differences in mathematics
learning is also supported in one way or another in the
studies reported in the literature.
Problem Solving Performance and Related Variables
Problem solving has been perhaps the most extensively
researched area in mathematics education. Published
reviews by Kilpatrick (1969)/ Riedesel (1969)/ and Suydam
and Weaver (1970-1975) attest to this. Much of the
research done has focused on identifying the
determinants of problem difficulty and the problem features
that influence the solution process.


109
Creswell/ J. L. (1982/ February). Sex-related differences
in problem solving in rural black; anglo and Chicago
adolescents. Paper presented at the annual meeting
of the Southwest Educational Research Association.
Austin/ TX. (ERIC Document Reproduction Service
No. ED 216 895)
Dwyer/ C. A. (1974). Influence of children's sex role
standards on reading and arithmentic achievement.
Journal of Educational Psychology/ ££/ 811-816.
Eccles/ J. ( 1983 ) Sex differences in mathematics
participation. In N. Steinkamp & M. Maehr (Eds.)/
Women in Science (pp. 80-100). Greenwich/ CT: JAI
Press.
Elton/ C. F./ & Rose/ H. A. (1967). Traditional sex
attitudes and discrepant ability measures in college
women. jQucpai of CQhhseling Psycology* Hr
538-548.
Ernest/ J. (1976). Mai.hem3tics and Sex. Santa Barbara:
University of California.
Exedisis/ R. H. (1983). An investigation of the relation-
ship of reading comprehension/ vocabulary/
mathematical concepts/ and computation on problem
solving among angloz black/ and Chicago male and
female middle school adolescents (Doctoral
Dissertation/ University of Houston/ 1982).
ni_s.sgrtatieii_Abstracts International/ 41/
2264A-2265A.
Fennema/ E. (1974). Mathematics learning and the sexes:
A review. Journal for Research in Mathematics
-Educfltioji/ 5./ 126-139.
Fennema/ E. (1975). Mathematics/ spatial ability and the
sexes. In E. Fennema/ (Ed.)/ Mathematics: What
r_esga-rch_savs about_sex differences (pp. 33-44) .
Columbus: Ohio State University/ College of Education.
(ERIC Document Reproduction Service No. ED 128-195)
Fennema/ E. (1976) Influences of selected cognitive
afcfcectiahd_educatpn,al_,variables on sex-related
_djj.fergji.ces in.mathematics learning and studying.
Madison: University of Wisconsin/ Department of
Curriculum and Instruction. (ERIC Document
Reproduction Service No. ED 142 906)


100
exception was that/ when equivalence was controlled/
sex-related differences in favor of females/ persisted.
The question of whether male-female differences in
problem solving depend on computational skills was answered
partially/ in the affirmative. Interactions/ indicating
the dependence/ were significant only for multiplication
and division in the first year and for subtraction in the
second year.


COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
SONIA FELICIANO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1986


Ill METHOD 54
The Sample 54
The Instrument 55
Analysis of the Data 57
Analysis of Sex by Option by Year
Cross Classifications 57
Comparison of Males and Females in
Problem Solving Performance 66
IV RESULTS 68
Introduction 68
Sex-related Differences in the
Selection of Incorrect Responses 68
Sex-related Differences in Problem
Solving Performance 81
Summary 93
V DISCUSSION 101
Summary and Interpretation of the
Results 101
Implications of the Findings and
Suggestions for Further Research 103
Sex-related Differences in
Incorrect Responses 104
Sex-related Differences in Problem
Solving Performance 105
REFERENCES 107
BIOGRAPHICAL SKETCH 116
iv


.UL
59
60
61
62
63
64
65
66
67
68
69
70
71
73
74
76
77
78
79
72
TABLE 4.1 Continued
X1 for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(b)
(c)
(d)
50.96
41/829
7/560
66.43
60/494
8/387
21.39
35/632
15/342
55.03
50/699
8/485
158.36
52/068
3/028
88.01
73/196
7/660
56.28
42/429
6/943
50.51
78/708
14/352
34.28
51/862
13/935
48.01
63/223
12/128
75.29
62/511
7/647
44.39
67/528
14/011
14/425 .97
84/368
54**
24.61
53/358
19/969
22.75
84/175
34/077
99.68
70/183
6/485
79.38
66/304
7/693
64.56
76/806
10/957
23.98
48/690
18/700


15
The test is composed of 23 items designed to yield 3
scores: a comprehension score an application score and a
problem solving score. The comprehension question
ascertains whether a child understands the information
given in the item stem. The application question assesses
the child's mastery of a prerequisite concept or skill of
the problem solving question. The problem solving question
poses a question whose solution is not immediately
available that is a situation which does not lend itself
to the immediate application of a rule or algorithm. The
application and problem solving parts of the test may refer
to a common unit of information (the item stem) but the
questions are independent in that the response to the
application question is not used to respond to the problem
solving question.
Meyer (1978) Whitaker (1976) Fennema and Sherman
(1978) and Sherman (1979) have used the Romberg-Wearne
test in their studies. Meyer (1978) investigated whether
males and females differ in problem solving performance and
examined their prerequisite computational skills and
mathematical concepts for the problem solving questions. A
sample of 179 students from the 4th grade were administered
19 "reference tests" for intellectual abilities and the
Romberg-Wearne test. The analysis showed that males and
females were not significantly different in the
comprehension application and problem solving questions


53
Tne studies reviewed have confirmed the relationship
between problem solving performance and attitudes toward
problem solving (Carey 1958; Lindgren et al. 1964;
Whitaker 1976). However the results reportea in the
studies that investigated the relationship between problem
solving performance and computation and between reading and
problem solving fail to be consistent in their conclusions.
Hansen (1944) Chase (I960) Balow (1964) Kmfong and
Holtan (1976 1977) and Zalewski (1974; concluded that
computation is more strongly related to problem solving
than is reading. Martin (1964) Creswell (1982) Ballew
and Cunningham (1982) and Muth (1984) concluded that
reading aoility and mathematical problem solving show a
stronger relationsnip than computation and problem solving.
Exedisis's (1983) findings led to the conclusion that the
effect or reading and computation in problem solving
performance is unimportant.


6
Purpose of the Study
The purpose of this study was twofold. The first was
to investigate sex-related differences in the selection of
incorrect responses in a multiple-choice achievement test
in mathematics. For each test item* the research questions
were as follows:
1. Is there a difference in the proportion of males
and females choosing each incorrect option?
2. Is the same pattern of differences found in data
obtained in three different administrations of the test?
The second objective was to investigate sex-related
differences in test scores in mathematics problem solving.
The following questions were studied:
1. Do males and females differ in problem solving
performance?
2. Do sex differences in problem solving persist
after controlling for computational skills* and does the
differential success of males and females on problem
solving items depend on their success on the computation
items?
significenge Qf the Study
Item response patterns are very useful techniques in
the assessment of mathematics learning and achievement.
Total test scores can be very misleading in the assessment
of student performance and provide no diagnostic
information about the nature and seriousness of student


34
achievement and interest in mathematics result from
identification with the masculine role.
A study related to the differential cultural
reinforcement hypothesis is that of Dwyer (1974).
Dwyer examined the relationship between sex role standards
(the extent to which an individual considers certain
activities appropriate to males or females) and
achievement in reading and arithmetic. Students from
grades 2/ 4* 6/ 8# 10/ and 12 participated in this study.
She found that sex role standards contributed significant
variance to reading and arithmetic achievement test scores
and that the effect was stronger for males than for
females. This led to her conclusion that sex-related
differences in reading and arithmetic are more a function
of the child's perception of these areas as sex-appropriate
or sex-inappropriate than of the child's biological sex#
individual preference for masculine and feminine sex roles/
or liking or disliking reading or mathematics.
In a study which agrees with the masculine
identification hypothesis/ Milton (1957) found that
individuals who had received strong masculine orientation
performed better in problem solving than individuals who
received less masculine orientation. Elton and Rose (1967)
found that women with high mathematical aptitude and
average verbal aptitude scored higher on the masculinity
scale of the Omnibus Personality Inventory (OPI) than those
with average scores on both tasks.


4
mathematics preparation. Males who were enrolled or had
completed algebra II outperformed the females in
computation and problem solving but not in algebra. Males
who studied beyond algebra II outscored females on all
three subtests: computation/ algebra/ and problem solving
(Armstrong/ 1979) .
Carpenter/ Lindquist/ Mathews/ and Silver (1984)
analyzed the results of the Third National Assessment of
Educational Progress (NAEP)/ and compared them with the
First and Second Surveys. Between 1978 and 1982/ the
differences between the average performance of males and
females remained stable at each age level. At ages 9 and
13/ the overall performance of males and females was not
significantly different. At age 17/ males scored higher
than females by about 3 percentage points. When course
background was held constant/ achievement differences still
existed at age 17. For each category of course background/
male achievement exceeded female achievement. Consistent
with previous assessments/ sex differences in problem
solving in favor of males were found for the 17-year-old
sample. At ages 9 and 13/ no large differences were found
between the sexes within any level of course background.
Marshall (1981/ 1984) investigated sex differences in
mathematics performance. She found that males and females
excel each other in solving different types of problems.


80
TABLE 4.2 Continued
Item
Num
ber
X2 for
Model
3.2
X2 for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
95
40 .59
68.41
27 .82*
83 /969
60/638
96
29.31
212.34
183.03*
74/287
8/154
97
8.60
126.96
118.36*
50/028
8,492
98
29.87
41.77
11.90*
69/052
116/576
99
9.44
9.94
0.50
72/647
2/918/956
100
11.35
41.48
30.13*
60/931
40 /627
101
23.77
315.62
291.85*
83 /782
5/767
102
61.06
437.09
376.03*
75/708
4/045
103
6.12
7.97
1.85
50/663
550/173
104
40 .47
283.94
243 .47*
72/517
5/984
105
28.66
48.59
19.93*
60/712
61/199
106
13 .47
1/682 .71
1/669.34*
71/946
866*
107
13.13
99.20
86.07*
72/482
16 ,918
108
26 .56
79.33
52.77*
86/261
32,840
109
5.73
18.33
12.60*
76/800
112/453
110
17.80
49.11
31.31*
70/980
45 ,544
111
30.71
477.44
446 .73*
86/190
3/876
pC.oi
Minimum sample size less than 3/000


30
The issue of sex-related differences in spatial
visualization ability as an explanation for sex differences
observed in mathematics achievement is less convincing and
the findings more contradictory than in the issue of sex
differences in formal education. Besides these cognitive
issues/ other issues/ mostly affective in nature/ have also
been studied in trying to explain the origin of these sex
differences in mathematics achievement and learning. The
studies dealing with these affective variables are reviewed
in the next section.
Differentiated Effect of Affective Variables
Researchers have attempted to explain the effect of
sex differences in internal beliefs/ interests/ and
attitudes (affective variables) on mathematics learning and
achievement. A brief statement of each explanation
precedes the summary of studies conducted that support the
explanation.
than males/ lack confidence in their ability to learn
mathematics and this affects their achievement in
mathematics and their election of more advanced mathematics
courses.
Maccoby and Jacklin (1974) reported that self-
confidence in terms of grade expectancy and success in
particular tasks was found to be consistently lower in
women than in men. In 1978/ Fennema and Sherman reported


8
achievement. Since there are no investigations reported in
sex differences in item response patterns in Puerto Rico*
research is needed.
Organization of the Study
A review of the literature on sex differences in
mathematics performance is reported in Chapter II. The
research methodology is presented in Chapter III. Research
questions sample instrument and data analysis are
discussed in that chapter. Chapter IV is an exposition of
the results of the study. Chapter V contains a summary and
interpretation of the results of the study and the
implications of the findings together with suggestions for
further research.


LIST OF FIGURES
Figure Page
3.1 Number of Questions by Skill Area in a
Basic Skills Test in Mathematics-6 56
4.1 Sex by Multiplication Interaction 87
4.2 Sex by Division Interaction 88
4.3 Sex by Subtraction Interaction 89
vii


29
females may use them in a different way. Meyer (1978) ,
with an elementary grade sample/ and Pennema and Tartree
(1983) / with an intermediate level sampler found that the
influence of spatial visualization on solving mathematics
problems is subtler and that males and females use their
spatial skills differently in solving word story problems
(problems that measure problem solving ability or
reasoning) Fennema and Tartree (1983) carried out a
three-year longitudinal study which showed that girls and
boys with equivalent spatial visualization skills did not
solve the same number of items/ nor did they use the same
processes in solving problems. The results also suggested
that a low level of spatial visualization skills was a more
debilitating factor for girls than for boys in problem
solving performance.
Landau (1984) also investigated the relationship
between spatial visualization and mathematics achievement.
She studied the performance of middle school children in
mathematical problems of varying difficulty/ and the extent
to which a diagramatic representation is likely to
facilitate solution. She found that spatial ability was
strongly correlated to mathematical problem solving and
that the effect of spatial ability was more influential for
females. Females made more use of diagrams in the solution
of problems/ reducing the advantage of males over females
in problem solving performance.


lal
8
13
23
25
28
29
30
31
34
36
37
42
45
48
49
52
54
56
57
71
TABLE 4 .1
Summary of Significant Tests of Model 3.2
X1 for
odel 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(b)
(c)
(d)
46 .15
16,796
3,352
25.74
15,718
5,624
29.25
60,377
19,011
35.90
26 ,707
6,852
46 .00
84 ,847
16,988
61.48
88,396
13,242
25.19
87 ,473
31,982
23 .26
32,870
13 ,015
58.25
82 ,257
13 ,006
84.36
86 ,137
9,404
23 .70
37,068
14,405
23 .26
59,140
23 ,417
22.41
63,098
25,932
179.25
97 ,593
5,014
27.82
16,766
5,550
26.01
39,858
14,114
41.41
74,502
16,570
21.62
44,107
18,789
28.51
53 ,075
17 ,146


69
are independent/ conditional upon sex,
log m i j k = Ui + Uj + Uk+Uik+Uj k* (3.2)
Substantively/ this model implies that the pattern of male
and female option choices is consistent over the three
years of test administration.
The second model indicates that option is independent
of sex/
log mj.jk = u i+ u j + uk+ ui k (3.3)
This model implies that the pattern of option choice is the
same for males and females.
In order to determine if males and females differed in
the selection of incorrect responses and if these
differences were stable across the three years of test
administration/ a two-step test was performed. First/ a
likelihood ratio test of the adequacy of model (3.2) was
conducted for each of the 111 items of the test. Under
this test a model fits the data when the value
obtained is nonsignificant at a specified alpha level. For
50 (45%) of the items/ model (3.2) fitted the data
adequately. For these items the pattern of male and female
option choices was consistent for the three years. The
values for the 61 items for which the likelihood ratio


109
Creswell/ J. L. (1982/ February). Sex-related differences
i..n_problem solving in rural black/ anglo and Chicago
adolescents. Paper presented at the annual meeting
of the Southwest Educational Research Association.
Austin/ TX. (ERIC Document Reproduction Service
No. ED 216 895)
Dwyer/ C. A. ( 197 4) Influence of children's sex role
standards on reading and arithmentic achievement.
Journal of Educational Psychology/ ££/ 811-816.
Eccles/ J. ( 1 9 83 ) Sex differences in mathematics
participation. In M. Steinkamp & M. Maehr (Eds.)/
Women in Science (pp. 80-100). Greenwich/ CT: JAI
Press.
Elton/ C. F. / & Rose/ H. A. ( 1967). Traditional sex
attitudes and discrepant ability measures in college
women. Journal of Counseling Psycology/ M*
538-548.
Ernest/ J. (1976). Mathematics and Sex. Santa Barbara:
University of California.
Exedisisz R. H. (1983). An investigation of the relation
ship of reading comprehension/ vocabulary/
mathematical concepts/ and computation on problem
solving among anglo/ black/ and Chicago male and
female middle school adolescents (Doctoral
Dissertation/ University of Houston/ 1982).
Dissertation,Abstracts International/ 42.,
2264A-2265A.
Fennema/ E. (1974). Mathematics learning and the sexes:
A review. Journal for Research in Mathematics
Education/ / 126-139.
Fennema/ E. (1975). Mathematics/ spatial ability and the
sexes. In E. Fennema/ (Ed.)/ Mathematics: What
research savs about sex differences (pp. 33-44).
Columbus: Ohio State University/ College of Education.
(ERIC Document Reproduction Service No. ED 128-195)
Fennema/ E. (1976) Influences of selected cognitive
affective/ and educational variables on sex-related
djff_exgnc.e.s^,n_mathematics learning and studying.
Madison: University of Wisconsin/ Department of
Curriculum and Instruction. (ERIC Document
Reproduction Service No. ED 142 906)


93
superiority in problem solving lost statistical
significance. Adjusted means, associated with the
analyses, are reported in Table 4.9.
Summaries of the analyses of covariance for the second
year are reported in Table 4.10. The analyses show that
females retained their superiority in problem solving after
controlling for equivalence. When the controlling
variables were addition, multiplication, division, addition
of fractions, and subtraction with decimals females
superiority in problem solving lost statistical
significance. Adjusted means are reported in Table 4.11.
The analyses of covariance for the third year are
summarized in Table 4.12. The results of the analyses show
that females maintained their superiority in problem
solving when performance on either subtraction or
equivalence was controlled. When addition, multiplication,
division, addition of fractions, and subtraction of
decimals were the controlling variables, female superiority
in problem solving lost statistical significance. Adjusted
means are reported in Table 4.13.
Summary
Sex-related differences in mathematics achievement
were the subject of this study. The first objective was to
investigate whether males and females differ in the
selection of incorrect responses, and if the pattern of


32
attributes success to an internal, stable attribute, such
as ability, then one is confident of being successful in
the future and will continue to strive in that area. If
one attributes success to an external factor such as a
teacher, or to an unstable one, such as effort, then one
will not be as confident of success in the future and will
cease to strive. Failure attribution patterns work this
way: if failure is attributed to unstable causes, such as
effort, failure can be avoided in the future and the
tendency will be to persist in the task. However, if
failure is attributed to a stable cause, such as ability,
the belief that one cannot avoid failure will remain.
Studies reported by Bar-Tal and Frieze (1977) suggest
that males and females tend to exhibit different
attributional patterns of success and failure. Males tend
to attribute their success to internal causes and their
failures to external or unstable ones. Females show a
different pattern; they tend to attribute success to
external or unstable causes and failures to internal ones.
The pattern of attributions, success attributed externally
and failure attributed internally, has become hypothesized
to show a strong effect on mathematics achievement in
females. Kloosterman's (1985) study supported these
findings. According to Kloosterman, attributional
variables appear to be more important achievement mediators
for females than for males, as measured by mathematics word
problems. More research is needed in this area.


50
The studies reviewed in this section show a positive
relationship between reading and problem solving
performance* but in the case of Ballew and Cunningham
(1982)* this relationship is not viewed singly but rather
as one among the interacting factors that produce
successful problem solving.
The third variable reviewed is the effect of student
attitudes toward problem solving on problem solving
performance. Many researchers have tried to demonstrate
that this variable is a determinant factor in problem
solving achievement.
Attitudes. Toward, Problem Solving and Problem Solving
performance
Research studies support the existence of positive and
rather stable relationships between student attitudes and
achievement in mathematics. Aiken (1970) has suggested
that an individual's attitude toward one aspect of the
discipline (mathematics) such as problem solving* may be
entirely different from his/her attitude toward another
phase of the discipline* such as computation.
Research* however* has been directed to the use of
single* global measures of attitudes toward mathematics
rather than to the investigation of attitudes toward a
particular phase of the discipline.
The studies described below are only part of the few
investigations which have examined the relationship between


45
He found that in the 4th grade the correlations
between problem solving and abstract verbal reasoning
reading comprehension/ arithmetic concepts/ and computation
were .61/ .64/ .66/ and .60 respectively/ and .56/ .68/
.69/ and .63 in the 8th grade. When computation was held
constant/ the correlation between problem solving and
reading was .52 in grade 4 and .54 in grade 8. When
reading was held constant the correlation between problem
solving and computation was .43 in grade 4 and .42 in
grade 8.
Creswell (1982) worked with a sample of anglo and
black adolescents from Chicago. Each subject was
administered the California Achievement test. Multiple
regression was used to analyze the data. The analysis
showed that reading is more important than computation in
predicting student performance in problem solving. Reading
accounted for 49.5% of the variance; computation accounted
for 14.6% of the variance.
Ballew and Cunningham (1982) worked with 6th grade
students in an attempt to find what proportion of students
have as their main source of difficulty with word problems
each of the following factors: a) computation skills/
b) interpretation of the problem/ c) reading and/
d) integrating these skills in the solution of problems.
They also wanted to know if a student can be efficiently
diagnosed as having one of the four categories as his/her
main difficulty with mathematics word problems.


49
number of problems set up correctly* and amount of time
spent taking the test.
Reading ability and computational ability were both
positively correlated with number of correct answers and
with number of problems correctly set up* and negatively
correlated with test-taking time. Presence of extraneous
information was negatively correlated with correct answers
and correct set ups and positively correlated with
test-taking time. Syntactic complexity was not
significantly correlated with any of the performance
measures.
Results of a multiple regression analysis showed that
reading accounted for 46% of the variance in total correct
answers and computation accounted for 8%. Reading ability
and computational ability uniquely accounted for 14% and 8%
of the variance in the number of correct answers*
respectively. Extraneous information added significantly
to the variance explained in the number of correct answers*
but syntactic structure did not. Reading ability accounted
for 5% of the variance in test-taking time* but computation
did not add significantly to the variance explained by
reading.
Muth concluded that reading and computation both
contribute significantly to success in solving arithmetic
word problems* but that reading plays a more significant
role than does computation.


COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
SONIA FELICIANO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1986


36
they show less interest in the subject than counterpart
males. These differences in interest are what Hilton and
Berglund (1974) suggest to account for sex-related
differences in mathematics achievement.
Although the perception of the usefulness of
mathematics is still an important predictor of course
taking for girls* there is a growing similarity between
males and females regarding the usefulness of mathematics
(Armstrong & Price* 1982; Fennema & Sherman* 1977; Moller*
1982/1983). Armstrong and Price investigated the relative
influence of selected factors in sex-related differences in
mathematics participation. Both males and females selected
usefulness of mathematics as the most important factor in
deciaing whether or not to take more mathematics in high
school. Moller's study revealed that both males and
females based mathematics course-taking decisions on career
usefulness. A Fennema and Sherman (1977) study showed only
slight differences between males and females in their
feelings about the usefulness of mathematics. In her study
of this variable among college students* Probert
(1983/1984) did not find any sex-related differences
either.
These have been the main affective variables
researched in attempting to explain the underlying causes
of sex-related differences observed in mathematics learning
and achievement. In spite of the great diversity of studies


39
The findings suggest that reading factors are not as
important as arithmetic and mental factors in problem
solving performance. However these findings should be
taken cautiously* as the content of the Gates tests (used
to measure reading) is literary and does not include
mathematical material.
Chase (1960) studied 15 variables in an effort to find
out which ones have significant influence on the aDility to
solve verbal mathematics problems. Only computation*
reading to note details* and fundamental knowledge were
primarily related to problem solving. Computation
accounted for 20.4% of the 32% variance directly associated
with problem solving.
Chase concluded that a pupil's ability in the
mechanics of computation* comprehension of the principles
that underline the number systems* and the extent to which
important items of information are noticed when reading*
are good predictors of the student's ability in solving
verbal problems.
Balow (1964) investigated the importance of reading
ability and computation ability in problem solving
performance. He objected to the approaches used by other
researchers who in their analyses dichotomized research
subjects as "poor" or "good" students* and who ignored the
recognizea effect of intelligence on reading and on
mathematics achievement
Balow administered the Stanford


33
Mathematics as a male domain. Mathematics is an
activity more closely related to the male sex domain than
to the female sex domain (Eccles et al.r 1983). Thus* the
mathematical achievement ot boys is higher than that of
gi rls.
According to John Ernest (1976) in his study
Mathematics and Sex> mathematics is a sexist discipline.
He attributed sex-related differences in mathematical
achievement to the creation by society of sexual
stereotypes and attitudes* restrictions* and constraints
that promote the idea of the superiority of boys in
mathematics. Ernest reported that boys* girls* and
teachers* all believe that boys are superior in
mathematics* at least by the time students reach
adolescence. Bern and Bern (1970) agree and argue that an
American woman is trained to "know her place" in society
because ot the pervasive sex-role concept which results in
differential expectations and socialization practices.
Plank and Plank (195 4) were more specific. They
discussed two hypotheses related to this view: the
differential cultural reinforcement hypothesis and the
masculine identification hypothesis. The differential
cultural reinforcement hypothesis states that society in
general perceives mathematics as a male domain* giving
females less encouragement for excelling in it. The
masculine identification hypothesis establishes that


80
81
82
83
84
85
86
87
88
89
90
91
93
94
95
96
98
101
102
104
73
TABLE 4.1 Continued
x* for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
lb)
(c)
(d)
75.01
55,277
6,787
95.51
57 *710
5,565
51.99
71/315
12/633
20.12
81/021
37/088
38.62
78/959
18,830
1240.78
41,180
306**
102.85
39/903
3/573
66.75
48,194
6,650
63.01
38/742
5,663
43.26
39,648
8,441
85.53
53/557
5/767
82.53
62,954
7,025
78.39
80/723
9,484
70.15
87,241
11,454
40.59
83,969
19,053
29.31
74,287
23/343
29.87
69/052
21,291
23.77
83,782
32,462
61.06
75/708
11,419
40.47
72,517
16/503


70
tests were significant at the .01 level are reported in
Table 4.1. Also reported in this table are the actual
sample sizes and the minimum sample sizes necessary for the
likelihood ratio tests to be significant. Of the 61 items*
59 had minimum sample sizes greater than 3*000. Thus*
although for the three years both males and females samples
had inconsistent option choices on these 61 items* on 59 of
the items the inconsistency was relatively minor.
Consequently* these 59 items were included in step 2 along
with the initial 50 items.
In step 2* a likelihood ratio test was performed
comparing the adequacy of the models in (3.2) and (3.3).
The values associated with models (3.2) and (3.3) for
each of the 109 items subjected to step 2 are reported in
columns b and c of Table 4.2. Also reported in Table 4.2
is the difference between the two values (see
column d). This latter figure is the test statistic for
comparing the adequacies of models (3.2) and (3.3).
Significant X^ statistics are indicated by asterisks.
The X^ statistics were significant for 100 of the items
indicating a male-female difference in option choice for
these items. In column e of Table 4.2 the actual sample
sizes are reported. In column f* the minimum sample sizes
necessary for significance are reported. For those 100
items which had significant X^ statistics reported in
column d* 94 (94%) had minimum sample sizes greater than
3*000.


CHAPTER I
INTRODUCTION
Sex differences in mathematics learning and
achievement have been the subject of intensive research.
Research done before 1974 has shown that male performance
on mathematical achievement tests is superior to female
performance by the time they reach upper elementary or
junior high school (Fennema# 1976# p. 2). The literature
strongly suggests that at the elementary level females
outperform males in computation and males excel in
mathematical reasoning (Glennon & Callahan# 1968; Jarvis#
1964; Maccoby# 1966).
Since 1974# research findings have been less
consistent. Fennema (1974)# after reviewing 36 studies#
found that during secondary school or earlier# sex-related
differences in mathematics achievement are not so evident#
but that when differences are found# they favor males in
high level cognitive tasks (problem solving) and females in
low level cognitive tasks (computation). As a result of a
further review of the literature# Fennema (1977) concluded
that at the elementary level# sex-related differences do
not exist at all cognitive levels# from computation to
problem solving.
1


COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
SONIA FELICIANO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1986

ACKNOWLEDGMENTS
I would like to acknowledge my gratitude to several
people who have influenced my formal education and/or made
this study possible. My special thanks to Dr. James
Alginar Chairman of the doctoral committee/ who contributed
to the development of my love for research and statistics.
He has been insuperable as professor and valued friend; his
help and guidance in the preparation and completion of this
study were invaluable. I extend my thanks to Dr. Linda
Crocker for her advice and help during my doctoral studies
at University of Florida. Thanks also go to Dr. Michael
Nunnery/ member of the doctoral committee. To Dr. Wilson
Guertin* who was a friend for me and my family/ I extend my
special thanks. Thanks are also extended to Dr. Amalia
Charneco/ past Undersecretary of Education of the Puerto
Rico Department of Education/ for her continuous support.
To my sister Nilda Santaella/ who typed the thesis/ I give
my sincere thanks. Special thanks go to my family and to
those friends who provided encouragement throughout this
critical period of my life.
ii

TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS Ü
LIST OF TABLES V
LIST OF FIGURES VÜ
ABSTRACT viii
CHAPTER
I INTRODUCTION 1
Purpose of the Study 6
Significance of the Study 6
Organization of the Study 8
II REVIEW OF THE LITERATURE 9
Sex-related Differences in Incorrect
Response Patterns 10
Sex-related Differences in Problem
Solving 14
Cognitive and Affective Variables that
Influence Sex Differences in Mathematics
Learning and Achievement 23
Differences in Formal Mathematics
Education 24
Differences in Spatial Ability 26
Differentiated Effect of Affective
Variables 30
Problem Solving Performance and Related
Variables 37
Computational Skills and Problem Solving
Performance 38
Reading and Problem Solving Perfor¬
mance 44
Attitudes Toward Problem Solving and
Problem Solving Performance 50
iii

IllMETHOD
54
The Sample 54
The Instrument 55
Analysis of the Data 57
Analysis of Sex by Option by Year
Cross Classifications 57
Comparison of Males and Females in
Problem Solving Performance 66
IV RESULTS 6 8
Introduction 68
Sex-related Differences in the
Selection of Incorrect Responses 68
Sex-related Differences in Problem
Solving Performance 81
Summary 93
V DISCUSSION 101
Summary and Interpretation of the
Results 101
Implications of the Findings and
Suggestions for Further Research 103
Sex-related Differences in
Incorrect Responses 104
Sex-related Differences in Problem
Solving Performance 105
REFERENCES 107
BIOGRAPHICAL SKETCH 116
IV

LIST OF TABLES
Table Page
3.1 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex 58
3.2 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex and
Arranged by Sex 58
3.3 Hypothetical Joint Probabilities of Year/
Option/ and Sex 64
3.4 Hypothetical Probabilities of Option
Choice Conditional on Sex and Year 66
4.1 Summary of Significant Tests of Model 3.2 71
4.2 Chi-square Values for the Comparison of 75
Models (3.2) and (3.3) with Actual Sample
Sizes and with the Corresponding Number
of Subjects Needed for Significant Results 75
4.3 Means/ Standard Deviations/ and t-Test for
the Eight Mathematical Variables 82
4.4 ANCOVA - First Year: Multiplication Covariate 85
4.5 ANCOVA - First Year: Division Covariate 85
4.6 ANCOVA - Second Year: Subtraction Covariate... 86
4.7 Reliability of the Covariates for Each
of the Three Years of Test Administra¬
tion 91
4.8 ANCOVA - First Year: Other Covariates 92
4.9 Adjusted Means on Problem Solving/ by
Covariate and Sex 94
4.10 ANCOVA - Second Year: Other Covariates 95
v

Table Page
4.11 Adjusted Means on Problem Solving*
by Covariate and Sex 96
4.12 ANCOVA - Third Year 97
4.13 Adjusted Means on Problem Solving*
by Covariate and Sex 98
vi

LIST OF FIGURES
Figure Page
3.1 Number of Questions by Skill Area in a
Basic Skills Test in Mathematics-6 56
4.1 Sex by Multiplication Interaction 87
4.2 Sex by Division Interaction 88
4.3 Sex by Subtraction Interaction 89
vii

Abstract of Dissertation Presented to the Graduate
School of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy
COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
Sonia Feliciano
August/ 1986
Chairman: James Algina
Major Department: Foundations of Education
The first objective of this study was to investigate
sex differences in the selection of incorrect responses on
a mathematics multiple-choice test, and to determine
whether these differences were consistent over three
consecutive administrations of the test. A second
objective was to compare male and female performance in
problem solving after controlling for computational skills.
The responses of all 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 ("Prueba de Destrezas Básicas en
Matemáticas-6") during three academic years were used in
the analyses relevant to the first objective.
VXll

Log-linear models were used in the analysis of
incorrect responses. The results of the analyses showed
that for 100 of the 111 items of the testr males and
females selected different incorrect options/ and this
pattern of responses was consistently found during the
three years of test administration. However/ for the vast
majority of the 100 items the male-female differences were
relatively small/ considering the fact that the number of
subjects needed to obtain statistical significance was very
large.
The responses of approximately 1/000 randomly selected
students per academic year were analyzed in the comparison
of male and female performance in problem solving. Females
outperformed males in problem solving and in six of the
seven computational variables. Males showed superiority in
equivalence in all the three years/ but statistical
significance was obtained in only one of the years.
Analysis of covariance (ANCOVA) was used in the
comparison of male and female performance in problem
solving after controlling for computational skills. Seven
analyses of covariance tests were conducted/ one for
each of the covariates. Estimated true scores for observed
scores were used in the analyses. The results tend to show
that for examinees with similar levels of computational
IX

skills/ sex-related differences in problem solving
performance do not exist. Females retained their
superiority in problem solving when equivalence (in all
three years) and subtraction (in one year) were the
controlling variables.
The question of whether male-female differences in
problem solving depend on computational skills was
answered/ partially/ in the affirmative.
x

CHAPTER I
INTRODUCTION
Sex differences in mathematics learning and
achievement have been the subject of intensive research.
Research done before 1974 has shown that male performance
on mathematical achievement tests is superior to female
performance by the time they reach upper elementary or
junior high school (Fennema/ 1976/ p. 2). The literature
strongly suggests that at the elementary level females
outperform males in computation and males excel in
mathematical reasoning (Glennon & Callahan/ 1968; Jarvis/
1964; Maccoby/ 1966).
Since 1974/ research findings have been less
consistent. Fennema (1974)/ after reviewing 36 studies/
found that during secondary school or earlier/ sex-related
differences in mathematics achievement are not so evident/
but that when differences are found/ they favor males in
high level cognitive tasks (problem solving) and females in
low level cognitive tasks (computation) . As a result of a
further review of the literature/ Fennema (1977) concluded
that at the elementary level/ sex-related differences do
not exist at all cognitive levels/ from computation to
problem solving.
1

2
Many variables* cognitive* affective* and educational*
have been investigated since 1974 in relation to sex
differences in mathematics learning and achievement.
Fennema and Sherman (1977) investigated the effect of
differential formal mathematics education. After
controlling for the number of years of exposure to the
study of mathematics* they found sex differences in only
two of the four schools under study. However* in those
schools where boys scored higher than girls* differences
were also found in their attitudes toward mathematics.
Hilton and Berglund (1974) found significant sex
differences after controlling for the number of mathematics
courses taken* and attributed them to sex differences in
interests. "As the boys' interests in science increase
related to the girls'* their achievement in mathematics
increases relative to that of the girls" (p. 234) .
Wise* Steel* and McDonald (1979) reanalyzed test data
collected in a longitudinal study of 400*000 high school
students (Project Talent). They found that when the effect
of the number of high school mathematics courses was not
controlled* no sex differences emerged for 9th graders* but
that gains made by boys during the next three years were
more than twice that of the girls. These differences
between the sexes disappeared when the number of
mathematics courses taken was controlled. Results of the
1978 Women and Mathematics National Survey* Survey I*

3
indicated no significant sex differences for 8th grade
students on measures of problem solving or algebra.
However/ females outperformed males in computation and
spatial visualization. For the 12th grade students/
statistically significant sex differences favoring males
were found in problem solving/ but not in algebra/
computation/ or spatial visualization. For males and
females who had enrolled in courses beyond general
mathematics and who had taken or were enrolled in courses
such as pre-calculus/ calculus/ or geometry/ differences in
problem solving or spatial visualization did not exist.
Sex differences favoring males were found on a total score
obtained summing across the computation/ problem solving/
and algebra subtests (Armstrong/ 1979). The mathematics
data collected in the second survey by the 1978 National
Assessment of Educational Progress showed significant sex
differences for both 13- and 17-year-old students. The
13-year-old females outperformed males in the computational
subtest and males outscored females by 1 1/2 percentage
points in problem solving (statistically significant) . No
statistically significant differences were found in
algebra. No sex differences were found for the 17-year-old
group either in the computation subtest or in the algebra
subtest. Males surpassed females in problem solving. A
reanalysis of the data from the 17-year-old group confirmed
male superiority in problem solving after controlling for

4
mathematics preparation. Males who were enrolled or had
completed algebra II outperformed the females in
computation and problem solving but not in algebra. Males
who studied beyond algebra II outscored females on all
three subtests: computation/ algebra/ and problem solving
(Armstrong/ 1979) .
Carpenter/ Lindquist/ Mathews/ and Silver (1984)
analyzed the results of the Third National Assessment of
Educational Progress (NAEP)/ and compared them with the
First and Second Surveys. Between 1978 and 1982/ the
differences between the average performance of males and
females remained stable at each age level. At ages 9 and
13/ the overall performance of males and females was not
significantly different. At age 17/ males scored higher
than females by about 3 percentage points. When course
background was held constant/ achievement differences still
existed at age 17. For each category of course background/
male achievement exceeded female achievement. Consistent
with previous assessments/ sex differences in problem
solving in favor of males were found for the 17-year-old
sample. At ages 9 and 13/ no large differences were found
between the sexes within any level of course background.
Marshall (1981/ 1984) investigated sex differences in
mathematics performance. She found that males and females
excel each other in solving different types of problems.

5
Females were better on items of computation and males were
more successful on word-story problem items (problem
solving). She also found that females successful
performance in the problem solving items was more dependent
on their successful performance in the computation items.
Males did not need/ as much as females/ to succeed in the
computation items in order to answer correctly the problem
solving items.
Although the general findings seem to support sex
differences in mathematics learning and achievement/ the
research done does not consistently support superiority for
either sex. Most of the research has been concerned with
how the sexes differ on subtests or total test scores in
mathematics. Moreover/ the great majority of the studies
deal with correct responses. Sex differences in incorrect
responses at the item level have not been fully researched.
Only two studies dealing with sex differences in incorrect
responses at the item level were found in the research
literature (Marshall/ 1981/ 1983). Marshall investigated
whether boys and girls made similar errors in computation
and story problems. She analyzed boys' and girls' answers
to six mathematics items and found that the sexes made
different errors/ possibly reflecting different problem
solving strategies. Her original findings were supported
when she studied the same problem using a large number of
items three years later.

6
Purpose of the Study
The purpose of this study was twofold. The first was
to investigate sex-related differences in the selection of
incorrect responses in a multiple-choice achievement test
in mathematics. For each test item* the research questions
were as follows:
1. Is there a difference in the proportion of males
and females choosing each incorrect option?
2. Is the same pattern of differences found in data
obtained in three different administrations of the test?
The second objective was to investigate sex-related
differences in test scores in mathematics problem solving.
The following questions were studied:
1. Do males and females differ in problem solving
perf ormance?
2. Do sex differences in problem solving persist
after controlling for computational skillsf and does the
differential success of males and females on problem
solving items depend on their success on the computation
items?
Significance of the Study
Item response patterns are very useful techniques in
the assessment of mathematics learning and achievement.
Total test scores can be very misleading in the assessment
of student performance and provide no diagnostic
information about the nature and seriousness of student

7
errors (Harnisch? 1983). Item response patterns are
valuable for the identification of large group differences/
including d i st rict-to-district/ school-to-school/ and
classroom-to-classroom variations on different subsets of
items. The response patterns can provide diagnostic
information about the type of understanding the student has
on various mathematics topics (e.g. / problem solving).
Marshall (1981/ 1984) has used the item response
pattern technique and her findings indicate sex differences
in mathematics performance at the item level. Females
outperformed males in computation and males outscored
females in problem solving. Also/ the success of girls in
the problem solving items was dependent upon their success
in the computation items; for boys/ success in the problem
solving items did not depend as much on their computational
performance. Marshall (1981/ 1983) has also reported that
males and females differ in the selection of incorrect
responses/ reflecting differences in reasoning abilities.
In Puerto Rico/ a high percentage of children promoted
to the 7th grade in the public schools does not master the
basic skills in mathematics. If 6th grade male and female
children can be diagnosed as having different problem
solving abilities/ as Marshall found with California
children/ teachers may need to provide tailor-made
mathematics instruction for each sex/ in order to ensure
equal access to formal education and enhance mathematics

8
achievement. Since there are no investigations reported in
sex differences in item response patterns in Puerto Rico*
research is needed.
Organization of the â– Study
A review of the literature on sex differences in
mathematics performance is reported in Chapter II. The
research methodology is presented in Chapter III. Research
questions» sample» instrument» and data analysis are
discussed in that chapter. Chapter IV is an exposition of
the results of the study. Chapter V contains a summary and
interpretation of the results of the study and the
implications of the findings together with suggestions for
further research.

CHAPTER II
REVIEW OF THE LITERATURE
Sex differences in mathematics learning and
achievement have been a subject of concern for educators
and psychologists. Many studies found in the literature
support the existence of these differences. Boys show
superiority in higher level cognitive tasks (problem
solving or mathematical reasoning) in the upper elementary
years and in the early high school years (Fennema/ 1974;
Maccoby & Jacklin/ 1974) .
Almost all the research carried out has dealt with
analysis of total correct scores in mathematics aptitude
and achievement tests or scores in subtests. The
literature related to sex differences in incorrect
responses/ the main subject of the present study/ is
surprisingly sparse. For the most part/ the studies have
investigated the differences between the sexes in
mathematics learning and achievement and the underlying
variables causing the differences. Cognitive and affective
variables have been the matter at issue in the
establishment of sex differences.
Although research in mathematics problem solving/ the
secondary subject of this investigation/ is extensive/ most
9

10
of the studies consider sex differences incidental to the
major study findings. The available literature offers very
little research directly related to the problem of sex
differences in this area.
The review of the literature has been divided in four
sections. The first section consists of a detailed summary
of the available research on sex diferences in incorrect
responses. The second section deals with sex-related
differences in problem solving performance. These sections
are directly related to the objectives of the study. The
third section is more peripheral» and contains a discussion
of the more prevalent issues about the influence of
cognitive and affective variables on sex differences in
mathematics learning and achievement. The fourth is a
summary of the research dealing with variables considered
as influential to mathematics problem solving performance.
Sex-related Differences in Incorrect Response Patterns
Research findings tend to suggest that boys and girls
may be approaching problem solving differently (e.g.»
Fennema and Sherman» 1978; Marshall» 1981» 1983; Meyer»
1978» among others). Marshall (1981) investigated whether
6th grade boys and girls approach mathematical problem
solving with different strategies. Her specific interest
was whether the sexes made the same errors.

11
She analyzed the responses of 9/000 boys and 9/000
girls to 6 selected items/ 2 computation items/ and 1 story
problem item from each of 2 of the 16 forms of the Survey
of Basic Skills test administered during the academic year
1978-79. The Survey is a 30-item achievement test
administered every year to all 6th grade children in
California through the California Assessment Program.
There are 16 forms of the test/ to which approximately
9 /000 boys and 9 /000 girls respond each year. Of the 160
mathematics items contained in the 16 forms of the test/ 32
are on measurement and graphing/ 28 on number concepts/ 28
on whole number arithmetics/ 20 on fraction arithmetics/ 20
on decimal arithmetics/ 20 on geometry/ and 12 on
probability and statistics.
The item analysis performed on the 1978-79 data showed
that boys and girls tended to select different incorrect
responses. In the first computation item (Form 1 of the
test) both sexes reflected similar mistakes in carrying/
but in different columns. In the second computation item/
both sexes ignored the decimal points and selected the same
incorrect response. However/ more girls than boys chose
this response.
In the first computation item (Form 2 of the test) the
incorrect choice of both sexes was option £/ but the
second most frequently selected option was for boys
and b for girls. In the second computation item of this

12
test form/ no sex differences were found in response
patterns. Approximately 45% of each sex selected option
c. The next popular choice for both sexes was option
df selected by approximately 35% of both boys and girls.
On the story problem of Form If males and females
responded alike. Their most popular incorrect response
choice was option a for both males and females. The
second most popular incorrect choice was option c for
both sexes.
Response to the story problem in Form 2 showed sex
differences in response choice. Including the correct
option/ 33% of the girls selected option a./ 20% chose
option C/ and 20% option d. For males/ approximately
25% selected option a and the same percent selected
option d.
Marshall concluded that although the analysis of
incorrect responses does not explain why boys and girls
differ in their responses/ the analysis shows that boys and
girls approach problems in different ways and these varying
strategies can be useful in indentifying how the sexes
differ in reasoning abilities.
Two years later/ Marshall (1983) analyzed the
responses of approximately 300/000 boys and girls to
mathematics items contained in the 16 test forms of the
Survey of Basic Skills during the years 1977/ 1978/ and
1979. She used log-linear models (explained in Chapter

13
III) to investigate sex related differences in the
selection of incorrect responses/ and the consistency of
such differences over three years of administration of the
test.
Based on her findings that sex differences were found
in 80% of the items/ Marshall classified the students'
errors according to Radatz' (1979) five-category error
classification. The categories are language (errors in
semantics)/ spatial visualization/ mastery/ association/
and use of irrelevant rules.
It was found that girls' errors are more likely to be
due to the misuse of spatial information/ the use of
irrelevant rules/ or the choice of an incorrect operation.
Girls also make relatively more errors of negative transfer
and keyword association. Boys seem more likely than girls
to make errors of perseverance and formula interference.
Both sexes make language-related errors/ but the errors are
not the same.
Available research is not extensive enough to make
definite judgments about the sex-related differences
observed in incorrect responses. Clearly more research is
needed.

14
Sex-related Differences in Problem Solving
It has already been acknowledged that the subject of
problem solving has been extensively researched. However*
as early as 1969* Kilpatrick criticized the fact that the
study of problem solving has not been systematic; some
researchers have studied the characteristics of the problem
while others have given their attention to the
characteristics of the problem solvers. Moreover*
differences in the tests used to measure problem solving
performance also constitute an obstacle when trying to
compare the results of the studies carried out.
In order to avoid this pitfall and provide a basis for
comparison* the studies reviewed in this section* dealing
with sex-related differences in problem solving* have been
divided in two groups. The first comprises those studies
that used the Romberg-Wearne Problem Solving Test. The
second contains other relevant studies in which problem
solving performance has been measured by means of other
instruments.
The Romberg-Wearne Problem Solving Test merits special
mention because it was the first attempt "to develop a test
to overcome the inadequacies of total test scores in
explaining the reasons why some students are successful
problem solvers and others are not" (Whitaker* 1976* pp. 9*
10) .

15
The test is composed of 23 items designed to yield 3
scores: a comprehension score/ an application score/ and a
problem solving score. The comprehension question
ascertains whether a child understands the information
given in the item stem. The application question assesses
the child's mastery of a prerequisite concept or skill of
the problem solving question. The problem solving question
poses a question whose solution is not immediately
available/ that is/ a situation which does not lend itself
to the immediate application of a rule or algorithm. The
application and problem solving parts of the test may refer
to a common unit of information (the item stem) but the
questions are independent in that the response to the
application question is not used to respond to the problem
solving question.
Meyer (1978)/ Whitaker (1976)/ Fennema and Sherman
(1978)/ and Sherman (1979) have used the Romberg-Wearne
test in their studies. Meyer (1978) investigated whether
males and females differ in problem solving performance and
examined their prerequisite computational skills and
mathematical concepts for the problem solving questions. A
sample of 179 students from the 4th grade were administered
19 "reference tests" for intellectual abilities and the
Romberg-Wearne test. The analysis showed that males and
females were not significantly different in the
comprehension/ application/ and problem solving questions

16
of the test. The sexes differed in only 2 of the 19
reference tests» Spatial Relations and Picture Group
Name-Selection. A factor analysis» however» showed
differences in the number and composition of the factors.
For females/ a general mathematics factor was determined by
mathematics computation/ comprehension/ application/ and
problem solving. For males/ the comprehension and
application parts determined one factor; problem solving
with two other reference tests (Gestalt and Omelet)
determined another factor.
Meyer concluded that comprehension of the data and
mastery of the prerequisite mathematical concepts did not
guarantee successful problem solving either for males or
for females. Problem solving scores for both sexes were
about one third their scores in comprehension and one
fourth their scores in application.
She also concluded that the sexes may have approached
the problem solving questions differently. The methods
used by females for solving problem situations may have
paralleled their approach to the application parts.
Males may have used established rules and algorithms for
the application parts/ but may have used more of a Gestalt
approach to the problem solving situation.
Whitaker (1976) investigated the relationship between
the mathematical problem performance of 4th grade children
and their attitudes toward problem solving/ their teachers'

17
attitude toward mathematical problem solving/ and related
sex and program-type differences. Although his main
objective was to construct an attitude scale to measure
attitudes toward problem solving/ his study is important
because his findings support Meyer's regarding the lack of
significant sex-related differences in problem solving
performance. Performance in the problem solving questions/
for both males and females/ was much lower than performance
in the application questions/ and much lower than
performance in the comprehension questions. In fact/ the
mean score for each part of the item/ for both males and
females/ was almost identical to the mean scores obtained
by males and females in Meyer's study. Whitaker noted that
each application item is more difficult than its preceding
comprehension item/ and that each problem solving item is
more difficult than its preceding application item. No
significant sex-related differences were found for any of
the three parts of the item (comprehension/ application/ or
problem solving).
Fennema and Sherman (1978) investigated sex-related
differences in mathematics achievement and cognitive and
affective variables related to such differences. They
administered the Romberg-Wearne Problem Solving Test to a
representative sample of 1320 students (grades 6-8) from
Madison/ Wisconsin/ predominantly middle-class/ but
including great diversity in SES. The sample consisted of

18
students who had taken a similar number of mathematics
courses and were in the top 85% of the class in mathematics
achievement. They were tested in 1976. Four high school
districts were included. In only one of the high school
districts were sex-related differences in application and
problem solving found# in favor of males. They concluded
that when relevant factors are controlled/ sex-related
differences in favor of males do not appear often/ and when
they do/ they are not large.
Sherman (1980) investigated the causes of the emerging
sex-related differences in mathematics performance/ in
favor of males/ during adolescence (grades 8-11). She
wanted to know if these differences emerge as a function of
sex-related differences in spatial visualization and
sociocultural influences that consider math as a male
domain. In grade 8/ she used the Romberg-Wearne Test and/
in grade 11/ a mathematical problem solving test derived
from the French Kit of Tests.
The analysis showed that for girls/ problem solving
performance remained stable across the years. Mean problem
solving performance for boys/ however/ was higher in grade
11 than in grade 8. No sex-related differences were found
in grade 8/ but boys outperformed girls in grade 11/ where
the Stafford test was used.
Sherman found that for both sexes problem solving
performance in grade 8 was the best predictor of problem

19
solving performance in grade 11. Spatial visualization was
a stronger predictor for girls than for boys. Mathematics
as a male domain was a good predictor for girls only; the
less a girl stereotyped mathematics as a male domain in
grade 8/ the higher her problem solving score in grade 11.
Attitudes toward success in mathematics in grade 8 was a
more positive predictor of problem solving performance for
boys than for girls; the more positive the attitudes toward
success in mathematics in grade 8/ the higher their
performance in problem solving in grade 11.
None of these four studies» all of which used the
Romberg-Wearne Mathematics Problem Solving Test» show
statistically significant sex-related differences in
problem solving performance. In later studies other tests
were used to measure this variable (Kaufman» 1984;
Marshall» 1981» 1984) .
Kaufman (1983/1984) investigated if sex differences in
problem solving» favoring males» exist in the 5th and 6th
grades and if these differences were more pronounced in
mathematically gifted students than in students of average
mathematical ability. The Iowa Test of Basic Skills and a
mathematics problem-solving test were administered to 504
subjects. Males in the average group as well as males in
the gifted group outperformed females» but only the gifted
group showed statistically significant differences.

20
As a result of her investigations/ Marshall (1981)
concluded that sex-related differences in mathematics
performance may be the result of comparing the sexes on
total test scores. If the test contains more computation
items than problem solving items/ girls will perform better
than boys/ but if the test contains more problem solving
items than computation ones/ boys will outperform girls.
With this in mind/ Marshall investigated sex-related
differences in computation and problem solving by analyzing
the responses of approximately 18/000 students from grade 6
who had been administered the Survey of Basic Skills Test:
Grade 6/ during the academic year 1978-79.
Two of the 16 test forms of the Survey were used to
assess skills such as concepts of whole numbers/ fractions/
and decimals. These skills were tested both as simple
computations and as story problems (problem solving).
Two computation items and one story problem item were
selected because they were particularly related; both
computation items required skills needed in solving the
corresponding story problem. It was assumed that correct
solution of the computation item correlates with solving
the story problem because the story problem requires a
similar computation.
Marshall found that girls were better in computation
and boys were better in problem solving. She also found
that boys were much more likely than girls to answer the

21
story problem item correctly after giving incorrect
responses to both computation items. Apparently* mastery
of the skills required by the computation items is more
important for girls than for boys. If girls cannot solve
the computation items* they have little chance of solving
the related story problem item. For girls* the
probability of success in the story problem item after
giving successful answers to both computation items is
almost 2 1/2 times the probability of success after giving
incorrect responses to both computation items. For boys*
the probability of success in the story problem item after
successful responses to the computation items is about
1 3/4 times the probability of success on the story problem
item after incorrect responses to the computation items.
Three years later* Marshall (1984) analyzed more in
depth these phenomena of sex-related differences. Her
interest was twofold. First* she wanted to know if there
were differences in the rate of success for boys and girls
in solving computation and story problem items. Second*
she examined additional factors that interact with sex to
influence mathematics performance* such as reading
achievement* socio-economic status (SES) * primary language*
and chronological age. Two questions were raised: Do the
probabilibities of successful solving of computation and
story problem items increase with reading score? Are these
probabilities different for the two sexes?

22
Approximately 270/000 students from the 6th grade were
administered the Survey of Basic Skills of the California
Assessment Program/ during the years 1977/ 1978/ and 1979.
Responses were analyzed using log-linear models.
Successful solving of computation items was positively
associated with successful solving of story problems.
Girls were more successful in computation than boys/ and
boys were more successful than girls in solving story
problems. This finding supports reports from the National
Assessment of Educational Progress (NAEP) (Armstrong/
1979) .
To investigate the effects of reading/ SES/ language/
and chronological age/ only those test forms containing 2
computation items and 2 story problems were considered for
analysis; 32 items from 8 test forms were included in the
analysis.
The results of these analyses showed that at every
level of reading score/ 6th grade children were more
successful in computation than in story problems.
Although the differences were not large/ at every reading
score boys consistently had higher probabilities of
success in story problems than did girls/ and girls
consistently showed higher probabilities of success in
computation than boys. Also/ as the reading score
increased/ the difference between the probability of
success in story problems and the probability of success in

23
computation grew larger. This difference grew larger for
girls than for boys.
Although SES was a major factor in solving computation
and story problem items successfully/ the effect was
similar for each sex. Sex-related differences by primary
language or chronological age were not large.
This research carried out by Marshall with elementary
grade children supports previous research findings that
males are better than females in mathematics problem
solving (a higher order skill) and females are better than
their counterpart males in computation (a lower level
skill). Marshall's research also brought out a different
aspect of this question: the notion that girls find it more
necessary than boys to succeed in the computation items in
order to successfully solve the story problem items.
Cognitive and Affective Factors That
Influence_Sex Differences in Mathematics Learning
and Achievement
The research reviewed in the literature does not
provide evidence of any unique variable that could serve
as an explanation for the observed sex-related differences
in mathematics learning and achievement. However/ some
issues have been discussed/ among which the most prevalent
are that sex-related differences in mathematics learning
and achievement are a result of differences in formal
education; that sex-related differences in mathematics

24
learning and achievement arise from sex differences in
spatial visualization; and that sex-related differences
result from a differentiated effect of affective variables
on the mathematics performance of males and females.
Differences in Formal Mathematics Education (Differential
Coursework Hypothesis)
The basis for the differential coursework hypothesis
is the fact that sex-related differences in mathematical
learning and achievement show up when comparing groups
which are not equal in previous mathematics learning.
After the 8th grade/ boys tend to select mathematics
courses more otten than girls. Therefore/ girls show lower
achievement scores in mathematics tests because their
mathematics experience is not as strong as the boys'
(Fennema/ 1975; Fennema & Sherman/ 1977; Sherman/ 1979).
Fennema and Sherman's study (1977) lends additional
support to the feasibility of viewing sex differences in
mathematics learning and achievement as reflecting
something other than a difference in mathematics aptitude.
After controlling for previous study of mathematics/ they
found significant sex differences in mathematics
achievement in only two of the four schools under study/
making the attribution to sex per se less likely.
Controlling for the number of space visualization-related
courses/ the sex-related differences which originally

25
emerged in spatial visualization scores became
non-significant. In the two schools where sex differences
in mathematics achievement were found/ differences between
the sexes were also found in their attitudes toward
mathematics.
Researchers like Backman (1972) / who analyzed data
from Project Talent/ and Allen and Chambers (1977) have
also hypothesized that sex-related differences in
mathematics achievement may be related to different
curricula followed by males and females. Allen and
Chambers attributed male superiority in mathematics problem
solving to differences in the number of mathematics courses
taken in high school.
This issue has been seriously questioned by Astin
(1974)/ Fox (1975a/ 1975b)/ and Benbow and Stanley (1980)/
among others. Astin and Fox have reported large
differences in favor of males among gifted students taking
the Scholastic Achievement Test. These differences occur
as early as grade 7/ when there are no sex differences in
the number of courses taken. Benbow and Stanley (1980)
compared mathematically precocious boys and girls in the
7th grade/ with similar mathematics background/ and found
sizeable sex-related differences favoring boys in
mathematical reasoning ability. Five years later/ they
conducted a follow-up study which showed that boys

26
maintained their superiority in mathematics ability during
high school. While Fox attributed sex-related differences
in mathematical achievement to differential exposure to
mathematical games and activities outside school/ Benbow
and Stanley suggested that sex-related differences in
mathematics performance stem from superior mathematical
ability in males» not from differences in mathematics
formal education.
The differential coursework hypothesis is not totally
convincing and/ as reporteu berore/ it has been challenged
by researchers such as Benbow and Stanley (1980).
However/ Pallas and Alexander (1983) have questioneu the
generalizaoility of Benbow and Stanley's findings based on
the fact that they used highly precocious learners. The
differential coursework hypothesis can be accepted only as
a partial explanation of differences in mathematics
performance found between the sexes.
Differenees in Spatial Ability
The basic premise in this issue is that males and
females differ in spatial visualization and this explains
differential mathematics learning and achievement. Until
recently/ sex differences in spatial ability in favor of
males were believed to be a fact and were thought by some
to be related to sex differences in mathematical
achievement.

27
Research findings in this area have been inconsistent.
In 1966 / Maccoby stated that "by early school years/ boys
consistently do better (than girls) on spatial tasks and
this difference continues through the high school and
college years" (p.26). In 1972/ Maccoby and Jacklin said
that the differences in spatial ability between the sexes
"remain minimal and inconsistent until approximately the
ages of 10 or 11/ when the superiority of boys becomes
consistent in a wide range of populations and tests"
(p- 41) - In 1974 / after a comprehensive literature search/
Maccoby and Jacklin concluded that sex differences in
spatial visualization become more pronounced between upper
elementary years and the last year of high school/ the
years when sex-related differences in mathematics
achievement favoring boys emerge.
Guay and McDaniel (1977) supported in part Maccoby and
Jacklin's 1974 findings. They found that among elementary
school children/ males had greater high level spatial
ability than females/ but that males and females were equal
in low level spatial ability. This finding is inconsistent
with that portion of Maccoby and Jacklin's review that
suggests that sex differences become evident only during
early adolescence. Cohen and Wilkie (1979) however/ stated
that in tests measuring distinct spatial tasks/ males
perform better than females in early adolescence and
throughout their life span. Most studies carried out after

28
1974 have failed to support these sex differences in
spatial abilities (Armstrong/ 1979; Connor/ Serbin, &
Schackmanr 1977; Fennema & Sherman/ 1978; Sherman/ 1979).
Fennema and Sherman (1978) and Sherman (1979) have
explored sex-related differences in mathematical
achievement and cognitive and affective variables related
to these differences. In a study involving students from
grades 6/ 7/ and 8/ from four school districts/ Fennema and
Sherman found that spatial visualization and problem
solving were highly correlated for both sexes (.59 and
.60). Even in the school district where sex differences
were found in problem solving/ no significant sex-related
differences were found in spatial visualization.
When Sherman (1980) compared groups of males and
females in two different grades/ 8 and 11/ she found no
sex-related differences in problem solving or in spatial
visualization in grade 8. In grade 11/ however/ although
the sexes differed in their problem solving performance/ no
sex-related differences were found in spatial
vizualization. Even though spatial visualization in grade
8 was the second best predictor of problem solving
performance in grade 11/ sex differences in grade 11 were
not a result of spatial visualization since no differences
were found in that skill.
In spite of the fact that no sex differences were
found in spatial abilities/ it is evident that males and

29
females may use them in a different way. Meyer (1978) »
with an elementary grade sample» and Fennema and Tartree
(1983) » with an intermediate level sample» found that the
influence of spatial visualization on solving mathematics
problems is subtle» and that males and females use their
spatial skills differently in solving word story problems
(problems that measure problem solving ability or
reasoning). Fennema and Tartree (1983) carried out a
three-year longitudinal study which showed that girls and
boys with equivalent spatial visualization skills did not
solve the same number of items» nor did they use the same
processes in solving problems. The results also suggested
that a low level of spatial visualization skills was a more
debilitating factor for girls than for boys in problem
solving performance.
Landau (1984) also investigated the relationship
between spatial visualization and mathematics achievement.
She studied the performance of middle school children in
mathematical problems of varying difficulty» and the extent
to which a diagramatic representation is likely to
facilitate solution. She found that spatial ability was
strongly correlated to mathematical problem solving and
that the effect of spatial ability was more influential for
females. Females made more use of diagrams in the solution
of problems» reducing the advantage of males over females
in problem solving performance.

30
The issue of sex-related differences in spatial
visualization ability as an explanation for sex differences
observed in mathematics achievement is less convincing and
the findings more contradictory than in the issue of sex
differences in formal education. Besides these cognitive
issues/ other issues/ mostly affective in nature/ have also
been studied in trying to explain the origin of these sex
differences in mathematics achievement and learning. The
studies dealing with these affective variables are reviewed
in the next section.
Differentiated Effect of Affective Variables
Researchers have attempted to explain the effect of
sex differences in internal beliefs/ interests/ and
attitudes (affective variables) on mathematics learning and
achievement. A brief statement of each explanation
precedes the summary of studies conducted that support the
explanation.
;. Females/ more
than males/ lack confidence in their ability to learn
mathematics and this affects their achievement in
mathematics and their election of more advanced mathematics
courses.
Maccoby and Jacklin (1974) reported that self-
confidence in terms of grade expectancy and success in
particular tasks was found to be consistently lower in
women than in men. In 1978/ Fennema and Sherman reported

31
that in their study involving students from grades 6
through 12/ boys showed a higher level of confidence in
mathematics at each grade level. These differences between
the sexes occurred in most instances even when no
sex-related differences in mathematics achievement were
found. The correlation between confidence in mathematics
performance and mathematics achievement in this study was
higher than for any other affective variable investigated.
Sherman reported a similar finding in 1980; in males/
the most important factor related to continuation in
theoretical mathematics courses was confidence in learning
mathematics. This variable weighed more than any of the
cognitive variables: mathematics achievement/ spatial
visualization/ general ability/ and verbal skill. In the
case of females/ among the affective variables/ confidence
in learning mathematics was found to be second in
importance to perceived usefulness of mathematics. Probert
(1983) supported these findings with college students.
A variable that needs discussion within the context of
sex differences in confidence as learners of mathematics is
causal attribution. Causal attribution models attempt to
classify those factors to which one attributes success or
failure. The model proposed by Weiner (1974) categorizes
four dimensions of attribution of success and failure:
stable and internal/ unstable and internal/ stable and
external/ and unstable and external. For example/ if one

32
attributes success to an internal, stable attribute, such
as ability, then one is confident of being successful in
the future and will continue to strive in that area. If
one attributes success to an external factor such as a
teacher, or to an unstable one, such as effort, then one
will not be as confident of success in the future and will
cease to strive. Failure attribution patterns work this
way: if failure is attributed to unstable causes, such as
effort, failure can be avoided in the future and the
tendency will be to persist in the task. However, if
failure is attributed to a stable cause, such as ability,
the belief that one cannot avoid failure will remain.
Studies reported by Bar-Tal and Frieze (1977) suggest
that males and females tend to exhibit different
attributional patterns of success and failure. Males tend
to attribute their success to internal causes and their
failures to external or unstable ones. Females show a
different pattern; they tend to attribute success to
external or unstable causes and failures to internal ones.
The pattern of attributions, success attributed externally
and failure attributed internally, has become hypothesized
to show a strong effect on mathematics achievement in
females. Kloosterman's (1985) study supported these
findings. According to Kloosterman, attributional
variables appear to be more important achievement mediators
for females than for males, as measured by mathematics word
problems. More research is needed in this area.

33
Mathematics as a male domain. Mathematics is an
activity more closely related to the male sex domain than
to the female sex domain (Eccles et al.r 1983). Thus* the
mathematical achievement ot boys is higher than that of
gi rls.
According to John Ernest (1976) * in his study
Mathematics and Sex> mathematics is a sexist discipline.
He attributed sex-related differences in mathematical
achievement to the creation by society of sexual
stereotypes and attitudes* restrictions* and constraints
that promote the idea of the superiority of boys in
mathematics. Ernest reported that boys* girls* and
teachers* all believe that boys are superior in
mathematics* at least by the time students reach
adolescence. Bern and Bern (1970) agree and argue that an
American woman is trained to "know her place" in society
because ot the pervasive sex-role concept which results in
differential expectations and socialization practices.
Plank and Plank (195 4) were more specific. They
discussed two hypotheses related to this view: the
differential cultural reinforcement hypothesis and the
masculine identification hypothesis. The differential
cultural reinforcement hypothesis states that society in
general perceives mathematics as a male domain* giving
females less encouragement for excelling in it. The
masculine identification hypothesis establishes that

34
achievement and interest in mathematics result from
identification with the masculine role.
A study related to the differential cultural
reinforcement hypothesis is that of Dwyer (1974).
Dwyer examined the relationship between sex role standards
(the extent to which an individual considers certain
activities appropriate to males or females) and
achievement in reading and arithmetic. Students from
grades 2» 4* 6» 8» 10» and 12 participated in this study.
She found that sex role standards contributed significant
variance to reading and arithmetic achievement test scores
and that the effect was stronger for males than for
females. This led to her conclusion that sex-related
differences in reading and arithmetic are more a function
of the child's perception of these areas as sex-appropriate
or sex-inappropriate than of the child's biological sex»
individual preference for masculine and feminine sex roles»
or liking or disliking reading or mathematics.
In a study which agrees with the masculine
identification hypothesis» Milton (1957) found that
individuals who had received strong masculine orientation
performed better in problem solving than individuals who
received less masculine orientation. Elton and Rose (1967)
found that women with high mathematical aptitude and
average verbal aptitude scored higher on the masculinity
scale of the Omnibus Personality Inventory (OPI) than those
with average scores on both tasks.

35
It is not until adolescence that sex differences in
the perception of mathematics as a male domain are found
(Fennemar 1976; Stein/ 1971; Stein & Smithless/ 1969;
Verbeke/ 1983). In a study with 2nd/ 6th and 12th
graders/ Stein and Smithless (1969) found that students'
perceptions of spatial/ mechanical/ and arithmetic skills
as masculine became more defined as these students got
older. Fennema (1976) considers that the influence each
sex exerts upon the other on all aspects of behavior is
stronger during adolescence. Since during these years
males stereotype mathematics as a male domain/ they send
this message to females who/ in turn/ tend to be influenced
in their willingness to study or not to study mathematics.
Before that stage/ girls consider arithmetic feminine/
while boys consider it appropriate for both sexes (Bobbe/
1971).
Usefulness of mathematics. Females perceive mathema¬
tics as less useful to them than males do/ and this
perception occurs at a very young age. As a result/
females exert less effort than males to learn or elect to
take advanced mathematics courses.
Many studies reported before 1976 found that the
perception of the usefulness of mathematics for one's
future differs for males and females/ and is related to
course taking plans and behavior (Fox/ 1977). If females
do not perceive mathematics as useful for their future/

36
they show less interest in the subject than counterpart
males. These differences in interest are what Hilton and
Berglund (1974) suggest to account for sex-related
differences in mathematics achievement.
Although the perception of the usefulness of
mathematics is still an important predictor of course
taking for girls* there is a growing similarity between
males and females regarding the usefulness of mathematics
(Armstrong & Price* 1982; Fennema & Sherman* 1977; Moller*
1982/1983). Armstrong and Price investigated the relative
influence of selected factors in sex-related differences in
mathematics participation. Both males and females selected
usefulness of mathematics as the most important factor in
deciaing whether or not to take more mathematics in high
school. Moller's study revealed that both males and
females based mathematics course-taking decisions on career
usefulness. A Fennema and Sherman (1977) study showed only
slight differences between males and females in their
feelings about the usefulness of mathematics. In her study
of this variable among college students* Probert
(1983/1984) did not find any sex-related differences
either.
These have been the main affective variables
researched in attempting to explain the underlying causes
of sex-related differences observed in mathematics learning
and achievement. In spite of the great diversity of studies

37
dealing with both cognitive and affective variables/ there
are no clear-cut findings to render unequivocal support to
a particular variable as accounting for these sex-related
differences. However/ everything seems to point to the
fact that affective/ rather than cognitive variables play a
more significant role in the sex-related differences
observed in mathematics performance and learning. In most
of the studies dealing with affective variables/ findings
consistently show that these factors influence mathematics
performance in females more than in males. In at least one
area/ confidence as learners of mathematics/ Sherman (1980)
found that this variable influenced course election more
than all the cognitive variables previously discussed.
The case for the societal influences on sex roles and
expectations to account for the differences in mathematics
learning is also supported in one way or another in the
studies reported in the literature.
Problem Solving Performance and Related Variables
Problem solving has been perhaps the most extensively
researched area in mathematics education. Published
reviews by Kilpatrick (1969)/ Riedesel (1969)/ and Suydam
and Weaver (1970-1975) attest to this. Much of the
research done has focused on identifying the
determinants of problem difficulty and the problem features
that influence the solution process.

38
At present» no set of variables has been clearly
established as a determinant of problem difficulty.
Several researchers have investigated the effect of reading
and computation on problem solving performance. Others
have studied the effect of student attitudes toward problem
solving in problem solving learning and achievement.
Typically/ correlational methods have been used to
investigate these questions.
Computational Skills and Problem Solving Performance
One of the first researchers to study the effect of
computation and reading on problem solving performance was
Hansen (1944). He investigated the relationship of
arithmetical factors/ mental factors/ and reading factors
to achievement in problem solving. Sixth grade students
were administered tests in problem solving and categorized
as superior achievers (best problem solvers) and inferior
archievers (poorest problem solvers). The two groups were
compared in selected factors believed to be related to
success in arithmetic problem solving: arithmetical/ mental
and reading factors. After controlling for mental and
chronological age/ the superior achievers in problem
solving surpassed the inferior achievers in mental and
arithmetical factors. The superior group did better in
only two of the six items under the reading factors:
general language ability and the reading of graphs/ charts/
and tables.

39
The findings suggest that reading factors are not as
important as arithmetic and mental factors in problem
solving performance. However» these findings should be
taken cautiously* as the content of the Gates tests (used
to measure reading) is literary and does not include
mathematical material.
Chase (1960) studied 15 variables in an effort to find
out which ones have significant influence on the aDility to
solve verbal mathematics problems. Only computation*
reading to note details* and fundamental knowledge were
primarily related to problem solving. Computation
accounted for 20.4% of the 32% variance directly associated
with problem solving.
Chase concluded that a pupil's ability in the
mechanics of computation* comprehension of the principles
that underline the number systems* and the extent to which
important items of information are noticed when reading*
are good predictors of the student's ability in solving
verbal problems.
Balow (1964) investigated the importance of reading
ability and computation ability in problem solving
performance. He objected to the approaches used by other
researchers who in their analyses dichotomized research
subjects as "poor" or "good" students* and who ignored the
recognizea effect of intelligence on reading and on
mathematics achievement
Balow administered the Stanford

40
Achievement Test (subtests of reading/ arithmetic/ and
reasoning) and the California Short-Form test of mental
ability to a group of 1/400 children from the 6th grade.
All levels of achievement were included in the analysis.
Analysis of variance and covariance were used and compared.
He confirmed the findings of other researchers to the
effect that there is a direct relationship between I.Q. and
reading ability/ and between I.Q. and computational skills.
The results of the analysis of variance revealed that
increases in computation ability were associated with
higher achievement in problem solving. A relationship
between reading ability and problem solving was also found/
but it was not as strong. Significant differences in
problem solving performance associated with computational
ability were found when intelligence was controlled.
Balow concluded that computation is a much more
important factor in problem solving than reading ability/
and that when I.Q. is taken into consideration/ the degree
of the relationship between reading and problem solving
ability becomes less pronounced. Intelligence tends to
confound the relationship between these two variables.
Knifong and Holtan (1976/ 1977) attempted to
investigate the types of difficulties children have in
solving word problems. They administered the word problem
section of the Metropolitan Achievement Test to 35 children
from the 6th grade. Errors were classified in two

41
categories. Category I included clerical and computational
errors. Category II included other types of errors» such
as average and area errors/ use of wrong operation/ no
response/ and erred responses offering no clues. It the
student's work indicated the correct procedure and yet the
problem was missed because of a computational or clerical
error/ it was assumed that the problem was read and
understood.
An analysis of frequencies showed that clerical errors
were responsible for 3% of the problems incorrectly solved/
computational errors accounted for 49%/ and other errors
for 48% of the erred problems.
Knifong and Holtan concluded that "improved
computational skills could have eliminated nearly half ot
the word problem errors" (p. Ill). These computational
errors were made in a context where other skills such as
reading/ interpretation of the problem/ and integration of
these skills necessary for the solution of word problems/
might interact. However/ Knifong and Holtan state that
their findings neither confirm nor deny that improvement ot
reading skills will lead to improvement in problem solving.
They conclude that "it is difficult to attribute major
importance to reading as a source of failure" (p. 111).
In a later analysis/ looking for evidence of poor
reading abilities affecting children's success in word
problems/ Knifong and Holtan (1977) interviewed the

42
children whose errors fell under the category or "other
errors." Students were asked to read each problem aloud
and answer these questions: What kind of situation does
the problem describe? What does the problem ask you to
find? How would you work the problem?
Ninety five percent of the students read the problem
correctly; 98% explained the kind of situation the problem
described in a correct manner; 92% correctly answered what
the problem was asking them to find* and 36% correctly
answered the question of how to work the problem.
The fact that a large percent of the students whose
errors were classified as "other errors" (in which reading
skills might have been a factor) correctly stated how to
work the problem* is strong evidence of their ability to
read and interpret the problems correctly. The errors made
by this group of students had a distinct origin* unrelated
to reading ability.
Zalewski (1974) investigated the relative contribution
of verbal intelligence* reading comprehension* vocabulary*
interpretation of graphs and tables* mathematical concepts*
number sentence selection* and computation to successful
mathematical word problem solution* and the relationship of
the dependent variable to the eight independent variables.
She worked with a group of 4th grade children who
were administered the subtests of the Iowa Test of Basic
Skills (ITBS) and the Weschler Intelligence Scale for

43
Children (WISC). Multiple regression analysis was
performed. A correlation of .769 was found between word
problem solving and the eight independent variables.
Correlations between word problem solving and the
independent variables ranged from .363 (verbal
intelligence) to .674 (mathematical concepts).
Correlations between the independent variables ranged from
.369 (verbal intelligence and computation) to .749 (reading
comprehension and vocabulary). Mathematical concepts/
computation/ and number sentence selection were almost as
effective as all eight independent variables in predicting
achievement in mathematical word problem solving.
Mathematical concepts/ computation/ number sentence
selection/ and reading comprehension accounted for 58% of
the variance/ whereas all eight predictors accounted for
59% of the variance. The two best predictors were
mathematical concepts and computation/ which accounted for
54% variance. Other variables accounted for about 40% of
the variance.
The author recommends that the findings of this study
be interpreted cautiously because the correlation between
the eight independent variables was high/ and/ according to
Zal ewski /
in a study of this nature where the interest is
primarily in the influence of several variables
on one dependent variable/ a low correlation
between the independent variables is required.
(p. 2804)

44
In a more recent investigation? Exedisis (1983)
studied the contribution of reading ability, vocabulary,
mathematical concepts, computation, sex, and race on
problem-solving performance. The Iowa Test of Basic
Skills was administered to a group of 6th, 7th, and 8th
grade anglo and black Chicago male and female adolescents.
Problem solving was highly correlated to an understanding
of basic mathematical concepts, somewhat correlated to
race, and weakly correlated to computational and
vocabulary skills, sex, and reading ability.
Although the findings of these studies show a
relationship between computational skills and problem
solving achievement, this relationship is not strong enough
to be considered the most determinant factor in problem
solving achievement, as some of the researchers have been
careful to point out. In spite of the dismissal of reading
as a determinant factor in problem solving achievement by
some of these same researchers, more recent studies in this
area have led others to hold different views.
Headin_g_and_J)r obiem_ Solving Performance
Martin (1964) studied the contribution of reading
comprehension, computation, abstract verbal reasoning, and
arithmetic concepts to arithmetic problem solving
performance. Fourth and 8th grade students were
administered the Iowa tests of Basic Skills and the
Lorge-Thorndlike intelligence test (verbal).

45
He found that in the 4th grade the correlations
between problem solving and abstract verbal reasoning»
reading comprehension/ arithmetic concepts/ and computation
were .61/ .64/ .66/ and .60 respectively/ and .56/ .68/
.69/ and .63 in the 8th grade. When computation was held
constant/ the correlation between problem solving and
reading was .52 in grade 4 and .54 in grade 8. When
reading was held constant the correlation between problem
solving and computation was .43 in grade 4 and .42 in
grade 8.
Creswell (1982) worked with a sample of anglo and
black adolescents from Chicago. Each subject was
administered the California Achievement test. Multiple
regression was used to analyze the data. The analysis
showed that reading is more important than computation in
predicting student performance in problem solving. Reading
accounted for 49.5% of the variance; computation accounted
for 14.6% of the variance.
Ballew and Cunningham (1982) worked with 6th grade
students in an attempt to find what proportion of students
have as their main source of difficulty with word problems
each of the following factors: a) computation skills/
b) interpretation of the problem/ c) reading and/
d) integrating these skills in the solution of problems.
They also wanted to know if a student can be efficiently
diagnosed as having one of the four categories as his/her
main difficulty with mathematics word problems.

46
Their study is important because it represents an
attempt to demonstrate that multiple factors can interact
in the correct solution of a mathematics word problem.
They constructed three graded tests from a basal
mathematics series for grades 3 through 8. For test 1 / the
problems were set in pure computational form (the effects
or reading/ interpretation/ as well as the necessity for
integration were removed in an effort to measure the
computational skills required by the word problems).
For test 2/ the effects of reading and computation
were removed by reading the problems to the students and by
giving scores based on whether or not the students set them
up properly/ in an attempt to measure problem
interpretation alone. For test 3/ the effect of
computation was removed. The test yielded two scores—one
by grading the students on whether or not they set up the
problems properly and another by grading on the basis of
the correct answer.
The tests were administered to all 244 students from
the 6th grade in two different schools. A diagnostic
profile was obtained for each of the 217 students for which
complete data were available: a computational score/ a
probiem-interpretation score/ a reading score/ and a
reading-problem solving score.
They assumed that if the reading-prob1em
interpretation score was lower (one or more levels lower)

47
than the problem-interpretation score? the difficulty was
due to reading ability. If the score of the lowest of the
three areas (computation? problem interpretation? and
reading-problem interpretation score) was the same as the
reading-problem solving score? the student's area of
greatest immediate need was either computation? problem
interpretation? or reading. If the reading-problem solving
score was lower than the lowest of the other three scores?
the student's area of greatest immediate need was
integration.
Analysis of the data revealed that for 19% of the
students? problem interpretation was their major
difficulty; for 26% of the students? integration (total
problem solving) was their greatest immediate need; for
another 26%? computation was the major weakness; and for
29%? reading was their greatest immediate need.
Seventy five percent of the students demonstrated
clear strength in computation? 21% in problem
interpretation? and 4% in reading-problem interpretation.
An analysis across all students (including those without
complete data) showed that 26% of the subjects could not
work word problems at a level as high as that at which they
could compute? interpret problems? and read and interpret
problems? when those areas were measured separately. This
led them to conclude that knowing the skills or the
components of solving word problems is not sufficient for

48
success* since the components must be integrated into a
whole process (mastery learning of the components cannot
assure mastery of the process).
Their analysis also led them to conclude that* in the
case of 6th graders* inability to read problems is a major
obstacle in solving word problems. Only 12% of the
subjects could read and set up problems correctly at a
higher level than they could compute* while 60% could
compute correctly at a higher level than they could read
and set up problems; 44% could set up problems better when
they heard them read than when they read the problem
themselves. Only 13% could set up problems better when
they read them than when they heard them read.
Muth ( 1984 ) investigated the role of reading and
computational skills in the solution of word problems. A
group of 200 students from the 6th grade were administered
a test of basic skills and a mathematics word problem test.
The word problem test consisted of 15 sample items supplied
by the National Assessment of Educational Progress. The
items were adapted to include some extraneous information
and complex syntactic structure. Four versions of the test
were constructed by combining two versions of problem
information (absence vs. presence of extraneous
information) with two versions of syntactic structure
(simple vs. complex syntax) . Task performance was measured
by means of the number of problems answered correctly*

49
number of problems set up correctly/ and amount of time
spent taking the test.
Reading ability and computational ability were both
positively correlated with number of correct answers and
with number of problems correctly set up/ and negatively
correlated with test-taking time. Presence of extraneous
information was negatively correlated with correct answers
and correct set ups and positively correlated with
test-taking time. Syntactic complexity was not
significantly correlated with any of the performance
measures.
Results of a multiple regression analysis showed that
reading accounted for 46% of the variance in total correct
answers and computation accounted for 8%. Reading ability
and computational ability uniquely accounted for 14% and 8%
of the variance in the number of correct answers/
respectively. Extraneous information added significantly
to the variance explained in the number of correct answers/
but syntactic structure did not. Reading ability accounted
for 5% of the variance in test-taking time/ but computation
did not add significantly to the variance explained by
reading.
Muth concluded that reading and computation both
contribute significantly to success in solving arithmetic
word problems/ but that reading plays a more significant
role than does computation.

50
The studies reviewed in this section show a positive
relationship between reading and problem solving
performance/ but in the case of Ballew and Cunningham
(1982)/ this relationship is not viewed singly but rather
as one among the interacting factors that produce
successful problem solving.
The third variable reviewed is the effect of student
attitudes toward problem solving on problem solving
performance. Many researchers have tried to demonstrate
that this variable is a determinant factor in problem
solving achievement.
Attitudes Toward Problem Solving and Problem Solving
Performance
Research studies support the existence of positive and
rather stable relationships between student attitudes and
achievement in mathematics. Aiken (1970) has suggested
that an individual's attitude toward one aspect of the
discipline (mathematics) / such as problem solving/ may be
entirely different from his/her attitude toward another
phase of the discipline/ such as computation.
Research/ however/ has been directed to the use of
single/ global measures of attitudes toward mathematics
rather than to the investigation of attitudes toward a
particular phase of the discipline.
The studies described below are only part of the few
investigations which have examined the relationship between

51
student attitudes and performance in the area of problem
solving.
Carey (1958) constructed a scale to measure attitudes
toward problem solving. Her interest was in general
problem solving rather than in mathematical problem
solving. Her work constitutes the first attempt to
construct a measure of attitudes toward problem solving.
The scale was used with a group of college students/ and
she foundf among other things/ that problem solving
performance is positively related to problem solving
attitudes and that/ in the case of females/ positive
modification of attitudes toward problem solving brings a
significant gain in problem solving performance.
Lindgren/ Silva/ Faracoz and DaRocha (1964) adapted
Carey's scale of attitudes toward problem solving and
applied it to a group of 4th grade Brazilian children.
Students also answered an arithmetic achievement test/ a
general intelligence test/ and a socioeconomic (SE) scale.
A low but significant positive correlation was found
between arithmetic achievement and attitudes toward problem
solving. A near zero correlation was found between
attitudes toward problem solving and intelligence. Since
problem solving is one aspect of the discipline of
mathematics/ this correlation between attitudes and
arithmetic achievement can lead to a conclusion or a strong

52
correlation between attitudes toward problem solving and
problem solving performance.
Whitaker (1976) constructed a student attitude scale
to measure some aspects of 4th grade student attitudes
toward mathematic problem solving. He included statements
reflecting children's beliefs about the nature of various
types of mathematical problems/ the nature of the problem
solving process/ the desirability or persevering when
solving a problem/ and the value of generating several
ideas for solving a problem.
He correlated student attitudes toward problem solving
with their scores in a mathematical test which yielded a
comprehension score/ an application score/ and a problem
solving score. He found a significant positive
relationship between problem solving performance and
student attitude scores on the subscale which measured
reactions to such things as problem solving techniques or
problem situations/ or to the frustation or anxiety
experienced when confronted with problem solving
situations.
In another part of this study/ Whitaker investigated
the relationship between the attitudes of 4th grade
teachers toward problem solving and their students'
performance in problem solving. A very weak and
nonsignificant negative correlation was found between the
teacher's attitudes toward problem solving and student
performance.

53
Tne studies reviewed have confirmed the relationship
between problem solving performance and attitudes toward
problem solving (Carey» 1958; Lindgren et al.» 1964;
Whitaker» 1976). However» the results reportea in the
studies that investigated the relationship between problem
solving performance and computation and between reading and
problem solving fail to be consistent in their conclusions.
Hansen (1944)» Chase (I960)» Balow (1964)» Kmfong and
Holtan (1976» 1977)» and Zalewski (1974; concluded that
computation is more strongly related to problem solving
than is reading. Martin (1964)» Creswell (1982)» Ballew
and Cunningham (1982)» and Muth (1984)» concluded that
reading aoility and mathematical problem solving show a
stronger relationsnip than computation and problem solving.
Exedisis's (1983) findings led to the conclusion that the
effect or reading and computation in problem solving
performance is unimportant.

CHAPTER III
METHOD
The first objective of this study was to investigate
sex differences in the selection of incorrect responses in
a mathematics multiple-choice achievement test, and to
determine whether these differences were consistent over
three consecutive administrations of the test. The second
objective was to investigate whether males and females
differ in problem solving performance* if these differences
persist after accounting for computational skills* and if
the male-female differences depend on the level of
computational skills. This chapter contains descriptions
of the sample* the test instrument* and the statistical
analysis used in achieving the above mentioned objectives.
The Sample
To achieve the first objective of the study* all the
students who took the Basic Skills Test in Mathematics-6
("Prueba de Destrezas Básicas en Matemáticas-6") in
each or the three years were included in the study. To
achieve the second objective of the study* approximately
54

55
1/000 students were selected randomly for each year. For
the first yearr 1/002 were selected (492 boys and
510 girls); for the second year/ 1/013 students were
selected (504 boys and 509 girls); and/ for the third year/
1/013 students were selected (509 boys and 504 girls).
The student population in Puerto Rico includes
children from the urban and rural zones and comprises
children from low and middle socioeconomic levels.
Findings can be generalized only to this population.
The Instrument
The Basic Skills Test in Mathematics-6 is a
criter ion-referenced test used in the Department of
Education of Puerto Rico as part of the annual assessment
program. The test measures academic achievement in
operations/ mathematical concepts/ and story problems. It
has a reported split half reliability or .95. The test was
designed specifically for Puerto Rico. Its contents and
the procedures followed for its development were formulated
and reviewed by educators from the mathematics department
of the Department of Education of Puerto Rico/ in
coordination with the Evaluation Center of the Department
of Education and mathematics teachers from the school
districts. The emphasis placed on each skill area is
depicted in Figure 3.1.

MATH CONCEPTS
Fractions j
2 items) j
/ Numbers
j
\
(18 items )
\
/
Fig. 3.1 Number of Questions by Skill Area in the Basic
Skills Test in Mathematics - 6

57
Analysis of the Data
Analysis of Sex bv Option by Year Cross Classifications
Log-linear models were used to analyze the sex by
option by year cross classifications for each item. Two
topics are addressed in this section/ the hypotheses tested
using log-linear models and a comparison of the hypotheses
tested in this study with those tested by Marshall (1981/
1983) .
The object of the analysis was to test two hypotheses:
1. The proportion of males and the proportion of
females choosing each incorrect option does not vary from
year to year. Note that this hypothesis is stated in the
null form.
2. Assuming that the first hypothesis is correct/ the
proportion of males who choose each incorrect option is
different from the proportion of females who choose each
incorrect option. This hypothesis is stated in the
alternate form.
In Table 3.1 a hypothetical cross classification of
sex/ option/ and year is presented. Hypothesis 1 is true
for this three dimensional contingency table. In Table 3.2
the three dimensional contingency table is rearranged to
show the year by option contingency table for each gender.

58
Table 3.1
Hypothetical Probabilities of Option Choice
Conditional on Year and Sex
Year
Sex
1
Option
2
3
First
M
.7
.2
.1
F
.5
.3
.2
Second
M
.7
.2
.1
F
.5
.3
.2
Third
M
.7
.2
.1
F
.5
.3
.2
Table 3.2
Hypothetical
Conditional on
Probabilities of
Year and Sex and
Option Choice
Arranged by Sex
Sex Year
1
Option
2 3
M
First
.7
.2
.1
Second
.7
.2
.1
Third
.7
.2
.1
F
First
.5
.3
.2
Second
.5
.3
.2
Third
.5
.3
.2

59
Inspection of the year by option contingency tables shows
that year and option are independent for each gender.
Thus/ hypothesis 1 is equivalent to the hypothesis that
year and option are independent conditional upon gender.
Hypothesis 2 is also true for Table 3.1 and Table 3.2.
Therefore/ when hypothesis 1 is correct/ hypothesis 2
is equivalent to the hypothesis that sex and option choice
are dependent.
In his discussion of the analysis of three dimensional
contingency tables/ Fienberg (1980) presents the following
saturated model for the data:
log m i j k = ü i + Uj + Uk+Uij + Uik+Ujk+U i j k. (3.1)
In this model/ mijk is the expected value of the
frequency in cell ijk of the three dimensional table. The
model states that all three classification factors for a
three dimensional contingency table are mutually dependent.
In the present research i is the year index/ j is the
option index/ and k is the sex index. Fienberg shows that
deleting the terms Uij and Uijk yields a model in which
year and option are independent conditional upon sex. This
model is
log mij k
Ui + Uj + Uk+Uik+Uj k.
(3.2)

60
Fienberg also shows that deleting the Ujterm from (3.2)
to obtain
log mij k = Ui + Uj+Uk + Uik. (3.3)
yields a model that specifies that option is independent of
sex.
Based on Fienberg1 s presentation» an appropriate
analysis for testing the hypotheses is
1. Conduct a likelihood ratio test of the adequacy of
model (3.2). If this test is nonsignificant» then the
data support the adequacy of the model» and the hypothesis
that conditional on gender» option» and year are
independent. Because model (3.1) is a saturated model»
testing the adequacy of model (3.2) is the same as
comparing the adequacies of models (3.1) and (3.2).
2. Conduct a likelihood ratio test comparing the
adequacies of models (3.2) and (3.3). If this test is
significant» then model (3.2) fits the data better than
model (3.3) and the data support the hypothesis that the
choice of incorrect option is dependent on sex.
To summarize» if the first test is nonsignificant and
the second test is significant» then the choice of option
is dependent on sex» and the pattern of dependency is the
same for all three years.

61
A problem arises in interpreting this analysis with
the data used in this research. Over the three years?
there were responses available from 135/340 students. Even
if only 5% of the students answered an item incorrectly/
the responses of 7/767 students would be used in analyzing
this item. On the other hand/ if 90% answered an item
incorrectly/ the responses of 121/806 students would be
used in analyzing the data. As a result of the large
sample size/ the tests described above are likely to be
very powerful. In step 1 of the analysis/ then/ even a
very small change from year to year in the proportion of
males or females who choose an option is apt to be
detected/ and the results will indicate that
hypothesis 1 is not supported. For step 2/ even a very
small dependence of option choice on sex is likely to be
detected and hypothesis 2 is likely to be supported. In
brief/ the problem caused by the large sample size is that
practically insignificant differences may yield statistical
significance.
Fortunately/ the form of the test statistic used in
the likelihood ratio test suggests a reasonable solution to
this problem. The test statistic is
- 2 ^ F0 l°9e (F0/Fe).
(3.4)

62
Here the summation is over all the cells in the contingency
table/ F0 refers to the observed frequency in a cell/ and
Fe refers to the estimated expected frequency in a cell.
Denoting the observed proportion in a cell by PG and the
estimated expected proportion in a cell by Pe/ the test
statistic can be written
G2 = 2 N £^P0 1 ° ge (Po/Pe5- (3.5)
where N is the total number of subjects. This form of the
test statistic suggests the following strategy. For any
significant G2/ using P0 and Pe calculated from the
total data set available for an item/ calculate the minimum
N required for G2 to be significant. If the minimum N is
very large/ this suggests that the statistically
significant result is not practically significant since it
can only be detected in very large samples. Of course/ the
question remains as to what can be considered a minimum
large N. Although there is room for argument/ it seems
reasonable to claim that if an average of 1000 subjects per
year is required to show significance/ then the result is
not likely to be practically significant. On the basis of
this reasoning/ it was proposed to ignore all significant
results that would be nonsignificant if there were less
than 3000 subjects available. In addition/ all log-linear

63
model tests were conducted using a .01 level of
significance.
Since this research is based on Marshall (1981/ 1983)/
it is important to compare the method of analysis used in
this study to the one used by Marshall. Marshall also used
a two-step analysis. In the first step of her analysis she
deleted the Uijk term from (3.1) and tested the adequacy
of the model/
1 ° g m j. j k = ü i + U j + o k+ U i j + Ui k+ Ü j k • (3.6)
Following this/ she deleted the Ujk term to obtain
log m ¿ j k = Ui + Uj + Uk+Uij + Uik. (3.7)
and compared these two models using a likelihood ratio
test. If the first test was nonsignificant and the second
significant/ Marshall claimed that option choice was
dependent on sex and the pattern of dependency was the same
from year to year.
This is the same claim that this study sought to
establish. However/ the approach used here was to
present evidence that model (3.2) fits the data while
Marshall tried to show that model (3.6) fits the data.
The major difference between the two approaches
concerns the operationalization of the concept that the

64
gender-option dependency is the same from year to year. In
this study/ a three dimensional table was considered to
exhibit the same year to year pattern of gender-option
dependency if / conditioned on gender/ the same proportion
of students chose each incorrrect response over the three
year period. This seems to be a straightforward and
natural way to operationalize the concept. To illustrate
how Marshall operationalized the concept in question/ a
hypothetical set of probabilities was constructed
conforming to the pattern specified by Marshall. This is
displayed in Table 3.3.
Table 3 .3
Hypothetical Joint Probabilities of Year/ Option/
and Sex
Year
Sex
1
Option
2
3
First
M
.120
.045 ¡
.015
F
.060
.060 ¡
.030
Second
M
.144
.022 |
.019
F
.072
.030 ¡
.039
Third
M
.132
.040 ¡
.016
F
.066
.054 ¡
J
.039
Note: Probabilities reported are truncated to three
places.

65
One important characteristic of this table involves
two-by-two subtables of sex cross classified with two of
the options for each year. For example* consider the three
two-by-two tables obtained by cross classifying sex and
option choices one and two for each year. These tables are
indicated by dotted lines on Table 3.3. For this table*
the ratio of the odds of a male choosing option one to the
odds of a female choosing option one is the same (within
truncating error) for each year. For example* this odds
ratio for the first year is (.120/.045) / (.060/.060) =
2.67. Within the error caused by reporting truncated
probabilities* the odds ratio for years two and three is
the same as that for year one. The equality of these odds
ratios is Marshall's operationalization stage of the year-
to-year gender-option dependency.
To show that the odds ratio can be constant over
years* but that the probabilities of option choice
conditional on sex and year can change from year to year*
for both males and females* the probabilities in Table 3.3
were converted to the probabilities of option choice
conditional upon sex and year. These conditional
probabilities are reported in Table 3.4. Unlike the
probabilities in Table 3.2* those in Table 3.4 change from
year to year for both males and females.

66
Table 3.4
Hypothetical Probabilities of Option Choice
Conditional on Sex and Year
Year
Sex
1
Option
2
3
First
M
.666
.250
.083
F
.400
.400
.200
Second
M
.774
.120
.104
F
.510
.212
.276
Third
M
.698
.214
.087
F
.431
.352
.215
Note: Probabilities reported are truncated to three
places.
Which procedure is more appropriate to test the claim
that the pattern of male and female option choice remains
the same from year to year? It seems more reasonable to
test for a pattern like that in Table 3.2 than to test for
a pattern like that in Table 3.4 / and/ consequently/ this
was the strategy adopted in this study.
Comparison of Males and Females in Problem Solving
Performance
One object of the study was to compare the performance
of males and females on problem solving. Two questions
were addressed. Firstr do males and females differ in
problem solving performance? Second/ do these differences
persist when computational skill is controlled for/ and do
these differences depend on the level of computational
skill? Seven analyses of covariance (ANCOVA) tests were

67
conducted/ one with each of addition/ subtraction/
multiplication/ division/ addition of fractions/ decimals
in subtraction/ and equivalence as covariates. A well
known problem that arises in the use of ANCOVA is that an
unreliable covariate can cause spurious differences between
the sex groups. To solve this problem/ Porter (1967/1968)
proposed the use of estimated true scores for observed
scores. Porter (1967/1968) conducted a simulation that
gave empirical support to the adequacy of this strategy.
Hunter and Cohen (1974) have provided theoretical support
for this strategy.

CHAPTER IV
RESULTS
Introduction
The data gathered from 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 during three consecutive years were
analyzed in this study. The first objective was to
investigate whether boys and girls differ in the
selection of incorrect responses, and if the pattern of
differences was consistent throughout the three years in
which the test was administered. The second objective was
to investigate whether males and females differ in problem
solving performance, if these differences persist after
accounting for computational skills, and if the male-female
differences depend on the level of computational skills.
The results are discussed in two sections. Study
findings in the area of sex differences in the selection of
incorrect responses are discussed in the first section.
The second section is devoted to an exposition of the
findings in the area of sex-related differences in problem
solving.
Sex-Related Differences in the Selection of
Incorrect Responses
As indicated in Chapter III, there are two models of
interest. The first model indicates that year and option
68

69
are independent/ conditional upon sex*
log m i j k = Ui + Uj + U^+Uij^+Uji^. (3.2)
Substantively/ this model implies that the pattern of male
and female option choices is consistent over the three
years of test administration.
The second model indicates that option is independent
of sexr
log = U i+ u j + uk+ ui k • (3.3)
This model implies that the pattern of option choice is the
same for males and females.
In order to determine if males and females differed in
the selection of incorrect responses and if these
differences were stable across the three years of test
administration/ a two-step test was performed. First/ a
likelihood ratio test of the adequacy of model (3.2) was
conducted for each of the 111 items of the test. Under
this test a model fits the data when the value
obtained is nonsignificant at a specified alpha level. For
50 (45%) of the items/ model (3.2) fitted the data
adequately. For these items the pattern of male and female
option choices was consistent for the three years. The
values for the 61 items for which the likelihood ratio

70
tests were significant at the .01 level are reported in
Table 4.1. Also reported in this table are the actual
sample sizes and the minimum sample sizes necessary for the
likelihood ratio tests to be significant. Of the 61 items*
59 had minimum sample sizes greater than 3*000. Thus*
although for the three years both males and females samples
had inconsistent option choices on these 61 items* on 59 of
the items the inconsistency was relatively minor.
Consequently* these 59 items were included in step 2 along
with the initial 50 items.
In step 2* a likelihood ratio test was performed
comparing the adequacy of the models in (3.2) and (3.3).
The X2 values associated with models (3.2) and (3.3) for
each of the 109 items subjected to step 2 are reported in
columns b and c of Table 4.2. Also reported in Table 4.2
is the difference between the two X2 values (see
column d). This latter figure is the test statistic for
comparing the adequacies of models (3.2) and (3.3).
Significant X2 statistics are indicated by asterisks.
The X2 statistics were significant for 100 of the items
indicating a male-female difference in option choice for
these items. In column e of Table 4.2 the actual sample
sizes are reported. In column f* the minimum sample sizes
necessary for significance are reported. For those 100
items which had significant X2 statistics reported in
column d* 94 (94%) had minimum sample sizes greater than
3*000.

lal
8
13
23
25
28
29
30
31
34
36
37
42
45
48
49
52
54
56
57
71
TABLE 4 .1
Summary of Significant Tests of Model 3.2
X1 for
odel 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(b)
(c)
(d)
46 .15
16,796
3,352
25.74
15,718
5,624
29.25
60,377
19,011
35.90
26 ,707
6,852
46 .00
84 ,847
16,988
61.48
88,396
13,242
25.19
87 ,473
31,982
23 .26
32,870
13 ,015
58.25
82 ,257
13 ,006
84.36
86 ,137
9,404
23 .70
37,068
14,405
23 .26
59,140
23 ,417
22.41
63,098
25,932
179.25
97 ,593
5,014
27.82
16,766
5,550
26.01
39,858
14,114
41.41
74,502
16,570
21.62
44,107
18,789
28.51
53 ,075
17 ,146

laL
59
60
61
62
63
64
65
66
67
68
69
70
71
73
74
76
77
78
79
72
TABLE 4.1 - Continued
X2 for Actual Sample Min. Sample
Model 3.2* Size Size for
Significance
(b)
(c)
LÚ)
50.96
41,829
7,560
66.43
60,494
8,387
21.39
35,632
15,342
55.03
50,699
8,485
158.36
52,068
3 ,028
88.01
73,196
7,660
56 .28
42,429
6 ,943
50.51
78,708
14,352
34.28
51,862
13,935
48.01
63,223
12,128
75.29
62,511
7 ,647
44.39
67 ,528
14,011
14,425.97
84 ,368
54**
24.61
53 ,358
19,969
22.75
84 ,175
34,077
99.68
70,183
6,485
79.38
66,304
7,693
64.56
76,806
10,957
23 .98
48,690
18,700

la)
80
81
82
83
84
85
86
87
88
89
90
91
93
94
95
96
98
101
102
104
73
TABLE 4.1 - Continued
X* for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(b)
(c)
(d)
75.01
55/277
6,787
95 .51
57 ,710
5,565
51.99
71/315
12,633
20.12
81/021
37 ,088
38.62
78/959
18,830
1240 .78
41,180
306**
102.85
39/903
3 ,573
66.75
48,194
6,650
63.01
38,742
5/663
43 .26
39,648
8,441
85 .53
53 ,557
5,767
82.53
62,954
7/025
78.39
80 ,723
9,484
70.15
87,241
11,454
40.59
83 ,969
19,053
29.31
74,287
23 ,343
29.87
69,052
21,291
23 .77
83 ,782
32/462
61.06
75/708
11/419
40.47
72 ,517
16 ,503

7 4
TABLE 4.1 - Continued
Item
Number
X2 for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
la)
(b)
(c)
(d)
105
28.66
60/712
19/510
108
26 .56
86/261
29/912
111
30.71
86/190
25/849
* p<.01
** Minimum sample size less than 3/000

75
TABLE 4.2
Chi-square Values for the Comparison of Models (3.2) and
(3.3) with Actual Sample Sizes and with the Corresponding
Number of Subjects Needed for Significant Results
Item
Num¬
ber
Xa- for
Model
3.2
X1 for
Model
3.3
Xa- for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
1
16.11
30.94
14.83*
7*427
10*061
2
17.54
103.22
85 .68*
11*018
2*583**
3
20.06
47.51
27.45*
11*497
8*733
4
7.54
21.87
14.33*
6*016
8*414
5
14.75
70.40
55.65*
7*706
2*782**
6
11.86
23 .13
11.27*
13*046
23 *256
7
4.60
35.73
31.13*
8*832
5*700
8
46.15
329.98
283.83*
16*796
1*189**
9
10.13
109.99
99.86*
23 *075
4*642
10
9.54
65.29
55.75*
35*060
12*634
11
5.89
19.06
13.17*
24*908
37 *996
12
12.58
198.46
185.88*
40 *423
4*369
13
25.74
62.37
36.63*
15*718
8*621
14
11.19
80 .50
69.31*
30*238
8*765
15
16.63
17.33
0.70
42*889
1*230*914
16
14.51
84.29
69.78*
35*114
10*109
17
17.90
44.20
26 .30*
43 *962
33 *582

76
TABLE 4.2. - Continued
Item
Num¬
ber
X2 for
Model
3.2
X2 for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
18
11.71
15.68
3.97
52/3 40
264/864
19
9.37
18.77
9.40*
41/363
88/402
20
12.37
13.46
1.09
55 /707
1/026 /7 46
21
11.51
19.03
7.52
49/240
131/547
22
9.63
83 .74
74.11*
51/7 24
14/022
23
29.25
153.74
124.49*
60/377
9/744
24
10.35
42.93
32.58*
68/426
42/194
25
35.90
147.40
111.50*
26 /707
4/812
26
18.84
47 .30
28.46*
56/717
40 /037
27
8.78
222.87
214.09*
78/353
7/353
28
46.00
242.78
196.78*
84/847
8/662
29
61.48
170.22
108.74*
88/396
16/331
30
25.19
53 .84
28.65*
87/473
61/338
31
23 .26
47.42
24.16*
32/870
27/333
32
15.51
41.69
26.18*
48/777
37/430
33
7.57
48.84
41.27*
39/470
19/214
34
58.25
324.67
266.42*
82 /257
6/203
35
6.91
225 .66
218.75*
88/069
8/088

77
TABLE 4.2 - Continued
Item
Num¬
ber
Xa for
Model
3.2
Xa for
Model
3 .3
Xa for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
36
84.36
134.56
50 .20*
86 f137
34/472
37
23.70
76.85
53 .15*
37 f06 8
14/011
38
13.64
84 .24
70.60*
38/870
11/0618
39
13.35
757.60
744.25*
61/057
1/648**
40
6.67
701.39
694.72*
56 /438
1/630**
41
12.65
603.42
590.77*
57 /145
1/943**
42
23 .26
189.11
165.85*
59/140
7/154
43
16.29
21.65
5.36
33/712
126 ,357
44
3.55
115.84
112.29*
67/192
12/021
45
22.41
55.90
33.49*
13/098
37/851
46
6.56
16.30
9.74*
56 ,121
117 /007
47
7.82
76.15
68.33*
56 /555
16/628
48
179.25
452.64
273.39*
97 /593
7/172
49
27 .82
59.75
31.93*
16/766
10/549
50
14.23
240.27
226.04*
25/410
2/258**
51
16.37
224.54
208.17*
24 /715
2/385**
52
26.01
96 .14
70.13*
39/858
11/418
53
6.69
27.49
20.80*
44/694
43 /168
54
41.41
95.41
54.00*
74/502
27/718

78
TABLE 4.2 - Continued
Item
Num¬
ber
Xz for
Model
3.2
X1 for
Model
3.3
Xa for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
55
19.53
90 .87
71.34*
41,7 96
11,770
56
21.62
70.14
48.52*
44,107
18,263
57
28.51
35.78
7.27
53 ,07 5
146,668
58
17.49
51.69
34.20*
44,044
25,873
59
50.96
73.68
22.72*
41,829
36 ,987
60
66.43
148.39
81.96*
60,494
14,828
61
21.39
394.54
373.15*
35,632
1,918**
62
55.03
300.95
245.92*
50,699
4,142
63
158.36
278.34
119.98*
52 ,068
8,719
64
88.01
246.48
158.47*
73,196
9,279
65
56 .28
94.11
37.83*
42,429
22,532
66
50.51
113.88
63.37*
78,708
24,953
67
34.28
180 .60
146.32*
51,862
7,121
68
48.01
225.46
177.45*
63,223
7,158
69
75.29
251.66
176.37*
62,511
7,121
70
44.39
271.35
226 .96*
67,528
5,977
72
17.41
208.90
191.49*
84 ,175
8,831
73
24.61
37.80
13.19*
53 ,358
81,271
74
22.75
163.01
140 .26*
84 ,175
12,057

79
TABLE 4.2 - Continued
Item
Num¬
ber
Xa for
Model
3.2
Xz for
Model
3 .3
Xz for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
75
8.69
237.07
228.38*
67 /783
5/963
76
99.68
167 .72
68.04*
70/183
20,723
77
79.38
432.36
352.98*
66/304
3,774
78
64.56
127 .04
62.48*
76/806
24,696
79
23.98
72.59
48.61*
48/690
20/123
80
75.01
128.13
53 .12*
55/277
20,906
81
95.51
106.82
11.31*
57 /710
102/511
82
51.99
108.12
56.13*
71/315
25/525
83
20.12
95.18
75.06*
81 /021
21/685
84
38.62
69.05
30.43*
78,959
52,129
86
102.85
433.47
330.62*
39/903
2/425**
87
66.75
430.40
363.65*
48/194
2/662**
88
63.01
126 .27
63.26*
38/742
12/304
89
43 .26
289.14
245.88*
39/648
3 ,240
90
85.53
405.04
319.51*
53 ,557
3,368
91
82 .53
85.48
2.95
62/954
428,727
92
16.36
93 .12
76.76*
88/231
23 ,092
93
78.39
229.02
150.63*
80 ,723
10,766
94
70.15
137 .27
67.12*
87 ,241
26 ,113

80
TABLE 4.2 - Continued
Item
Num¬
ber
X2 for
Model
3.2
X2 for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
95
40 .59
68.41
27 .82*
83 /969
60/638
96
29.31
212.34
183.03*
74/287
8/154
97
8.60
126.96
118.36*
50/028
8/492
98
29.87
41.77
11.90*
69/052
116/576
99
9.44
9.94
0.50
72/647
2/918/956
100
11.35
41.48
30.13*
60/931
40 /627
101
23.77
315.62
291.85*
83 /782
5/767
102
61.06
437.09
376.03*
75/708
4/045
103
6.12
7.97
1.85
50/663
550/173
104
40 .47
283.94
243 .47*
72/517
5/984
105
28.66
48.59
19.93*
60/712
61/199
106
13 .47
1/682 .71
1/669.34*
71/946
866*'
107
13.13
99.20
86.07*
72/482
16 /918
108
26 .56
79.33
52.77*
86/261
32/840
109
5.73
18.33
12.60*
76/800
112/453
110
17.80
49.11
31.31*
70/980
45 /5 44
111
30.71
477.44
446 .73*
86/190
3/876
pC.oi
Minimum sample size less than 3/000

Sex-related Differences in Problem
Solving Performance
81
In this part/ results for the two questions related to
the second objective of the study are presented. These
questions serve as the framework for the presentation.
Each question is stated/ followed by the results pertaining
to that question.
Question 1: Do males and females differ in problem solving
performance?
The responses of 492 males and 510 females who took
the Puerto Rico Basic Skills Test-6 during the spring of
the first year were analyzed in this study. Also/ data
from 504 males and 509 females tested in the second year
and from 509 males and 504 females tested in the third year
were included in the analysis.
The mean performance scores and the standard
deviations for each of the eight variables are presented in
Table 4.3. Results of t-tests are also presented in this
table. Females outperformed males in problem solving/ a
finding consistently present in all the three years of test
administration. Over the three-year period the mean
differences favored females in all variables except
equivalence. The sex-related differences in problem
solving were significant (p
82
TABLE 4.3
Means* Standard Deviations* and t-Tests for the Eight
Mathematical Subtests
Year/
Subtest
Males
X
SD
Females
X
SD
t
First
Problem Solving
3.436
2.32
3.862
2.45
2.82*
Addition
5.412
1.00
5.592
.86
3.05*
Subtraction
4.550
1.66
4.852
1.56
1.22
Multi plication
4.014
1.80
4.433
1.68
3.81*
Division
3.136
1.87
3.580
1.91
3.72*
Fracadd
2.475
1.75
2.743
1.81
2.32**
Decsub
2.648
1.83
3.321
1.86
5.77*
Equivalence
.77
.77
.73
.76
- .83
N =
492
N =
510
Second
Problem Solving
3.632
2.28
4.015
2.36
2.65*
Addition
5.470
.98
5.603
.86
2.29**
Subtraction
4.843
1.52
4.923
1.44
.86
Multi plication
4.148
1.73
4.550
1.67
3.76*
Division
3.242
1.77
3.632
1.86
3.42*
Fracadd
2.565
1.74
2.903
1.91
2.94*

83
TABLE 4.3 - Continued
Year/
Subtest
Males
X
SD
Females
X
SD
t
Decsub
2.863
1.87
3.440
1.80
5.00*
Equivalence
.76
.76
.65
.75
-2.32**
N =
504
N =
509
Third
Problem Solving
3.927
2.49
4.341
2.44
2.67*
Addition
5 .536
.84
5.704
.64
3.59*
Subtraction
4.836
1.52
4.958
1.46
1.30
Multiplication
4.168
1.74
4.541
1.68
3 .47*
Division
3.343
1.88
3.795
1.83
3.87*
Fracadd
2.819
1.89
3 .117
1.85
2.53**
Decsub
3.021
1.93
3.448
1.85
3.60*
Equivalanee
.830
.82
.800
.78
- .60
N =
509
N =
504
Note: The number of
items in
the problem solving
subtest
was 9 .
In each computation subtestr the number of items
was 6. An item was included in the computation subtest
only if it measured a computation skill required to solve
a problem solving item.
*
**
p <.01
P <.05

84
Consistent significant differences were also found for
addition/ multi plication/ division addition of fractions/
and subtraction of decimals. For subtraction the
difference was not statistically significant.
Question 2: Do sex-related differences in problem solving
persist when computational skills are
controlled for» and is the male-female
differences in problem solving dependent on
level of computational skills?
To address the question of dependence of male-female
problem solving differences in computational skills for
each year and computation subtest/ the possibility of an
interaction was investigated. For the first year/
statistically significant interactions were found between
sex and multiplication/ F (1/998) = 8.59/ p and division/ F (1/998) = 4.25/ p interaction was found between sex and subtraction in the
second year/ F (1 /1009) = 6 .39 / p<].05. No significant
interactions were found in the third year. Analysis of
covariance summary tables are shown as Tables 4.4/ 4.5/ and
4.6. Also/ the three interactions are depicted in
Figures 4.1/ 4.2/ and 4.3. Each figure indicates that at
lower levels of computational skills/ males outperformed
females in problem solving/ with the reverse happening at
higher levels of computational skills.

85
TABLE 4.4
ANCOVA Summary Table: Multiplication Covariate
First Year
Source
df
SS
MS
F
Multiplication
(M) 1
1195.90
1195.90
265.66*
Sex (S)
1
31.95
31.95
7.07*
M x S
1
38.80
38.80
8.59*
Error
998
4509.00
4.51
* p <.01
TABLE 4.5
ANCOVA Summary
Table: Divison Covariate
First Year
Source
df
SS
MS
F
Division (D)
1
1195.90
1195.90
264.66*
Sex (S)
1
13.94
13.94
3.22
D x S
1
18.39
18.39
4.25*
Error
998
4317.00
4.32
* p<.01

86
TABLE 4.6
ANCOVA Summary Table: Subtraction Covariate
Second Year
Source
df
SS
MS
F
Subtraction (S)
1
93 .14
593 .14
122.80*
Sex (S)
1
18.58
18.58
3 .85*
SxS
1
30.85
30.85
6.39*
Error
1009
4873.40
4.83
*P <• 05

Problem Solving
87
9 „
8 -
6
5 -
4 -
3 -
2 -
1 -
T
1
“I 1 1 1 1
2 3 4 5 6
Multiplication
Fig. 4.1 Sex by Multiplication Interaction.

Problem Solving
88
9
8.
7-
5-
4-
3-
2.
1 ' r r t
1 2 3
Division
Fig. 4.2 Sex by Division Interaction
4
5
6

Problem Solving
89
Subtraction
Fig. 4.3 Sex by Subtraction Interaction

90
The significant interactions found between sex and
multiplication^ and sex and division (first year)/ and sex
and subtraction (second year)/ answered/ in part/ the
question of whether male-female differences in problem
solving performance depend on computational ability.
However/ the evidence is quite weak. Of 21 possible
interactions/ only three were significant. No variable
exhibited a significant interaction for each of the three
years.
The analysis of covariance was also used to determine
if sex-related differences exist after controlling for
computational skills. Analyses were conducted for
those variables that did not exhibit significant
interactions with sex. As discussed in Chapter III/
estimated true scores were used for observed scores to
adjust for unreliability of the covariates (the
computational subtests). Reliability coefficients
calculated for each covariate are shown in Table 4.7.
Summaries of the analyses of covariance for the first
year are reported in Table 4.8. The results show that
females retained their superiority in problem solving
performance when equivalence was the controlling variable
in the analysis of covariance/ the only variable in which
males outperformed females (nonsignificant). When the
controlling variables were addition/ subtraction/ addition
of fractions/ and subtraction with decimals/ female

91
TABLE 4.7
Reliability
Three
of the
Years of
Covariates for Each of
Test Administration
the
Year
Covariate
First
Second
Third
Addition
.579
.610
.454
Subtraction
.756
.724
.737
Multiplication
.729
.729
.734
Division
.716
.680
.714
Fracadd
.709
.728
.748
Decsub
.720
.716
.742
Equivalence
.421
.394
.488

92
TABLE 4.8
ANCOVA Summary Table: Other Covariates
First Year
Source
df
SS
MS
F
Addition
1
271.57
271.57
49.74*
Sex
1
15.54
15.54
2.85
Error
999
5453.90
5.46
Subtraction
1
771.78
771.78
155.60*
Sex
1
10.73
10.73
2.16
Error
999
4953.70
4.96
Fracadd
1
1412.11
1412.11
326.87*
Sex
1
7.54
7.54
1.75
Error
999
4313.30
4.32
Decsub
1
1281.10
1281.10
287.89*
Sex
1
4.94
4.94
1.11
Error
999
4444.30
4.45
Equivalence
1
811.44
811.44
164.93*
Sex
1
69.69
69.69
14.16*
Error
999
4914.00
4.92
* p .05

93
superiority in problem solving lost statistical
significance. Adjusted means, associated with the
analyses, are reported in Table 4.9.
Summaries of the analyses of covariance for the second
year are reported in Table 4.10. The analyses show that
females retained their superiority in problem solving after
controlling for equivalence. When the controlling
variables were addition, multiplication, division, addition
of fractions, and subtraction with decimals females
superiority in problem solving lost statistical
significance. Adjusted means are reported in Table 4.11.
The analyses of covariance for the third year are
summarized in Table 4.12. The results of the analyses show
that females maintained their superiority in problem
solving when performance on either subtraction or
equivalence was controlled. When addition, multiplication,
division, addition of fractions, and subtraction of
decimals were the controlling variables, female superiority
in problem solving lost statistical significance. Adjusted
means are reported in Table 4.13.
Summary
Sex-related differences in mathematics achievement
were the subject of this study. The first objective was to
investigate whether males and females differ in the
selection of incorrect responses, and if the pattern of

94
TABLE 4.9
Adjusted Means on Problem Solvingf
by Covariate and Sex
First Year
Covariate
Adjusted Means
Males
Females
Addition
3 .53
3 .78
Subtraction
3 .55
3.76
Fracadd
3.56
3.74
Decsub
3.73
3.58
Equivalence
3 .38
3.91

95
TABLE 4.10
ANCOVA Summary Table: Other Covariates
Second Year
Source
df
ss
MS
F
Addition
1
206.97
206.97
39.57*
Sex
1
19.12
19.12
3.66
Error
1010
5277 .90
5.23
Multiplication
1
1026.68
1026.68
232.80*
Sex
1
.80
.80
.18
Error
1010
4458.30
4.41
Division
1
1117 .46
1117.46
258.67*
Sex
1
.64
.64
.15
Error
1010
4367.50
4.32
Fracadd
1
1341.13
1341.13
327.10*
Sex
1
2.03
2.03
.50
Error
1010
4143.80
4.10
Decsub
1
1453 .91
1453.91
364.39*
Sex
1
4.88
4.88
1.22
Error
1010
4031.10
3 .99
Equivalence
1
797.11
797.11
171.79*
Sex
1
127.33
127.33
27.44*
Error
1010
4687.90
4.64
*
P<-05

96
TABLE 4 .11
Adjusted Means on Problem Solving/
by Covariate and Sex
Second Year
Covariate
Adjusted Means
Males
Females
Addition
3 .69
3 .96
Multiplication
3 .80
3.85
Division
3.80
3.85
Fracadd
3.78
3.87
Decsub
3.90
3 .75
Equivalence
3.46
44.18

97
TABLE 4.12
ANCOVA Summary Table
Third Year
Source
df
ss
MS
F
Addition
1
205.96
205 .96
34.85*
Sex
1
1.88
1.88
.32
Error
1010
5967.60
5.91
Subtraction
1
886.42
886.42
169.48*
Sex
1
24.39
24.39
4.66*
Error
1010
5287.2
5.23
Multi pi icati on
1
1402.42
1402.42
297.12*
Sex
1
1.05
1.05
.22
Error
1010
4771.20
4.72
Division
1
2028.49
2028.49
494.75*
Sex
1
1.11
1.11
.27
Error
1010
4145 .10
4.10
Fracadd
1
1733.72
1733.72
393.97*
Sex
1
4.64
4.64
1.05
Error
1010
4439.89
4 .40
Decsub
1
2116.84
2116.84
526 .57*
Sex
1
.15
.15
.04
Error
1010
4056 .76
4.02
Equivalence
1
1186.92
1186.92
240.26*
Sex
1
59.85
59.85
12.12*
Error
1010
4986.69
4.94
*
p<. 05

98
TABLE 4.13
Adjusted Means on Problem Solving/
by Covariate and Sex
Third Year
Covariate
Adjusted Means
Males
Females
Addition
4.09
4.18
Subtraction
3.98
4.29
Multiplication
4.10
4.17
Division
4.17
4.16
Addf rac
4.07
4.20
Subdec
4 .15
4.12
Equivalence
3 .79
4.45

99
responses was consistent throughout the years in which the
test was administered. Males and females selected
different incorrect responses in 100 of the 111 items of
the test. The pattern of male-female differences in the
selection of incorrect responses was consistently found in
each of the years of test administration. However/ for the
vast majority of the 100 items/ male-female differences
were relatively small in magnitude/ considering the fact
that the number of subjects needed to obtain significance
was very large. Therefore/ these findings lack educational
significance.
A second objective of the study was to investigate
whether males and females differ in problem solving
performance/ if these differences persist after controlling
for computational skills/ and if the male-female
differences depend on the level of computational skills.
Results of the analyses reported in this section
generally indicated that females outperformed males not
only in problem solving/ but in six of the seven
computational variables. Males surpassed females in
equivalence/ but statistical significance was obtained for
only one of the three years covered in the study.
The results also tended to show that/ for examinees
with similar levels of computational skills/ sex
differences in problem solving did not exist. The only

100
exception was that/ when equivalence was controlled/
sex-related differences in favor of females/ persisted.
The question of whether male-female differences in
problem solving depend on computational skills was answered
partially/ in the affirmative. Interactions/ indicating
the dependence/ were significant only for multiplication
and division in the first year and for subtraction in the
second year.

CHAPTER V
DISCUSSION
This cnapter comprises two sections: a) summary and
interpretation of the results of the study/ and
b) implications of the findings/ with suggestions for
further research.
Summary and Interpretation of the Results
The main objective of this study was to investigate
whether there were sex-related differences in incorrect
responses/ and if the pattern of response was consistent
over three consecutive administrations of the mathematics
test. Males and females selected different incorrect
responses in 100 of the 111 items of the test and this
pattern of response was consistenly found in each of the
three years of test administration.
The results tend to support/ in part/ Marshall's
findings (1981/ 1983) of observed sex-related differences
in incorrect responses/ but in no way do they point to a
conclusion that 6th grade boys and girls from the public
schools in Puerto Rico exhibit different problem solving
strategies. The significant differentiated pattern of
101

102
response was related to the use of a large number of
subjects in the analysis.
One must question whether analysis of incorrect
responses has any educational value. Although the results
of this study do not support the use of this analysis
(error analysis) for the study of sex-related differences*
it continues to be a promising technique in the discovery
of how children approach mathematics problem solving* more
so than analyses that look only at the items answered
correctly or at total test scores.
The second objective of the study was to investigate
sex-related differences in problem solving performance* to
find out if significant differences persisted after
controlling for computational skills* and if the
differences depended on the level of computational skills.
The results showed that females outperformed males in
problem solving and in six of the seven computational
skills. Males showed superiority in equivalence in each of
the three years* but statistical significance was obtained
only for the second year. The results also showed that
females retained their superiority in problem solving* only
when equivalence (for the three years data) and subtraction
(for the second year) were the controlling variables.
Analysis of covariance was also used to answer the
question of whether male-female differences in problem
solving depend on the level of computational skills. The

103
question was answered/ partially/ in the affirmative.
Interactions/ indicating the dependence/ were significant
for multiplication and division in the first year and for
subtration in the second year.
In general/ the sex-related differences in mathematics
problem-solving/ in favor of males performance/ reported by
Marshall (1981/ 1984)/ and Fennema and Tartree (1983) were
not supported in this investigation. Female performance
was equal or higher than male performance in problem
solving.
The findings of the study have something in common
with other investigations where mathematics problem solving
has been the subject of interest/ the fact that males and
females do poorly in word-story problem items (problem
solving) . The mean score in problem solving for males and
females in the 9-item problem solving subtest was 3.436 and
3.862 respectively for the first year/ 3.632 and 4.015 for
the second year/ and 3.927 and 4.341 for the third year.
Implications of the Findings and
Suggestions for Further Research
The main purpose of this study was to investigate
sex-related differences in the selection of incorrect
responses. Another objective was to compare male and
female performance after accounting for computational
ability. This type of research is conducted in an attempt

104
to contribute to the body of knowledge regarding such
differences in mathematical learning and achievement. This
section of Chapter V comprises implications of the findings
and recommendations evolving from the present
investigation.
Sex-related Differences in Incorrect Responses
The main purpose of education is to impart formal
education equally for the sexes. Analysis of the data on
incorrect responses showed that for 100 of the 111 items of
the test/ males and females selected different incorrect
responses/ and this pattern of responses was consistently
found during the three years of test administration.
However/ with a sample of less than 3/000 subjects/
statistical significant results were obtained for only
6 items. Therefore/ it seems that/ at least for 6th grade
children/ educators need not concern themselves with
varying teaching techniques for the sexes.
This study/ however/ has value apart from the
investigation of sex-related differences. It was found
that/ year after year/ 6th graders from the public schools
in Puerto Rico tend to make the same errors in the solution
of mathematical problems. It is important/ therefore/ that
teachers provide the learning strategies necessary to
enable students to obtain a better understanding of the
concepts of numbers.

105
¿exr_related.,Differences in Problem Solving Performance
The results of this study do not give strong support
to Marshall's (1981/ 1984) findings that the effect of
computational skills on problem solving performance is
different for each sex. However/ the findings of the
present investigation are consistent with those found by
Meyer (1978) and Whitaker (1976): female performance in
problem solving is not significantly different from male
performance. Females showed superiority in problem solving
performance in all three years of test administration
although their superiority was retained only when
performance in equivalence was controlled for (for the
three years data) and when performance in subtraction was
controlled for (second year) (statistically significant
only after accounting for equivalence) .
The findings of this study are also in agreement with
those reported by the National Assessment of Educational
Progress (NAEP) (Carpenter et al.z 1980). Related to
problem solving/ they conclude/
if it were necessary to single out one area that
demands urgent attention/ it would clearly be
problem solving. At all age levels/ and in vir¬
tually every content area/ performance was extre¬
mely low on exercises requiring problem solving
or application of mathematical skills. In gene¬
ral/ respondents demonstrated a lack of the most
basic problem-solving skills. Rather than attempt¬
ing to think through a problem and figure out what
needed to be done to solve the problem/ most res¬
pondents simply tried to apply a single arithmetic

106
operation to the numbers in the problem. The re¬
sults indicate that students are not familiar with
such basic problem-solving strategies as drawing
a picture of a figure described in a problem or
checking the reasonableness of a result, (p. 338)
The fact that, in general» after controlling for
computational skills» there were no statistically
significant differences between the problem solving
performance of the males and females who participated in
this study would seem to imply that teachers need not worry
about designing different teaching strategies for the
sexes. However^ because the performance in problem-solving
is low for both sexes» teachers may be encouraged to
emphasize problem solving in their daily teaching.
This study was designed to investigate sex-related
differences in mathematical performance. The findings
clearly demonstrated that 6th grade females and males
from the public schools in Puerto Rico are equally good or
equally bad in the solution of word-story items (problem
solving). Also/ in general/ the effect of computational
skills is similar for each sex. Based on these findings/
it appears that greater differences exist within each sex
than between the sexes. In conclusion/ the findings of the
present study highlight the need for investigations of
different approaches to instructional designs aimed at
improving both male and female performance in problem
solving.

REFERENCES
Aiken» L. R. (1970). Attitudes toward mathematics.
Review of Educational Resarch. IQ./ 551-596 .
Alien» R. H.» & Chambers» D. L. (1977). A comparison of
the mathematics achievement of males and females
(Bulletin No. 9194). Madison: Wisconsin State
Department of Public Instruction» Division for
Management and Planning Service. (ERIC Document
Reproduction Services No. ED 159 073)
Armstrong» J. M. (1979) . Achievement and participation of
W.pme_n_in_.math.emati£sj ¿Ln_Q_y.ejyi-e.W-. (Report of a
two year study). Denver» CO: Educational Commission
of the States» National Assessment of Educational
Progress. (ERIC Document Reproduction Service
No. ED 184 878)
Armstrong» J. M.» & Price» R. A. (1982). Correlates and
predictors of women's mathematics participation.
Journal for Research in Mathematics Education» 11»
99-109.
Astin» H. S. ( 197 4) . Overview of the findings. In H.
Astin» H. Suniewick» & S. Dweck (Eds.)» Women: A
Bibliography on their Education and Careers.
(pp. 1-10). New York: Behavorial Publications.
Backman» M. E. (1972). Patterns of mental abilities:
Ethnic» socioeconomic and sex differences. American
Educational Research Journal» 1» 1-11.
Ballew» H.» & Cunningham» W. (1982). Diagnosing strengths
and weaknesses of sixth-grade students in solving
word problems. Journal for Research in Mathematics
Education» 13» 202-210.
Balow» I. H. (1964). Reading and computation ability as
determinants of problem solving. The Arithmetic
Teacher» 11» 18-22.
Bar-Tal» D. » & Frieze» J. H. ( 1977 ) . Achievement
motivation for males and females as a determinant of
attribution for success and failure. Sex roles.
1» 301-313.
107

108
Bern* S. L. / & Bern D. V. (1970). We are all nonconscious
sexists. Psychology Todays 22-24/ 26 / 115-116.
Benbowf C. P./ & Stanley/ J. C. (1980). Sex differences
in mathematical ability: Fact or artifact?
Science/ 210/ 1262-1264.
Benbow/ C. P./ & Stanley/ J. C. (1982). Consequences in
high school and college of sex differences in mathema¬
tical reasoning ability: A longitudinal perspective.
American Educational Research Journal/ 12/
598 - 622.
Bobbie/ C. N. (1971). Sex-role preference and academic
achievement (Doctoral dissertation/ Yeshiva Univer¬
sity) . £iss.ejtation Abstracts International/ 22/
1818B-1819B.
Carey/ G. L. (1958). Sex differences in problem solving
performance as a function of attitude differences.
Journal of Abnormal and Social Psychology/ 2£./
256-260.
Carpenter/ T. P./ Corbitt/ M. K./ Kepner/ H. S./ Lindquist/
M. M. / & Reys/ R. E. (1980). Results of the second
NAEP mathematics assessment: Secondary school.
Mathematics Teacher/ 73, 329-338.
Carpenter/ T. P. / Lindquist/ M. M. / Mathews/ W. / & Silver/
E. A. (1984). Results of the third NAED mathematics
assessment: Secondary school. Mathematics Teacher/
2£/ 652-659.
Chase/ C. I. (1960). The position of certain variables in
the prediction of problem solving in arithmetic.
Journal of Educational Research/ 21/ 9-15.
Cohen/ D./ & Wilkie/ F. (1979). Sex-related differences in
cognition among the elderly. In M. A. Witting and
A. C. Petersen (Eds.)/ Sex-related differences in
cognitive functioning: Developmental Issues/ New
York: Academic Press.
Connor/ J. M. / Serbin/ L. A./ & Schakman/ M. (1977). Sex
differences in children's response to training on
a visual-spatial test. Developmental Psychology/
12./ 392-394 .

109
Creswell/ J. L. (1982/ February). Sex-related differences
i..n_problem solving in rural black/ anglo and Chicago
adolescents. Paper presented at the annual meeting
of the Southwest Educational Research Association.
Austin/ TX. (ERIC Document Reproduction Service
No. ED 216 895)
Dwyer/ C. A. ( 197 4) . Influence of children's sex role
standards on reading and arithmentic achievement.
Journal of Educational Psychology/ ££/ 811-816.
Eccles/ J. ( 1 9 83 ) . Sex differences in mathematics
participation. In M. Steinkamp & M. Maehr (Eds.)/
Women in Science (pp. 80-100). Greenwich/ CT: JAI
Press.
Elton/ C. F. / & Rose/ H. A. ( 1967). Traditional sex
attitudes and discrepant ability measures in college
women. Journal of Counseling Psycology/ M*
538-548.
Ernest/ J. (1976). Mathematics and Sex. Santa Barbara:
University of California.
Exedisisz R. H. (1983). An investigation of the relation¬
ship of reading comprehension/ vocabulary/
mathematical concepts/ and computation on problem
solving among anglo/ black/ and Chicago male and
female middle school adolescents (Doctoral
Dissertation/ University of Houston/ 1982).
Dissertation,Abstracts International/ 42.,
226 4A-226 5A.
Fennema/ E. (1974). Mathematics learning and the sexes:
A review. Journal for Research in Mathematics
Education/ ¿/ 126-139.
Fennema/ E. (1975). Mathematics/ spatial ability and the
sexes. In E. Fennema/ (Ed.)/ Mathematics: What
research savs about sex differences (pp. 33-44).
Columbus: Ohio State University/ College of Education.
(ERIC Document Reproduction Service No. ED 128-195)
Fennema/ E. (1976) . Influences of selected cognitive
affective/ and educational variables on sex-related
djff_exgnc.e.s^,.n-mathematics learning and studying.
Madison: University of Wisconsin/ Department of
Curriculum and Instruction. (ERIC Document
Reproduction Service No. ED 142 906)

110
Fennemar E. * & Shennanr J. (1977). Sex-related
differences in mathematics achievements spatial
visualizations and affective factors. American
Educational.Research Journals 14' 51-71.
Fennemas E.s & Shermans J. A. (1978). Sex-related
differences in mathematics achievement and related
factors. A further study. Journal for Research in
Mathematics .Education/ 9s 189-204.
Fennemas E.s & Tartrees L. A. (1983). Research on
relationship of spatial visualization and confidence
to male/female mathematics achievement in grades 6-8.
Washingtons DC: National Science Foundation. (ERIC
Document Reproduction Service No. ED 232 853)
Fienbergs S. E. (1980). The Analysis of .Cross-Classified
Categorical Data. Cambridges MA: The MIT Press.
Foxs L. H. (1975a). Mathematically precocious: Male or
female. In E. Fennema (Ed.) Mathematics learning:
What research says.about sex differences (pp. 1-12).
Columbus: Ohio State Universitys College of
Education. (ERIC Document Reproduction Service No. ED
128 195)
Foxs L. H. (1975b). Sex differences: Implications
for program planning for the academically gifted.
Paper presented at the Lewis M. Terman Memorial
Symposium on Intellectual Talent held at the Johns
Hopkins University* Baltimore* MD.
Fox* L. H. (1977). The effects of sex role socialization
on mathematics participation and achievement. In J.
Shoemaker (Ed.)* Women_and_Mathe.mati,cs• Research
Perspectives for Change* (N.I.E. Papers in Educa¬
tion and Work: No. 8). Washington* D. C. : Eduation
and Work Group* The National Institute of Education.
Glennon* V. J.* & Callahan* L. G. (1968) Elementary school
mathematics: A_guide._tp_..cur rent research*
Washington* D. C.: Association for Supervision
and Curriculum Development.
Guay* R. B.* & McDaniel* D. (1977). The relationship
between mathematics achievement and space ability
among elementary school children. Journal for
Research in Mathematics Education* 8* 211-215.

Ill
Hansen/ C. W. (1944). Factors associated with successful
achievement in problem solving in sixth grade
arithmentic. Journal of Educational Research# 38 #
111-119.
Harnisch# D. L. (1983). Item response patterns:
Application for educational practice. Journal of
Educational Measurement# 21/ 191-205.
Hilton# T. L. # & Berglund# G. W. (1974). Sex differences
in mathematics achievement: A longitudinal study.
Journal of. Educational Research# £2# 231-237.
Hunter# J. E. # & Cohen# S. H. ( 1974). Correcting for
unreliability in nonlinear models of attitude change.
Psychometrika# 39 # 445-468.
Jarvis# 0. T. (1964). Boy-girl ability differences in
elementary school arithmentic. School Science and
Mathematics# 64 # 657-659.
Kaufman# A. (1984). Sex differences in mathematics
reasoning in mathematically gifted and average fifth
and sixth graders (Doctoral dissertation# Hofstra
University# 1983) . Dissertation Abstracts Inter¬
national# 45 # 1094A-1095A.
Kilpatrick I. (1969). Problem solving and creative
behavior in mathematics. In J. W. Wilson# & L. R.
Carry (Eds.)# Studies in mathematics (Vol. No. 19).
Stanford# CA: School Mathematics Study Group.
Kloosterman# P. W. (1985). Attributional theory# learned
helplessness# and achievement in ninth grade
mathematics (Doctoral dissertation# University of
Wisconsin# 1984) . Dissertation Abstracts Internatio¬
nal# 41/ 919A-920A.
Knifong# J. D. # & Holtan# B. ( 1976) . An analysis of
children's written solutions to word problems.
Journal for Research in Mathematics_Education# 2#
106-112.
Knifong# J. D.# & Holtan# B. (1977). A search for reading
difficulties among erred word problems. Journal for
Research in Mathematics Education# 1# 227-230.

112
LandaUf M. (1984). The effects of spatial ability and
problem presentation format in mathematical problem
solving performance on middle school students
(Doctoral dissertation/ Northwestern University).
Dissertation Abstracts International/ 4¿/ 442A-
443 A.
Lindgren/ H. C. / Silva/ I./ Faraco» I./ & DaRocha/ N. S.
(1964). Attitudes toward problem solving as a
function of success in arithmetic in Brazilian
elementary schools. Journal of Educational Research/
¿2., 44-45 .
Maccoby/ E. E. (1966). Sex differences in intellectual
functioning. In E. E. Maccoby (Ed.)/ The Development
of Sex Differences (pp. 25-55). Stanford/ CA: The
Stanford University Press.
Maccoby/ E. E. / & Jacklin/ C. N. (1972). Sex differences
in intellectual functioning. In A. Anastasi/ (Ed.)/
Assessment in a pluralistic society, (pp. 37-55) .
Proceedings of the 1972 Invitational Conference on
Testing Problems/ Educational Testing Service/
Princeton/ NJ.
Maccoby/ E. E./ & Jacklin/ C. N. (1974) . The Psychology of
Sex Differences/ Stanford/ CA: Stanford University
Press.
Marshall/ S. P. (1981) . Sex Differences in Sixth Grade
Children's Problem Solving. Paper presented at the
annual meeting of the American Educational Research
Association/ Los Angeles/ CA.
Marshall/ S. P. (1983). Sex differences in mathematical
errors: Analysis of distractor choice. Journal for
Research in Mathematics Education/ 14./ 325-336 .
Marshall/ S. 'p. (1984). Sex differences in children's
mathematics achievement: Solving computations and
story problems/ Journal of Educational Psychology/
lAf 194-204.
Martin/ M. D. (1964). Reading comprehension/ abstract
verbal reasoning and computation as factors in
arithmetic problem solving (Doctoral dissertation/
University of Iowa/ 1963) . Dissertation Abstracts
International/ 24/ 4547A-4548A.

113
Meyer? R. A. (1978). Sex-related differences in mathema¬
tical problem solving performance and intellectual
abilities. Madison: University of Wisconsin?
Research and Development Center for Cognitive
Learning. (ERIC Document Reproduction Service
No. ED 166 023).
Milton? G. A. (1957) . The effects of sex-role identifica-
cation upon problem solving skill. Journal of
Abnormal & Social,Psychology? 5¿? 208-212.
Moller? N. J. (1983). The impact of gender? masculinity?
and feminity on math achievement and course decisions
(Doctoral dissertion? Purdue University? 1982).
£is_sertation_Abstracts International? 42/ 2584A.
Muth? K. D. (1984). Solving arithmetic word problems:
Role of reading and computational skills. Journal of
Educational Psychology? 76 ? 205-210.
Pallas? A. M.? & Alexander? K. L. (1983). Sex differences
in quantitative SAT performance: New evidence on
the differential coursework hypothesis. American
Educational Research Journal? 21/ 165-182.
Plank? E.? & Plank? R. (1954). Emotional components in
arithmetical learning as seen through autobiographies.
Psychological Studies of the Child? 2.? 274-296.
Porter? A. C. (1968). The effects of using fallible
variables in the analysis of covariance (Doctoral
dissertation? University of Wisconsin? 1967).
Dissertation Abstract International? 2£? 3517B.
Probert? B. S. (1984). Math confidence workshops: A
multimodal group intervention strategy in mathematics
anxiety/avoidance. (Doctoral dissertation? University
of Florida? 1983). Dissertation Abstracts
International? 44 ? 2231B.
Puerto Rico Department of Education. (1980). Basic
skills test in mathematics-6. Hato Rey: Center of
Evaluation.
Radatz? H. (1979) . Error analysis in mathematics educa-
tion. Journal for Research in Mathematics Educa¬
tion ? 10 ? 163—172.
Riedesel? A. C. (1969). Problem solving: Some suggestions
from research. Arithmetic Teacher? 16 ? 54-58.

114
Sherman/ J. (1979) . Predicting mathematics performance in
high school girls and boys. Journal of Educational
Psychology/ 71/ 242-249.
Sherman/ J. (1980). Mathematics/ spatial visualization/
and related factors: Changes in girls and boys/
grades 8-11. Journal of Educational Psychology/
22/ 476-482.
Stein/ A. H. (1971). The effects of sex role standards
for achievement and sex role preference on three
determinants of achievement motiviation.
Developmental Psychology/ 4./ 219-231.
Stein/ A. H./ & Smithless/ J. (1969). Age and sex
differences in children's sex role standards about
achievement/ Developmental Psychology/ 2/ 252-259.
Suydam/ M. N. (1971) . Research on mathematics education
(K-12) reported in 1970. Journal for Research in
Mathematics Education/ 2/ 257-298.
Suydam/ M. N. (1972). Research on mathematics education
(K-12) reported in 1971. Journal for Research in
Mathematics Education/ 2, 196-232.
Suydam/ M. N. (1973) . Research on mathematics education
(K-12) reported in 1972. Journal for Research in
Mathematics Education/ 1/ 205-242.
Suydam/ M. N. (1974). Research on mathematics education
(K-12) recorded in 1973. Journal for Research in
Mathematics Education/ ¿/ 238-272.
Suydam/ M. N. / & Weaver/ J. F. ( 1975) . Research on
mathematics education (K-12) recorded in 1974.
Journal for Research in Mathematics Education/ £/
253-282.
Verbeke/ K. A. (1983). Sex-related differences in
mathematically gifted secondary students: An
investigation of selected cognitive/ affective/ and
educational factors (Doctoral dissertation/
University of Maryland/ 1982) . Dissertation Abstracts
International/ 43/ 2267A-2268A.
Weiner/ B. (1974). Achievement motivation and attribution
theory. Morristown/ NJ: General Learning Press.

115
Whitaker, D. R. (1976). A study of the relationship
between selected noncognitive factors and the problem
solving performance of fourth-grade children. (Tech.
Rep. No. 396). Madison: University of Wisconsin/
Research and Development Center for Cognitive
Learning.
Wiser L. K.r Steel» L.r & MacDonald/ C. (1979).
Origins and career consequences of sex differences in
high school mathematics achievement (Grant No. NEI-G
7 8-001 ) . Palo Alto/ CA : American Institutes for
Research.
Zalewski/ D. L. (1974). An exploratory study to compare
two performance measures: An interview-coding
scheme of mathematical problem solving and a written
test (Doctoral Dissertation/ University or Wisconsin/
Madison). Dissertation Abstracts International/
1£# 3797A.

BIOGRAPHICAL SKETCH
Sonia Feliciano was born on March 17/ 1938/ in
Mayaguez/ Puerto Rico. She graduated from Eugenio Ma. de
Hostos High School/ in her hometown. She received a degree
in business administration/ with a major in finance/ from
the University of Puerto Rico in June/ 1961.
After graduation she worked as junior accountant for
Pittsburgh Plate Glass/ International. During the academic
year of 1969-70 she started her career in education/
obtaining the degree of Master in Education from the
University of Puerto Rico in June/ 1975.
From January/ 1971 to June/ 1985/ she served in the
public school system of Puerto Rico in different positions.
She worked as educational researcher until July/ 1977/ when
was appointed Director of the Research Center of the
Commonwealth Department of Education.
During the academic year of 1978-79/ she initiated
doctoral studies in the Faculty of Philosophy and
Educational Sciences of the Universidad Complutense de
Madrid/ in Madrid/ Spain. She completed the coursework
requirements and then transferred to University of Florida
during the fall of 1979.
116

117
From May* 1982 to July* 1983 / she worked as graduate
research assistant at the Foundations of Education
Department/ University of Florida. She returned to Puerto
Rico in August/ 1983 / to serve as Special Aide to the
Assistant Secretary for the Vocational/ Technical/ and High
Skills Educational Programs of the Commonwealth Department
of Education.

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequatef in scope and qualityf
as a dissertation for the degree of Doctor of Philosophy.
Jam\es Algina* Chairman
Processor of Foundations of Education
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequatef in scope and qualityf
as a dissertation for the degree of Doctor of Philosophy.
Linda Crocker
Professor of Foundations of Education
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequatef in scope and qualityf
as a dissertation for the degree of Doctor of Philosophy.
, Jüyfrnl
njaery
of Educational Leadership
lichael
Professor
This dissertation was submitted to the Graduate Faculty of
the College of Education and to the Graduate Schoolf and
was accepted as partial fulfillment for the degree of
Doctor of Philosophy.
Augustf 1986
Chairmanf Foundations of Education
Deanf College of Education
Deanf Graduate School

UNIVERSITY OF FLORIDA
3 1262 08554 6660



84
Consistent significant differences were also found for
addition* multiplication* division* addition of fractions*
and subtraction of decimals. For subtraction the
difference was not statistically significant.
Question 2: Do sex-related differences in problem solving
persist when computational skills are
controlled for* and is the male-female
differences in problem solving dependent on
level of computational skills?
To address the question of dependence of male-female
problem solving differences in computational skills for
each year and computation subtest* the possibility of an
interaction was investigated. For the first year*
statistically significant interactions were found between
sex and multiplication* F (1*998) = 8.59* p and division* F (1*998) = 4.25* p interaction was found between sex and subtraction in the
second year* F (1 *1009) = 6 .39 p<].05. No significant
interactions were found in the third year. Analysis of
covariance summary tables are shown as Tables 4.4* 4.5* and
4.6. Also* the three interactions are depicted in
Figures 4.1* 4.2* and 4.3. Each figure indicates that at
lower levels of computational skills* males outperformed
females in problem solving* with the reverse happening at
higher levels of computational skills.


98
TABLE 4.13
Adjusted Means on Problem Solving*
by Covariate and Sex
Third Year
Covariate
Adjusted
Males
Means
Females
Addition
4.09
4.18
Subtraction
3.98
4.29
Multiplication
4.10
4.17
Division
4.17
4.16
Addfrac
4.07
4.20
Subdec
4.15
4.12
Equivalence
3.79
4.45


18
students who had taken a similar number of mathematics
courses and were in the top 85% of the class in mathematics
achievement. They were tested in 1976. Four high school
districts were included. In only one of the high school
districts were sex-related differences in application and
problem solving found# in favor of males. They concluded
that when relevant factors are controlled/ sex-related
differences in favor of males do not appear often/ and when
they do/ they are not large.
Sherman (1980) investigated the causes of the emerging
sex-related differences in mathematics performance/ in
favor of males/ during adolescence (grades 8-11). She
wanted to know if these differences emerge as a function of
sex-related differences in spatial visualization and
sociocultural influences that consider math as a male
domain. In grade 8/ she used the Romberg-Wearne Test and/
in grade 11/ a mathematical problem solving test derived
from the French Kit of Tests.
The analysis showed that for girls/ problem solving
performance remained stable across the years. Mean problem
solving performance for boys/ however/ was higher in grade
11 than in grade 8. No sex-related differences were found
in grade 8/ but boys outperformed girls in grade 11/ where
the Stafford test was used.
Sherman found that for both sexes problem solving
performance in grade 8 was the best predictor of problem


40
Achievement Test (subtests of reading/ arithmetic/ and
reasoning) and the California Short-Form test of mental
ability to a group of 1/400 children from the 6th grade.
All levels of achievement were included in the analysis.
Analysis of variance and covariance were used and compared.
He confirmed the findings of other researchers to the
effect that there is a direct relationship between I.Q. and
reading ability/ and between I.Q. and computational skills.
The results of the analysis of variance revealed that
increases in computation ability were associated with
higher achievement in problem solving. A relationship
between reading ability and problem solving was also found/
but it was not as strong. Significant differences in
problem solving performance associated with computational
ability were found when intelligence was controlled.
Balow concluded that computation is a much more
important factor in problem solving than reading ability/
and that when I.Q. is taken into consideration/ the degree
of the relationship between reading and problem solving
ability becomes less pronounced. Intelligence tends to
confound the relationship between these two variables.
Knifong and Holtan (1976/ 1977) attempted to
investigate the types of difficulties children have in
solving word problems. They administered the word problem
section of the Metropolitan Achievement Test to 35 children
from the 6th grade. Errors were classified in two


74
TABLE 4.1 Continued
Item
Number
Xa for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(a)
... (b)
Cc)
(d)
105
28.66
60/712
19/510
108
26.56
86/261
29/912
111
30.71
86/190
25/849
* p<.01
** Minimum sample size less than 3/000


laL
59
60
61
62
63
64
65
66
67
68
69
70
71
73
74
76
77
78
79
72
TABLE 4.1 Continued
X2 for Actual Sample Min. Sample
Model 3.2* Size Size for
Significance
(b)
(c)
L)
50.96
41,829
7,560
66.43
60,494
8,387
21.39
35,632
15,342
55.03
50,699
8,485
158.36
52,068
3 ,028
88.01
73,196
7,660
56 .28
42,429
6 ,943
50.51
78,708
14,352
34.28
51,862
13,935
48.01
63,223
12,128
75.29
62,511
7 ,647
44.39
67 ,528
14,011
14,425.97
84 ,368
54**
24.61
53 ,358
19,969
22.75
84 ,175
34,077
99.68
70,183
6,485
79.38
66,304
7,693
64.56
76,806
10,957
23 .98
48,690
18,700


61
A problem arises in interpreting this analysis with
the data used in this research. Over the three yearsr
there were responses available from 135*340 students. Even
if only 5% of the students answered an item incorrectly*
the responses of 7*767 students would be used in analyzing
this item. On the other hand* if 90% answered an item
incorrectly* the responses of 121*806 students would be
used in analyzing the data. As a result of the large
sample size* the tests described above are likely to be
very powerful. In step 1 of the analysis* then* even a
very small change from year to year in the proportion of
males or females who choose an option is apt to be
detected* and the results will indicate that
hypothesis 1 is not supported. For step 2* even a very
small dependence of option choice on sex is likely to be
detected and hypothesis 2 is likely to be supported. In
brief* the problem caused by the large sample size is that
practically insignificant differences may yield statistical
significance.
Fortunately* the form of the test statistic used in
the likelihood ratio test suggests a reasonable solution to
this problem. The test statistic is
(3.4)


15
The test is composed of 23 items designed to yield 3
scores: a comprehension scorer an application scorer and a
problem solving score. The comprehension question
ascertains whether a child understands the information
given in the item stem. The application question assesses
the child's mastery of a prerequisite concept or skill of
the problem solving question. The problem solving question
poses a question whose solution is not immediately
availabler that isr a situation which does not lend itself
to the immediate application of a rule or algorithm. The
application and problem solving parts of the test may refer
to a common unit of information (the item stem) but the
questions are independent in that the response to the
application question is not used to respond to the problem
solving question.
Meyer (1978) / Whitaker (1976) / Fennema and Sherman
(1978)/ and Sherman (1979) have used the Romberg-Wearne
test in their studies. Meyer (1978) investigated whether
males and females differ in problem solving performance and
examined their prerequisite computational skills and
mathematical concepts for the problem solving questions. A
sample of 179 students from the 4th grade were administered
19 "reference tests" for intellectual abilities and the
Romberg-Wearne test. The analysis showed that males and
females were not significantly different in the
comprehension/ application/ and problem solving questions


55
1/000 students were selected randomly for each year. For
the first yearr 1/002 were selected (492 boys and
510 girls); for the second year/ 1/013 students were
selected (504 boys and 509 girls); and/ for the third year/
1/013 students were selected (509 boys and 504 girls).
The student population in Puerto Rico includes
children from the urban and rural zones and comprises
children from low and middle socioeconomic levels.
Findings can be generalized only to this population.
The Instrument
The Basic Skills Test in Mathematics-6 is a
criter ion-referenced test used in the Department of
Education of Puerto Rico as part of the annual assessment
program. The test measures academic achievement in
operations/ mathematical concepts/ and story problems. It
has a reported split half reliability or .95. The test was
designed specifically for Puerto Rico. Its contents and
the procedures followed for its development were formulated
and reviewed by educators from the mathematics department
of the Department of Education of Puerto Rico/ in
coordination with the Evaluation Center of the Department
of Education and mathematics teachers from the school
districts. The emphasis placed on each skill area is
depicted in Figure 3.1.


Problem Solving
88
9
8.
7-
5-
4-
3-
2.
1 r r t
1 2 3
Division
Fig. 4.2 Sex by Division Interaction
4
5
6


5
Females were better on items of computation and males were
more successful on word-story problem items (problem
solving). She also found that females successful
performance in the problem solving items was more dependent
on their successful performance in the computation items.
Males did not need as much as females to succeed in the
computation items in order to answer correctly the problem
solving items.
Although the general findings seem to support sex
differences in mathematics learning and achievement the
research done does not consistently support superiority for
either sex. Most of the research has been concerned with
how the sexes differ on subtests or total test scores in
mathematics. Moreover the great majority of the studies
deal with correct responses. Sex differences in incorrect
responses at the item level have not been fully researched.
Only two studies dealing with sex differences in incorrect
responses at the item level were found in the research
literature (Marshall 1981 1983). Marshall investigated
whether boys and girls made similar errors in computation
and story problems. She analyzed boys' and girls' answers
to six mathematics items and found that the sexes made
different errors possibly reflecting different problem
solving strategies. Her original findings were supported
when she studied the same problem using a large number of
items three years later.


70
tests were significant at the .01 level are reported in
Table 4.1. Also reported in this table are the actual
sample sizes and the minimum sample sizes necessary for the
likelihood ratio tests to be significant. Of the 61 items#
59 had minimum sample sizes greater than 3#000. Thus#
although for the three years both males and females samples
had inconsistent option choices on these 61 items# on 59 of
the items the inconsistency was relatively minor.
Consequently# these 59 items were included in step 2 along
with the initial 50 items.
In step 2# a likelihood ratio test was performed
comparing the adequacy of the models in (3.2) and (3.3).
The X2 values associated with models (3.2) and (3.3) for
each of the 109 items subjected to step 2 are reported in
columns b and c of Table 4.2. Also reported in Table 4.2
is the difference between the two X2 values (see
column d) This latter figure is the test statistic for
comparing the adequacies of models (3.2) and (3.3).
Significant X2 statistics are indicated by asterisks.
The X2 statistics were significant for 100 of the items
indicating a male-female difference in option choice for
these items. In column e of Table 4.2 the actual sample
sizes are reported. In column f# the minimum sample sizes
necessary for significance are reported. For those 100
items which had significant X2 statistics reported in
column d# 94 (94%) had minimum sample sizes greater than
3 #000


100
exception was that/ when equivalence was controlled/
sex-related differences in favor of females/ persisted.
The question of whether male-female differences in
problem solving depend on computational skills was answered
partially/ in the affirmative. Interactions/ indicating
the dependence/ were significant only for multiplication
and division in the first year and for subtraction in the
second year.


2
Many variables* cognitive* affective* and educational*
have been investigated since 1974 in relation to sex
differences in mathematics learning and achievement.
Fennema and Sherman (1977) investigated the effect of
differential formal mathematics education. After
controlling for the number of years of exposure to the
study of mathematics* they found sex differences in only
two of the four schools under study. However* in those
schools where boys scored higher than girls* differences
were also found in their attitudes toward mathematics.
Hilton and Berglund (1974) found significant sex
differences after controlling for the number of mathematics
courses taken* and attributed them to sex differences in
interests. "As the boys' interests in science increase
related to the girls'* their achievement in mathematics
increases relative to that of the girls" (p. 234) .
Wise* Steel* and McDonald (1979) reanalyzed test data
collected in a longitudinal study of 400*000 high school
students (Project Talent). They found that when the effect
of the number of high school mathematics courses was not
controlled* no sex differences emerged for 9th graders* but
that gains made by boys during the next three years were
more than twice that of the girls. These differences
between the sexes disappeared when the number of
mathematics courses taken was controlled. Results of the
1978 Women and Mathematics National Survey* Survey I*


13
III) to investigate sex related differences in the
selection of incorrect responses/ and the consistency of
such differences over three years of administration of the
test.
Based on her findings that sex differences were found
in 80% of the items/ Marshall classified the students'
errors according to Radatz' (1979) five-category error
classification. The categories are language (errors in
semantics)/ spatial visualization/ mastery/ association/
and use of irrelevant rules.
It was found that girls' errors are more likely to be
due to the misuse of spatial information/ the use of
irrelevant rules/ or the choice of an incorrect operation.
Girls also make relatively more errors of negative transfer
and keyword association. Boys seem more likely than girls
to make errors of perseverance and formula interference.
Both sexes make language-related errors/ but the errors are
not the same.
Available research is not extensive enough to make
definite judgments about the sex-related differences
observed in incorrect responses. Clearly more research is
needed.


CHAPTER IV
RESULTS
Introduction
The data gathered from 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 during three consecutive years were
analyzed in this study. The first objective was to
investigate whether boys and girls differ in the
selection of incorrect responses, and if the pattern of
differences was consistent throughout the three years in
which the test was administered. The second objective was
to investigate whether males and females differ in problem
solving performance, if these differences persist after
accounting for computational skills, and if the male-female
differences depend on the level of computational skills.
The results are discussed in two sections. Study
findings in the area of sex differences in the selection of
incorrect responses are discussed in the first section.
The second section is devoted to an exposition of the
findings in the area of sex-related differences in problem
solving.
Sex-Related Differences in the Selection of
Incorrect Responses
As indicated in Chapter III, there are two models of
interest. The first model indicates that year and option
68


86
TABLE 4.6
ANCOVA Summary Table: Subtraction Covariate
Second Year
Source
df
SS
MS
F
Subtraction
(S) 1
93.14
593.14
122.80*
Sex (S)
1
18.58
18.58
3.85*
SxS
1
30.85
30.85
6.39*
Error
1009
4873.40
4.83
*p<.05


25
emerged in spatial visualization scores became
non-significant. In the two schools where sex differences
in mathematics achievement were found/ differences between
the sexes were also found in their attitudes toward
mathematics.
Researchers like Backman (1972) / who analyzed data
from Project Talent/ and Allen and Chambers (1977) have
also hypothesized that sex-related differences in
mathematics achievement may be related to different
curricula followed by males and females. Allen and
Chambers attributed male superiority in mathematics problem
solving to differences in the number of mathematics courses
taken in high school.
This issue has been seriously questioned by Astin
(1974)/ Fox (1975a/ 1975b)/ and Benbow and Stanley (1980)/
among others. Astin and Fox have reported large
differences in favor of males among gifted students taking
the Scholastic Achievement Test. These differences occur
as early as grade 7/ when there are no sex differences in
the number of courses taken. Benbow and Stanley (1980)
compared mathematically precocious boys and girls in the
7th grade/ with similar mathematics background/ and found
sizeable sex-related differences favoring boys in
mathematical reasoning ability. Five years later/ they
conducted a follow-up study which showed that boys


95
TABLE 4.10
ANCOVA Summary Table: Other Covariates
Second Year
Source
df
ss
MS
F
Addition
1
206.97
206.97
39.57*
Sex
1
19.12
19.12
3.66
Error
1010
5277 .90
5.23
Multiplication
1
1026.68
1026.68
232.80*
Sex
1
.80
.80
.18
Error
1010
4458.30
4.41
Division
1
1117 .46
1117.46
258.67*
Sex
1
.64
.64
.15
Error
1010
4367.50
4.32
Fracadd
1
1341.13
1341.13
327.10*
Sex
1
2.03
2.03
.50
Error
1010
4143.80
4.10
Decsub
1
1453 .91
1453.91
364.39*
Sex
1
4.88
4.88
1.22
Error
1010
4031.10
3 .99
Equivalence
1
797.11
797.11
171.79*
Sex
1
127.33
127.33
27.44*
Error
1010
4687.90
4.64
*
P<-05


25
emerged in spatial visualization scores became
non-significant. In the two schools where sex differences
in mathematics achievement were found differences between
the sexes were also found in their attitudes toward
mathematics.
Researchers like Backman (1972) r who analyzed data
from Project Talent and Allen and Chambers (1977) have
also hypothesized that sex-related differences in
mathematics achievement may be related to different
curricula followed by males and females. Allen and
Chambers attributed male superiority in mathematics problem
solving to differences in the number of mathematics courses
taken in high school.
This issue has been seriously questioned by Astin
(1974) Fox (1975a 1975b) and Benbow and Stanley (1980)
among others. Astin and Fox have reported large
differences in favor of males among gifted students taking
the Scholastic Achievement Test. These differences occur
as early as grade 7 when there are no sex differences in
the number of courses taken. Benbow and Stanley (1980)
compared mathematically precocious boys and girls in the
7th grade with similar mathematics background and found
sizeable sex-related differences favoring boys in
mathematical reasoning ability. Five years later they
conducted a follow-up study which showed that boys


skills* sex-related differences in problem solving
performance do not exist. Females retained their
superiority in problem solving when equivalence (in all
three years) and subtraction (in one year) were the
controlling variables.
The question of whether male-female differences in
problem solving depend on computational skills was
answered* partially* in the affirmative.
x


22
Approximately 270/000 students from the 6th grade were
administered the Survey of Basic Skills of the California
Assessment Program/ during the years 1977/ 1978/ and 1979.
Responses were analyzed using log-linear models.
Successful solving of computation items was positively
associated with successful solving of story problems.
Girls were more successful in computation than boys/ and
boys were more successful than girls in solving story
problems. This finding supports reports from the National
Assessment of Educational Progress (NAEP) (Armstrong/
1979).
To investigate the effects of reading/ SES/ language/
and chronological age/ only those test forms containing 2
computation items and 2 story problems were considered for
analysis; 32 items from 8 test forms were included in the
analysis.
The results of these analyses showed that at every
level of reading score/ 6th grade children were more
successful in computation than in story problems.
Although the differences were not large/ at every reading
score boys consistently had higher probabilities of
success in story problems than did girls/ and girls
consistently showed higher probabilities of success in
computation than boys. Also/ as the reading score
increased/ the difference between the probability of
success in story problems and the probability of success in


ACKNOWLEDGMENTS
I would like to acknowledge ray gratitude to several
people who have influenced my formal education and/or made
this study possible. My special thanks to Dr. James
Algina* Chairman of the doctoral committee* who contributed
to the development of my love for research and statistics.
He has been insuperable as professor and valued friend; his
help and guidance in the preparation and completion of this
study were invaluable. I extend my thanks to Dr. Linda
Crocker for her advice and help during my doctoral studies
at University of Florida. Thanks also go to Dr. Michael
Nunnery* member of the doctoral committee. To Dr. Wilson
Guertin* who was a friend for me and my family* I extend my
special thanks. Thanks are also extended to Dr. Amalia
Charneco* past Undersecretary of Education of the Puerto
Rico Department of Education* for her continuous support.
To my sister Nilda Santaella* who typed the thesis* I give
my sincere thanks. Special thanks go to my family and to
those friends who provided encouragement throughout this
critical period of my life.
ii


42
children whose errors fell under the category or "other
errors." Students were asked to read each problem aloud
and answer these questions: What kind of situation does
the problem describe? What does the problem ask you to
find? How would you work the problem?
Ninety five percent of the students read the problem
correctly; 98% explained the kind of situation the problem
described in a correct manner; 92% correctly answered what
the problem was asking them to find* and 36% correctly
answered the question of how to work the problem.
The fact that a large percent of the students whose
errors were classified as "other errors" (in which reading
skills might have been a factor) correctly stated how to
work the problem is strong evidence of their ability to
read and interpret the problems correctly. The errors made
by this group of students had a distinct origin unrelated
to reading ability.
Zalewski (1974) investigated the relative contribution
of verbal intelligence reading comprehension vocabulary
interpretation of graphs and tables mathematical concepts
number sentence selection and computation to successful
mathematical word problem solution and the relationship of
the dependent variable to the eight independent variables.
She worked with a group of 4th grade children who
were administered the subtests of the Iowa Test of Basic
Skills (ITBS) and the Weschler Intelligence Scale for


64
gender-option dependency is the same from year to year. In
this study/ a three dimensional table was considered to
exhibit the same year to year pattern of gender-option
dependency if / conditioned on gender/ the same proportion
of students chose each incorrrect response over the three
year period. This seems to be a straightforward and
natural way to operationalize the concept. To illustrate
how Marshall operationalized the concept in question/ a
hypothetical set of probabilities was constructed
conforming to the pattern specified by Marshall. This is
displayed in Table 3.3.
Table 3 .3
Hypothetical Joint Probabilities of Year/ Option/
and Sex
Year
Sex
1
Option
2
3
First
M
.120
.045 ¡
.015
F
.060
.060 ¡
.030
Second
M
.144
.022 |
.019
F
.072
.030 ¡
.039
Third
M
.132
.040 !
.016
F
.066
.054 ¡
J
.039
Note: Probabilities reported are truncated to three
places.


67
conducted/ one with each of addition/ subtraction/
multiplication/ division/ addition of fractions/ decimals
in subtraction/ and equivalence as covariates. A well
known problem that arises in the use of ANCOVA is that an
unreliable covariate can cause spurious differences between
the sex groups. To solve this problem/ Porter (1967/1968)
proposed the use of estimated true scores for observed
scores. Porter (1967/1968) conducted a simulation that
gave empirical support to the adequacy of this strategy.
Hunter and Cohen (1974) have provided theoretical support
for this strategy.


104
to contribute to the body of knowledge regarding such
differences in mathematical learning and achievement. This
section of Chapter V comprises implications of the findings
and recommendations evolving from the present
investigation.
Sex-related Differences in Incorrect Responses
The main purpose of education is to impart formal
education equally for the sexes. Analysis of the data on
incorrect responses showed that for 100 of the 111 items of
the test/ males and females selected different incorrect
responses/ and this pattern of responses was consistently
found during the three years of test administration.
However/ with a sample of less than 3/000 subjects/
statistical significant results were obtained for only
6 items. Therefore/ it seems that/ at least for 6th grade
children/ educators need not concern themselves with
varying teaching techniques for the sexes.
This study/ however/ has value apart from the
investigation of sex-related differences. It was found
that/ year after year/ 6th graders from the public schools
in Puerto Rico tend to make the same errors in the solution
of mathematical problems. It is important/ therefore/ that
teachers provide the learning strategies necessary to
enable students to obtain a better understanding of the
concepts of numbers.


Table Page
4.11 Adjusted Means on Problem Solving*
by Covariate and Sex 96
4.12 ANCOVA Third Year 97
4.13 Adjusted Means on Problem Solving*
by Covariate and Sex 98
vi


90
The significant interactions found between sex and
multiplication* and sex and division (first year)* and sex
and subtraction (second year)* answered* in part* the
question of whether male-female differences in problem
solving performance depend on computational ability.
However* the evidence is quite weak. Of 21 possible
interactions* only three were significant. No variable
exhibited a significant interaction for each of the three
years.
The analysis of covariance was also used to determine
if sex-related differences exist after controlling for
computational skills. Analyses were conducted for
those variables that did not exhibit significant
interactions with sex. As discussed in Chapter III*
estimated true scores were used for observed scores to
adjust for unreliability of the covariates (the
computational subtests). Reliability coefficients
calculated for each covariate are shown in Table 4.7.
Summaries of the analyses of covariance for the first
year are reported in Table 4.8. The results show that
females retained their superiority in problem solving
performance when equivalence was the controlling variable
in the analysis of covariance* the only variable in which
males outperformed females (nonsignificant). When the
controlling variables were addition* subtraction* addition
of fractions* and subtraction with decimals* female


59
Inspection of the year by option contingency tables shows
that year and option are independent for each gender.
Thus/ hypothesis 1 is equivalent to the hypothesis that
year and option are independent conditional upon gender.
Hypothesis 2 is also true for Table 3.1 and Table 3.2.
Therefore/ when hypothesis 1 is correct/ hypothesis 2
is equivalent to the hypothesis that sex and option choice
are dependent.
In his discussion of the analysis of three dimensional
contingency tables/ Fienberg (1980) presents the following
saturated model for the data:
log m i j k = i + Uj + Uk+Uij + Uik+Ujk+U i j k. (3.1)
In this model/ mijk is the expected value of the
frequency in cell ijk of the three dimensional table. The
model states that all three classification factors for a
three dimensional contingency table are mutually dependent.
In the present research i is the year index/ j is the
option index/ and k is the sex index. Fienberg shows that
deleting the terms Uij and Uijk yields a model in which
year and option are independent conditional upon sex. This
model is
log mij k
Ui + Uj + Uk+Uik+Uj k.
(3.2)


75
TABLE 4.2
Chi-square Values for the Comparison of Models (3.2) and
(3.3) with Actual Sample Sizes and with the Corresponding
Number of Subjects Needed for Significant Results
Item
Num
ber
Xa- for
Model
3.2
X1 for
Model
3.3
Xa- for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
1
16.11
30.94
14.83*
7*427
10*061
2
17.54
103.22
85 .68*
11*018
2*583**
3
20.06
47.51
27.45*
11*497
8*733
4
7.54
21.87
14.33*
6*016
8*414
5
14.75
70.40
55.65*
7*706
2*782**
6
11.86
23 .13
11.27*
13*046
23 *256
7
4.60
35.73
31.13*
8*832
5*700
8
46.15
329.98
283.83*
16*796
1*189**
9
10.13
109.99
99.86*
23 *075
4*642
10
9.54
65.29
55.75*
35*060
12*634
11
5.89
19.06
13.17*
24*908
37 *996
12
12.58
198.46
185.88*
40 *423
4*369
13
25.74
62.37
36.63*
15*718
8*621
14
11.19
80 .50
69.31*
30*238
8*765
15
16.63
17.33
0.70
42*889
1*230*914
16
14.51
84.29
69.78*
35*114
10*109
17
17.90
44.20
26 .30*
43 *962
33 *582


41
categories. Category I included clerical and computational
errors. Category II included other types of errors such
as average and area errors/ use of wrong operation/ no
response/ and erred responses offering no clues. It the
student's work indicated the correct procedure and yet the
problem was missed because of a computational or clerical
error/ it was assumed that the problem was read and
understood.
An analysis of frequencies showed that clerical errors
were responsible for 3% of the problems incorrectly solved/
computational errors accounted for 49%/ and other errors
for 48% of the erred problems.
Knifong and Holtan concluded that "improved
computational skills could have eliminated nearly half ot
the word problem errors" (p. Ill). These computational
errors were made in a context where other skills such as
reading/ interpretation of the problem/ and integration of
these skills necessary for the solution of word problems/
might interact. However/ Knifong and Holtan state that
their findings neither confirm nor deny that improvement ot
reading skills will lead to improvement in problem solving.
They conclude that "it is difficult to attribute major
importance to reading as a source of failure" (p. 111).
In a later analysis/ looking for evidence of poor
reading abilities affecting children's success in word
problems/ Knifong and Holtan (1977) interviewed the


69
are independent/ conditional upon sex*
log m j j k = U + U j + U k+U k+U j k (3.2)
Substantively/ this model implies that the pattern of male
and female option choices is consistent over the three
years of test administration.
The second model indicates that option is independent
of sex/
log mijk = Ui+Uj+Uk+Uik. (3.3)
This model implies that the pattern of option choice is the
same for males and females.
In order to determine if males and females differed in
the selection of incorrect responses and if these
differences were stable across the three years of test
administration/ a two-step test was performed. First/ a
likelihood ratio test of the adequacy of model (3.2) was
conducted for each of the 111 items of the test. Under
this test a model fits the data when the value
obtained is nonsignificant at a specified alpha level. For
50 (45%) of the items/ model (3.2) fitted the data
adequately. For these items the pattern of male and female
option choices was consistent for the three years. The X^
values for the 61 items for which the likelihood ratio


97
TABLE 4.12
ANCOVA Summary Table
Third Year
Source
df
ss
MS
F
Addition
1
205.96
205.96
34.85*
Sex
1
1.88
1.88
.32
Error
1010
5967.60
5.91
Subtraction
1
886.42
886.42
169.48*
Sex
1
24.39
24.39
4.66*
Error
1010
5287.2
5.23
Multiplication
1
1402.42
1402.42
297.12*
Sex
1
1.05
1.05
.22
Error
1010
4771.20
4.72
Division
1
2028.49
2028.49
494.75*
Sex
1
1.11
1.11
.27
Error
1010
4145.10
4.10
Fracadd
1
1733.72
1733.72
393.97*
Sex
1
4.64
4.64
1.05
Error
1010
4439.89
4.40
Decsub
1
2116.84
2116.84
526.57*
Sex
1
.15
.15
.04
Error
1010
4056.76
4.02
Equivalence
1
1186.92
1186.92
240.26*
Sex
1
59.85
59.85
12.12*
Error
1010
4986.69
4.94
* p<. 05


66
Table 3.4
Hypothetical Probabilities of Option Choice
Conditional on Sex and Year
Year
Sex
1
Option
2
3
First
M
.666
.250
.083
F
.400
.400
.200
Second
M
.774
.120
.104
F
.510
.212
.276
Third
M
.698
.214
.087
F
.431
.352
.215
Note: Probabilities reported are truncated to three
places.
Which procedure is more appropriate to test the claim
that the pattern of male and female option choice remains
the same from year to year? It seems more reasonable to
test for a pattern like that in Table 3.2 than to test for
a pattern like that in Table 3.4 / and/ consequently/ this
was the strategy adopted in this study.
Comparison of Males and Females in Problem Solving
Performance
One object of the study was to compare the performance
of males and females on problem solving. Two questions
were addressed. Firstr do males and females differ in
problem solving performance? Second/ do these differences
persist when computational skill is controlled for/ and do
these differences depend on the level of computational
skill? Seven analyses of covariance (ANCOVA) tests were


85
TABLE 4.4
ANCOVA Summary Table: Multiplication Covariate
First Year
Source
df
SS
MS
F
Multiplication
(M) 1
1195.90
1195.90
265.66*
Sex (S)
1
31.95
31.95
7.07*
M x S
1
38.80
38.80
8.59*
Error
998
4509.00
4.51
* p <;.oi
TABLE 4.5
ANCOVA Summary
Table: Divison Covariate
First Year
Source
df
SS
MS
F
Division (D)
1
1195.90
1195.90
264.66*
Sex (S)
1
13.94
13.94
3.22
D x S
1
18.39
18.39
4.25*
Error
998
4317.00
4.32
* p<.01


64
gender-option dependency is the same from year to year. In
this study/ a three dimensional table was considered to
exhibit the same year to year pattern of gender-option
dependency if/ conditioned on gender/ the same proportion
of students chose each incorrrect response over the three
year period. This seems to be a straightforward and
natural way to operationalize the concept. To illustrate
how Marshall operationalized the concept in question/ a
hypothetical set of probabilities was constructed
conforming to the pattern specified by Marshall. This is
displayed in Table 3.3.
Table 3.3
Hypothetical Joint Probabilities of Year/ Option/
and Sex
Year
Sex
1
Option
2
3
First
M
.120
.045 |
.015
F
.060
.060 ¡
.030
Second
M
.144
.022**
.019
F
.072
.030 ;
.039
Third
M
.132
.040*1
.016
F
.066
.054 ¡
.039
Note: Probabilities reported are truncated to three
places.


102
response was related to the use of a large number of
subjects in the analysis.
One must question whether analysis of incorrect
responses has any educational value. Although the results
of this study do not support the use of this analysis
(error analysis) for the study of sex-related differences*
it continues to be a promising technique in the discovery
of how children approach mathematics problem solving* more
so than analyses that look only at the items answered
correctly or at total test scores.
The second objective of the study was to investigate
sex-related differences in problem solving performance* to
find out if significant differences persisted after
controlling for computational skills* and if the
differences depended on the level of computational skills.
The results showed that females outperformed males in
problem solving and in six of the seven computational
skills. Males showed superiority in equivalence in each of
the three years* but statistical significance was obtained
only for the second year. The results also showed that
females retained their superiority in problem solving* only
when equivalence (for the three years data) and subtraction
(for the second year) were the controlling variables.
Analysis of covariance was also used to answer the
question of whether male-female differences in problem
solving depend on the level of computational skills. The


12
test form/ no sex differences were found in response
patterns. Approximately 45% of each sex selected option
c. The next popular choice for both sexes was option
d/ selected by approximately 35% of both boys and girls.
On the story problem of Form If males and females
responded alike. Their most popular incorrect response
choice was option a for both males and females. The
second most popular incorrect choice was option c for
both sexes.
Response to the story problem in Form 2 showed sex
differences in response choice. Including the correct
option/ 33% of the girls selected option a./ 20% chose
option C/ and 20% option d. For males/ approximately
25% selected option a and the same percent selected
option d.
Marshall concluded that although the analysis of
incorrect responses does not explain why boys and girls
differ in their responses/ the analysis shows that boys and
girls approach problems in different ways and these varying
strategies can be useful in indentifying how the sexes
differ in reasoning abilities.
Two years later/ Marshall (1983) analyzed the
responses of approximately 300/000 boys and girls to
mathematics items contained in the 16 test forms of the
Survey of Basic Skills during the years 1977/ 1978/ and
1979. She used log-linear models (explained in Chapter


42
children whose errors fell under the category or "other
errors." Students were asked to read each problem aloud
and answer these questions: What kind of situation does
the problem describe? What does the problem ask you to
find? How would you work the problem?
Ninety five percent of the students read the problem
correctly; 98% explained the kind of situation the problem
described in a correct manner; 92% correctly answered what
the problem was asking them to find* and 36% correctly
answered the question of how to work the problem.
The fact that a large percent of the students whose
errors were classified as "other errors" (in which reading
skills might have been a factor) correctly stated how to
work the problem* is strong evidence of their ability to
read and interpret the problems correctly. The errors made
by this group of students had a distinct origin* unrelated
to reading ability.
Zalewski (1974) investigated the relative contribution
of verbal intelligence* reading comprehension* vocabulary*
interpretation of graphs and tables* mathematical concepts*
number sentence selection* and computation to successful
mathematical word problem solution* and the relationship of
the dependent variable to the eight independent variables.
She worked with a group of 4th grade children who
were administered the subtests of the Iowa Test of Basic
Skills (ITBS) and the Weschler Intelligence Scale for


34
achievement and interest in mathematics result from
identification with the masculine role.
A study related to the differential cultural
reinforcement hypothesis is that of Dwyer (1974).
Dwyer examined the relationship between sex role standards
(the extent to which an individual considers certain
activities appropriate to males or females) and
achievement in reading and arithmetic. Students from
grades 2/ 4 > 6 > 81 10 r and 12 participated in this study.
She found that sex role standards contributed significant
variance to reading and arithmetic achievement test scores
and that the effect was stronger for males than for
females. This led to her conclusion that sex-related
differences in reading and arithmetic are more a function
of the child's perception of these areas as sex-appropriate
or sex-inappropriate than of the child's biological sex
individual preference for masculine and feminine sex roles/
or liking or disliking reading or mathematics.
In a study which agrees with the masculine
identification hypothesis* Milton (1957) found that
individuals who had received strong masculine orientation
performed better in problem solving than individuals who
received less masculine orientation. Elton and Rose (1967)
found that women with high mathematical aptitude and
average verbal aptitude scored higher on the masculinity
scale of the Omnibus Personality Inventory (OPI) than those
with average scores on both tasks.


67
conducted* one with each of addition* subtraction*
multiplication* division* addition of fractions* decimals
in subtraction* and equivalence as covariates. A well
known problem that arises in the use of ANCOVA is that an
unreliable covariate can cause spurious differences between
the sex groups. To solve this problem* Porter (1967/1968)
proposed the use of estimated true scores for observed
scores. Porter (1967/1968) conducted a simulation that
gave empirical support to the adequacy of this strategy.
Hunter and Cohen (1974) have provided theoretical support
for this strategy.


103
question was answered/ partially/ in the affirmative.
Interactions/ indicating the dependence/ were significant
for multiplication and division in the first year and for
subtration in the second year.
In general/ the sex-related differences in mathematics
problem-solving/ in favor of males performance/ reported by
Marshall (1981/ 1984)/ and Fennema and Tartree (1983) were
not supported in this investigation. Female performance
was equal or higher than male performance in problem
solving.
The findings of the study have something in common
with other investigations where mathematics problem solving
has been the subject of interest/ the fact that males and
females do poorly in word-story problem items (problem
solving) The mean score in problem solving for males and
females in the 9-item problem solving subtest was 3.436 and
3.862 respectively for the first year/ 3.632 and 4.015 for
the second year/ and 3.927 and 4.341 for the third year.
Implications of the Findings and
Suggestions for Further Research
The main purpose of this study was to investigate
sex-related differences in the selection of incorrect
responses. Another objective was to compare male and
female performance after accounting for computational
ability. This type of research is conducted in an attempt


46
Their study is important because it represents an
attempt to demonstrate that multiple factors can interact
in the correct solution of a mathematics word problem.
They constructed three graded tests from a basal
mathematics series for grades 3 through 8. For test lr the
problems were set in pure computational form (the effects
or reading/ interpretation/ as well as the necessity for
integration were removed in an effort to measure the
computational skills required by the word problems) .
For test 2/ the effects of reading and computation
were removed by reading the problems to the students and by
giving scores based on whether or not the students set them
up properly/ in an attempt to measure problem
interpretation alone. For test 3/ the effect of
computation was removed. The test yielded two scoresone
by grading the students on whether or not they set up the
problems properly and another by grading on the basis of
the correct answer.
The tests were administered to all 244 students from
the 6th grade in two different schools. A diagnostic
profile was obtained for each of the 217 students for which
complete data were available: a computational score/ a
probiem-interpretation score/ a reading score/ and a
reading-problem solving score.
They assumed that if the reading-prob1em
interpretation score was lower (one or more levels lower)


106
operation to the numbers in the problem. The re
sults indicate that students are not familiar with
such basic problem-solving strategies as drawing
a picture of a figure described in a problem or
checking the reasonableness of a result, (p. 338)
The fact that, in general after controlling for
computational skills there were no statistically
significant differences between the problem solving
performance of the males and females who participated in
this study would seem to imply that teachers need not worry
about designing different teaching strategies for the
sexes. However^ because the performance in problem-solving
is low for both sexes teachers may be encouraged to
emphasize problem solving in their daily teaching.
This study was designed to investigate sex-related
differences in mathematical performance. The findings
clearly demonstrated that 6th grade females and males
from the public schools in Puerto Rico are equally good or
equally bad in the solution of word-story items (problem
solving). Also/ in general/ the effect of computational
skills is similar for each sex. Based on these findings/
it appears that greater differences exist within each sex
than between the sexes. In conclusion/ the findings of the
present study highlight the need for investigations of
different approaches to instructional designs aimed at
improving both male and female performance in problem
solving.


32
attributes success to an internal, stable attribute, such
as ability, then one is confident of being successful in
the future and will continue to strive in that area. If
one attributes success to an external factor such as a
teacher, or to an unstable one, such as effort, then one
will not be as confident of success in the future and will
cease to strive. Failure attribution patterns work this
way: if failure is attributed to unstable causes, such as
effort, failure can be avoided in the future and the
tendency will be to persist in the task. However, if
failure is attributed to a stable cause, such as ability,
the belief that one cannot avoid failure will remain.
Studies reported by Bar-Tal and Frieze (1977) suggest
that males and females tend to exhibit different
attributional patterns of success and failure. Males tend
to attribute their success to internal causes and their
failures to external or unstable ones. Females show a
different pattern; they tend to attribute success to
external or unstable causes and failures to internal ones.
The pattern of attributions, success attributed externally
and failure attributed internally, has become hypothesized
to show a strong effect on mathematics achievement in
females. K1oosterman' s (1985) study supported these
findings. According to Kloosterman, attributional
variables appear to be more important achievement mediators
for females than for males, as measured by mathematics word
problems. More research is needed in this area.


108
Bern/ S. L. / & Bern D. V. (1970). We are all nonconscious
sexists. Psychology Today/ 22-24, 261 115-116.
Benbow/ C. P./ & Stanley/ J. C. (1980). Sex differences
in mathematical ability: Fact or artifact?
Science/ 210/ 1262-1264.
Benbow/ C. P./ & Stanley/ J. C. (1982). Consequences in
high school and college of sex differences in mathema
tical reasoning ability: A longitudinal perspective.
American .Educational Research Journal/ 12./
598 622.
Bobbie/ C. N. (1971) Sex-role preference and academic
achievement (Doctoral dissertation/ Yeshiva Univer
sity) Dissertation Abstg^ 1818B-1819B.
Carey/ G. L. (1958) Sex differences in problem solving
performance as a function of attitude differences.
Journal Of Abnormal. $tn.d Social psychology/ 22/
256-260.
Carpenter/ T. P./ Corbitt/ M. K./ Kepner/ H. S./ Lindquist/
M. M./ & Reys/ R. E. (1980). Results of the second
NAEP mathematics assessment: Secondary school.
Mathematics Teacher/ 22/ 329-338.
Carpenter/ T. P. / Lindquist/ M. M./ Mathews/ W./ & Silver/
E. A. (1984). Results of the third NAED mathematics
assessment: Secondary school. Mathematics Teacher/
22/ 652-659.
Chase/ C. I. (1960). The position of certain variables in
the prediction of problem solving in arithmetic.
Journal of Educational Research/ 14./ 9-15.
Cohen/ D./ & Wilkie/ F. (1979). Sex-related differences in
cognition among the elderly. In M. A. Witting and
A. C. Petersen (Eds.)/ Sex-related differences in
C.ognjtjve_J.unct,iojiin.qj Developmental Issues/ New
York: Academic Press.
Connor/ J. M. / Serbin/ L. A./ & Schakmanz M. (1977). Sex
differences in children's response to training on
a visual-spatial test. Developmental Psychology/
12/ 392-394.


CHAPTER IV
RESULTS
Introduction
The data gathered from 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 during three consecutive years were
analyzed in this study. The first objective was to
investigate whether boys and girls differ in the
selection of incorrect responses/ and if the pattern of
differences was consistent throughout the three years in
which the test was administered. The second objective was
to investigate whether males and females differ in problem
solving performance/ if these differences persist after
accounting for computational skills/ and if the male-female
differences depend on the level of computational skills.
The results are discussed in two sections. Study
findings in the area of sex differences in the selection of
incorrect responses are discussed in the first section.
The second section is devoted to an exposition of the
findings in the area of sex-related differences in problem
solving.
Sex-Related Differences in the Selection of
,Incorr_egt.. Response
As indicated in Chapter III/ there are two models of
interest. The first model indicates that year and option
68


94
TABLE 4.9
Adjusted Means on Problem Solving*
by Covariate and Sex
First Year
Covariate
Adjusted Means
Males
Females
Addition
3.53
3.78
Subtraction
3.55
3.76
Fracadd
3.56
3.74
Decsub
3.73
3.58
Equivalence
3.38
3.91


REFERENCES
Aiken/ L. R. (1970). Attitudes toward mathematics.
Review of Educational Resarch/ IQ./ 551-596 .
Allen/ R. H./ & Chambers/ D. L. (1977). A comparison of
the mathematics achievement of males and females
(Bulletin No. 9194). Madison: Wisconsin State
Department of Public Instruction/ Division for
Management and Planning Service. (ERIC Document
Reproduction Services No. ED 159 073)
Armstrong/ J. M. (1979) Achievement and participation of
W.pme_n_in_.math.emati£sj Ln_Q_y.ejyi-e.W-. (Report of a
two year study). Denver/ CO: Educational Commission
of the States/ National Assessment of Educational
Progress. (ERIC Document Reproduction Service
No. ED 184 878)
Armstrong/ J. M./ & Price/ R. A. (1982). Correlates and
predictors of women's mathematics participation.
Journal for Research in Mathematics Education/ 12/
99-109.
Astin/ H. S. ( 197 4) Overview of the findings. In H.
Astin/ H. Suniewick/ & S. Dweck (Eds.) / Women: A
Bibliography on their Education and Careers.
(pp. 1-10). New York: Behavorial Publications.
Backman/ M. E. (1972). Patterns of mental abilities:
Ethnic/ socioeconomic and sex differences. American
Educational Research Journal/ 1/ 1-11.
Ballew/ H./ & Cunningham/ W. (1982). Diagnosing strengths
and weaknesses of sixth-grade students in solving
word problems. Journal for Research in Mathematics
Education/ 13/ 202-210.
Balow/ I. H. (1964). Reading and computation ability as
determinants of problem solving. The Arithmetic
Teacher/ 11/ 18-22.
Bar-Tal / D. / & Frieze/ J. H. ( 1977 ) Achievement
motivation for males and females as a determinant of
attribution for success and failure. Sex roles.
1/ 301-313.
107


83
TABLE 4.3 Continued
Year/
Subtest
Males
X
SD
Females
X
SD
t
Decsub
2.863
1.87
3.440
1.80
5.00*
Equivalence
.76
.76
.65
.75
-2.32**
N =
504
N =
509
Third
Problem Solving
3.927
2.49
4.341
2.44
2.67*
Addition
5 .536
.84
5.704
.64
3.59*
Subtraction
4.836
1.52
4.958
1.46
1.30
Multiplication
4.168
1.74
4.541
1.68
3 .47*
Division
3.343
1.88
3.795
1.83
3.87*
Fracadd
2.819
1.89
3 .117
1.85
2.53**
Decsub
3.021
1.93
3.448
1.85
3.60*
Equivalanee
.830
.82
.800
.78
- .60
N =
509
N =
504
Note: The number of
items in
the problem solving
subtest
was 9 .
In each computation subtestr the number of items
was 6. An item was included in the computation subtest
only if it measured a computation skill required to solve
a problem solving item.
*
**
p <.01
P <.05


38
At present# no set of variables has been clearly
established as a determinant of problem difficulty.
Several researchers have investigated the effect of reading
and computation on problem solving performance. Others
have studied the effect of student attitudes toward problem
solving in problem solving learning and achievement.
Typically# correlational methods have been used to
investigate these questions.
Computational_S_k_U.ls, and_.Pr,QbJ.etn Solving Performance
One of the first researchers to study the effect of
computation and reading on problem solving performance was
Hansen (1944). He investigated the relationship of
arithmetical factors# mental factors# and reading factors
to achievement in problem solving. Sixth grade students
were administered tests in problem solving and categorized
as superior achievers (best problem solvers) and inferior
archievers (poorest problem solvers) The two groups were
compared in selected factors believed to be related to
success in arithmetic problem solving: arithmetical# mental
and reading factors. After controlling for mental and
chronological age# the superior achievers in problem
solving surpassed the inferior achievers in mental and
arithmetical factors. The superior group did better in
only two of the six items under the reading factors:
general language ability and the reading of graphs# charts#
and tables.


58
Table 3.1
Hypothetical Probabilities of Option Choice
Conditional on Year and Sex
Year
Sex
1
Option
2
3
First
M
.7
.2
.1
F
.5
.3
.2
Second
M
.7
.2
.1
F
.5
.3
.2
Third
M
.7
.2
.1
F
.5
.3
.2
Hypothetical
Conditional on
Table 3.2
Probabilities of
Year and Sex and
Option Choice
Arranged by Sex
Option
Sex
Year
1
2
3
M
First
.7
.2
.1
Second
.7
.2
.1
Third
.7
.2
.1
F
First
.5
.3
.2
Second
.5
.3
.2
Third
.5
.3
.2


5
Females were better on items of computation and males were
more successful on word-story problem items (problem
solving). She also found that females successful
performance in the problem solving items was more dependent
on their successful performance in the computation items.
Males did not need/ as much as females/ to succeed in the
computation items in order to answer correctly the problem
solving items.
Although the general findings seem to support sex
differences in mathematics learning and achievement/ the
research done does not consistently support superiority for
either sex. Most of the research has been concerned with
how the sexes differ on subtests or total test scores in
mathematics. Moreover/ the great majority of the studies
deal with correct responses. Sex differences in incorrect
responses at the item level have not been fully researched.
Only two studies dealing with sex differences in incorrect
responses at the item level were found in the research
literature (Marshall/ 1981/ 1983). Marshall investigated
whether boys and girls made similar errors in computation
and story problems. She analyzed boys' and girls' answers
to six mathematics items and found that the sexes made
different errors/ possibly reflecting different problem
solving strategies. Her original findings were supported
when she studied the same problem using a large number of
items three years later.


Problem Solving
87
9 .
8
7
6 .
5 -
4 -
3 -
2 -
1 -
T 1 1 1 I
2 3 4 5 6
Multiplication
Fig. 4.1 Sex by Multiplication Interaction


78
TABLE 4.2 Continued
Item
Num
ber
X1 for
Model
3.2
Xa- for
Model
3.3
X* for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
. (b)
(c)
(d)
(e)
(f)
55
19.53
90.87
71.34*
41/796
11/770
56
21.62
70.14
48.52*
44/107
18/263
57
28.51
35.78
7.27
53/075
146/668
58
17.49
51.69
34.20*
44/044
25/873
59
50.96
73.68
22.72*
41/829
36/987
60
66.43
148.39
81.96*
60/494
14/828
61
21.39
394.54
373.15*
35/632
1/918**
62
55.03
300.95
245.92*
50/699
4/142
63
158.36
278.34
119.98*
52/068
8/719
64
88.01
246.48
158.47*
73/196
9/279
65
56.28
94.11
37.83*
42/429
22/532
66
50.51
113.88
63.37*
78/708
24/953
67
34.28
180.60
146.32*
51/862
7/121
68
48.01
225.46
177.45*
63/223
7/158
69
75.29
251.66
176.37*
62/511
7/121
70
44.39
271.35
226.96*
67/528
5/977
72
17.41
208.90
191.49*
84/175
8/831
73
24.61
37.80
13.19*
53/358
81/271
74
22.75
163.01
140.26*
84/175
12,057


66
Table 3.4
Hypothetical Probabilities of Option Choice
Conditional on Sex and Year
Year
Sex
1
Option
2
3
First
M
.666
.250
.083
F
.400
.400
.200
Second
M
.774
.120
.104
F
.510
.212
.276
Third
M
.698
.214
.087
F
.431
.352
.215
Note: Probabilities reported are truncated to three
places.
Which procedure is more appropriate to test the claim
that the pattern of male and female option choice remains
the same from year to year? it seems more reasonable to
test for a pattern like that in Table 3.2 than to test for
a pattern like that in Table 3.4* and* consequently* this
was the strategy adopted in this study.
C-Omparisoa of_Males_and Females in Problem Solving
Pejfbrmance
One object of the study was to compare the performance
of males and females on problem solving. Two questions
were addressed. First* do males and females differ in
problem solving performance? Second* do these differences
persist when computational skill is controlled for* and do
these differences depend on the level of computational
skill? Seven analyses of covariance (ANCOVA) tests were


BIOGRAPHICAL SKETCH
Sonia Feliciano was born on March 17/ 1938/ in
Mayaguez/ Puerto Rico. She graduated from Eugenio Ma. de
Hostos High School/ in her hometown. She received a degree
in business administration/ with a major in finance/ from
the University of Puerto Rico in June/ 1961.
After graduation she worked as junior accountant for
Pittsburgh Plate Glass/ International. During the academic
year of 1969-70 she started her career in education/
obtaining the degree of Master in Education from the
University of Puerto Rico in June/ 1975.
From January/ 1971 to June/ 1985/ she served in the
public school system of Puerto Rico in different positions.
She worked as educational researcher until July/ 1977/ when
was appointed Director of the Research Center of the
Commonwealth Department of Education.
During the academic year of 1978-79/ she initiated
doctoral studies in the Faculty of Philosophy and
Educational Sciences of the Universidad Complutense de
Madrid/ in Madrid/ Spain. She completed the coursework
requirements and then transferred to University of Florida
during the fall of 1979.
116


27
Research findings in this area have been inconsistent.
In 1966 / Maccoby stated that "by early school years/ boys
consistently do better (than girls) on spatial tasks and
this difference continues through the high school and
college years" (p.26). In 1972/ Maccoby and Jacklin said
that the differences in spatial ability between the sexes
"remain minimal and inconsistent until approximately the
ages of 10 or 11/ when the superiority of boys becomes
consistent in a wide range of populations and tests"
(p- 41) In 1974 / after a comprehensive literature search/
Maccoby and Jacklin concluded that sex differences in
spatial visualization become more pronounced between upper
elementary years and the last year of high school/ the
years when sex-related differences in mathematics
achievement favoring boys emerge.
Guay and McDaniel (1977) supported in part Maccoby and
Jacklin's 1974 findings. They found that among elementary
school children/ males had greater high level spatial
ability than females/ but that males and females were equal
in low level spatial ability. This finding is inconsistent
with that portion of Maccoby and Jacklin's review that
suggests that sex differences become evident only during
early adolescence. Cohen and Wilkie (1979) however/ stated
that in tests measuring distinct spatial tasks/ males
perform better than females in early adolescence and
throughout their life span. Most studies carried out after


36
they show less interest in the subject than counterpart
males. These differences in interest are what Hilton and
Berglund (1974) suggest to account for sex-related
differences in mathematics achievement.
Although the perception of the usefulness of
mathematics is still an important predictor of course
taking for girls there is a growing similarity between
males and females regarding the usefulness of mathematics
(Armstrong & Price/ 1982; Fennema & Sherman/ 1977; Moller/
1982/1983) Armstrong and Price investigated the relative
influence of selected factors in sex-related differences in
mathematics participation. Both males and females selected
usefulness of mathematics as the most important factor in
deciaing whether or not to take more mathematics in high
school. Moller's study revealed that both males and
females based mathematics course-taking decisions on career
usefulness. A Fennema and Sherman (1977) study showed only
slight differences between males and females in their
feelings about the usefulness of mathematics. In her study
of this variable among college students/ Probert
(1983/1984) did not find any sex-related differences
either.
These have been the main affective variables
researched in attempting to explain the underlying causes
of sex-related differences observed in mathematics learning
and achievement. In spite of the great diversity of studies


TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS
LIST OF TABLES V
LIST OF FIGURES V
ABSTRACT viii
CHAPTER
I INTRODUCTION 1
Purpose of the Study 6
Significance of the Study 6
Organization of the Study 8
II REVIEW OF THE LITERATURE 9
Sex-related Differences in Incorrect
Response Patterns 10
Sex-related Differences in Problem
Solving 14
Cognitive and Affective Variables that
Influence Sex Differences in Mathematics
Learning and Achievement 23
Differences in Formal Mathematics
Education 24
Differences in Spatial Ability 26
Differentiated Effect of Affective
Variables 30
Problem Solving Performance and Related
Variables 37
Computational Skills and Problem Solving
Performance 38
Reading and Problem Solving Perfor
mance 44
Attitudes Toward Problem Solving and
Problem Solving Performance 50
iii


CHAPTER I
INTRODUCTION
Sex differences in mathematics learning and
achievement have been the subject of intensive research.
Research done before 1974 has shown that male performance
on mathematical achievement tests is superior to female
performance by the time they reach upper elementary or
junior high school (Fennema/ 1976/ p. 2). The literature
strongly suggests that at the elementary level females
outperform males in computation and males excel in
mathematical reasoning (Glennon & Callahan/ 1968; Jarvis/
1964; Maccoby/ 1966).
Since 1974/ research findings have been less
consistent. Fennema (1974)/ after reviewing 36 studies/
found that during secondary school or earlier/ sex-related
differences in mathematics achievement are not so evident/
but that when differences are found/ they favor males in
high level cognitive tasks (problem solving) and females in
low level cognitive tasks (computation) As a result of a
further review of the literature/ Fennema (1977) concluded
that at the elementary level/ sex-related differences do
not exist at all cognitive levels/ from computation to
problem solving.
1


S^x^-elajted-Diiferences in Problem
Solving Performance
81
In this part* results for the two questions related to
the second objective of the study are presented. These
questions serve as the framework for the presentation.
Each question is stated* followed by the results pertaining
to that question.
Question 1: Do males and females differ in problem solving
performance?
The responses of 492 males and 510 females who took
the Puerto Rico Basic Skills Test-6 during the spring of
the first year were analyzed in this study. Also* data
from 504 males and 509 females tested in the second year
and from 509 males and 504 females tested in the third year
were included in the analysis.
The mean performance scores and the standard
deviations for each of the eight variables are presented in
Table 4.3. Results of t-tests are also presented in this
table. Females outperformed males in problem solving* a
finding consistently present in all the three years of test
administration. Over the three-year period the mean
differences favored females in all variables except
equivalence. The sex-related differences in problem
solving were significant (p<.01) for all three years.


86
TABLE 4.6
ANCOVA Summary Table: Subtraction Covariate
Second Year
Source
df
SS
MS
F
Subtraction (S)
1
93 .14
593 .14
122.80*
Sex (S)
1
18.58
18.58
3 .85*
SxS
1
30.85
30.85
6.39*
Error
1009
4873.40
4.83
*P < 05


14
jLexrxe-lat£_d_Differ enees in Problem Solving
It has already been acknowledged that the subject of
problem solving has been extensively researched. However/
as early as 1969/ Kilpatrick criticized the fact that the
study of problem solving has not been systematic? some
researchers have studied the characteristics of the problem
while others have given their attention to the
characteristics of the problem solvers. Moreover/
differences in the tests used to measure problem solving
performance also constitute an obstacle when trying to
compare the results of the studies carried out.
In order to avoid this pitfall and provide a basis for
comparison/ the studies reviewed in this section/ dealing
with sex-related differences in problem solving/ have been
divided in two groups. The first comprises those studies
that used the Romberg-Wearne Problem Solving Test. The
second contains other relevant studies in which problem
solving performance has been measured by means of other
instruments.
The Romberg-Wearne Problem Solving Test merits special
mention because it was the first attempt "to develop a test
to overcome the inadequacies of total test scores in
explaining the reasons why some students are successful
problem solvers and others are not" (Whitaker/ 1976/ pp. 9/
10)


63
model tests were conducted using a .01 level of
significance.
Since this research is based on Marshall (1981/ 1983)/
it is important to compare the method of analysis used in
this study to the one used by Marshall. Marshall also used
a two-step analysis. In the first step of her analysis she
deleted the Uijk term from (3.1) and tested the adequacy
of the model/
1 g m j. j k = i + U j + o k+ U i j + Ui k+ j k (3.6)
Following this/ she deleted the Ujk term to obtain
log m j k = (Ji + Uj + Uk+Ui j + Uik. (3.7)
and compared these two models using a likelihood ratio
test. If the first test was nonsignificant and the second
significant/ Marshall claimed that option choice was
dependent on sex and the pattern of dependency was the same
from year to year.
This is the same claim that this study sought to
establish. However/ the approach used here was to
present evidence that model (3.2) fits the data while
Marshall tried to show that model (3.6) fits the data.
The major difference between the two approaches
concerns the operationalization of the concept that the


16
of the test. The sexes differed in only 2 of the 19
reference tests Spatial Relations and Picture Group
Name-Selection. A factor analysis however showed
differences in the number and composition of the factors.
For females/ a general mathematics factor was determined by
mathematics computation/ comprehension/ application/ and
problem solving. For males/ the comprehension and
application parts determined one factor; problem solving
with two other reference tests (Gestalt and Omelet)
determined another factor.
Meyer concluded that comprehension of the data and
mastery of the prerequisite mathematical concepts did not
guarantee successful problem solving either for males or
for females. Problem solving scores for both sexes were
about one third their scores in comprehension and one
fourth their scores in application.
She also concluded that the sexes may have approached
the problem solving questions differently. The methods
used by females for solving problem situations may have
paralleled their approach to the application parts.
Males may have used established rules and algorithms for
the application parts/ but may have used more of a Gestalt
approach to the problem solving situation.
Whitaker (1976) investigated the relationship between
the mathematical problem performance of 4th grade children
and their attitudes toward problem solving/ their teachers'


45
He found that in the 4th grade the correlations
between problem solving and abstract verbal reasoning/
reading comprehension/ arithmetic concepts/ and computation
were .61/ .64/ .66/ and .60 respectively/ and .56/ .68/
.69/ and .63 in the 8th grade. When computation was held
constant/ the correlation between problem solving and
reading was .52 in grade 4 and .54 in grade 8. When
reading was held constant the correlation between problem
solving and computation was .43 in grade 4 and .42 in
grade 8.
Creswell (1982) worked with a sample of anglo and
black adolescents from Chicago. Each subject was
administered the California Achievement test. Multiple
regression was used to analyze the data. The analysis
showed that reading is more important than computation in
predicting student performance in problem solving. Reading
accounted for 49.5% of the variance; computation accounted
for 14.6% of the variance.
Ballew and Cunningham (1982) worked with 6th grade
students in an attempt to find what proportion of students
have as their main source of difficulty with word problems
each of the following factors: a) computation skills/
b) interpretation of the problem/ c) reading and/
d) integrating these skills in the solution of problems.
They also wanted to know if a student can be efficiently
diagnosed as having one of the four categories as his/her
main difficulty with mathematics word problems.


112
Landau* M. (1984). The effects of spatial ability and
problem presentation format in mathematical problem
solving performance on middle school students
(Doctoral dissertation* Northwestern University).
Dissertation. Abstracts., mteEP3ti.onal* * 442a-
443 A.
Lindgren* H. C. Silva* I.* Faraco* I.* & DaRocha* N. S.
(1964). Attitudes toward problem solving as a
function of success in arithmetic in Brazilian
elementary schools. Journal of Educational Research*
5&* 44-45.
Maccoby* E. E. (1966) Sex differences in intellectual
functioning. In E. E. Maccoby (Ed.)* The Development
of Sex Differences (pp. 25-55) Stanford* CA: The
Stanford University Press.
Maccoby* E. E. & Jacklin* C. N. (1972). Sex differences
in intellectual functioning. In A. Anastasi* (Ed.)*
Asseestnent.in a pluralistic, sopi-elx* (pp. 37-55).
Proceedings of the 1972 Invitational Conference on
Testing Problems* Educational Testing Service*
Princeton* NJ.
Maccoby* E. E.* & Jacklin* C. N. (1974). The Psychology of
Sex Differences* Stanford* CA: Stanford University
Press.
Marshall* S. P. (1981) Sex Differences in Sixth Grade
Children's Problem Solving. Paper presented at the
annual meeting of the American Educational Research
Association* Los Angeles* CA.
Marshall* S. P. (1983). Sex differences in mathematical
errors: Analysis of distractor choice. Journal for
Research in Mathematics Education* 14.* 325-336.
Marshall* S. *P. (1984). Sex differences in children's
mathematics achievement: Solving computations and
story problems* Journal. Qf Ed.ucati.Q.P.^l_Pgy.ch.Q.l.P.gy*
2£* 194-204.
Martin* M. D. (1964). Reading comprehension* abstract
verbal reasoning and computation as factors in
arithmetic problem solving (Doctoral dissertation*
University of Iowa* 1963) pissertatiPn_Ab.strac.t.5
International* 24 4547A-4548A.


Problem Solving
87
9
8 -
6
5 -
4 -
3 -
2 -
1 -
T
1
I 1 1 1 1
2 3 4 5 6
Multiplication
Fig. 4.1 Sex by Multiplication Interaction.


24
learning and achievement arise from sex differences in
spatial visualization; and that sex-related differences
result from a differentiated effect of affective variables
on the mathematics performance of males and females.
D.,iff£JC-epcgg.. in. F,Qfmal_ M3th.em.3tics_J¡ducation (Differential
Coursework Hypothesis)
The basis for the differential coursework hypothesis
is the fact that sex-related differences in mathematical
learning and achievement show up when comparing groups
which are not equal in previous mathematics learning.
After the 8th grade/ boys tend to select mathematics
courses more often than girls. Therefore/ girls show lower
achievement scores in mathematics tests because their
mathematics experience is not as strong as the boys'
(Fennema/ 1975; Fennema & Sherman/ 1977; Sherman/ 1979).
Fennema and Sherman's study (1977) lends additional
support to the feasibility of viewing sex differences in
mathematics learning and achievement as reflecting
something other than a difference in mathematics aptitude.
After controlling for previous study of mathematics/ they
found significant sex differences in mathematics
achievement in only two of the four schools under study/
making the attribution to sex per se less likely.
Controlling for the number of space visualization-related
courses/ the sex-related differences which originally


79
TABLE 4.2 Continued
Item
Num
ber
X1 for
Model
3.2
Xz for
Model
3 .3
Xz for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
75
8.69
237.07
228.38*
67 /783
5/963
76
99.68
167 .72
68.04*
70/183
20/723
77
79.38
432.36
352.98*
66/304
3/774
78
64.56
127 .04
62.48*
76/806
24/696
79
23.98
72.59
48.61*
48/690
20/123
80
75.01
128.13
53 .12*
55/277
20 /906
81
95.51
106.82
11.31*
57 /710
102/511
82
51.99
108.12
56.13*
71/315
25/525
83
20.12
95.18
75.06*
81 /021
21/685
84
38.62
69.05
30.43*
78/959
52/129
86
102.85
433.47
330.62*
39/903
2/425**
87
66.75
430.40
363.65*
48/194
2/662**
88
63.01
126 .27
63.26*
38/742
12/304
89
43 .26
289.14
245.88*
39/648
3 /240
90
85.53
405.04
319.51*
53 /557
3/368
91
82 .53
85.48
2.95
62/954
428/727
92
16.36
93 .12
76.76*
88/231
23 /092
93
78.39
229.02
150.63*
80 ,122
10/766
94
70.15
137 .27
67.12*
87 /241
26 /113


80
TABLE 4.2 Continued
Item
Num
ber
Xa for
Model
3.2
Xa for
Model
3.3
X2 for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
la)
lb)
(c)
(d)
lell .
Cf)
95
40.59
68.41
27.82*
83/969
60/638
96
29.31
212.34
183.03*
74/287
8/154
97
8.60
126.96
118.36*
50/028
8/492
98
29.87
41.77
11.90*
69/052
116/576
99
9.44
9.94
0.50
72/647
2/918/956
100
11.35
41.48
30.13*
60/931
40/627
101
23.77
315.62
291.85*
83/782
5/767
102
61.06
437.09
376.03*
75/708
4/045
103
6.12
7.97
1.85
50/663
550/173
104
40.47
283.94
243.47*
72/517
5/984
105
28.66
48.59
19.93*
60/712
61/199
106
13.47
1/682.71
1/669.34*
71/946
866*1
107
13.13
99.20
86.07*
72/482
16/918
108
26.56
79.33
52.77*
86/261
32/840
109
5.73
18.33
12.60*
76/800
112/453
110
17.80
49.11
31.31*
70/980
45/544
111
30.71
477.44
446.73*
86/190
3/876
PC.01
Minimum sample size less than 3/000


112
LandaUf M. (1984). The effects of spatial ability and
problem presentation format in mathematical problem
solving performance on middle school students
(Doctoral dissertation* Northwestern University).
Dissertation Abstracts International? 4/ 442A-
443 A.
Lindgren/ H. C./ Silva/ I./ Faraco I./ & DaRocha/ N. S.
(1964). Attitudes toward problem solving as a
function of success in arithmetic in Brazilian
elementary schools. Journal of Educational Research/
2., 44-45 .
Maccoby/ E. E. (1966). Sex differences in intellectual
functioning. In E. E. Maccoby (Ed.)/ The Development
of Sex Differences (pp. 25-55). Stanford/ CA: The
Stanford University Press.
Maccoby/ E. E. / & Jacklin/ C. N. (1972). Sex differences
in intellectual functioning. In A. Anastasi/ (Ed.)/
Assessment in a pluralistic society, (pp. 37-55) .
Proceedings of the 1972 Invitational Conference on
Testing Problems/ Educational Testing Service/
Princeton/ NJ.
Maccoby/ E. E./ & Jacklin/ C. N. (1974) The Psychology of
Sex Differences/ Stanford/ CA: Stanford University
Press.
Marshall/ S. P. (1981) Sex Differences in Sixth Grade
Children's Problem Solving. Paper presented at the
annual meeting of the American Educational Research
Association/ Los Angeles/ CA.
Marshall/ S. P. (1983). Sex differences in mathematical
errors: Analysis of distractor choice. Journal for
Research in Mathematics Education/ 14./ 325-336 .
Marshall/ S. 'p. (1984). Sex differences in children's
mathematics achievement: Solving computations and
story problems/ Journal of Educational Psychology/
lAf 194-204.
Martin/ M. D. (1964). Reading comprehension/ abstract
verbal reasoning and computation as factors in
arithmetic problem solving (Doctoral dissertation/
University of Iowa/ 1963) Dissertation Abstracts
International/ 24/ 4547A-4548A.


47
than the problem-interpretation score? the difficulty was
due to reading ability. If the score of the lowest of the
three areas (computation? problem interpretation? and
reading-problem interpretation score) was the same as the
reading-problem solving score? the student's area of
greatest immediate need was either computation? problem
interpretation? or reading. If the reading-problem solving
score was lower than the lowest of the other three scores?
the student's area of greatest immediate need was
integration.
Analysis of the data revealed that for 19% of the
students? problem interpretation was their major
difficulty; for 26% of the students? integration (total
problem solving) was their greatest immediate need; for
another 26%? computation was the major weakness; and for
29%? reading was their greatest immediate need.
Seventy five percent of the students demonstrated
clear strength in computation? 21% in problem
interpretation? and 4% in reading-problem interpretation.
An analysis across all students (including those without
complete data) showed that 26% of the subjects could not
work word problems at a level as high as that at which they
could compute? interpret problems? and read and interpret
problems? when those areas were measured separately. This
led them to conclude that knowing the skills or the
components of solving word problems is not sufficient for


6
Purpose of the Study
The purpose of this study was twofold. The first was
to investigate sex-related differences in the selection of
incorrect responses in a multiple-choice achievement test
in mathematics. For each test item* the research questions
were as follows:
1. Is there a difference in the proportion of males
and females choosing each incorrect option?
2. Is the same pattern of differences found in data
obtained in three different administrations of the test?
The second objective was to investigate sex-related
differences in test scores in mathematics problem solving.
The following questions were studied:
1. Do males and females differ in problem solving
perf ormance?
2. Do sex differences in problem solving persist
after controlling for computational skillsf and does the
differential success of males and females on problem
solving items depend on their success on the computation
items?
Significance of the Study
Item response patterns are very useful techniques in
the assessment of mathematics learning and achievement.
Total test scores can be very misleading in the assessment
of student performance and provide no diagnostic
information about the nature and seriousness of student


65
One important characteristic of this table involves
two-by-two subtables of sex cross classified with two of
the options for each year. For example* consider the three
two-by-two tables obtained by cross classifying sex and
option choices one and two for each year. These tables are
indicated by dotted lines on Table 3.3. For this table*
the ratio of the odds of a male choosing option one to the
odds of a female choosing option one is the same (within
truncating error) for each year. For example* this odds
ratio for the first year is (.120/.045) / (.060/.060) *
2.67. Within the error caused by reporting truncated
probabilities* the odds ratio for years two and three is
the same as that for year one. The equality of these odds
ratios is Marshall's operationalization stage of the year-
to-year gender-option dependency.
To show that the odds ratio can be constant over
years* but that the probabilities of option choice
conditional on sex and year can change from year to year*
for both males and females* the probabilities in Table 3.3
were converted to the probabilities of option choice
conditional upon sex and year. These conditional
probabilities are reported in Table 3.4. Unlike the
probabilities in Table 3.2* those in Table 3.4 change from
year to year for both males and females.


77
TABLE 4.2 Continued
Item
Num
ber
Xa for
Model
3.2
Xa for
Model
3 .3
Xa for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
36
84.36
134.56
50 .20*
86 f137
34*472
37
23.70
76.85
53 .15*
37 *06 8
14*011
38
13.64
84 .24
70.60*
38*870
11*0618
39
13.35
757.60
744.25*
61*057
1*648**
40
6.67
701.39
694.72*
56 *438
1*630**
41
12.65
603.42
590.77*
57 *145
1*943**
42
23 .26
189.11
165.85*
59*140
7*154
43
16.29
21.65
5.36
33*712
126 *357
44
3.55
115.84
112.29*
67*192
12*021
45
22.41
55.90
33.49*
13*098
37*851
46
6.56
16.30
9.74*
56 *727
117 *007
47
7.82
76.15
68.33*
56 *555
16*628
48
179.25
452.64
273.39*
97 *593
7*172
49
27 .82
59.75
31.93*
16*766
10*549
50
14.23
240.27
226.04*
25*410
2*258**
51
16.37
224.54
208.17*
24 *715
2*385**
52
26.01
96 .14
70.13*
39*858
11*418
53
6.69
27.49
20.80*
44*694
43 *168
54
41.41
95.41
54.00*
74*502
27*718


92
TABLE 4.8
ANCOVA Summary Table: Other Covariates
First Year
Source
df
SS
MS
F
Addition
1
271.57
271.57
49.74*
Sex
1
15.54
15.54
2.85
Error
999
5453.90
5.46
Subtraction
1
771.78
771.78
155.60*
Sex
1
10.73
10.73
2.16
Error
999
4953.70
4.96
Fracadd
1
1412.11
1412.11
326.87*
Sex
1
7.54
7.54
1.75
Error
999
4313.30
4.32
Decsub
1
1281.10
1281.10
287.89*
Sex
1
4.94
4.94
1.11
Error
999
4444.30
4.45
Equivalence
1
811.44
811.44
164.93*
Sex
1
69.69
69.69
14.16*
Error
999
4914.00
4.92
* p .05


UNIVERSITY OF FLORIDA
3 1262 08554 6660


77
TABLE 4.2 Continued
Item
Num
ber
X* for
Model
3.2
Xa for
Model
3.3
X2- for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
. (d) .
(e)
(f)
36
84.36
134.56
50.20*
86,137
34,472
37
23.70
76.85
53.15*
37,068
14,011
38
13.64
84.24
70.60*
38,870
11,0618
39
13.35
757.60
744.25*
61,057
1,648**
40
6.67
701.39
694.72*
56,438
1,630**
41
12.65
603.42
590.77*
57,145
1,943**
42
23 .26
189.11
165.85*
59,140
7,154
43
16.29
21.65
5.36
33,712
126,357
44
3.55
115.84
112.29*
67,192
12,021
45
22.41
55.90
33.49*
13,098
37,851
46
6.56
16.30
9.74*
56,727
117,007
47
7.82
76.15
68.33*
56,555
16,628
48
179.25
452.64
273.39*
97,593
7,172
49
27.82
59.75
31.93*
16,766
10,549
50
14.23
240.27
226.04*
25,410
2,258**
51
16.37
224.54
208.17*
24,715
2,385**
52
26.01
96.14
70.13*
39,858
11,418
53
6.69
27.49
20.80*
44,694
43,168
54
41.41
95.41
54.00*
74,502
27,718


CHAPTER II
REVIEW OF THE LITERATURE
Sex differences in mathematics learning and
achievement have been a subject of concern for educators
and psychologists. Many studies found in the literature
support the existence of these differences. Boys show
superiority in higher level cognitive tasks (problem
solving or mathematical reasoning) in the upper elementary
years and in the early high school years (Fennema/ 1974;
Maccoby & Jacklin/ 1974) .
Almost all the research carried out has dealt with
analysis of total correct scores in mathematics aptitude
and achievement tests or scores in subtests. The
literature related to sex differences in incorrect
responses/ the main subject of the present study/ is
surprisingly sparse. For the most part/ the studies have
investigated the differences between the sexes in
mathematics learning and achievement and the underlying
variables causing the differences. Cognitive and affective
variables have been the matter at issue in the
establishment of sex differences.
Although research in mathematics problem solving/ the
secondary subject of this investigation/ is extensive/ most
9


48
success* since the components must be integrated into a
whole process (mastery learning of the components cannot
assure mastery of the process).
Their analysis also led them to conclude that* in the
case of 6th graders* inability to read problems is a major
obstacle in solving word problems. Only 12% of the
subjects could read and set up problems correctly at a
higher level than they could compute* while 60% could
compute correctly at a higher level than they could read
and set up problems; 44% could set up problems better when
they heard them read than when they read the problem
themselves. Only 13% could set up problems better when
they read them than when they heard them read.
Muth ( 1984 ) investigated the role of reading and
computational skills in the solution of word problems. A
group of 200 students from the 6th grade were administered
a test of basic skills and a mathematics word problem test.
The word problem test consisted of 15 sample items supplied
by the National Assessment of Educational Progress. The
items were adapted to include some extraneous information
and complex syntactic structure. Four versions of the test
were constructed by combining two versions of problem
information (absence vs. presence of extraneous
information) with two versions of syntactic structure
(simple vs. complex syntax) Task performance was measured
by means of the number of problems answered correctly*


104
to contribute to the body of knowledge regarding such
differences in mathematical learning and achievement. This
section of Chapter V comprises implications of the findings
and recommendations evolving from the present
investigation.
ex-related,,Differences jp incorrect Responses
The main purpose of education is to impart formal
education equally for the sexes. Analysis of the data on
incorrect responses showed that for 100 of the 111 items of
the test/ males and females selected different incorrect
responses/ and this pattern of responses was consistently
found during the three years of test administration.
However/ with a sample of less than 3/000 subjects/
statistical significant results were obtained for only
6 items. Therefore/ it seems that/ at least for 6th grade
children/ educators need not concern themselves with
varying teaching techniques for the sexes.
This study/ however/ has value apart from the
investigation of sex-related differences. It was found
that/ year after year/ 6th graders from the public schools
in Puerto Rico tend to make the same errors in the solution
of mathematical problems. It is important/ therefore/ that
teachers provide the learning strategies necessary to
enable students to obtain a better understanding of the
concepts of numbers.


23
computation grew larger. This difference grew larger for
girls than for boys.
Although SES was a major factor in solving computation
and story problem items successfully/ the effect was
similar for each sex. Sex-related differences by primary
language or chronological age were not large.
This research carried out by Marshall with elementary
grade children supports previous research findings that
males are better than females in mathematics problem
solving (a higher order skill) and females are better than
their counterpart males in computation (a lower level
skill). Marshall's research also brought out a different
aspect of this question: the notion that girls find it more
necessary than boys to succeed in the computation items in
order to successfully solve the story problem items.
Cognitive and Affective Factors That
Influence_Sex Differences in Mathematics Learning
and Achievement
The research reviewed in the literature does not
provide evidence of any unique variable that could serve
as an explanation for the observed sex-related differences
in mathematics learning and achievement. However/ some
issues have been discussed/ among which the most prevalent
are that sex-related differences in mathematics learning
and achievement are a result of differences in formal
education; that sex-related differences in mathematics


30
The issue of sex-related differences in spatial
visualization ability as an explanation for sex differences
observed in mathematics achievement is less convincing and
the findings more contradictory than in the issue of sex
differences in formal education. Besides these cognitive
issues/ other issues/ mostly affective in nature/ have also
been studied in trying to explain the origin of these sex
differences in mathematics achievement and learning. The
studies dealing with these affective variables are reviewed
in the next section.
Differentiated Effect of Affective Variables
Researchers have attempted to explain the effect of
sex differences in internal beliefs/ interests/ and
attitudes (affective variables) on mathematics learning and
achievement. A brief statement of each explanation
precedes the summary of studies conducted that support the
explanation.
;. Females/ more
than males/ lack confidence in their ability to learn
mathematics and this affects their achievement in
mathematics and their election of more advanced mathematics
courses.
Maccoby and Jacklin (1974) reported that self-
confidence in terms of grade expectancy and success in
particular tasks was found to be consistently lower in
women than in men. In 1978/ Fennema and Sherman reported


43
Children (WISC). Multiple regression analysis was
performed. A correlation of .769 was found between word
problem solving and the eight independent variables.
Correlations between word problem solving and the
independent variables ranged from .363 (verbal
intelligence) to .674 (mathematical concepts).
Correlations between the independent variables ranged from
.369 (verbal intelligence and computation) to .749 (reading
comprehension and vocabulary). Mathematical concepts*
computation* and number sentence selection were almost as
effective as all eight independent variables in predicting
achievement in mathematical word problem solving.
Mathematical concepts* computation* number sentence
selection* and reading comprehension accounted for 58% of
the variance* whereas all eight predictors accounted for
59% of the variance. The two best predictors were
mathematical concepts and computation* which accounted for
54% variance. Other variables accounted for about 40% of
the variance.
The author recommends that the findings of this study
be interpreted cautiously because the correlation between
the eight independent variables was high* and* according to
Zalewski*
in a study of this nature where the interest is
primarily in the influence of several variables
on one dependent variable* a low correlation
between the independent variables is required.
(p. 2804)


99
responses was consistent throughout the years in which the
test was administered. Males and females selected
different incorrect responses in 100 of the 111 items of
the test. The pattern of male-female differences in the
selection of incorrect responses was consistently found in
each of the years of test administration. However/ for the
vast majority of the 100 items/ male-female differences
were relatively small in magnitude/ considering the fact
that the number of subjects needed to obtain significance
was very large. Therefore/ these findings lack educational
significance.
A second objective of the study was to investigate
whether males and females differ in problem solving
performance/ if these differences persist after controlling
for computational skills/ and if the male-female
differences depend on the level of computational skills.
Results of the analyses reported in this section
generally indicated that females outperformed males not
only in problem solving/ but in six of the seven
computational variables. Males surpassed females in
equivalence/ but statistical significance was obtained for
only one of the three years covered in the study.
The results also tended to show that/ for examinees
with similar levels of computational skills/ sex
differences in problem solving did not exist. The only


103
question was answered/ partially/ in the affirmative.
Interactions/ indicating the dependence/ were significant
for multiplication and division in the first year and for
subtration in the second year.
In general/ the sex-related differences in mathematics
problem-solving/ in favor of males performance/ reported by
Marshall (1981/ 1984)/ and Fennema and Tartree (1983) were
not supported in this investigation. Female performance
was equal or higher than male performance in problem
solving.
The findings of the study have something in common
with other investigations where mathematics problem solving
has been the subject of interest/ the fact that males and
females do poorly in word-story problem items (problem
solving) The mean score in problem solving for males and
females in the 9-item problem solving subtest was 3.436 and
3.862 respectively for the first year/ 3.632 and 4.015 for
the second year/ and 3.927 and 4.341 for the third year.
Implications of the Findings and
Suggestions for Further Research
The main purpose of this study was to investigate
sex-related differences in the selection of incorrect
responses. Another objective was to compare male and
female performance after accounting for computational
ability. This type of research is conducted in an attempt


43
Children (WISC). Multiple regression analysis was
performed. A correlation of .769 was found between word
problem solving and the eight independent variables.
Correlations between word problem solving and the
independent variables ranged from .363 (verbal
intelligence) to .674 (mathematical concepts).
Correlations between the independent variables ranged from
.369 (verbal intelligence and computation) to .749 (reading
comprehension and vocabulary). Mathematical concepts
computation and number sentence selection were almost as
effective as all eight independent variables in predicting
achievement in mathematical word problem solving.
Mathematical concepts computation number sentence
selection and reading comprehension accounted for 58% of
the variance whereas all eight predictors accounted for
59% of the variance. The two best predictors were
mathematical concepts and computation which accounted for
54% variance. Other variables accounted for about 40% of
the variance.
The author recommends that the findings of this study
be interpreted cautiously because the correlation between
the eight independent variables was high and according to
Zal ewski
in a study of this nature where the interest is
primarily in the influence of several variables
on one dependent variable a low correlation
between the independent variables is required.
(p. 2804)


65
One important characteristic of this table involves
two-by-two subtables of sex cross classified with two of
the options for each year. For example* consider the three
two-by-two tables obtained by cross classifying sex and
option choices one and two for each year. These tables are
indicated by dotted lines on Table 3.3. For this table*
the ratio of the odds of a male choosing option one to the
odds of a female choosing option one is the same (within
truncating error) for each year. For example* this odds
ratio for the first year is (.120/.045) / (.060/.060) =
2.67. Within the error caused by reporting truncated
probabilities* the odds ratio for years two and three is
the same as that for year one. The equality of these odds
ratios is Marshall's operationalization stage of the year-
to-year gender-option dependency.
To show that the odds ratio can be constant over
years* but that the probabilities of option choice
conditional on sex and year can change from year to year *
for both males and females* the probabilities in Table 3.3
were converted to the probabilities of option choice
conditional upon sex and year. These conditional
probabilities are reported in Table 3.4. Unlike the
probabilities in Table 3.2* those in Table 3.4 change from
year to year for both males and females.


MATH CONCEPTS
Fractions j
2 items) j
/ Numbers
j
\
\
(18 items )
\
/
Fig. 3.1 Number of Questions by Skill Area in the Basic
Skills Test in Mathematics 6


CHAPTER III
METHOD
The first objective of this study was to investigate
sex differences in the selection of incorrect responses in
a mathematics multiple-choice achievement test* and to
determine whether these differences were consistent over
three consecutive administrations of the test. The second
objective was to investigate whether males and females
differ in problem solving performance* if these differences
persist after accounting for computational skills* and if
the male-female differences depend on the level of
computational skills. This chapter contains descriptions
of the sample* the test instrument* and the statistical
analysis used in achieving the above mentioned objectives.
The Sample
To achieve the first objective of the study* all the
students who took the Basic Skills Test in Mathematics-6
("Prueba de Destrezas Bsicas en Matemticas-6") in
each or the three years were included in the study. To
achieve the second objective of the study* approximately
54


REFERENCES
Aiken/ L. R. (1970). Attitudes toward mathematics.
Re.Y,i£M_Qf. Education2_Reg.ar.£h/ 10, 551-596.
Allen, R. H., & Chambers, D. L. (1977). A comparison of
the mathematics achievement of males and females
(Bulletin No. 9194). Madison: Wisconsin State
Department of Public Instruction, Division for
Management and Planning Service. (ERIC Document
Reproduction Services No. ED 159 073)
Armstrong, j. m. (1979). ftgh.ie-yejnent_and..participatiQn pf
women in mathematics: An overview. (Report of a
two year study). Denver, CO: Educational Commission
of the States, National Assessment of Educational
Progress. (ERIC Document Reproduction Service
No. ED 184 878)
Armstrong, J. M., & Price, R. A. (1982). Correlates and
predictors of women's mathematics participation,
journal f. 99-109.
Astin, H. S. (1974). Overview of the findings. In H.
Astin, H. Suniewick, & S. Dweck (Eds.), Women: A
Bibliography on their Education and Careers.
(pp. 1-10). New York: Behavorial Publications.
Backman, M. E. (1972). Patterns of mental abilities:
Ethnic, socioeconomic and sex differences. American
Educational Research Journal, £, 1-11.
Ballew, H., & Cunningham, W. (1982). Diagnosing strengths
and weaknesses of sixth-grade students in solving
word problems. Journal for Research in Mathematics
Education, 11, 202-210.
Balow, I. H. (1964). Reading and computation ability as
determinants of problem solving. The Arithmetic
Teacher, 11, 18-22.
Bar-Tal, D., & Frieze, J. H. (1977). Achievement
motivation for males and females as a determinant of
attribution for success and failure. Sex roles,
1, 301-313.
107


Table Page
4.11 Adjusted Means on Problem Solving*
by Covariate and Sex 96
4.12 ANCOVA Third Year 97
4.13 Adjusted Means on Problem Solving*
by Covariate and Sex 98
vi


35
It is not until adolescence that sex differences in
the perception of mathematics as a male domain are found
(Fennema* 1976; Stein* 1971? Stein & Smithless* 1969;
Verbeke* 1983). In a study with 2nd* 6th and 12th
graders* Stein and Smithless (1969) found that students'
perceptions of spatial* mechanical* and arithmetic skills
as masculine became more defined as these students got
older. Fennema (1976) considers that the influence each
sex exerts upon the other on all aspects of behavior is
stronger during adolescence. Since during these years
males stereotype mathematics as a male domain* they send
this message to females who* in turn* tend to be influenced
in their willingness to study or not to study mathematics.
Before that stage* girls consider arithmetic feminine*
while boys consider it appropriate for both sexes (Bobbe*
1971) .
Usefulness of mathematics. Females perceive mathema
tics as less useful to them than males do* and this
perception occurs at a very young age. As a result*
females exert less effort than males to learn or elect to
take advanced mathematics courses.
Many studies reported before 1976 found that the
perception of the usefulness of mathematics for one's
future differs for males and females* and is related to
course taking plans and behavior (Fox* 1977). If females
do not perceive mathematics as useful for their future*


92
TABLE 4.8
ANCOVA Summary Table: Other Covariates
First Year
Source
df
SS
MS
F
Addition
1
271.57
271.57
49.74*
Sex
1
15.54
15.54
2.85
Error
999
5453.90
5.46
Subtraction
1
771.78
771.78
155.60*
Sex
1
10.73
10.73
2.16
Error
999
4953.70
4.96
Fracadd
1
1412.11
1412.11
326.87*
Sex
1
7.54
7.54
1.75
Error
999
4313.30
4.32
Decsub
1
1281.10
1281.10
287.89*
Sex
1
4.94
4.94
1.11
Error
999
4444.30
4.45
Equivalence
1
811.44
811.44
164.93*
Sex
1
69.69
69.69
14.16*
Error
999
4914.00
4.92
* p .05


114
Sherman/ J. (1979) Predicting mathematics performance in
high school girls and boys. Journal of Educational
Psychology/ 71/ 242-249.
Sherman/ J. (1980). Mathematics/ spatial visualization/
and related factors: Changes in girls and boys/
grades 8-11. Journal of Educational Psychology/
22./ 476-482.
Stein/ A. H. (1971). The effects of sex role standards
for achievement and sex role preference on three
determinants of achievement motiviation.
Developmental Psychology/ 4./ 219-231.
Stein/ A. H./ & Smithless/ J. (1969). Age and sex
differences in children's sex role standards about
achievement/ Developmental Psychology/ 1/ 252-259.
Suydam/ M. N. (1971) Research on mathematics education
(K-12) reported in 1970. Journal for Research in
Mathematics Education/ Z> 257-298.
Suydam/ M. N. (1972). Research on mathematics education
(K-12) reported in 1971. Journal for Research in
Mathematics Education/ 196-232.
Suydam/ M. N. (1973) Research on mathematics education
(K-12) reported in 1972. Journal for Research in
Mathematics Education/ 1/ 205-242.
Suydam/ M. N. (1974). Research on mathematics education
(K-12) recorded in 1973. Journal for Research in
Mathematics Education/ / 238-272.
Suydam/ M. N. / & Weaver/ J. F. ( 1975) Research on
mathematics education (K-12) recorded in 1974.
Journal for Research in Mathematics Education/ £/
253-282.
Verbeke/ K. A. (1983). Sex-related differences in
mathematically gifted secondary students: An
investigation of selected cognitive/ affective/ and
educational factors (Doctoral dissertation/
University of Maryland/ 1982) Dissertation Abstracts
International/ 43/ 2267A-2268A.
Weiner/ B. (1974). Achievement motivation and attribution
theory. Morristown/ NJ: General Learning Press.


Ill
Hansen/ C. W. (1944). Factors associated with successful
achievement in problem solving in sixth grade
arithmentic. Journal of Educational Research# 38 #
111-119.
Harnisch# D. L. (1983). Item response patterns:
Application for educational practice. Journal of
Educational Measurement# 2SL' 191-205.
Hilton# T. L. # & Berglund# G. W. (1974). Sex differences
in mathematics achievement: A longitudinal study.
Journal of. Educational Research# £2# 231-237.
Hunter# J. E. # & Cohen# S. H. ( 1974). Correcting for
unreliability in nonlinear models of attitude change.
Psychometrika# 39 # 445-468.
Jarvis# 0. T. (1964). Boy-girl ability differences in
elementary school arithmentic. School Science and
Mathematics# 64 # 657-659.
Kaufman# A. (1984). Sex differences in mathematics
reasoning in mathematically gifted and average fifth
and sixth graders (Doctoral dissertation# Hofstra
University# 1983) Dissertation Abstracts Inter
national# 45 # 1094A-1095A.
Kilpatrick I. (1969). Problem solving and creative
behavior in mathematics. In J. W. Wilson# & L. R.
Carry (Eds.)# Studies in mathematics (Vol. No. 19).
Stanford# CA: School Mathematics Study Group.
Kloosterman# P. W. (1985). Attributional theory# learned
helplessness# and achievement in ninth grade
mathematics (Doctoral dissertation# University of
Wisconsin# 1984) Dissertation Abstracts Internatio
nal# 4£/ 919A-920A.
Knifong# J. D. # & Holtan# B. ( 1976) An analysis of
children's written solutions to word problems.
Journal for Research in Mathematics_Education# 2#
106-112.
Knifong# J. D.# & Holtan# B. (1977). A search for reading
difficulties among erred word problems. Journal for
Research in Mathematics Education# £# 227-230.


24
learning and achievement arise from sex differences in
spatial visualization; and that sex-related differences
result from a differentiated effect of affective variables
on the mathematics performance of males and females.
Differences in Formal Mathematics Education (Differential
Coursework Hypothesis)
The basis for the differential coursework hypothesis
is the fact that sex-related differences in mathematical
learning and achievement show up when comparing groups
which are not equal in previous mathematics learning.
After the 8th grade/ boys tend to select mathematics
courses more often than girls. Therefore/ girls show lower
achievement scores in mathematics tests because their
mathematics experience is not as strong as the boys'
(Fennema/ 1975; Fennema & Sherman/ 1977; Sherman/ 1979).
Fennema and Sherman's study (1977) lends additional
support to the feasibility of viewing sex differences in
mathematics learning and achievement as reflecting
something other than a difference in mathematics aptitude.
After controlling for previous study of mathematics/ they
found significant sex differences in mathematics
achievement in only two of the four schools under study/
making the attribution to sex per se less likely.
Controlling for the number of space visualization-related
courses/ the sex-related differences which originally


Problem Solving
88
T
T
T
1
1 2 3
Division
Fig. 4.2 Sex by Division Interaction
4
5
6


95
TABLE 4.10
ANCOVA Summary Table: Other Covariates
Second Year
Source
df
ss
MS
F
Addition
1
206.97
206.97
39.57*
Sex
1
19.12
19.12
3.66
Error
1010
5277.90
5.23
Multiplication
1
1026.68
1026.68
232.80*
Sex
1
.80
.80
.18
Error
1010
4458.30
4.41
Division
1
1117 .46
1117.46
258.67*
Sex
1
.64
.64
.15
Error
1010
4367.50
4.32
Fracadd
1
1341.13
1341.13
327.10*
Sex
1
2.03
2.03
.50
Error
1010
4143.80
4.10
Decsub
1
1453 .91
1453.91
364.39*
Sex
1
4.88
4.88
1.22
Error
1010
4031.10
3.99
Equivalence
1
797.11
797.11
171.79*
Sex
1
127.33
127.33
27.44*
Error
1010
4687.90
4.64
*
P<.05


33
Mathematics.as. a male domain. Mathematics is an
activity more closely related to the male sex domain than
to the female sex domain (Eccles et al.# 1983). Thus* the
mathematical achievement ot boys is higher than that of
girls.
According to John Ernest (1976) # in his study
Mathematics and Sex/ mathematics is a sexist discipline.
He attributed sex-related differences in mathematical
achievement to the creation by society of sexual
stereotypes and attitudes# restrictions# and constraints
that promote the idea of the superiority of boys in
mathematics. Ernest reported that boys# girls# and
teachers# all believe that boys are superior in
mathematics# at least by the time students reach
adolescence. Bern and Bern (1970) agree and argue that an
American woman is trained to "know her place" in society
because ot the pervasive sex-role concept which results in
differential expectations and socialization practices.
Plank and Plank (1954) were more specific. They
discussed two hypotheses related to this view: the
differential cultural reinforcement hypothesis and the
masculine identification hypothesis. The differential
cultural reinforcement hypothesis states that society in
general perceives mathematics as a male domain# giving
females less encouragement for excelling in it. The
masculine identification hypothesis establishes that


CHAPTER V
DISCUSSION
This cnapter comprises two sections: a) summary and
interpretation of the results of the study/ and
b) implications of the findings/ with suggestions for
further research.
Summary and Interpretation of the Results
The main objective of this study was to investigate
whether there were sex-related differences in incorrect
responses/ and if the pattern of response was consistent
over three consecutive administrations of the mathematics
test. Males and females selected different incorrect
responses in 100 of the 111 items of the test and this
pattern of response was consistenly found in each of the
three years of test administration.
The results tend to support/ in part/ Marshall's
findings (1981/ 1983) of observed sex-related differences
in incorrect responses/ but in no way do they point to a
conclusion that 6th grade boys and girls from the public
schools in Puerto Rico exhibit different problem solving
strategies. The significant differentiated pattern of
101


44
In a more recent investigation# Exedisis (1983)
studied the contribution of reading ability# vocabulary#
mathematical concepts# computation# sex# and race on
problem-solving performance. The Iowa Test of Basic
Skills was administered to a group of 6th# 7th# and 8th
grade anglo and black Chicago male and female adolescents.
Problem solving was highly correlated to an understanding
of basic mathematical concepts# somewhat correlated to
race# and weakly correlated to computational and
vocabulary skills# sex# and reading ability.
Although the findings of these studies show a
relationship between computational skills and problem
solving achievement# this relationship is not strong enough
to be considered the most determinant factor in problem
solving achievement# as some of the researchers have been
careful to point out. In spite of the dismissal of reading
as a determinant factor in problem solving achievement by
some of these same researchers# more recent studies in this
area have led others to hold different views.
Reading end problem. SQlvinfl-JLgj.f.Qirniangg
Martin (1964) studied the contribution of reading
comprehension# computation# abstract verbal reasoning# and
arithmetic concepts to arithmetic problem solving
performance. Fourth and 8th grade students were
administered the Iowa tests of Basic Skills and the
Lorge-Thorndlike intelligence test (verbal).


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate/ in scope and quality/
as a dissertation for the degree of Doctor of Philosophy.
Jaimes Algina/Chai rman
Processor of Foundations of Education
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate/ in scope and quality/
as a dissertation for the degree of Doctor of Philosophy.
2U
J-
Linda Crocker
Professor of Foundations of Education
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate/ in scope and quality/
as a dissertation for the degree of Doctor of Philosophy.
Leadership
This dissertation was submitted to the Graduate Faculty of
the College of Education and to the Graduate School/ and
was accepted as partial fulfillment for the degree of
Doctor of Philosophy.
August/ 1986
Cnairman/ Founp
lations. of Education
A/uJ Cl zzUfh
Dean/ College of EducSrtion
Dean/ Graduate School


(a)
80
81
82
83
84
85
86
87
88
89
90
91
93
94
95
96
98
101
102
104
73
TABLE 4.1 Continued
X* for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(b)
(c)
(d)
75.01
55/277
6,787
95 .51
57 ,710
5,565
51.99
71/315
12,633
20.12
81/021
37 ,088
38.62
78/959
18,830
1240 .78
41,180
306**
102.85
39/903
3 ,573
66.75
48,194
6,650
63.01
38,742
5/663
43 .26
39,648
8,441
85 .53
53 ,557
5,767
82.53
62,954
7/025
78.39
80 ,723
9,484
70.15
87,241
11,454
40.59
83 ,969
19,053
29.31
74,287
23 ,343
29.87
69,052
21,291
23 .77
83 ,782
32/462
61.06
75/708
11/419
40.47
72 ,517
16 ,503


CHAPTER III
METHOD
The first objective of this study was to investigate
sex differences in the selection of incorrect responses in
a mathematics multiple-choice achievement test, and to
determine whether these differences were consistent over
three consecutive administrations of the test. The second
objective was to investigate whether males and females
differ in problem solving performance* if these differences
persist after accounting for computational skills* and if
the male-female differences depend on the level of
computational skills. This chapter contains descriptions
of the sample* the test instrument* and the statistical
analysis used in achieving the above mentioned objectives.
The Sample
To achieve the first objective of the study* all the
students who took the Basic Skills Test in Mathematics-6
("Prueba de Destrezas Bsicas en Matemticas-6") in
each or the three years were included in the study. To
achieve the second objective of the study* approximately
54


7
errors (Harnisch? 1983). Item response patterns are
valuable for the identification of large group differences/
including d i st rict-to-district/ school-to-school/ and
classroom-to-classroom variations on different subsets of
items. The response patterns can provide diagnostic
information about the type of understanding the student has
on various mathematics topics (e.g. / problem solving).
Marshall (1981/ 1984) has used the item response
pattern technique and her findings indicate sex differences
in mathematics performance at the item level. Females
outperformed males in computation and males outscored
females in problem solving. Also/ the success of girls in
the problem solving items was dependent upon their success
in the computation items; for boys/ success in the problem
solving items did not depend as much on their computational
performance. Marshall (1981/ 1983) has also reported that
males and females differ in the selection of incorrect
responses/ reflecting differences in reasoning abilities.
In Puerto Rico/ a high percentage of children promoted
to the 7th grade in the public schools does not master the
basic skills in mathematics. If 6th grade male and female
children can be diagnosed as having different problem
solving abilities/ as Marshall found with California
children/ teachers may need to provide tailor-made
mathematics instruction for each sex/ in order to ensure
equal access to formal education and enhance mathematics


49
number of problems set up correctly/ and amount of time
spent taking the test.
Reading ability and computational ability were both
positively correlated with number of correct answers and
with number of problems correctly set up/ and negatively
correlated with test-taking time. Presence of extraneous
information was negatively correlated with correct answers
and correct set ups and positively correlated with
test-taking time. Syntactic complexity was not
significantly correlated with any of the performance
measures.
Results of a multiple regression analysis showed that
reading accounted for 46% of the variance in total correct
answers and computation accounted for 8%. Reading ability
and computational ability uniquely accounted for 14% and 8%
of the variance in the number of correct answers/
respectively. Extraneous information added significantly
to the variance explained in the number of correct answers/
but syntactic structure did not. Reading ability accounted
for 5% of the variance in test-taking time/ but computation
did not add significantly to the variance explained by
reading.
Muth concluded that reading and computation both
contribute significantly to success in solving arithmetic
word problems/ but that reading plays a more significant
role than does computation.


39
The findings suggest that reading factors are not as
important as arithmetic and mental factors in problem
solving performance. However/ these findings should be
taken cautiously as the content of the Gates tests (used
to measure reading) is literary and does not include
mathematical material.
Chase (1960) studied 15 variables in an effort to find
out which ones have significant influence on the anility to
solve verbal mathematics problems. Only computation
reading to note details and fundamental knowledge were
primarily related to problem solving. Computation
accounted for 20.4% of the 32% variance directly associated
with problem solving.
Chase concluded that a pupil's ability in the
mechanics of computation comprehension of the principles
that underline the number systems and the extent to which
important items of information are noticed when reading
are good predictors of the student's ability in solving
verbal problems.
Balow (1964) investigated the importance of reading
ability and computation ability in problem solving
performance. He objected to the approaches used by other
researchers who in their analyses dichotomized research
subjects as "poor" or "good" students and who ignored the
recognizea effect of intelligence on reading and on
mathematics achievement. Balow administered the Stanford


31
that in their study involving students from grades 6
through 12, boys showed a higher level of confidence in
mathematics at each grade level. These differences between
the sexes occurred in most instances even when no
sex-related differences in mathematics achievement were
found. The correlation between confidence in mathematics
performance and mathematics achievement in this study was
higher than for any other affective variable investigated.
Sherman reported a similar finding in 1980; in males,
the most important factor related to continuation in
theoretical mathematics courses was confidence in learning
mathematics. This variable weighed more than any of the
cognitive variables: mathematics achievement, spatial
visualization, general ability, and verbal skill. In the
case of females, among the affective variables, confidence
in learning mathematics was found to be second in
importance to perceived usefulness of mathematics. Probert
(1983) supported these findings with college students.
A variable that needs discussion within the context of
sex differences in confidence as learners of mathematics is
causal attribution. Causal attribution models attempt to
classify those factors to which one attributes success or
failure. The model proposed by Weiner (1974) categorizes
four dimensions of attribution of success and failure:
stable and internal, unstable and internal, stable and
external, and unstable and external. For example, if one


93
superiority in problem solving lost statistical
significance. Adjusted meansr associated with the
analyses/ are reported in Table 4.9.
Summaries of the analyses of covariance for the second
year are reported in Table 4.10. The analyses show that
females retained their superiority in problem solving after
controlling for equivalence. When the controlling
variables were additionr multiplication/ division/ addition
of fractions/ and subtraction with decimals females
superiority in problem solving lost statistical
significance. Adjusted means are reported in Table 4.11.
The analyses of covariance for the third year are
summarized in Table 4.12. The results of the analyses show
that females maintained their superiority in problem
solving when performance on either subtraction or
equivalence was controlled. When addition/ multiplication/
division/ addition of fractions/ and subtraction of
decimals were the controlling variables/ female superiority
in problem solving lost statistical significance. Adjusted
means are reported in Table 4.13.
Summary
Sex-related differences in mathematics achievement
were the subject of this study. The first objective was to
investigate whether males and females differ in the
selection of incorrect responses/ and if the pattern of


98
TABLE 4.13
Adjusted Means on Problem Solving/
by Covariate and Sex
Third Year
Covariate
Adjusted Means
Males
Females
Addition
4.09
4.18
Subtraction
3.98
4.29
Multiplication
4.10
4.17
Division
4.17
4.16
Addf rac
4.07
4.20
Subdec
4 .15
4.12
Equivalence
3 .79
4.45


Problem Solving
89
Subtraction
Fig. 4.3 Sex by Subtraction Interaction


51
student attitudes and performance in the area of problem
solving.
Carey (1958) constructed a scale to measure attitudes
toward problem solving. Her interest was in general
problem solving rather than in mathematical problem
solving. Her work constitutes the first attempt to
construct a measure of attitudes toward problem solving.
The scale was used with a group of college students/ and
she foundr among other things/ that problem solving
performance is positively related to problem solving
attitudes and that/ in the case of females/ positive
modification of attitudes toward problem solving brings a
significant gain in problem solving performance.
Lindgren/ Silva/ Faracoz and DaRocha (1964) adapted
Carey's scale of attitudes toward problem solving and
applied it to a group of 4th grade Brazilian children.
Students also answered an arithmetic achievement test/ a
general intelligence test/ and a socioeconomic (SE) scale.
A low but significant positive correlation was found
between arithmetic achievement and attitudes toward problem
solving. A near zero correlation was found between
attitudes toward problem solving and intelligence. Since
problem solving is one aspect of the discipline of
mathematics/ this correlation between attitudes and
arithmetic achievement can lead to a conclusion or a strong


2
Many variables# cognitive/ affective# and educational#
have been investigated since 1974 in relation to sex
differences in mathematics learning and achievement.
Fennema and Sherman (1977) investigated the effect of
differential formal mathematics education. After
controlling for the number of years of exposure to the
study of mathematics# they found sex differences in only
two of the four schools under study. However# in those
schools where boys scored higher than girls# differences
were also found in their attitudes toward mathematics.
Hilton and Berglund (1974) found significant sex
differences after controlling for the number of mathematics
courses taken# and attributed them to sex differences in
interests. "As the boys' interests in science increase
related to the girls'# their achievement in mathematics
increases relative to that of the girls" (p. 234).
Wise# Steel# and McDonald (1979) reanalyzed test data
collected in a longitudinal study of 400#000 high school
students (Project Talent). They found that when the effect
of the number of high school mathematics courses was not
controlled# no sex differences emerged for 9th graders# but
that gains made by boys during the next three years were
more than twice that of the girls. These differences
between the sexes disappeared when the number of
mathematics courses taken was controlled. Results of the
1978 Women and Mathematics National Survey# Survey I#


28
1974 have failed to support these sex differences in
spatial abilities (Armstrong* 1979; Connor* Serbin* &
Schackman* 1977; Fennema & Sherman* 1978; Sherman* 1979).
Fennema and Sherman (1978) and Sherman (1979) have
explored sex-related differences in mathematical
achievement and cognitive and affective variables related
to these differences. In a study involving students from
grades 6* 7* and 8* from four school districts* Fennema and
Sherman found that spatial visualization and problem
solving were highly correlated for both sexes (.59 and
.60). Even in the school district where sex differences
were found in problem solving* no significant sex-related
differences were found in spatial visualization.
When Sherman (1980) compared groups of males and
females in two different grades* 8 and 11* she found no
sex-related differences in problem solving or in spatial
visualization in grade 8. In grade 11* however* although
the sexes differed in their problem solving performance* no
sex-related differences were found in spatial
visualization. Even though spatial visualization in grade
8 was the second best predictor of problem solving
performance in grade 11* sex differences in grade 11 were
not a result of spatial visualization since no differences
were found in that skill.
In spite of the fact that no sex differences were
found in spatial abilities* it is evident that males and


83
TABLE
4.3 -
Continued
Year/
Subtest
Males
X
SD
Females
X
SD
t
Decsub
2.863
1.87
3.440
1.80
5.00*
Equivalence
.76
.76
.65
.75
-2.32**
N =
504
N =
509
Third
Problem Solving
3.927
2.49
4.341
2.44
2.67*
Addition
5.536
.84
5.704
.64
3.59*
Subtraction
4.836
1.52
4.958
1.46
1.30
Multiplication
4.168
1.74
4.541
1.68
3.47*
Division
3.343
1.88
3.795
1.83
3.87*
Fracadd
2.819
1.89
3.117
1.85
2.53**
Decsub
3.021
1.93
3.448
1.85
3.60*
Equivalance
.830
.82
.800
.78
- .60
N =
509
N =
504
Note: The number of items in the problem solving subtest was 9.
In each computation subtest the number of items
was 6. An item was included in the computation subtest
only if it measured a computation skill required to solve
a problem solving item.
* p <.01
** p <.05


12
test form/ no sex differences were found in response
patterns. Approximately 45% of each sex selected option
£. The next popular choice for both sexes was option
d/ selected by approximately 35% of both boys and girls.
On the story problem of Form 1/ males and females
responded alike. Their most popular incorrect response
choice was option a. for both males and females. The
second most popular incorrect choice was option js for
both sexes.
Response to the story problem in Form 2 showed sex
differences in response choice. Including the correct
option/ 33% of the girls selected option ji/ 20% chose
option c / and 20% option d. For males/ approximately
25% selected option s and the same percent selected
option d.
Marshall concluded that although the analysis of
incorrect responses does not explain why boys and girls
differ in their responses/ the analysis shows that boys and
girls approach problems in different ways and these varying
strategies can be useful in indentifying how the sexes
differ in reasoning abilities.
Two years later/ Marshall (1983) analyzed the
responses of approximately 300/000 boys and girls to
mathematics items contained in the 16 test forms of the
Survey of Basic Skills during the years 1977/ 1978/ and
1979. She used log-linear models (explained in Chapter


106
operation to the numbers in the problem. The re
sults indicate that students are not familiar with
such basic problem-solving strategies as drawing
a picture of a figure described in a problem or
checking the reasonableness of a result, (p. 338)
The fact that/ in general/ after controlling for
computational skills/ there were no statistically
significant differences between the problem solving
performance of the males and females who participated in
this study would seem to imply that teachers need not worry
about designing different teaching strategies for the
sexes. However/ because the performance in problem-solving
is low for both sexes/ teachers may be encouraged to
emphasize problem solving in their daily teaching.
This study was designed to investigate sex-related
differences in mathematical performance. The findings
clearly demonstrated that 6th grade females and males
from the public schools in Puerto Rico are equally good or
equally bad in the solution of word-story items (problem
solving). Also/ in general/ the effect of computational
skills is similar for each sex. Based on these findings/
it appears that greater differences exist within each sex
than between the sexes. In conclusion/ the findings of the
present study highlight the need for investigations of
different approaches to instructional designs aimed at
improving both male and female performance in problem
solving.


82
TABLE 4.3
Means* Standard Deviations* and t-Tests for the Eight
Mathematical Subtests
Year/
Subtest
Males
X
SD
Females
X
SD
t
First
Problem Solving
3.436
2.32
3.862
2.45
2.82*
Addition
5.412
1.00
5.592
.86
3.05*
Subtraction
4.550
1.66
4.852
1.56
1.22
Multi plicati on
4.014
1.80
4.433
1.68
3.81*
Division
3.136
1.87
3.580
1.91
3.72*
Fracadd
2.475
1.75
2.743
1.81
2.32**
Decsub
2.648
1.83
3.321
1.86
5.77*
Equivalence
.77
.77
.73
.76
- .83
N =
492
N =
510
Second
Problem Solving
3.632
2.28
4.015
2.36
2.65*
Addition
5.470
.98
5.603
.86
2.29**
Subtraction
4.843
1.52
4.923
1.44
.86
Multiplication
4.148
1.73
4.550
1.67
3.76*
Division
3.242
1.77
3.632
1.86
3.42*
Fracadd
2.565
1.74
2.903
1.91
2.94*


20
As a result of her investigations/ Marshall (1981)
concluded that sex-related differences in mathematics
performance may be the result of comparing the sexes on
total test scores. If the test contains more computation
items than problem solving items/ girls will perform better
than boys/ but if the test contains more problem solving
items than computation ones/ boys will outperform girls.
With this in mind/ Marshall investigated sex-related
differences in computation and problem solving by analyzing
the responses of approximately 18/000 students from grade 6
who had been administered the Survey of Basic Skills Test:
Grade 6/ during the academic year 1978-79.
Two of the 16 test forms of the Survey were used to
assess skills such as concepts of whole numbers/ fractions/
and decimals. These skills were tested both as simple
computations and as story problems (problem solving).
Two computation items and one story problem item were
selected because they were particularly related; both
computation items required skills needed in solving the
corresponding story problem. It was assumed that correct
solution of the computation item correlates with solving
the story problem because the story problem requires a
similar computation.
Marshall found that girls were better in computation
and boys were better in problem solving. She also found
that boys were much more likely than girls to answer the


51
student attitudes and performance in the area of problem
solving.
Carey (1958) constructed a scale to measure attitudes
toward problem solving. Her interest was in general
problem solving rather than in mathematical problem
solving. Her work constitutes the first attempt to
construct a measure of attitudes toward problem solving.
The scale was used with a group of college students* and
she found* among other things* that problem solving
performance is positively related to problem solving
attitudes and that* in the case of females* positive
modification of attitudes toward problem solving brings a
significant gain in problem solving performance.
Lindgren* Silva* Faraco* and DaRocha (1964) adapted
Carey's scale of attitudes toward problem solving and
applied it to a group of 4th grade Brazilian children.
Students also answered an arithmetic achievement test* a
general intelligence test* and a socioeconomic (SE) scale.
A low but significant positive correlation was found
between arithmetic achievement and attitudes toward problem
solving. A near zero correlation was found between
attitudes toward problem solving and intelligence. Since
problem solving is one aspect of the discipline of
mathematics* this correlation between attitudes and
arithmetic achievement can lead to a conclusion or a strong


47
than the problem-interpretation scorer the difficulty was
due to reading ability. If the score of the lowest of the
three areas (computationr problem interpretation/ and
reading-problem interpretation score) was the same as the
reading-problem solving score/ the student's area of
greatest immediate need was either computation/ problem
interpretation/ or reading. If the reading-problem solving
score was lower than the lowest of the other three scores/
the student's area of greatest immediate need was
integration.
Analysis of the data revealed that for 19% of the
students/ problem interpretation was their major
difficulty; for 26% of the students/ integration (total
problem solving) was their greatest immediate need; for
another 26%/ computation was the major weakness; and for
29%/ reading was their greatest immediate need.
Seventy five percent of the students demonstrated
clear strength in computation/ 21% in problem
interpretation/ and 4% in reading-problem interpretation.
An analysis across all students (including those without
complete data) showed that 26% of the subjects could not
work word problems at a level as high as that at which they
could compute/ interpret problems/ and read and interpret
problems/ when those areas were measured separately. This
led them to conclude that knowing the skills or the
components of solving word problems is not sufficient for


62
Here the summation is over all the cells in the contingency
table/ F0 refers to the observed frequency in a cell/ and
Fe refers to the estimated expected frequency in a cell.
Denoting the observed proportion in a cell by P0 and the
estimated expected proportion in a cell by Pe/ the test
statistic can be written
G2 = 2 N J2Po 1 9e (Po/pe}- (3.5)
where N is the total number of subjects. This form of the
test statistic suggests the following strategy. For any
significant G2/ using P0 and Pe calculated from the
total data set available for an item/ calculate the minimum
N required for G2 to be significant. If the minimum N is
very large/ this suggests that the statistically
significant result is not practically significant since it
can only be detected in very large samples. Of course/ the
question remains as to what can be considered a minimum
large N. Although there is room for argument/ it seems
reasonable to claim that if an average of 1000 subjects per
year is required to show significance/ then the result is
not likely to be practically significant. On the basis of
this reasoning/ it was proposed to ignore all significant
results that would be nonsignificant if there were less
than 3000 subjects available. In addition/ all log-linear


BIOGRAPHICAL SKETCH
Sonia Feliciano was born on March 17/ 1938/ in
Mayaguez/ Puerto Rico. She graduated from Eugenio Ma. de
Hostos High School/ in her hometown. She received a degree
in business administration/ with a major in finance/ from
the University of Puerto Rico in June/ 1961.
After graduation she worked as junior accountant for
Pittsburgh Plate Glass/ International. During the academic
year of 1969-70 she started her career in education/
obtaining the degree of Master in Education from the
University of Puerto Rico in June/ 1975.
From January/ 1971 to June/ 1985/ she served in the
public school system of Puerto Rico in different positions.
She worked as educational researcher until July/ 1977/ when
was appointed Director of the Research Center of the
Commonwealth Department of Education.
During the academic year of 1978-79/ she initiated
doctoral studies in the Faculty of Philosophy and
Educational Sciences of the Universidad Complutense de
Madrid/ in Madrid/ Spain. She completed the coursework
requirements and then transferred to University of Florida
during the fall of 1979.
116


52
correlation between attitudes toward problem solving and
problem solving performance.
Whitaker (1976) constructed a student attitude scale
to measure some aspects of 4th grade student attitudes
toward mathematic problem solving. He included statements
reflecting children's beliefs about the nature of various
types of mathematical problems the nature of the problem
solving process the desirability or persevering when
solving a problem and the value of generating several
ideas for solving a problem.
He correlated student attitudes toward problem solving
with their scores in a mathematical test which yielded a
comprehension score an application score and a problem
solving score. He found a significant positive
relationship between problem solving performance and
student attitude scores on the subscale which measured
reactions to such things as problem solving techniques or
problem situations or to the frustation or anxiety
experienced when confronted with problem solving
situations.
In another part of this study Whitaker investigated
the relationship between the attitudes of 4th grade
teachers toward problem solving and their students'
performance in problem solving. A very weak and
nonsignificant negative correlation was found between the
teacher's attitudes toward problem solving and student
performance.


20
As a result of her investigations Marshall (1981)
concluded that sex-related differences in mathematics
performance may be the result of comparing the sexes on
total test scores. If the test contains more computation
items than problem solving items girls will perform better
than boys but if the test contains more problem solving
items than computation ones boys will outperform girls.
With this in mind Marshall investigated sex-related
differences in computation and problem solving by analyzing
the responses of approximately 18000 students from grade 6
who had been administered the Survey of Basic Skills Test:
Grade 6 during the academic year 1978-79.
Two of the 16 test forms of the Survey were used to
assess skills such as concepts of whole numbers fractions
and decimals. These skills were tested both as simple
computations and as story problems (problem solving).
Two computation items and one story problem item were
selected because they were particularly related; both
computation items required skills needed in solving the
corresponding story problem. It was assumed that correct
solution of the computation item correlates with solving
the story problem because the story problem requires a
similar computation.
Marshall found that girls were better in computation
and boys were better in problem solving. She also found
that boys were much more likely than girls to answer the


91
TABLE 4.7
Reliability
Three
of the
Years of
Covariates for Each of
Test Administration
the
Year
Covariate
First
Second
Third
Addition
.579
.610
.454
Subtraction
.756
.724
.737
Multiplication
.729
.729
.734
Division
.716
.680
.714
Fracadd
.709
.728
.748
Decsub
.720
.716
.742
Equivalence
.421
.394
.488


LIST OF TABLES
Table Page
3.1 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex 58
3.2 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex and
Arranged by Sex 58
3.3 Hypothetical Joint Probabilities of Year/
Option/ and Sex 64
3.4 Hypothetical Probabilities of Option
Choice Conditional on Sex and Year 66
4.1 Summary of Significant Tests of Model 3.2 71
4.2 Chi-square Values for the Comparison of 75
Models (3.2) and (3.3) with Actual Sample
Sizes and with the Corresponding Number
of Subjects Needed for Significant Results 75
4.3 Means/ Standard Deviations/ and t-Test for
the Eight Mathematical Variables 82
4.4 ANCOVA First Year: Multiplication Covariate 85
4.5 ANCOVA First Year: Division Covariate 85
4.6 ANCOVA Second Year: Subtraction Covariate... 86
4.7 Reliability of the Covariates for Each
of the Three Years of Test Administra
tion 91
4.8 ANCOVA First Year: Other Covariates 92
4.9 Adjusted Means on Problem Solving/ by
Covariate and Sex 94
4.10 ANCOVA Second Year: Other Covariates 95
v


LIST OF TABLES
Table Page
3.1 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex 58
3.2 Hypothetical Probabilities of Option
Choice Conditional on Year and Sex and
Arranged by Sex 58
3.3 Hypothetical Joint Probabilities of Yearr
Option/ and Sex 64
3.4 Hypothetical Probabilities of Option
Choice Conditional on Sex and Year 66
4.1 Summary of Significant Tests of Model 3.2 71
4.2 Chi-square Values for the Comparison of 75
Models (3.2) and (3.3) with Actual Sample
Sizes and with the Corresponding Number
of Subjects Needed for Significant Results 75
4.3 Means/ Standard Deviations/ and t-Test for
the Eight Mathematical Variables 82
4.4 ANCOVA First Year: Multiplication Covariate 85
4.5 ANCOVA First Year: Division Covariate 85
4.6 ANCOVA Second Year: Subtraction Covariate... 86
4.7 Reliability of the Covariates for Each
of the Three Years of Test Administra
tion 91
4.8 ANCOVA First Year: Other Covariates 92
4.9 Adjusted Means on Problem Solving/ by
Covariate and Sex 94
4.10 ANCOVA Second Year: Other Covariates 95
v


22
Approximately 270/000 students from the 6th grade were
administered the Survey of Basic Skills of the California
Assessment Program/ during the years 1977/ 1978/ and 1979.
Responses were analyzed using log-linear models.
Successful solving of computation items was positively
associated with successful solving of story problems.
Girls were more successful in computation than boys/ and
boys were more successful than girls in solving story
problems. This finding supports reports from the National
Assessment of Educational Progress (NAEP) (Armstrong/
1979) .
To investigate the effects of reading/ SES/ language/
and chronological age/ only those test forms containing 2
computation items and 2 story problems were considered for
analysis; 32 items from 8 test forms were included in the
analysis.
The results of these analyses showed that at every
level of reading score/ 6th grade children were more
successful in computation than in story problems.
Although the differences were not large/ at every reading
score boys consistently had higher probabilities of
success in story problems than did girls/ and girls
consistently showed higher probabilities of success in
computation than boys. Also/ as the reading score
increased/ the difference between the probability of
success in story problems and the probability of success in


94
TABLE 4.9
Adjusted Means on Problem Solvingf
by Covariate and Sex
First Year
Covariate
Adjusted Means
Males
Females
Addition
3 .53
3 .78
Subtraction
3 .55
3.76
Fracadd
3.56
3.74
Decsub
3.73
3.58
Equivalence
3 .38
3.91


17
attitude toward mathematical problem solving/ and related
sex and program-type differences. Although his main
objective was to construct an attitude scale to measure
attitudes toward problem solving/ his study is important
because his findings support Meyer's regarding the lack of
significant sex-related differences in problem solving
performance. Performance in the problem solving questions/
for both males and females/ was much lower than performance
in the application questions/ and much lower than
performance in the comprehension questions. In fact/ the
mean score for each part of the item/ for both males and
females/ was almost identical to the mean scores obtained
by males and females in Meyer's study. Whitaker noted that
each application item is more difficult than its preceding
comprehension item/ and that each problem solving item is
more difficult than its preceding application item. No
significant sex-related differences were found for any of
the three parts of the item (comprehension/ application/ or
problem solving).
Fennema and Sherman (1978) investigated sex-related
differences in mathematics achievement and cognitive and
affective variables related to such differences. They
administered the Romberg-Wearne Problem Solving Test to a
representative sample of 1320 students (grades 6-8) from
Madison/ Wisconsin/ predominantly middle-class/ but
including great diversity in SES. The sample consisted of


19
solving performance in grade 11. Spatial visualization was
a stronger predictor for girls than for boys. Mathematics
as a male domain was a good predictor for girls only; the
less a girl stereotyped mathematics as a male domain in
grade 8/ the higher her problem solving score in grade 11.
Attitudes toward success in mathematics in grade 8 was a
more positive predictor of problem solving performance for
boys than for girls; the more positive the attitudes toward
success in mathematics in grade 8/ the higher their
performance in problem solving in grade 11.
None of these four studies all of which used the
Romberg-Wearne Mathematics Problem Solving Test show
statistically significant sex-related differences in
problem solving performance. In later studies other tests
were used to measure this variable (Kaufman 1984;
Marshall 1981 1984) .
Kaufman (1983/1984) investigated if sex differences in
problem solving favoring males exist in the 5th and 6th
grades and if these differences were more pronounced in
mathematically gifted students than in students of average
mathematical ability. The Iowa Test of Basic Skills and a
mathematics problem-solving test were administered to 504
subjects. Males in the average group as well as males in
the gifted group outperformed females but only the gifted
group showed statistically significant differences.


7 4
TABLE 4.1 Continued
Item
Number
X2 for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
la)
(b)
(c)
(d)
105
28.66
60/712
19/510
108
26 .56
86/261
29/912
111
30.71
86/190
25/849
* p<.01
** Minimum sample size less than 3/000


Abstract of Dissertation Presented to the Graduate
School of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy
COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
Sonia Feliciano
August/ 1986
Chairman: James Algina
Major Department: Foundations of Education
The first objective of this study was to investigate
sex differences in the selection of incorrect responses on
a mathematics multiple-choice test/ and to determine
whether these differences were consistent over three
consecutive administrations of the test. A second
objective was to compare male and female performance in
problem solving after controlling for computational skills.
The responses of all 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 ("Prueba de Destrezas Bsicas en
Matemticas-6") during three academic years were used in
the analyses relevant to the first objective.
VXll


14
Sex-related Differences in Problem Solving
It has already been acknowledged that the subject of
problem solving has been extensively researched. However*
as early as 1969* Kilpatrick criticized the fact that the
study of problem solving has not been systematic; some
researchers have studied the characteristics of the problem
while others have given their attention to the
characteristics of the problem solvers. Moreover*
differences in the tests used to measure problem solving
performance also constitute an obstacle when trying to
compare the results of the studies carried out.
In order to avoid this pitfall and provide a basis for
comparison* the studies reviewed in this section* dealing
with sex-related differences in problem solving* have been
divided in two groups. The first comprises those studies
that used the Romberg-Wearne Problem Solving Test. The
second contains other relevant studies in which problem
solving performance has been measured by means of other
instruments.
The Romberg-Wearne Problem Solving Test merits special
mention because it was the first attempt "to develop a test
to overcome the inadequacies of total test scores in
explaining the reasons why some students are successful
problem solvers and others are not" (Whitaker* 1976* pp. 9*
10) .


57
Analysis of the Data
Analysis of Sex bv Option by Year Cross Classifications
Log-linear models were used to analyze the sex by
option by year cross classifications for each item. Two
topics are addressed in this section/ the hypotheses tested
using log-linear models and a comparison of the hypotheses
tested in this study with those tested by Marshall (1981/
1983) .
The object of the analysis was to test two hypotheses:
1. The proportion of males and the proportion of
females choosing each incorrect option does not vary from
year to year. Note that this hypothesis is stated in the
null form.
2. Assuming that the first hypothesis is correct/ the
proportion of males who choose each incorrect option is
different from the proportion of females who choose each
incorrect option. This hypothesis is stated in the
alternate form.
In Table 3.1 a hypothetical cross classification of
sex/ option/ and year is presented. Hypothesis 1 is true
for this three dimensional contingency table. In Table 3.2
the three dimensional contingency table is rearranged to
show the year by option contingency table for each gender.


82
TABLE 4.3
Means* Standard Deviations* and t-Tests for the Eight
Mathematical Subtests
Year/
Subtest
Males
X
SD
Females
X
SD
t
First
Problem Solving
3.436
2.32
3.862
2.45
2.82*
Addition
5.412
1.00
5.592
.86
3.05*
Subtraction
4.550
1.66
4.852
1.56
1.22
Multi plication
4.014
1.80
4.433
1.68
3.81*
Division
3.136
1.87
3.580
1.91
3.72*
Fracadd
2.475
1.75
2.743
1.81
2.32**
Decsub
2.648
1.83
3.321
1.86
5.77*
Equivalence
.77
.77
.73
.76
- .83
N =
492
N =
510
Second
Problem Solving
3.632
2.28
4.015
2.36
2.65*
Addition
5.470
.98
5.603
.86
2.29**
Subtraction
4.843
1.52
4.923
1.44
.86
Multi plication
4 .148
1.73
4.550
1.67
3.76*
Division
3.242
1.77
3.632
1.86
3.42*
Fracadd
2.565
1.74
2.903
1.91
2.94*


CHAPTER V
DISCUSSION
This cnapter comprises two sections: a) summary and
interpretation of the results of the study; and
b) implications of the findings/ with suggestions for
further research.
Summary and Interpretation of the Results
The main objective of this study was to investigate
whether there were sex-related differences in incorrect
responses/ and if the pattern of response was consistent
over three consecutive administrations of the mathematics
test. Males and females selected different incorrect
responses in 100 of the 111 items of the test and this
pattern of response was consistenly found in each of the
three years of test administration.
The results tend to support/ in part/ Marshall's
findings (1981/ 1983) of observed sex-related differences
in incorrect responses/ but in no way do they point to a
conclusion that 6th grade boys and girls from the public
schools in Puerto Rico exhibit different problem solving
strategies. The significant differentiated pattern of
101


31
that in their study involving students from grades 6
through 12 boys showed a higher level of confidence in
mathematics at each grade level. These differences between
the sexes occurred in most instances even when no
sex-related differences in mathematics achievement were
found. The correlation between confidence in mathematics
performance and mathematics achievement in this study was
higher than for any other affective variable investigated.
Sherman reported a similar finding in 1980; in males
the most important factor related to continuation in
theoretical mathematics courses was confidence in learning
mathematics. This variable weighed more than any of the
cognitive variables: mathematics achievement spatial
visualization general ability and verbal skill. In the
case of females among the affective variables confidence
in learning mathematics was found to be second in
importance to perceived usefulness of mathematics. Probert
(1983) supported these findings with college students.
A variable that needs discussion within the context of
sex differences in confidence as learners of mathematics is
causal attribution. Causal attribution models attempt to
classify those factors to which one attributes success or
failure. The model proposed by Weiner (1974) categorizes
four dimensions of attribution of success and failure:
stable and internal unstable and internal stable and
external and unstable and external. For example if one


97
TABLE 4.12
ANCOVA Summary Table
Third Year
Source
df
ss
MS
F
Addition
1
205.96
205 .96
34.85*
Sex
1
1.88
1.88
.32
Error
1010
5967.60
5.91
Subtraction
1
886.42
886.42
169.48*
Sex
1
24.39
24.39
4.66*
Error
1010
5287.2
5.23
Multi pi icati on
1
1402.42
1402.42
297.12*
Sex
1
1.05
1.05
.22
Error
1010
4771.20
4.72
Division
1
2028.49
2028.49
494.75*
Sex
1
1.11
1.11
.27
Error
1010
4145 .10
4.10
Fracadd
1
1733.72
1733.72
393.97*
Sex
1
4.64
4.64
1.05
Error
1010
4439.89
4 .40
Decsub
1
2116.84
2116.84
526 .57*
Sex
1
.15
.15
.04
Error
1010
4056 .76
4.02
Equivalence
1
1186.92
1186.92
240.26*
Sex
1
59.85
59.85
12.12*
Error
1010
4986.69
4.94
*
p<. 05


35
It is not until adolescence that sex differences in
the perception of mathematics as a male domain are found
(Fennemar 1976; Stein/ 1971; Stein & Smithless/ 1969;
Verbeke/ 1983). In a study with 2nd/ 6th and 12th
graders/ Stein and Smithless (1969) found that students'
perceptions of spatial/ mechanical/ and arithmetic skills
as masculine became more defined as these students got
older. Fennema (1976) considers that the influence each
sex exerts upon the other on all aspects of behavior is
stronger during adolescence. Since during these years
males stereotype mathematics as a male domain/ they send
this message to females who/ in turn/ tend to be influenced
in their willingness to study or not to study mathematics.
Before that stage/ girls consider arithmetic feminine/
while boys consider it appropriate for both sexes (Bobbe/
1971).
Usefulness of mathematics. Females perceive mathema
tics as less useful to them than males do/ and this
perception occurs at a very young age. As a result/
females exert less effort than males to learn or elect to
take advanced mathematics courses.
Many studies reported before 1976 found that the
perception of the usefulness of mathematics for one's
future differs for males and females/ and is related to
course taking plans and behavior (Fox/ 1977). If females
do not perceive mathematics as useful for their future/


52
correlation between attitudes toward problem solving and
problem solving performance.
Whitaker (1976) constructed a student attitude scale
to measure some aspects of 4th grade student attitudes
toward mathematic problem solving. He included statements
reflecting children's beliefs about the nature of various
types of mathematical problems/ the nature of the problem
solving process/ the desirability or persevering when
solving a problem/ and the value of generating several
ideas for solving a problem.
He correlated student attitudes toward problem solving
with their scores in a mathematical test which yielded a
comprehension score/ an application score/ and a problem
solving score. He found a significant positive
relationship between problem solving performance and
student attitude scores on the subscale which measured
reactions to such things as problem solving techniques or
problem situations/ or to the frustation or anxiety
experienced when confronted with problem solving
situations.
In another part of this study/ Whitaker investigated
the relationship between the attitudes of 4th grade
teachers toward problem solving and their students'
performance in problem solving. A very weak and
nonsignificant negative correlation was found between the
teacher's attitudes toward problem solving and student
performance.


I certify that
opinion it conforms
presentation and is
as a dissertation for
I have read this study and that in my
to acceptable standards of scholarly
fully adequatef in scope and qualityf
the degree of Doctor of Philosophy.
is Algina/ Cl
iessor of Foi
irman
dations
of Education
I certify that
opinion it conforms
presentation and is
as a dissertation for
I have read this study and that in my
to acceptable standards of scholarly
fully adequatef in scope and qualityf
the degree of Doctor of Philosophy.
Linda Crocker
Professor of Foundations of Education
I certify that
opinion it conforms
presentation and is
as a dissertation for
I have read this study and that in my
to acceptable standards of scholarly
fully adequatef in scope and qualityf
the degree of Doctor of Philosophy.
.chael
Professor
r
njary
of Educational Leadership
This dissertation was submitted to the Graduate Faculty of
the College of Education and to the Graduate Schoolf and
was accepted as partial fulfillment for the degree of
Doctor
Augustf
Deanf Graduate School


10
of the studies consider sex differences incidental to the
major study findings. The available literature offers very
little research directly related to the problem of sex
differences in this area.
The review of the literature has been divided in four
sections. The first section consists of a detailed summary
of the available research on sex diferences in incorrect
responses. The second section deals with sex-related
differences in problem solving performance. These sections
are directly related to the objectives of the study. The
third section is more peripheral/ and contains a discussion
of the more prevalent issues about the influence of
cognitive and affective variables on sex differences in
mathematics learning and achievement. The fourth is a
summary of the research dealing with variables considered
as influential to mathematics problem solving performance.
Sex-related Differences in incorrect^Response Patterns
Research findings tend to suggest that boys and girls
may be approaching problem solving differently (e.g./
Fennema and Sherman# 1978; Marshall# 1981# 1983; Meyer#
1978# among others). Marshall (1981) investigated whether
6th grade boys and girls approach mathematical problem
solving with different strategies. Her specific interest
was whether the sexes made the same errors.


60
Fienberg also shows that deleting the Ujk term from (3.2)
to obtain
log mij k = Ui + Uj+Uk + Uik. (3.3)
yields a model that specifies that option is independent of
sex.
Based on Fienberg1 s presentation an appropriate
analysis for testing the hypotheses is
1. Conduct a likelihood ratio test of the adequacy of
model (3.2). If this test is nonsignificant then the
data support the adequacy of the model and the hypothesis
that conditional on gender option and year are
independent. Because model (3.1) is a saturated model
testing the adequacy of model (3.2) is the same as
comparing the adequacies of models (3.1) and (3.2).
2. Conduct a likelihood ratio test comparing the
adequacies of models (3.2) and (3.3). If this test is
significant then model (3.2) fits the data better than
model (3.3) and the data support the hypothesis that the
choice of incorrect option is dependent on sex.
To summarize if the first test is nonsignificant and
the second test is significant then the choice of option
is dependent on sex and the pattern of dependency is the
same for all three years.


117
From May 1982 to July* 1983 she worked as graduate
research assistant at the Foundations of Education
Department University of Florida. She returned to Puerto
Rico in August 1983 to serve as Special Aide to the
Assistant Secretary for the Vocational Technical and High
Skills Educational Programs of the Commonwealth Department
of Education.


110
Fennema E.r & Sherman/ J. (1977). Sex-related
differences in mathematics achievement/ spatial
visualization/ and affective factors. American
Educational Research Journal/ 14/ 51-71.
Fennema/ E. / & Sherman/ J. A. (1978). Sex-related
differences in mathematics achievement and related
factors. A further study. Journal for Research in
Mathematics...Education* h 189-204.
Fennema/ E./ & Tartree/ L. A. (1983). Research on
relationship of._spatisl...visualizatiQD,_and.confidence
to male/female mathematics achievement in grades 6-8.
Washington/ DC: National Science Foundation. (ERIC
Document Reproduction Service No. ED 232 853)
Fienberg/ s. e. (1980). Th.e...Analys.is..pf.CcQS.s=.ciaasified
Categorical Data. Cambridge/ MA: The MIT Press.
Fox/ L. H. (1975a). Mathematically precocious: Male or
female. In E. Fennema (Ed.) Mathematics learning:
what_r£S£arch--say-S-3bQut sea;_differences (pp. 1-12).
Columbus: Ohio State University/ College of
Education. (ERIC Document Reproduction Service No. ED
128 195)
Fox/ L. H. (1975b). Sex differences: Implications
for program_planning for..the academically giftedt
Paper presented at the Lewis M. Terman Memorial
Symposium on Intellectual Talent held at the Johns
Hopkins University/ Baltimore/ MD.
Fox/ L. H. (1977) The effects of sex role socialization
on mathematics participation and achievement. In J.
shoemaker (Ed.)/ women_and...Mathema.tics: Research
Perspectives for Change/ (N.I.E. Papers in Educa
tion and Work: No. 8). Washington/ D. C.: Eduation
and Work Group/ The National Institute of Education.
Glennon/ V. J./ & Callahan/ L. G. (1968) Elementary school
mathematics; A^guide tc current research/
Washington/ D. C.: Association for Supervision
and Curriculum Development.
Guay/ R. B. / & McDaniel/ D. (1977). The relationship
between mathematics achievement and space ability
among elementary school children. Journal for
Research.-in.-Mathamatics Education* £* 211-215.


18
students who had taken a similar number of mathematics
courses and were in the top 85% of the class in mathematics
achievement. They were tested in 1976. Four high school
districts were included. In only one of the high school
districts were sex-related differences in application and
problem solving found* in favor of males. They concluded
that when relevant factors are controlled* sex-related
differences in favor of males do not appear often* and when
they do* they are not large.
Sherman (1980) investigated the causes of the emerging
sex-related differences in mathematics performance* in
favor of males* during adolescence (grades 8-11). She
wanted to know if these differences emerge as a function of
sex-related differences in spatial visualization and
sociocultural influences that consider math as a male
domain. In grade 8* she used the Romberg-Wearne Test and*
in grade 11* a mathematical problem solving test derived
from the French Kit of Tests.
The analysis showed that for girls* problem solving
performance remained stable across the years. Mean problem
solving performance for boys* however* was higher in grade
11 than in grade 8. No sex-related differences were found
in grade 8* but boys outperformed girls in grade 11* where
the Stafford test was used.
Sherman found that for both sexes problem solving
performance in grade 8 was the best predictor of problem


TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS ii
LIST OF TABLES V
LIST OF FIGURES vii
ABSTRACT viii
CHAPTER
I INTRODUCTION 1
Purpose of the Study 6
Significance of the Study 6
Organization of the Study 8
II REVIEW OF THE LITERATURE 9
Sex-related Differences in Incorrect
Response Patterns 10
Sex-related Differences in Problem
Solving 14
Cognitive and Affective Variables that
Influence Sex Differences in Mathematics
Learning and Achievement 23
Differences in Formal Mathematics
Education 24
Differences in Spatial Ability 26
Differentiated Effect of Affective
Variables 30
Problem Solving Performance and Related
Variables 37
Computational Skills and Problem Solving
Performance 38
Reading and Problem Solving Perfor
mance 44
Attitudes Toward Problem Solving and
Problem Solving Performance 50
iii


MATH CONCEPTS
Fig. 3.1 Number of Questions by Skill Area in the Basic
Skills Test in Mathematics 6


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E7GIIQLHZ_4RJU7A INGEST_TIME 2012-09-24T12:49:43Z PACKAGE AA00011802_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


28
1974 have failed to support these sex differences in
spatial abilities (Armstrong/ 1979; Connor/ Serbin, &
Schackmanr 1977; Fennema & Sherman/ 1978; Sherman/ 1979).
Fennema and Sherman (1978) and Sherman (1979) have
explored sex-related differences in mathematical
achievement and cognitive and affective variables related
to these differences. In a study involving students from
grades 6/ 7/ and 8/ from four school districts/ Fennema and
Sherman found that spatial visualization and problem
solving were highly correlated for both sexes (.59 and
.60). Even in the school district where sex differences
were found in problem solving/ no significant sex-related
differences were found in spatial visualization.
When Sherman (1980) compared groups of males and
females in two different grades/ 8 and 11/ she found no
sex-related differences in problem solving or in spatial
visualization in grade 8. In grade 11/ however/ although
the sexes differed in their problem solving performance/ no
sex-related differences were found in spatial
vizualization. Even though spatial visualization in grade
8 was the second best predictor of problem solving
performance in grade 11/ sex differences in grade 11 were
not a result of spatial visualization since no differences
were found in that skill.
In spite of the fact that no sex differences were
found in spatial abilities/ it is evident that males and


3
indicated no significant sex differences for 8th grade
students on measures of problem solving or algebra.
However* females outperformed males in computation and
spatial visualization. For the 12th grade students*
statistically significant sex differences favoring males
were found in problem solving* but not in algebra*
computation* or spatial visualization. For males and
females who had enrolled in courses beyond general
mathematics and who had taken or were enrolled in courses
such as pre-calculus* calculus* or geometry* differences in
problem solving or spatial visualization did not exist.
Sex differences favoring males were found on a total score
obtained summing across the computation* problem solving*
and algebra subtests (Armstrong* 1979). The mathematics
data collected in the second survey by the 1978 National
Assessment of Educational Progress showed significant sex
differences for both 13- and 17-year-old students. The
13-year-old females outperformed males in the computational
subtest and males outscored females by 1 1/2 percentage
points in problem solving (statistically significant). No
statistically significant differences were found in
algebra. No sex differences were found for the 17-year-old
group either in the computation subtest or in the algebra
subtest. Males surpassed females in problem solving. A
reanalysis of the data from the 17-year-old group confirmed
male superiority in problem solving after controlling for


113
Meyer# R. A. (1978). £g.xrJg.la.ted_diffgJ.gi3c.es.j,p_math.ema:-
ti,cal_problem_ solving performance and intellectual
abilities. Madison: University of Wisconsin#
Research and Development Center for Cognitive
Learning. (ERIC Document Reproduction Service
No. ED 166 023) .
Milton# G. A. (1957) The effects of sex-role identifica-
cation upon problem solving skill. Journal of
Abnormal & Social.Psychology# 55.# 208-212.
Moller# N. J. (1983). The impact of gender# masculinity#
and feminity on math achievement and course decisions
(Doctoral dissertion# Purdue University# 1982).
Hiss ej.t a.tion,Abstrac.te_ International# l# 2584A.
Muth# K. D. (1984). Solving arithmetic word problems:
Role of reading and computational skills. Journal of
Educational. Psychology# 15# 205-210.
Pallas# A. M.# & Alexander# K. L. (1983). Sex differences
in quantitative SAT performance: New evidence on
the differential coursework hypothesis. American
Idji^atj^p.aJ.,Res.ea_rc.li.Journal* 21# 165-182.
Plank# E.# & Plank# R. (1954). Emotional components in
arithmetical learning as seen through autobiographies.
Psychological Studies of the Child# £# 274-296.
Porter# A. C. (1968). The effects of using fallible
variables in the analysis of covariance (Doctoral
dissertation# University of Wisconsin# 1967).
Di.s.s.er.tatifih-^abgtjact.,.Inte.r.P,a-t.ip.nai# 25# 3517B.
Probert# B. S. (1984). Math confidence workshops: A
multimodal group intervention strategy in mathematics
anxiety/avoidance. (Doctoral dissertation# University
of Florida# 1983). nis£.er.tatiQp_Abgtracts
International# 2231B.
Puerto Rico Department of Education. (1980). Basic
skills,test in mathematics-6. Hato Rey: Center of
Evaluation.
Radatz# H. (1979) Error analysis in mathematics educa
tion. jQumai fo.r.jtene.arch .in Majieinatl_c,s_fcjuc.a-
tion# lfi.# 163-172.
Riedesel# A. C. (1969) Problem solving: Some suggestions
from research. Arithmetic Teacher# 15# 54-58.


13
III) to investigate sex related differences in the
selection of incorrect responses/ and the consistency of
such differences over three years of administration of the
test.
Based on her findings that sex differences were found
in 80% of the items/ Marshall classified the students'
errors according to Radatz' (1979) five-category error
classification. The categories are language (errors in
semantics)/ spatial visualization/ mastery/ association/
and use of irrelevant rules.
It was found that girls' errors are more likely to be
due to the misuse of spatial information/ the use of
irrelevant rules/ or the choice of an incorrect operation.
Girls also make relatively more errors of negative transfer
and keyword association. Boys seem more likely than girls
to make errors of perseverance and formula interference.
Both sexes make language-related errors/ but the errors are
not the same.
Available research is not extensive enough to make
definite judgments about the sex-related differences
observed in incorrect responses. Clearly more research is
needed.


105
exr_related.,Differences in Problem Solving Performance
The results of this study do not give strong support
to Marshall's (1981/ 1984) findings that the effect of
computational skills on problem solving performance is
different for each sex. However/ the findings of the
present investigation are consistent with those found by
Meyer (1978) and Whitaker (1976): female performance in
problem solving is not significantly different from male
performance. Females showed superiority in problem solving
performance in all three years of test administration
although their superiority was retained only when
performance in equivalence was controlled for (for the
three years data) and when performance in subtraction was
controlled for (second year) (statistically significant
only after accounting for equivalence) .
The findings of this study are also in agreement with
those reported by the National Assessment of Educational
Progress (NAEP) (Carpenter et al.z 1980). Related to
problem solving/ they conclude/
if it were necessary to single out one area that
demands urgent attention/ it would clearly be
problem solving. At all age levels/ and in vir
tually every content area/ performance was extre
mely low on exercises requiring problem solving
or application of mathematical skills. In gene
ral/ respondents demonstrated a lack of the most
basic problem-solving skills. Rather than attempt
ing to think through a problem and figure out what
needed to be done to solve the problem/ most res
pondents simply tried to apply a single arithmetic


17
attitude toward mathematical problem solving* and related
sex and program-type differences. Although his main
objective was to construct an attitude scale to measure
attitudes toward problem solving* his study is important
because his findings support Meyer's regarding the lack of
significant sex-related differences in problem solving
performance. Performance in the problem solving questions*
for both males and females* was much lower than performance
in the application questions* and much lower than
performance in the comprehension questions. In fact* the
mean score for each part of the item* for both males and
females* was almost identical to the mean scores obtained
by males and females in Meyer's study. Whitaker noted that
each application item is more difficult than its preceding
comprehension item* and that each problem solving item is
more difficult than its preceding application item. No
significant sex-related differences were found for any of
the three parts of the item (comprehension* application* or
problem solving).
Fennema and Sherman (1978) investigated sex-related
differences in mathematics achievement and cognitive and
affective variables related to such differences. They
administered the Romberg-Wearne Problem Solving Test to a
representative sample of 1320 students (grades 6-8) from
Madison* Wisconsin* predominantly middle-class* but
including great diversity in SES. The sample consisted of


26
maintained their superiority in mathematics ability during
high school. While Fox attributed sex-related differences
in mathematical achievement to differential exposure to
mathematical games and activities outside school Benbow
and Stanley suggested that sex-related differences in
mathematics performance stem from superior mathematical
ability in males not from differences in mathematics
formal education.
The differential coursework hypothesis is not totally
convincing and as reporteu berore it has been challenged
by researchers such as Benbow and Stanley (1980).
However Pallas and Alexander (1983) have questioneu the
generalizanility of Benbow and Stanley's findings based on
the fact that they used highly precocious learners. The
differential coursework hypothesis can be accepted only as
a partial explanation of differences in mathematics
performance found between the sexes.
JUJ-££j^g£.s-Jn_Sp.fltial Ability
The basic premise in this issue is that males and
females differ in spatial visualization and this explains
differential mathematics learning and achievement. Until
recently sex differences in spatial ability in favor of
males were believed to be a fact and were thought by some
to be related to sex differences in mathematical
achievement


CHAPTER II
REVIEW OF THE LITERATURE
Sex differences in mathematics learning and
achievement have been a subject of concern for educators
and psychologists. Many studies found in the literature
support the existence of these differences. Boys show
superiority in higher level cognitive tasks (problem
solving or mathematical reasoning) in the upper elementary
years and in the early high school years (Fennema/ 1974;
Maccoby & Jacklin/ 1974) .
Almost all the research carried out has dealt with
analysis of total correct scores in mathematics aptitude
and achievement tests or scores in subtests. The
literature related to sex differences in incorrect
responses the main subject of the present study/ is
surprisingly sparse. For the most part/ the studies have
investigated the differences between the sexes in
mathematics learning and achievement and the underlying
variables causing the differences. Cognitive and affective
variables have been the matter at issue in the
establishment of sex differences.
Although research in mathematics problem solving/ the
secondary subject of this investigation/ is extensive/ most
9


3
indicated no significant sex differences for 8th grade
students on measures of problem solving or algebra.
However/ females outperformed males in computation and
spatial visualization. For the 12th grade students/
statistically significant sex differences favoring males
were found in problem solving/ but not in algebra/
computation/ or spatial visualization. For males and
females who had enrolled in courses beyond general
mathematics and who had taken or were enrolled in courses
such as pre-calculus/ calculus/ or geometry/ differences in
problem solving or spatial visualization did not exist.
Sex differences favoring males were found on a total score
obtained summing across the computation/ problem solving/
and algebra subtests (Armstrong/ 1979). The mathematics
data collected in the second survey by the 1978 National
Assessment of Educational Progress showed significant sex
differences for both 13- and 17-year-old students. The
13-year-old females outperformed males in the computational
subtest and males outscored females by 1 1/2 percentage
points in problem solving (statistically significant) No
statistically significant differences were found in
algebra. No sex differences were found for the 17-year-old
group either in the computation subtest or in the algebra
subtest. Males surpassed females in problem solving. A
reanalysis of the data from the 17-year-old group confirmed
male superiority in problem solving after controlling for


79
TABLE 4.2 Continued
Item
Num
ber
Xa for
Model
3.2
Xz for
Model
3.3
Xz for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
lb) .
(c) .
(d)...
(e)
(f)
75
8.69
237.07
228.38*
67/783
5/963
76
99.68
167.72
68.04*
70/183
20/723
77
79.38
432.36
352.98*
66/304
3/774
78
64.56
127.04
62.48*
76/806
24/696
79
23.98
72.59
48.61*
48/690
20/123
80
75.01
128.13
53.12*
55/277
20/906
81
95.51
106.82
11.31*
57/710
102/511
82
51.99
108.12
56.13*
71/315
25/525
83
20.12
95.18
75.06*
81/021
21/685
84
38.62
69.05
30.43*
78/959
52/129
86
102.85
433.47
330.62*
39/903
2/425**
87
66.75
430.40
363.65*
48/194
2/662**
88
63.01
126.27
63.26*
38/742
12/304
89
43.26
289.14
245.88*
39/648
3/240
90
85.53
405.04
319.51*
53/557
3/368
91
82.53
85.48
2.95
62/954
428/727
92
16.36
93.12
76.76*
88/231
23/092
93
78.39
229.02
150.63*
80/723
10/766
94
70-15
137.27
67.12*
87/241
26/113


41
categories. Category I included clerical and computational
errors. Category II included other types of errors* such
as average and area errors/ use of wrong operation* no
response* and erred responses offering no clues. It the
student's work indicated the correct procedure and yet the
problem was missed because of a computational or clerical
error* it was assumed that the problem was read and
understood.
An analysis of frequencies showed that clerical errors
were responsible for 3% of the problems incorrectly solved*
computational errors accounted for 49%* and other errors
for 48% of the erred problems.
Knifong and Holtan concluded that "improved
computational skills could have eliminated nearly half of
the word problem errors" (p. Ill). These computational
errors were made in a context where other skills such as
reading* interpretation of the problem* and integration of
these skills necessary for the solution of word problems*
might interact. However* Knifong and Holtan state that
their findings neither confirm nor deny that improvement of
reading skills will lead to improvement in problem solving.
They conclude that "it is difficult to attribute major
importance to reading as a source of failure" (p. 111).
In a later analysis* looking for evidence of poor
reading abilities affecting children's success in word
problems* Knifong and Holtan (1977) interviewed the


59
Inspection of the year by option contingency tables shows
that year and option are independent for each gender.
Thus/ hypothesis 1 is equivalent to the hypothesis that
year and option are independent conditional upon gender.
Hypothesis 2 is also true for Table 3.1 and Table 3.2.
Therefore/ when hypothesis 1 is correct/ hypothesis 2
is equivalent to the hypothesis that sex and option choice
are dependent.
In his discussion of the analysis of three dimensional
contingency tables/ Fienberg (1980) presents the following
saturated model for the data:
log mijk = Ui + U j + jc+i j + Uik+Uj k+Ui j k- (3.1)
In this model/ mjk is the expected value of the
frequency in cell ijk of the three dimensional table. The
model states that all three classification factors for a
three dimensional contingency table are mutually dependent.
In the present research i is the year index/ j is the
option index/ and k is the sex index. Fienberg shows that
deleting the terms Uj and U^jk yields a model in which
year and option are independent conditional upon sex. This
model is
log mij k
Ui+Uj+Uk+Uik+Ujk
(3.2)


29
females may use them in a different way. Meyer (1978)
with an elementary grade sample and Fennema and Tartree
(1983) with an intermediate level sample found that the
influence of spatial visualization on solving mathematics
problems is subtle and that males and females use their
spatial skills differently in solving word story problems
(problems that measure problem solving ability or
reasoning). Fennema and Tartree (1983) carried out a
three-year longitudinal study which showed that girls and
boys with equivalent spatial visualization skills did not
solve the same number of items nor did they use the same
processes in solving problems. The results also suggested
that a low level of spatial visualization skills was a more
debilitating factor for girls than for boys in problem
solving performance.
Landau (1984) also investigated the relationship
between spatial visualization and mathematics achievement.
She studied the performance of middle school children in
mathematical problems of varying difficulty and the extent
to which a diagramatic representation is likely to
facilitate solution. She found that spatial ability was
strongly correlated to mathematical problem solving and
that the effect of spatial ability was more influential for
females. Females made more use of diagrams in the solution
of problems reducing the advantage of males over females
in problem solving performance.


Problem Solving
89
Subtraction
Fig. 4.3 Sex by Subtraction Interaction


44
In a more recent investigation? Exedisis (1983)
studied the contribution of reading ability* vocabulary*
mathematical concepts* computation* sex* and race on
problem-solving performance. The Iowa Test of Basic
Skills was administered to a group of 6th* 7th* and 8th
grade anglo and black Chicago male and female adolescents.
Problem solving was highly correlated to an understanding
of basic mathematical concepts* somewhat correlated to
race* and weakly correlated to computational and
vocabulary skills* sex* and reading ability.
Although the findings of these studies show a
relationship between computational skills and problem
solving achievement* this relationship is not strong enough
to be considered the most determinant factor in problem
solving achievement* as some of the researchers have been
careful to point out. In spite of the dismissal of reading
as a determinant factor in problem solving achievement by
some of these same researchers* more recent studies in this
area have led others to hold different views.
Headin_g_and_J)r obiem_ Solving Performance
Martin (1964) studied the contribution of reading
comprehension* computation* abstract verbal reasoning* and
arithmetic concepts to arithmetic problem solving
performance. Fourth and 8th grade students were
administered the Iowa tests of Basic Skills and the
Lorge-Thorndlike intelligence test (verbal).


38
At present no set of variables has been clearly
established as a determinant of problem difficulty.
Several researchers have investigated the effect of reading
and computation on problem solving performance. Others
have studied the effect of student attitudes toward problem
solving in problem solving learning and achievement.
Typically/ correlational methods have been used to
investigate these questions.
Computational Skills and Problem.Solving Performance
One of the first researchers to study the effect of
computation and reading on problem solving performance was
Hansen (1944). He investigated the relationship of
arithmetical factors/ mental factors/ and reading factors
to achievement in problem solving. Sixth grade students
were administered tests in problem solving and categorized
as superior achievers (best problem solvers) and inferior
archievers (poorest problem solvers). The two groups were
compared in selected factors believed to be related to
success in arithmetic problem solving: arithmetical/ mental
and reading factors. After controlling for mental and
chronological age/ the superior achievers in problem
solving surpassed the inferior achievers in mental and
arithmetical factors. The superior group did better in
only two of the six items under the reading factors:
general language ability and the reading of graphs/ charts/
and tables.


26
maintained their superiority in mathematics ability during
high school. While Fox attributed sex-related differences
in mathematical achievement to differential exposure to
mathematical games and activities outside school/ Benbow
and Stanley suggested that sex-related differences in
mathematics performance stem from superior mathematical
ability in males not from differences in mathematics
formal education.
The differential coursework hypothesis is not totally
convincing and/ as reporteu berore/ it has been challenged
by researchers such as Benbow and Stanley (1980).
However/ Pallas and Alexander (1983) have questioneu the
generalizaoility of Benbow and Stanley's findings based on
the fact that they used highly precocious learners. The
differential coursework hypothesis can be accepted only as
a partial explanation of differences in mathematics
performance found between the sexes.
Differenees in Spatial Ability
The basic premise in this issue is that males and
females differ in spatial visualization and this explains
differential mathematics learning and achievement. Until
recently/ sex differences in spatial ability in favor of
males were believed to be a fact and were thought by some
to be related to sex differences in mathematical
achievement.


108
Bern* S. L. & Bern D. V. (1970). We are all nonconscious
sexists. Psychology Todays 22-24* 26/ 115-116.
Benbow* C. P.* & Stanley* J. C. (1980). Sex differences
in mathematical ability: Fact or artifact?
Science* 210 1262-1264.
Benbow* C. P.* & Stanley* J. C. (1982). Consequences in
high school and college of sex differences in mathema
tical reasoning ability: A longitudinal perspective.
American Educational Research Journal* 12*
598 622.
Bobbie* C. N. (1971). Sex-role preference and academic
achievement (Doctoral dissertation* Yeshiva Univer
sity) £iss.ejtation Abstracts International* 22*
1818B-1819B.
Carey* G. L. (1958). Sex differences in problem solving
performance as a function of attitude differences.
Journal of Abnormal and Social Psychology* 2£*
256-260.
Carpenter* T. P.* Corbitt* M. K.* Kepner* H. S.* Lindquist*
M. M. & Reys* R. E. (1980). Results of the second
NAEP mathematics assessment: Secondary school.
Mathematics Teacher* 73 329-338.
Carpenter* T. P. Lindquist* M. M. Mathews* W. & Silver*
E. A. (1984). Results of the third NAED mathematics
assessment: Secondary school. Mathematics Teacher*
16.* 652-659.
Chase* C. I. (1960). The position of certain variables in
the prediction of problem solving in arithmetic.
Journal of Educational Research* 24/ 9-15.
Cohen* D.* & Wilkie* F. (1979). Sex-related differences in
cognition among the elderly. In M. A. Witting and
A. C. Petersen (Eds.)* Sex-related differences in
cognitive functioning: Developmental Issues* New
York: Academic Press.
Connor* J. M. Serbin* L. A.* & Schakman* M. (1977). Sex
differences in children's response to training on
a visual-spatial test. Developmental Psychology*
12* 392-394.


37
dealing with both cognitive and affective variables* there
are no clear-cut findings to render unequivocal support to
a particular variable as accounting for these sex-related
differences. However* everything seems to point to the
fact that affective* rather than cognitive variables play a
more significant role in the sex-related differences
observed in mathematics performance and learning. In most
of the studies dealing with affective variables* findings
consistently show that these factors influence mathematics
performance in females more than in males. In at least one
area* confidence as learners of mathematics* Sherman (1980)
found that this variable influenced course election more
than all the cognitive variables previously discussed.
The case for the societal influences on sex roles and
expectations to account for the differences in mathematics
learning is also supported in one way or another in the
studies reported in the literature.
Problem Solving Performance _and_Rela.ted Variables
Problem solving has been perhaps the most extensively
researched area in mathematics education. Published
reviews by Kilpatrick (1969)* Riedesel (1969)* and Suydam
and Weaver (1970-1975) attest to this. Much of the
research done has focused on identifying the
determinants of problem difficulty and the problem features
that influence the solution process.


LIST OF FIGURES
Figure Page
3.1 Number of Questions by Skill Area in a
Basic Skills Test in Mathematics-6 56
4.1 Sex by Multiplication Interaction 87
4.2 Sex by Division Interaction 88
4.3 Sex by Subtraction Interaction 89
vii


8
achievement. Since there are no investigations reported in
sex differences in item response patterns in Puerto Rico
research is needed.
Organization of the Study
A review of the literature on sex differences in
mathematics performance is reported in Chapter II. The
research methodology is presented in Chapter III. Research
questions sample instrument and data analysis are
discussed in that chapter. Chapter IV is an exposition of
the results of the study. Chapter V contains a summary and
interpretation of the results of the study and the
implications of the findings together with suggestions for
further research.


23
computation grew larger. This difference grew larger for
girls than for boys.
Although SES was a major factor in solving computation
and story problem items successfully/ the effect was
similar for each sex. Sex-related differences by primary
language or chronological age were not large.
This research carried out by Marshall with elementary
grade children supports previous research findings that
males are better than females in mathematics problem
solving (a higher order skill) and females are better than
their counterpart males in computation (a lower level
skill). Marshall's research also brought out a different
aspect of this question: the notion that girls find it more
necessary than boys to succeed in the computation items in
order to successfully solve the story problem items.
£ognitive_and_Affective Factors That
Influence Sex Differences in Mathematics Learning
and Achievement
The research reviewed in the literature does not
provide evidence of any unique variable that could serve
as an explanation for the observed sex-related differences
in mathematics learning and achievement. However/ some
issues have been discussed/ among which the most prevalent
are that sex-related differences in mathematics learning
and achievement are a result of differences in formal
education; that sex-related differences in mathematics


115
Whitaker, D. R. (1976). A study of the relationship
between selected noncognitive factors and the problem
solving performance of fourth-grade children. (Tech.
Rep. No. 396). Madison: University of Wisconsin/
Research and Development Center for Cognitive
Learning.
Wiser L. K.r Steel L.r & MacDonald/ C. (1979).
Origins and career consequences of sex differences in
high school mathematics achievement (Grant No. NEI-G
7 8-001 ) Palo Alto/ CA : American Institutes for
Research.
Zalewski/ D. L. (1974). An exploratory study to compare
two performance measures: An interview-coding
scheme of mathematical problem solving and a written
test (Doctoral Dissertation/ University or Wisconsin/
Madison). Dissertation Abstracts International/
15./ 3797A.


91
TABLE 4.7
Reliability of the Covariates for Each of the
Three Years of Test Administration
Covariate
First
Year
Second
Third
Addition
.579
.610
.454
Subtraction
.756
.724
.737
Multiplication
.729
.729
.734
Division
.716
.680
.714
Fracadd
.709
.728
.748
Decsub
.720
.716
.742
Equivalence
.421
.394
.488


skills/ sex-related differences in problem solving
performance do not exist. Females retained their
superiority in problem solving when equivalence (in all
three years) and subtraction (in one year) were the
controlling variables.
The question of whether male-female differences in
problem solving depend on computational skills was
answered/ partially/ in the affirmative.
x


110
Fennemar E. & Shermanr J. (1977). Sex-related
differences in mathematics achievements spatial
visualizations and affective factors. American
Educational.Research Journals 14' 51-71.
Fennemas E.s & Shermans J. A. (1978). Sex-related
differences in mathematics achievement and related
factors. A further study. Journal for Research in
Mathematics .Education/ 9s 189-204.
Fennemas E.s & Tartrees L. A. (1983). Research on
relationship of spatial visualization and confidence
to male/female mathematics achievement in grades 6-8.
Washingtons DC: National Science Foundation. (ERIC
Document Reproduction Service No. ED 232 853)
Fienbergs S. E. (1980). The Analysis of .Cross-Classified
Categorical Data. Cambridges MA: The MIT Press.
Foxs L. H. (1975a). Mathematically precocious: Male or
female. In E. Fennema (Ed.) Mathematics learning:
What research says.about sex differences (pp. 1-12).
Columbus: Ohio State Universitys College of
Education. (ERIC Document Reproduction Service No. ED
128 195)
Foxs L. H. (1975b). Sex differences: Implications
for program planning for the academically gifted.
Paper presented at the Lewis M. Terman Memorial
Symposium on Intellectual Talent held at the Johns
Hopkins University* Baltimore* MD.
Fox* L. H. (1977). The effects of sex role socialization
on mathematics participation and achievement. In J.
Shoemaker (Ed.)* Women_and_Mathe.mati,csResearch
Perspectives for Change* (N.I.E. Papers in Educa
tion and Work: No. 8). Washington* D. C. : Eduation
and Work Group* The National Institute of Education.
Glennon* V. J.* & Callahan* L. G. (1968) Elementary school
mathematics: A_guide._tp_..cur rent research*
Washington* D. C.: Association for Supervision
and Curriculum Development.
Guay* R. B.* & McDaniel* D. (1977). The relationship
between mathematics achievement and space ability
among elementary school children. Journal for
Research in Mathematics Education* 8* 211-215.


61
A problem arises in interpreting this analysis with
the data used in this research. Over the three years?
there were responses available from 135/340 students. Even
if only 5% of the students answered an item incorrectly/
the responses of 7/767 students would be used in analyzing
this item. On the other hand/ if 90% answered an item
incorrectly/ the responses of 121/806 students would be
used in analyzing the data. As a result of the large
sample size/ the tests described above are likely to be
very powerful. In step 1 of the analysis/ then/ even a
very small change from year to year in the proportion of
males or females who choose an option is apt to be
detected/ and the results will indicate that
hypothesis 1 is not supported. For step 2/ even a very
small dependence of option choice on sex is likely to be
detected and hypothesis 2 is likely to be supported. In
brief/ the problem caused by the large sample size is that
practically insignificant differences may yield statistical
significance.
Fortunately/ the form of the test statistic used in
the likelihood ratio test suggests a reasonable solution to
this problem. The test statistic is
- 2 ^ F0 l9e (F0/Fe).
(3.4)


Ill
Hansen* C. W. (1944) Factors associated with successful
achievement in problem solving in sixth grade
arithmentic. journal q£ Educational,-Researoh' 22.1
111-119.
Harnisch* D. L. (1983). Item response patterns:
Application for educational practice. Journal of
Educational Measurement* 22* 191-205.
Hilton* T. L. & Berglund* G. W. (1974). Sex differences
in mathematics achievement: A longitudinal study.
Journal of Educational Research* £2* 231-237.
Hunter* J. E. & Cohen* S. H. (1974). Correcting for
unreliability in nonlinear models of attitude change.
Psvchometrika* 22* 445-468.
Jarvis* O. T. (1964). Boy-girl ability differences in
elementary school arithmentic. School Science and
Mtkemati_c_s/ 4, 657-659.
Kaufman* A. (1984). Sex differences in mathematics
reasoning in mathematically gifted and average fifth
and sixth graders (Doctoral dissertation* Hofstra
University* 1983) Dissertation Abstracts Inter-
JjafcionflJU 15-' 1094A-1095A.
Kilpatrick I. (1969). Problem solving and creative
behavior in mathematics. In J. W. Wilson* & L. R.
Carry (Eds.)* Studies in mathematics (Vol. No. 19).
Stanford* CA: School Mathematics Study Group.
Kloosterman* P. W. (1985). Attributional theory* learned
helplessness* and achievement in ninth grade
mathematics (Doctoral dissertation* University of
Wisconsin* 1984) Dissertation Abstracts Internatio-
nal, 4£/ 919A-920A.
Knifong* J. D.* & Holtan* B. (1976). An analysis of
children's written solutions to word problems.
Journal .for Research in Mathematics Education* 1*
106-112.
Knifong* J. D.* & Holtan* B. (1977). A search for reading
difficulties among erred word problems. Journal for
2es£.arch_, in Mathematics .Education* 2* 227-230.


50
The studies reviewed in this section show a positive
relationship between reading and problem solving
performance/ but in the case of Ballew and Cunningham
(1982)/ this relationship is not viewed singly but rather
as one among the interacting factors that produce
successful problem solving.
The third variable reviewed is the effect of student
attitudes toward problem solving on problem solving
performance. Many researchers have tried to demonstrate
that this variable is a determinant factor in problem
solving achievement.
Attitudes Toward Problem Solving and Problem Solving
Performance
Research studies support the existence of positive and
rather stable relationships between student attitudes and
achievement in mathematics. Aiken (1970) has suggested
that an individual's attitude toward one aspect of the
discipline (mathematics) / such as problem solving/ may be
entirely different from his/her attitude toward another
phase of the discipline/ such as computation.
Research/ however/ has been directed to the use of
single/ global measures of attitudes toward mathematics
rather than to the investigation of attitudes toward a
particular phase of the discipline.
The studies described below are only part of the few
investigations which have examined the relationship between


96
TABLE 4.11
Adjusted Means on Problem Solving*
by Covariate and Sex
Second Year
Adjusted Means
Covariate
Males
Females
Addition
3.69
3.96
Multiplication
3.80
3.85
Division
3.80
3.85
Fracadd
3.78
3.87
Decsub
3.90
3.75
Equivalence
3.46
44.18


Log-linear models were used in the analysis of
incorrect responses. The results of the analyses showed
that for 100 of the 111 items of the testr males and
females selected different incorrect options/ and this
pattern of responses was consistently found during the
three years of test administration. However/ for the vast
majority of the 100 items the male-female differences were
relatively small/ considering the fact that the number of
subjects needed to obtain statistical significance was very
large.
The responses of approximately 1/000 randomly selected
students per academic year were analyzed in the comparison
of male and female performance in problem solving. Females
outperformed males in problem solving and in six of the
seven computational variables. Males showed superiority in
equivalence in all the three years/ but statistical
significance was obtained in only one of the years.
Analysis of covariance (ANCOVA) was used in the
comparison of male and female performance in problem
solving after controlling for computational skills. Seven
analyses of covariance tests were conducted/ one for
each of the covariates. Estimated true scores for observed
scores were used in the analyses. The results tend to show
that for examinees with similar levels of computational
IX


11
She analyzed the responses of 9>000 boys and 9>000
girls to 6 selected items> 2 computation items> and 1 story
problem item from each of 2 of the 16 forms of the Survey
of Basic Skills test administered during the academic year
1978-79. The Survey is a 30-item achievement test
administered every year to all 6th grade children in
California through the California Assessment Program.
There are 16 forms of the test> to which approximately
9 >000 boys and 9 >000 girls respond each year. Of the 160
mathematics items contained in the 16 forms of the test> 32
are on measurement and graphing> 28 on number concepts> 28
on whole number arithmetics> 20 on fraction arithmetics> 20
on decimal arithmeti cs > 20 on geometry > and 12 on
probability and statistics.
The item analysis performed on the 1978-79 data showed
that boys and girls tended to select different incorrect
responses. In the first computation item (Form 1 of the
test) both sexes reflected similar mistakes in carrying>
but in different columns. In the second computation item>
both sexes ignored the decimal points and selected the same
incorrect response. However> more girls than boys chose
this response.
In the first computation item (Form 2 of the test) the
incorrect choice of both sexes was option £> but the
second most frequently selected option was for boys
and b for girls. In the second computation item of this


63
model tests were conducted using a .01 level of
significance.
Since this research is based on Marshall (1981/ 1983)/
it is important to compare the method of analysis used in
this study to the one used by Marshall. Marshall also used
a two-step analysis. In the first step of her analysis she
deleted the Uijk term from (3.1) and tested the adequacy
of the model/
1 g m j. j k = Ui + Uj + Ufc+Uij + Uik+Ujk. (3.6)
Following this/ she deleted the Ujk term to obtain
log mij = Ui+Uj + Uk+ij + Uik. (3.7)
and compared these two models using a likelihood ratio
test. If the first test was nonsignificant and the second
significant/ Marshall claimed that option choice was
dependent on sex and the pattern of dependency was the same
from year to year.
This is the same claim that this study sought to
establish. However/ the approach used here was to
present evidence that model (3.2) fits the data while
Marshall tried to show that model (3.6) fits the data.
The major difference between the two approaches
concerns the operationalization of the concept that the


115
Whitaker/ D. R. (1976). A study of the relationship
between selected noncognitive factors and the problem
solving performance of fourth-grade children. (Tech.
Rep. No. 396). Madison: University of Wisconsin/
Research and Development Center for Cognitive
Learning.
Wise/ L. K./ Steel/ L./ & MacDonald/ C. (1979).
Origins and career conseguences of sex differences in
bigh__s_chool mathematics achievement (Grant No. NEI-G
78-001). Palo Alto/ CA : American Institutes for
Research.
Zalewski/ D. L. (1974). An exploratory study to compare
two performance measures: An interview-coding
scheme of mathematical problem solving and a written
test (Doctoral Dissertation/ University or Wisconsin/
Madison). Dissertation Abstracts International/
15./ 3797A.


27
Research findings in this area have been inconsistent.
In 1966 / Maccoby stated that "by early school years/ boys
consistently do better (than girls) on spatial tasks and
this difference continues through the high school and
college years" (p.26). In 1972/ Maccoby and Jacklin said
that the differences in spatial ability between the sexes
"remain minimal and inconsistent until approximately the
ages of 10 or 11/ when the superiority of boys becomes
consistent in a wide range of populations and tests"
(p.41). In 1974/ after a comprehensive literature search/
Maccoby and Jacklin concluded that sex differences in
spatial visualization become more pronounced between upper
elementary years and the last year of high school/ the
years when sex-related differences in mathematics
achievement favoring boys emerge.
Guay and McDaniel (1977) supported in part Maccoby and
Jacklin's 1974 findings. They found that among elementary
school children/ males had greater high level spatial
ability than females/ but that males and females were equal
in low level spatial ability. This finding is inconsistent
with that portion of Maccoby and Jacklin's review that
suggests that sex differences become evident only during
early adolescence. Cohen and Wilkie (1979) however/ stated
that in tests measuring distinct spatial tasks/ males
perform better than females in early adolescence and
throughout their life span. Most studies carried out after


78
TABLE 4.2 Continued
Item
Num
ber
Xz for
Model
3.2
X1 for
Model
3.3
Xa for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
(b)
(c)
(d)
(e)
(f)
55
19.53
90 .87
71.34*
41,7 96
11,770
56
21.62
70.14
48.52*
44,107
18,263
57
28.51
35.78
7.27
53 ,075
146,668
58
17.49
51.69
34.20*
44,044
25,873
59
50.96
73.68
22.72*
41,829
36 ,987
60
66.43
148.39
81.96*
60,494
14,828
61
21.39
394.54
373.15*
35,632
1,918**
62
55.03
300.95
245.92*
50,699
4,142
63
158.36
278.34
119.98*
52 ,068
8,719
64
88.01
246.48
158.47*
73,196
9,279
65
56 .28
94.11
37.83*
42,429
22,532
66
50.51
113.88
63.37*
78,708
24,953
67
34.28
180 .60
146.32*
51,862
7,121
68
48.01
225.46
177.45*
63,223
7,158
69
75.29
251.66
176.37*
62,511
7,121
70
44.39
271.35
226 .96*
67,528
5,977
72
17.41
208.90
191.49*
84 ,175
8,831
73
24.61
37.80
13.19*
53 ,358
81,271
74
22.75
163.01
140 .26*
84,175
12,057


75
TABLE 4.2
Chi-square Values for the Comparison of Models (3.2) and
(3.3) with Actual Sample Sizes and with the Corresponding
Number of Subjects Needed for Significant Results
Item
Num
ber
Xa- for
Model
3.2
Xa- for
Model
3.3
Xa for
Difference
Actual
Sample
Size
Minimum
Sample
Size for
Signif.
(a)
lb)
(c)
(d)
(e)
(f)
1
16.11
30.94
14.83*
7 r 427
10,061
2
17.54
103.22
85.68*
11,018
2,583**
3
20.06
47.51
27.45*
11,497
8,733
4
7.54
21.87
14.33*
6,016
8,414
5
14.75
70.40
55.65*
7,706
2,782**
6
11.86
23.13
11.27*
13,046
23,256
7
4.60
35.73
31.13*
8,832
5,700
8
46.15
329.98
283.83*
16,796
1,189**
9
10.13
109.99
99.86*
23,075
4,642
10
9.54
65.29
55.75*
35,060
12,634
11
5.89
19.06
13.17*
24,908
37,996
12
12.58
198.46
185.88*
40 ,423
4,369
13
25.74
62.37
36.63*
15,718
8,621
14
11.19
80.50
69.31*
30,238
8,765
15
16.63
17.33
0.70
42,889
1,230,914
16
14.51
84.29
69.78*
35,114
10,109
17
17.90
44.20
26.30*
43,962
33,582


113
Meyer? R. A. (1978). Sex-related differences in mathema
tical problem solving performance and intellectual
abilities. Madison: University of Wisconsin?
Research and Development Center for Cognitive
Learning. (ERIC Document Reproduction Service
No. ED 166 023).
Milton? G. A. (1957) The effects of sex-role identifica-
cation upon problem solving skill. Journal of
Abnormal & Social,Psychology? 5? 208-212.
Moller? N. J. (1983). The impact of gender? masculinity?
and feminity on math achievement and course decisions
(Doctoral dissertion? Purdue University? 1982).
nis_sertation_Abstracts International? 41/ 2584A.
Muth? K. D. (1984). Solving arithmetic word problems:
Role of reading and computational skills. Journal of
Educational Psychology? 76 ? 205-210.
Pallas? A. M.? & Alexander? K. L. (1983). Sex differences
in quantitative SAT performance: New evidence on
the differential coursework hypothesis. American
Educational Research Journal? 21/ 165-182.
Plank? E.? & Plank? R. (1954). Emotional components in
arithmetical learning as seen through autobiographies.
Psychological Studies of the Child? 2.? 274-296.
Porter? A. C. (1968). The effects of using fallible
variables in the analysis of covariance (Doctoral
dissertation? University of Wisconsin? 1967).
Dissertation Abstract International? 2£? 3517B.
Probert? B. S. (1984). Math confidence workshops: A
multimodal group intervention strategy in mathematics
anxiety/avoidance. (Doctoral dissertation? University
of Florida? 1983). Dissertation Abstracts
International? 44 ? 2231B.
Puerto Rico Department of Education. (1980). Basic
skills test in mathematics-6. Hato Rey: Center of
Evaluation.
Radatz? H. (1979) Error analysis in mathematics educa-
tion. Journal for Research in Mathematics Educa
tion ? 10 ? 163172.
Riedesel? A. C. (1969). Problem solving: Some suggestions
from research. Arithmetic Teacher? 16 ? 54-58.


48
success since the components must be integrated into a
whole process (mastery learning of the components cannot
assure mastery of the process).
Their analysis also led them to conclude that* in the
case of 6th graders inability to read problems is a major
obstacle in solving word problems. Only 12% of the
subjects could read and set up problems correctly at a
higher level than they could compute while 60% could
compute correctly at a higher level than they could read
and set up problems; 44% could set up problems better when
they heard them read than when they read the problem
themselves. Only 13% could set up problems better when
they read them than when they heard them read.
Muth (1984) investigated the role of reading and
computational skills in the solution of word problems. A
group of 200 students from the 6th grade were administered
a test of basic skills and a mathematics word problem test.
The word problem test consisted of 15 sample items supplied
by the National Assessment of Educational Progress. The
items were adapted to include some extraneous information
and complex syntactic structure. Four versions of the test
were constructed by combining two versions of problem
information (absence vs. presence of extraneous
information) with two versions of syntactic structure
(simple vs. complex syntax). Task performance was measured
by means of the number of problems answered correctly


102
response was related to the use of a large number of
subjects in the analysis.
One must question whether analysis of incorrect
responses has any educational value. Although the results
of this study do not support the use of this analysis
(error analysis) for the study of sex-related differences#
it continues to be a promising technique in the discovery
of how children approach mathematics problem solving# more
so than analyses that look only at the items answered
correctly or at total test scores.
The second objective of the study was to investigate
sex-related differences in problem solving performance# to
find out if significant differences persisted after
controlling for computational skills# and if the
differences depended on the level of computational skills.
The results showed that females outperformed males in
problem solving and in six of the seven computational
skills. Males showed superiority in equivalence in each of
the three years# but statistical significance was obtained
only for the second year. The results also showed that
females retained their superiority in problem solving# only
when equivalence (for the three years data) and subtraction
(for the second year) were the controlling variables.
Analysis of covariance was also used to answer the
question of whether male-female differences in problem
solving depend on the level of computational skills. The


117
From May* 1982 to July* 1983 / she worked as graduate
research assistant at the Foundations of Education
Department/ University of Florida. She returned to Puerto
Rico in August/ 1983 / to serve as Special Aide to the
Assistant Secretary for the Vocational/ Technical/ and High
Skills Educational Programs of the Commonwealth Department
of Education.


46
Their study is important because it represents an
attempt to demonstrate that multiple factors can interact
in the correct solution of a mathematics word problem.
They constructed three graded tests from a basal
mathematics series for grades 3 through 8. For test 1* the
problems were set in pure computational form (the effects
or reading* interpretation* as well as the necessity for
integration were removed in an effort to measure the
computational skills required by the word problems).
For test 2* the effects of reading and computation
were removed by reading the problems to the students and by
giving scores based on whether or not the students set them
up properly* in an attempt to measure problem
interpretation alone. For test 3* the effect of
computation was removed. The test yielded two scoresone
by grading the students on whether or not they set up the
problems properly and another by grading on the basis of
the correct answer.
The tests were administered to all 244 students from
the 6th grade in two different schools. A diagnostic
profile was obtained for each of the 217 students for which
complete data were available: a computational score* a
problem-interpretation score* a reading score* and a
reading-problem solving score.
They assumed that if the reading-problem
interpretation score was lower (one or more levels lower)


71
TABLE 4.1
Summary of Significant Tests of Model 3.2
Item
Number
X* for
Model 3.2*
Actual Sample
Size
Min. Sample
Size for
Significance
(a)
(b)
(c)
(d)
8
46.15
16/796
3/352
13
25.74
15/718
5/624
23
29.25
60/377
19/011
25
35.90
26/707
6/852
28
46.00
84/847
16/988
29
61.48
88/396
13/242
30
25.19
87/473
31/982
31
23.26
32/870
13/015
34
58.25
82/257
13/006
36
84.36
86/137
9/404
37
23.70
37/068
14/405
42
23.26
59/140
23/417
45
22.41
63/098
25/932
48
179.25
97/593
5,014
49
27.82
16/766
5/550
52
26.01
39/858
14/114
54
41.41
74/502
16/570
56
21.62
44/107
18/789
57
28.51
53/075
17/146


58
Table 3.1
Hypothetical Probabilities of Option Choice
Conditional on Year and Sex
Year
Sex
1
Option
2
3
First
H
.7
.2
.1
F
.5
.3
.2
Second
H
.7
.2
.1
F
.5
.3
.2
Third
M
.7
.2
.1
F
.5
.3
.2
Table 3.2
Hypothetical
Conditional on
Probabilities of
Year and Sex and
Option Choice
Arranged by Sex
Sex Year
1
Option
2 3
M
First
.7
.2
.1
Second
.7
.2
.1
Third
.7
.2
.1
F
First
.5
.3
.2
Second
.5
.3
.2
Third
.5
.3
.2


90
The significant interactions found between sex and
multiplication^ and sex and division (first year)/ and sex
and subtraction (second year)/ answered/ in part/ the
question of whether male-female differences in problem
solving performance depend on computational ability.
However/ the evidence is quite weak. Of 21 possible
interactions/ only three were significant. No variable
exhibited a significant interaction for each of the three
years.
The analysis of covariance was also used to determine
if sex-related differences exist after controlling for
computational skills. Analyses were conducted for
those variables that did not exhibit significant
interactions with sex. As discussed in Chapter III/
estimated true scores were used for observed scores to
adjust for unreliability of the covariates (the
computational subtests). Reliability coefficients
calculated for each covariate are shown in Table 4.7.
Summaries of the analyses of covariance for the first
year are reported in Table 4.8. The results show that
females retained their superiority in problem solving
performance when equivalence was the controlling variable
in the analysis of covariance/ the only variable in which
males outperformed females (nonsignificant). When the
controlling variables were addition/ subtraction/ addition
of fractions/ and subtraction with decimals/ female


ACKNOWLEDGMENTS
I would like to acknowledge my gratitude to several
people who have influenced my formal education and/or made
this study possible. My special thanks to Dr. James
Alginar Chairman of the doctoral committee/ who contributed
to the development of my love for research and statistics.
He has been insuperable as professor and valued friend; his
help and guidance in the preparation and completion of this
study were invaluable. I extend my thanks to Dr. Linda
Crocker for her advice and help during my doctoral studies
at University of Florida. Thanks also go to Dr. Michael
Nunnery/ member of the doctoral committee. To Dr. Wilson
Guertin* who was a friend for me and my family/ I extend my
special thanks. Thanks are also extended to Dr. Amalia
Charneco/ past Undersecretary of Education of the Puerto
Rico Department of Education/ for her continuous support.
To my sister Nilda Santaella/ who typed the thesis/ I give
my sincere thanks. Special thanks go to my family and to
those friends who provided encouragement throughout this
critical period of my life.
ii


10
of the studies consider sex differences incidental to the
major study findings. The available literature offers very
little research directly related to the problem of sex
differences in this area.
The review of the literature has been divided in four
sections. The first section consists of a detailed summary
of the available research on sex diferences in incorrect
responses. The second section deals with sex-related
differences in problem solving performance. These sections
are directly related to the objectives of the study. The
third section is more peripheral and contains a discussion
of the more prevalent issues about the influence of
cognitive and affective variables on sex differences in
mathematics learning and achievement. The fourth is a
summary of the research dealing with variables considered
as influential to mathematics problem solving performance.
Sex-related Differences in Incorrect Response Patterns
Research findings tend to suggest that boys and girls
may be approaching problem solving differently (e.g.
Fennema and Sherman 1978; Marshall 1981 1983; Meyer
1978 among others). Marshall (1981) investigated whether
6th grade boys and girls approach mathematical problem
solving with different strategies. Her specific interest
was whether the sexes made the same errors.


UNIVERSITY OF FLORIDA
3 1262 08554 6660


Abstract of Dissertation Presented to the Graduate
School of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy
COMPARISON OF MALES AND FEMALES
ON MATH ITEM PERFORMANCE:
ANALYSIS OF RESPONSE PATTERNS
By
Sonia Feliciano
Augustf 1986
Chairman: James Algina
Major Department: Foundations of Education
The first objective of this study was to investigate
sex differences in the selection of incorrect responses on
a mathematics multiple-choice test* and to determine
whether these differences were consistent over three
consecutive administrations of the test. A second
objective was to compare male and female performance in
problem solving after controlling for computational skills.
The responses of all 6th grade students from the
public schools in Puerto Rico who took the Basic Skills
Test in Mathematics-6 ("Prueba de Destrezas Bsicas en
Matemticas-6") during three academic years were used in
the analyses relevant to the first objective.
viii


53
The studies reviewed have confirmed the relationship
between problem solving performance and attitudes toward
problem solving (Carey 1958? Lindgren et al. 1964?
Whitaker 1976). However the results reportea in the
studies that investigated the relationship between problem
solving performance and computation and between reading and
problem solving fail to be consistent in their conclusions.
Hansen (1944) Chase (I960) Balow (1964) Knifong and
Holtan (1976 1977) and Zalewski (1974? concluded that
computation is more strongly related to problem solving
than is reading. Martin (1964) Creswell (1982) Ballew
and Cunningham (1982) and Muth (1984) concluded that
reading aoility and mathematical problem solving show a
stronger relationship than computation and problem solving.
Exedisis's (1983) findings led to the conclusion that the
effect or reading and computation in problem solving
performance is unimportant.


IllMETHOD
54
The Sample 54
The Instrument 55
Analysis of the Data 57
Analysis of Sex by Option by Year
Cross Classifications 57
Comparison of Males and Females in
Problem Solving Performance 66
IV RESULTS 6 8
Introduction 68
Sex-related Differences in the
Selection of Incorrect Responses 68
Sex-related Differences in Problem
Solving Performance 81
Summary 93
V DISCUSSION 101
Summary and Interpretation of the
Results 101
Implications of the Findings and
Suggestions for Further Research 103
Sex-related Differences in
Incorrect Responses 104
Sex-related Differences in Problem
Solving Performance 105
REFERENCES 107
BIOGRAPHICAL SKETCH 116
IV


Sex-related Differences in Problem
Solving Performance
81
In this part/ results for the two questions related to
the second objective of the study are presented. These
questions serve as the framework for the presentation.
Each question is stated/ followed by the results pertaining
to that question.
Question 1: Do males and females differ in problem solving
performance?
The responses of 492 males and 510 females who took
the Puerto Rico Basic Skills Test-6 during the spring of
the first year were analyzed in this study. Also/ data
from 504 males and 509 females tested in the second year
and from 509 males and 504 females tested in the third year
were included in the analysis.
The mean performance scores and the standard
deviations for each of the eight variables are presented in
Table 4.3. Results of t-tests are also presented in this
table. Females outperformed males in problem solving/ a
finding consistently present in all the three years of test
administration. Over the three-year period the mean
differences favored females in all variables except
equivalence. The sex-related differences in problem
solving were significant (p

4
mathematics preparation. Males who were enrolled or had
completed algebra II outperformed the females in
computation and problem solving but not in algebra. Males
who studied beyond algebra II outscored females on all
three subtests: computation? algebra? and problem solving
(Armstrong? 1979).
Carpenter? Lindquist? Mathews? and Silver (1984)
analyzed the results of the Third National Assessment of
Educational Progress (NAEP)? and compared them with the
First and Second Surveys. Between 1978 and 1982? the
differences between the average performance of males and
females remained stable at each age level. At ages 9 and
13? the overall performance of males and females was not
significantly different. At age 17? males scored higher
than females by about 3 percentage points. When course
background was held constant? achievement differences still
existed at age 17. For each category of course background?
male achievement exceeded female achievement. Consistent
with previous assessments? sex differences in problem
solving in favor of males were found for the 17-year-old
sample. At ages 9 and 13? no large differences were found
between the sexes within any level of course background.
Marshall (1981? 1984) investigated sex differences in
mathematics performance. She found that males and females
excel each other in solving different types of problems.


105
Sexrrelated Dif fgrgnce_s_in. Problem Solving Performance
The results of this study do not give strong support
to Marshall's (1981/ 1984) findings that the effect of
computational skills on problem solving performance is
different for each sex. However/ the findings of the
present investigation are consistent with those found by
Meyer (1978) and Whitaker (1976): female performance in
problem solving is not significantly different from male
performance. Females showed superiority in problem solving
performance in all three years of test administration
although their superiority was retained only when
performance in equivalence was controlled for (for the
three years data) and when performance in subtraction was
controlled for (second year) (statistically significant
only after accounting for equivalence).
The findings of this study are also in agreement with
those reported by the National Assessment of Educational
Progress (NAEP) (Carpenter et al.z 1980). Related to
problem solving/ they conclude/
if it were necessary to single out one area that
demands urgent attention/ it would clearly be
problem solving. At all age levels/ and in vir
tually every content area/ performance was extre
mely low on exercises requiring problem solving
or application of mathematical skills. In gene
ral/ respondents demonstrated a lack of the most
basic problem-solving skills. Rather than attempt
ing to think through a problem and figure out what
needed to be done to solve the problem/ most res
pondents simply tried to apply a single arithmetic


40
Achievement Test (subtests of reading, arithmetic, and
reasoning) and the California Short-Form test of mental
ability to a group of 1,400 children from the 6th grade.
All levels of achievement were included in the analysis.
Analysis of variance and covariance were used and compared.
He confirmed the findings of other researchers to the
effect that there is a direct relationship between I.Q. and
reading ability, and between I.Q. and computational skills.
The results of the analysis of variance revealed that
increases in computation ability were associated with
higher achievement in problem solving. A relationship
between reading ability and problem solving was also found,
but it was not as strong. Significant differences in
problem solving performance associated with computational
ability were found when intelligence was controlled.
Balow concluded that computation is a much more
important factor in problem solving than reading ability,
and that when I.Q. is taken into consideration, the degree
of the relationship between reading and problem solving
ability becomes less pronounced. Intelligence tends to
confound the relationship between these two variables.
Knifong and Holtan (1976, 1977) attempted to
investigate the types of difficulties children have in
solving word problems. They administered the word problem
section of the Metropolitan Achievement Test to 35 children
from the 6th grade. Errors were classified in two


99
responses was consistent throughout the years in which the
test was administered. Males and females selected
different incorrect responses in 100 of the 111 items of
the test. The pattern of male-female differences in the
selection of incorrect responses was consistently found in
each of the years of test administration. However* for the
vast majority of the 100 items* male-female differences
were relatively small in magnitude* considering the fact
that the number of subjects needed to obtain significance
was very large. Therefore* these findings lack educational
significance.
A second objective of the study was to investigate
whether males and females differ in problem solving
performance* if these differences persist after controlling
for computational skills* and if the male-female
differences depend on the level of computational skills.
Results of the analyses reported in this section
generally indicated that females outperformed males not
only in problem solving* but in six of the seven
computational variables. Males surpassed females in
equivalence* but statistical significance was obtained for
only one of the three years covered in the study.
The results also tended to show that* for examinees
with similar levels of computational skills* sex
differences in problem solving did not exist. The only