Title: Association between student reliance upon no-penalty retaking of module tests and final examination scores
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00098117/00001
 Material Information
Title: Association between student reliance upon no-penalty retaking of module tests and final examination scores in flexibly paced engineering mechanics courses
Physical Description: xi, 122 leaves : ill. ; 28 cm.
Language: English
Creator: Doby, William Clifford, 1926-
Publication Date: 1976
Copyright Date: 1976
Subject: Students -- Rating of -- Examinations, questions, etc   ( lcsh )
Curriculum and Instruction thesis Ph. D
Dissertations, Academic -- Curriculum and Instruction -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 116-120.
Statement of Responsibility: by William Clifford Doby.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00098117
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000176311
oclc - 03067413
notis - AAU2791


This item has the following downloads:

PDF ( 4 MBs ) ( PDF )

Full Text








To Marcia and Ann ..


The writer wishes to express his gratitude to Dr. James Hensel

for his help and direction, to Dr. Martin Eisenberg for his assistance

and sharing of records, and to Dr. Robert Ramey for his guidance and

positive attitude.

Appreciation is expressed to Professor Walter Bond of the

University of North Florida for his assistance in statistical analysis.

To all the above the writer gives thanks for the generous amount

of freely given time.


ACKNOWLEDGEMENTS . . . . . . . . .

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

ABSTRACT . . . . . . . . . .. ..


Background . . . . . . . . . .

Statement of the Problem . . . . . ..

Need for the Study . . . . . . .

Assumptions. . . . . . . . . .

Definition of Terms . . . . . .

Boundaries of this Investigation . .. ....

Method . . . . . . . . . . .

Summary . . . . . . . .. . .


Introduction . . . . . . . . .

Predictors of Academic Success . . . . .

Comparisons of Various Teaching Methods . .

Additional Topics . . .. .......

Summary . . . . . . . . . ..








THE TEST POPULATION. . . . . . .... ... . 22

Introduction . . . . . . . . . . . 22

The Modular Instruction System . . . . .. ... 22

The Test Population. . . ... ... . . 28

Summary. . . . . . . . . . . . .. 30


Introduction . . .... .... ....... .. 31

Nature of Study. . . . . ... .. .. .. .. 31

List of Hypotheses . . . . . . . . ... 32

Rationale for Using "Modified Junior Year GPA" . . 33

Analysis of the Uniformity of the Modules and Units. .. 35

Data Collection and Selection. . . . . . ... .36

Characteristics of the Data. . . . . . . ... 37

Graphical Presentation of Selected Data. . . . ... 37

Summary of an Initial Chi Square Analysis. . . .. 38

Summary of a Revised Chi Square Analysis . . ... .39

Multiple Linear Regression Analysis for All Ten Sections 40

Effect of Date of Completion of the Modular Path . .. 52

Student Attitude . . . . . . . .... . 52

Summary . .... . . . .... 54

CHAPTER V SUMMARY. . . . . . . . . ... . 56

Introduction . . . . . . . . .. . . 56

The Instructional System . . . . . . . .. 56

Need for the Study . . . . . . .. . . 57

Nature of the Study. . . . . ... .. . . . 58

Review of the Literature . . . . . . . .. 59

Uniformity of Difficulty of the Module Tests . . .. 60

Data Collection and Selection. . . . . . . .. 60

Results of Chi Square Analysis . . . . .... .. . 61

Results of Linear Regression Analysis. . . . . ... 62


Introduction . . . . . . . . . . . 66

Conclusions. . . . . . . . . . . . 66

Recommendations. . . . .. . . . . . 68




STUDENT RECORD SHEET. . . . . . . . . . 88






WITHIN EACH MODULE . . . . . . . .... . 94







Student Attitude Questionnaire . . . . . ... 115

REFERENCES. ............... ... . . 116

BIOGRAPHICAL SKETCH . . . . . . . . ... . 121



PROGRAMS. . . . . . . . . .. . . 24

BY SUB-DISCIPLINE . . . . . . . . .. 26



SUMMARY . . . . . . . . . . . 45




INCLUDED. . . . . . . . . . . . 54


WITHIN EACH MODULE. . . . . . . . .. 94


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



William Clifford Doby

August, 1976

Chairman: James W. Hensel
Major Department: Curriculum and Instruction

This research investigated a modularized instructional system to

determine if there is an association between the extent of retaking of

module tests and the score on the final examination of a course. The

instructional system consisted of three junior level engineering mechan-

ics courses in statics, dynamics, and strength of materials. The

courses were taught at the University of Florida in a flexible modu-

larized form that offered several student options and individualization

by student engineering specialty. Several end-of-module proficiency

tests were provided for each module. One of these tests was selected

at random when requested by a student. A student failing a module test

could without penalty retake other tests until mastery of the module

was demonstrated. Only the passing grade was recorded. A different

option involving intracourse examinations could be followed by students

not selecting or continuing the modular proficiency test option. All

students took the same end-of-course examination. No sections of tra-

ditionally taught students were available as a control group.

The number of retakes referred to how many times a student attempted

to pass a module proficiency test after failing an initial attempt. The

number of retakes was used as an independent variable for multiple

regression analysis.

Score on the final examination was the dependent variable. Grade-

point average and number of retakes were used to predict the observed

final examination score. GPA was assumed to be the major determinant.

With its effects statistically controlled,the effect of the number of

retakes was sought as a second-order effect.

Ten sections of the modularized courses yielded usable data. GPA

was a statistically significant predictor of final examination score

in six sections. In three of these six sections, the number of retakes

was also statistically significant at the 95% level of confidence.

In two of these three cases, an increasing number of retakes was

associated with a decreasing score on the final examination. In the

third case, the opposite was unexpectedly found to be true. Students

who seemingly had difficulty mastering the modules did unexpectedly

well on the final examination

Investigation of circumstances relating to the three sections

yielded the finding that students in the first two sections frequently

took two, three, or in some cases four versions of a module completion

test on the same day. Their demonstrated mastery of the subject matter

could reasonably be attributed to happenstance rather than to restudying

and increased learning. Moreover, the subject matter was statics, which

was the easiest of the three courses. The students were beginning engi-

neering students and many of them were recent transfers from the junior

colleges. Circumstances were different for the third section, the one

in which greater retaking of module tests was associated with increased

achievement. Changes in administrative procedure hindered immediate

retaking. Students almost never took two tests of a module on the same

day. Moreover, the subject matter was mechanics of materials, which

was the most difficult of the three courses. Mastery upon initial

studying might well not be expected. The students were more mature.

Although this section was small, it yielded a highly statistically

significant result that could not be disregarded.

The researcher concluded that there is a small but discernible

association between the extent of module test retaking and the score on

the final examination of a course. The manner of retaking rather than

the amount of retaking appeared to be the condition that influenced

end-of-course achievement.




Academic course offerings at all levels have tended to change

from rigid to flexible formats in recent years. One of the innov-

ations has been the use of modularized packaging of course content.

This innovation has provided opportunities for individualization,

flexible pacing, and mastery learning. Use of these techniques at

the college level has been less extensive than at lower levels.

Three introductory engineering courses at the University of

Florida were restructured and offered in a modularized format that

permitted self-pacing and a degree of individualization. For a period

of approximately two years, an unlimited retaking of randomly selected

module tests was permitted, thus allowing students to maintain an

average of A or B up to the point of taking the final examination.

Statement of the Problem

The College of Engineering of the University of Florida in 1973

offered three basic undergraduate engineering courses in a modularized

form. The three courses were engineering mechanics courses in statics,

dynamics, and mechanics of materials and are described in Chapter III.

Minor variations in the selection of modules assigned to students pro-

vided a type of individualization of course content. These variations

provided a specialization according to a student's major within the

several departments of the College of Engineering. Student self-pacing

and unlimited retaking of module tests were the main student options

provided. This dissertation was undertaken to provide an in-depth

study of the no-penalty retaking of module tests. This dissertation

primarily seeks to answer the question, Is student use of the unlimited

no-penalty retaking of module tests associated with student achievement

an the final examination when the effects of other reference or causal

variables are statistically controlled? No causal relationship between

achievement and extent of module test retaking was hypothesized.

Need for the Study

The modularized sequence of engineering mechanics courses at the

University of Florida was implemented after considerable curriculum

development effort. The unlimited retaking of module tests without

penalty was a feature that had been included because of the belief

that it was a desirable liberalization from the previous instructional

method. In 1975, this feature was discontinued because reduced funding

curtailed personnel services necessary to maintain the testing system.

The question remained open as to the relationship between student

achievement and the no-penalty retaking of module tests. This dis-

sertation provided an in-depth investigation of results associated

with retaking of module tests. The associations that were found permit

prediction of achievement based upon known patterns. Counseling and

early remedial guidance can he facilitated.


The assumption of this investigation was that a student's academic

achievement is the best predictor of that student's expected academic

achievement in other courses under identical circumstances. Achievement

in standard courses was used to calculate the expected achievement in

other courses taken concurrently. The effects of ability as well as

environment were thus assumedly controlled by using known representative

achievement during a given time period to estimate an achievement that

would be expected in separate courses taken during the same time period.

Selections from the literature concerning this assumption are presented

in Chapter II.

Definition of Terms

The expression "modified junior year GPA" refers to a specially

calculated grade-point average earned in traditionally taught courses

taken during the same time period as were the modularized engineering

mechanics courses. The normal time for taking these engineering

mechanics courses was the junior year. Some students began the course

sequence early, during their sophomore year. Other students finished

the sequence late, during their senior year. Some students began

early and finished late. The many variations required the researcher

to select three representative quarters for each student. These

quarters were those in which the student took the three engineering

mechanics courses or a portion of them. The selected three quarters

approximate the junior year time period and are so labeled.

The term "no-penalty" in the expression "no-penalty retaking of

module tests" refers to the unlimited retaking of randomly selected

end-of-module tests until a score of 80% or better was obtained. Only

this last score was recorded for the module concerned. The scores of

previous attempts had no weight in determining the letter grade for

the course.

A retaking of an end-of-module completion test occurred after a

student failed in his first attempt to pass one of the randomly

selected tests. Each subsequent attempt to pass another of the tests

for that module was a retake. In an eight module version of a course,

a student requiring a total of 10 attempts had two retakes. When

used as an independent variable in regression equations, the number

of retakes is represented by the capitalized word "Retakes."

This investigation was an associational study. It was not an

experiment based on random assignment. A cause and effect relation-

ship was not being sought. Somewhat more general words like "result"

and "outcome" were usually used herein in lieu of "effect."

The term "independent variable" referred to the number of module

test retakes during a course. The term "dependent variable" referred

to a student's score on the final examination of the course. The

latter definition was augmented when chi square analysis was used. In

that situation, a student's final examination score as compared to his

expected score was used as the dependent variable. The standard math-

ematical terms "independent variable" and "dependent variable" were

used in this associational study although some researchers prefer to

use these terms only for experiments seeking cause and effect relation-


Boundaries of this Investigation

A multitude of factors influence or are associated with student

achievement. This investigation examined one of these, the number of

times individual students relied upon the no-penalty retaking of mod-

ule tests. This study was limited to those students who took one or

more of the three modularized engineering mechanics courses at the

University of Florida during the period from September 1973 to March

1975. This study considered only those students who pursued the mod-

ular path to completion and did not investigate dropouts or students

who chose to follow the alternate path involving scheduled intracourse



Student achievement is multi-determined. This investigation des-

ignated student achievement in the modularized instructional system

as the dependent variable. The extent of retaking of no-penalty mod-

ule tests was one independent variable. The many other influences

were grouped into a single variable termed modified junior year grade-

point average. Statistical analyses were performed to see if indi-

vidual student reliance upon the no-penalty retaking of module tests

was significantly associated with student achievement when the other

influences were controlled. The design and methodology are described

in Chapter IV.


A modularized instructional system was developed and implemented

at the University of Florida for three junior level engineering

mechanics courses. The effectiveness of the innovations is a relevant


question. An analysis was made of the way that academic achievement of

students was associated with an instructional feature allowing unlimited

no-penalty retaking of module tests.




An abundance of literature directly relates to predictors of

academic success. Because this dissertation compared observed student

performance with a calculated expected (predicted) performance, the

subject of methods of prediction of academic success is relevant.

Selections from the literature are reported herein.

A meager amount of literature relates to flexibly paced modu-

larized courses at the college level. The effects of unlimited no-

penalty retaking of module tests in college level courses is an

aspect that has seldom been reported.

A computer search of The Educational Resources Information Center

(ERIC) showed a listing of 31,238 articles on colleges, universities

and higher education of which 1180 related to testing and related

index terms such as testing methods. When the index term "modular"

or related words were added to those involving "college level" and

"testing" for an inclusive search specification, the number of articles

decreased to three. The index word "retake" and its variation when

used alone produced 10 references of which only two related to college

level testing. A character by character stringing was used to produce

the composite index term "retake module tests." No articles were

found for this term. It is noted that the Thesaurus of ERIC


Descriptors does not list the key words "modular" or "modularized" in

uses related to course organization or the key words "retake" or

"repeatable" in uses related to testing.

A computer search of dissertation abstracts showed 194 entries

having the key word "modular" or "modularized" and 4701 having the

key word "test," "testing" or variations. When variations of "mod-

ular" and "test" were used together, the number of entries decreased

to three, and the subject matter of one of these pertained to the

testing of electronic modules.

A manual search of dissertation abstracts and the ERIC system

was therefore the primary means of reviewing the literature.

Predictors of Academic Success

Prediction of academic success is one of the most extensively

explored areas of educational research. A large number of variables

have been investigated for their predictive efficacy. Sophisticated

statistical methods with use of computer processing of extensive data

have been used.

Of particular interest to this dissertation is its use of a

modified junior year grade-point average to estimate expected

achievement. Mann (21) found that the best single variable for use

in determining admission to a professional engineering program was

sophomore grade-point average. His study had examined 26 predictors.

Chapman (8) in a study of engineering students found that combi-

nations of several psychometric predictors offered no improvement

over prediction by any of his single predictors alone. Of interest

to this dissertation is that Chapman's predictors could establish

excellent group distinctions even though individual student predict-

ability was small.

For community college graduates transferring to the University

of Florida, Sitzman (31) found that the most powerful predictor of

success was the grade-point average earned prior to transfer. Post-

transfer variables such as age, marital status, and local residence

were included among the predictors tested.

In observing the validity of the 1966 precollege testing program

for students who entered Walla Walla College, Wagner (39) found that

the test score was a valid predictor of the all-college GPA, espe-

cially for female students.. The cumulative GPA at the end of the

third quarter was better predicted than the final GPA.

Schroeder and Sledge (30) found that intellective variables were

better predictors of college achievement than nonintellective vari-

ables. Ronald G. Taylor (35) supported the conclusions of many when

he found that ability factors were the best determinants of student

success in collegiate programs, especially so for academically

oriented curricula rather than for vocationally oriented programs.

Stone (33) reported that the first semester grade-point average

had a statistically significant relationship with continuance in col-

lege. Fairchild (13) found that the total grade-point average was the

better predictor of academic performance as compared to grade-point

average in the student's major.

For students within the community colleges in the state of Washington,

VanDruff (37) found that high school GPA and initial GPA at the com-

munity college were the best predictors of success in calculus. Only

a few predictor variables were studied. The multiple regression

equations derived were better predictors of the "A" grade or of the

no-credit grade than of the intermediate grades.

Another study to determine predictors of success in calculus was

conducted by Sommers (32) at Hope College. He related several pre-

course factors to the score obtained on the final examination in the

calculus course. The best precourse factor found was high school

GPA. A locally prepared test as well as the verbal and mathematics

scores from the SAT were also found to be valid predictors of success

in calculus.

For students transferring as juniors to the College of Engi-

neering at Oklahoma State University, Mouser (23) compared previous

academic aptitude variables and previous academic achievement vari-

ables to see which better predicted success in undergraduate engi-

neering courses. The previous aptitude variables were scores from

the ACT. The previous achievement variables were overall GPA and

subject GPA. The findings suggested that previous academic achieve-

ment is more closely related to subsequent GPA than aptitude as mea-

sured by the ACT. Use of both in multiple regression equations

produced the best predictive capacity.

Other studies have reached rather different conclusions about

efficacy of predictors of academic success. Elkins (12) showed that

the mathematics portion but not the verbal portion of the SAT could

discriminate between persisters and dropouts among freshman engi-

neering students at the University of Maryland. In studying achieve-

ment in the general educational requirements of lower division college

students, Cloninger (9) found that noncognitive factors play an

important role in predicting academic success. These factors

accounted for most of the variance between students with respect to

the regression equations predicting their achievement.

Comparisons of Various Teaching Methods

Another much researched area is the comparison of two teaching

methods. Attempts to reveal significant differences by statistical

analysis of observed data often do not indicate a difference of

learning effectiveness attributable to teaching method. Some

researchers finding this result conclude that student ability rather

than teaching method is what matters.

An individually paced instructional system in an engineering

college was compared by Venable (38) with group paced classes pro-

ceeding in a more traditional manner. The classes involved sophomore

engineering courses of statics and dynamics. Twenty-seven programmed

instructional units were prepared for each course. The classes

taught in the more traditional manner proceeded on a published

schedule. Students in the self-paced classes took quizzes available

40 hours per week. Examinations were given in sequence when an

individual student was ready. This research found no difference

between the instructional methods as indicated by the examination

scores. It also showed no clear relationship between the number of

attempts which a student in the self-paced classes made on unit

quizzes and his examination performance.

Eide (11) studied the effect of two different methods of

teaching engineering graphics. His experimental study compared the

learning-achievement of freshman students in a conventional "lecture

and problem" class with achievement of similar students using a

series of 25 learning packages with audio-visual tapes. Students in

the latter group could control the amount of time for individual study

used to complete a unit. No significant difference in the amount of

learning was found between the two instructional methods. High school

scholastic achievement was a good indicator of achievement for students

using either method.

In another study comparing two methods of teaching engineering

graphics, Walker (40) compared the effectiveness of an idea-communi-

cation method with that of the traditional method. No significant

differences in any of three dependent variables (results) were found.

Walker concluded only that students with higher levels of critical

thinking ability attained a greater degree of general drafting knowl-

edge than did students with lower levels of critical thinking ability.

Harris (17) compared student performance in a college engineering

science course in which students were randomly assigned to classes

using one of two different teaching methods. One method used an

audio-tutorial approach. The other used a printed transcript of the

same material. There was no significant difference between the

learning achievement of the two types of classes.

A similar finding was reported by Otten (26) who compared three

instructional models for teaching a sophomore electrical engineering

AC circuit course. One method used the traditional lecture approach

without written objectives or use of computer. A second method utilized

measurable behavioral objectives. A third method used the computer as

a computational tool to illustrate the material and motivate the stu-

dents. All three methods were found to be effective. No statistically

significant differences in achievement appeared among the groups

instructed by the three different strategies.

A different conclusion was reached by Aird (1) in comparing tradi-

tionally taught engineering students with students using self-study

involving computer based instruction. Mechanics of solids was the

engineering courses taught by the two methods. The findings seemed to

justify the conclusion that the computer based instructional mode pro-

duced students who performed better than traditionally taught students.

Related to these comparisons of teaching methods is the finding by

Tovey (36) that postadmission factors have little influence on perfor-

mance of incoming high school graduates and that the best predictor of

college success is rank in high school rather than postadmission events.

Simarily, a study of the use of repeatable testing for college

chemistry students by Donovan (10) revealed no difference in final

achievement as compared with the control group students who took non-

repeatable tests.

Compulsory attendance in an audio-tutorial college biology course

was compared by Nord (25) with noncompulsory attendance. He found no

significant difference in achievement between students participating

in the two methods.

Compulsory homework assignments for college mathematics students

was compared to no homework assignments in a study by Hasen (18). He

found no significant difference between group achievement means.

Comparisons of teaching methods involving self-pacing and mastery

learning are included in the following section.

Additional Topics

Self-pacing as well as repeatable testing is a part of the modu-

larized course offerings studied in this dissertation. Bass (4) in

his study of engineering graphics taught with self-pacing as compared

to the traditional method could reject no null hypotheses and concluded

that the new method was similar in effectiveness to the traditional


Gallegos (14) investigated pacing and found that forced pacing at

a rate greater than students would choose for themselves was less effec-

tive than self-pacing or slow prescribed pacing. He found that self-

pacing was particularly beneficial for high ability students. On the

other hand, slow but prescribed pacing was better for low ability stu-

dents than self-pacing alone.

Lasco (20) reported that external pacing (self-pacing) caused an

increase of student time devoted to study in a learning laboratory but

did not affect end-of-course group achievement. His experiment used

an individualized instructional system consisting of nine units in

audio-tutorial format for the teaching of college level geology. As a

student's characteristic work rate became slower, total unit achieve-

ment tended to increase but end-of-course achievement did not. As a

student's characteristic work rate became slower and in addition his

general ability increased, total unit achievement as well as end-of-

course achievement tended to increase. The above findings suggest

that those engineering students of this dissertation who finished the

modules early and hence seemingly without difficulty and those stu-

dents who progressed mainly by persistence will not necessarily be

high achievers on the final examination.

Self-pacing has the inherent disadvantage of permitting procrasti-

nation. Some personality types would be expected to be more prone to


difficulties caused by procrastination than other students. Gehlausen

(15) analyzed the personality types of beginning engineering students

at Tri-State College. Despite many similarities between the successful

students and unsuccessful students regarding their backgrounds and

interests, the former as compared to the latter were found to have a

history of academic success, higher aptitudes, higher expectancies,

and better study habits. The high achievers in engineering seem to do

less procrastinating, waste less time, have fewer distractions, and in

general have better study orientation.

A self-paced instructional system also using an open learning

laboratory and repeatable mastery examinations like the instructional

system of this dissertation was investigated by Naegele (24). The

subject matter was introductory college physics. His findings support

Ausbelian learning theory that the most important factor influencing

learning is the learner's possession of those concepts and skills

which have a clear and direct relationship to the subject matter under


Mastery learning is important in the instructional system inves-

tigated by this dissertation. It is a strategy that permits variations

of the kind, quality, and duration of instruction so as to fit an indi-

vidual's need as measured by frequent evaluations with immediate feed-


Mastery learning was compared with the traditional method of

teaching freshman mathematics in a study by Price (28). He found no

significant difference in his experimental study. Student profile did

not matter either. It appeared, however, that as students had time to

adjust to the mastery learning procedure, their performance on achieve-

ment tests improved. Price felt that the middle ability student in

particular reflected this trend.

Caponigri (7) reached a different conclusion in an investigation

of mastery learning methods for teaching college statistics. He com-

pared two methods of mastery learning with the traditional lecture and

demonstration method. Both methods of mastery learning showed a sig-

nificant improvement over the traditional method. Nonetheless, even

though mastery learning had been obtained, the end-of-course exami-

nation scores correlated directly with a precourse mathematics aptitude

test. This finding conflicts with the hypothesis that in a mastery

learning system the relationship between aptitude and achievement

should approach zero.

Rowberry (29) studied an adjunctive auto-instruction method for

dental students and found that every student failing a topic of his

course but willing to spend sufficient time could achieve mastery by

means of the adjunct method. Although failers became achievers, some

of the initial achievers failed to retain mastery to the time of a

review test. Apparently some of the initial achievers did not review

because pressure no longer existed. This indicates that retention is

a problem. Of interest to this dissertation is the question suggested

by the above that some seemingly best achieving engineering students

who finish the modules early with few retakes of module tests may not

necessarily perform best on the course final-examination.

This dissertation combined junior college transfer students with

native university students in the same test population. Wermers (41)

compared the achievement of both types within the upper division stu-

dents at the Urbana-Champaign campus of the University of Illinois.


He found no difference in general between these two types of students.

Similarly, Philip W. Taylor (34) concluded that transfer students

at East Carolina University experienced relatively the same diffi-

culties during their junior and senior years as native students.

The student population used for this dissertation was composed

almost exclusively of engineering students. These students controlled

their own efforts and pace. Achievement was a personal matter and

would obviously be influenced by the personality profile of each indi-

vidual. Brown (6) has studied the personality characteristics of engi-

neering students who succeed. He reports that the successful engi-

neering student sets high goals for himself and is motivated to attain

them. He tends to be orderly and self-sufficient. He relies on per-

sonal resources rather than looking to others. The engineering stu-

dent is aggressive and satisfies this drive through personal exploit

rather than by engaging in activities which involve group social or

political action.

Personality factors were considered by Kirkpatrick (19) for begin-

ning students of electrical and electronics technology students. He

investigated how personality traits and personality types (introversion

versus extroversion) are related to academic achievement in two dif-

ferent methods of instruction. One method was an individually paced

type; the other involved lectures, discussion, and demonstrations for

groups of students proceeding at a group pace. A general finding not

statistically significant was that the individually paced method was

best suited for students with introversion tendencies. Students with

extroversion tendencies generally achieved better in the group method.

When personality traits such as Order, Abasement, Change and Endurance

were included with personality types, certain grouping were found to

be statistically significant. This research like others shows that

individual student achievement is multidetermined.

Braun (5) in his study of engineering and engineering technology

students supported the contention that the self-concept of individuals

is related to their personal behavior and that measurement of this

self-concept should be useful as an aid to curriculum choice.

Peterson (27) considered the hypothesis that factors other than

those of an intellectual nature contribute to persistence in an under-

graduate engineering program. His study concluded that nonpersisters

in engineering tend toward greater independence and nonconformity than

do persisters. This finding suggests that the self-pacing feature of

the instructional system herein would decrease attrition.

A related finding by Augustine (3) was that both persisters and

nonpersisters among academically proficient engineering students are

frequently dissatisfied with highly structured inflexible engineering


Student attitude is a matter of interest in applying the findings

of this dissertation. A few additional comments about attitude follow.

A secondary finding of the previously reported research by Eide

(11) was that freshman engineering students preferred the modularized

version of an engineering graphics course and that those students

experiencing it had a lower attrition that did similar students taking

the traditionally taught version.

A contrary finding appeared in the previously reported research

by Venable (38). In his study of certain engineering students, fewer

students in a self-paced instructional system completed the courses


successfully as compared with students in regularly scheduled progression

through the identical subject matter.

The Arizona State Department of Education (2) sponsored research

that reported results of using modularized versions of five vocationally

oriented courses at Eastern Arizona College at Thatcher. Included

were engineering related courses in drafting and in electronics. One

conclusion was that students liked the ability to finish a self-paced

modularized course early but disliked the lack of instructor pressure

in setting deadlines.

Student attitude is a matter of interest in applying the finding

of this dissertation. In the use of repeatable tests in college chem-

istry, the previously reported research by Donovan (10) concluded that

students feel they learned in the process of repeating a test, that

pressure was relieved, and that cheating was reduced. Students did

not mind the extra work if an improved grade seemed to be almost a


Student attitude concerning attendance at lectures was reported

by Nord (25), whose research has been previously mentioned. He stated

that students who had attended a noncompulsory attendance version of a

college course recommended use of noncompulsory attendance far more

than did students who had taken the compulsory attendance version.

Harris (17) in his previously reported comparison of two methods

of media supplements to engineering courses found that students pre-

ferred to have both options available. One method involved an audio-

tutorial approach and the other used printed transcripts.

The previously discussed research by Otten (26) measured attitude

among students proceeding in three versions of an engineering course.


No statistically significant difference in attitude attributed to the

method of instruction were shown to occur among the three groups of


Various researchers have selected a multitude of variables that

offer promise of predicting academic achievement. Harding (16) used

multiple regression analysis to examine 38 endogenous (in school) and

exogenous (out of school) variables to determine their ability to pre-

dict academic achievement of students at Illinois State University.

High school rank and ACT scores were the preadmission variables that

had predictive value. For students already in college, those who had

greater than average amounts of class cutting or participation in ath-

letics and recreation or in television watching tended to have lower

grades. Students who spent greater than the average amount of time

with members of the administration tended to have lower grades, but

students who had more than the usual amount of out of class contact

with faculty members tended to have higher grades. This research

illustrates that student achievement is multidetermined.

Regression equations are used in the analysis of this disser-

tation. In the use of regression equations to predict academic suc-

cess of Tennessee community college transfer students, McCook (22)

found that a separate regression equation was necessary for the grad-

uates of each community college. Similarly, Mouser (23) found pre-

dictive capacity was improved when engineering transfer students from

two year and four year colleges were considered separately when using

regression procedure.


The literature reports many efforts to predict academic success

of individual students. Tests, accumulated grade-point averages, and

noncognitive factors have been investigated. Various degrees of suc-

cess have been obtained in various situations. A previously accumulated

grade-point average, usually an overall GPA, has often been found to be

a valid predictor and the best predictor of future academic success.

Referring to upper division matriculation the literature suggests

that the differences between transfer and native students are small.

For college level courses little has been published about the effect of

no-penalty repeatable testing in self-paced modularized courses. Inves-

tigations of student attitude suggest that students prefer to have

alternative methods of learning available.




This chapter outlines the instructional system which served as a

base for this study. The courses, their flexible packaging, the insti-

tutional setting, and the student body are described. This chapter

outlines the general procedures used for selecting appropriate stu-

dents to form a suitable base for statistical analysis.

The Modular Instruction System

The College of Engineering of the University of Florida has devel-

oped a modularized curriculum for three junior level introductory engi-

neering mechanics courses taken by most undergraduate engineering stu-

dents. These engineering core courses are Statics (ESM 301), Dynamics

(ESM 302), and Mechanics of Materials (ESM 303). They are referred to

herein as the ESM courses.

The integrated modular instructional system was developed during

the 1972-73 academic year by Professor Martin A. Eisenberg, Ph.D., of

the Department of Engineering Sciences. Beginning in the 1973-74

academic year, the modularized system was offered to students in lieu

of the traditionally taught versions of the three courses. The modu-

larized versions were the primary instructional mode but scheduled

examinations were available as an alternative to the end-of-module

tests. During its inception and initial use, the module system provided

for self-pacing and unlimited no-penalty retaking of module tests. As

the system evolved, various incentives were used to control the self-

pacing feature in order to discourage procrastination. The no-penalty

retake feature remained essentially constant until the spring quarter

of the 1974-75 academic year. During the proceeding six quarters, the

instructional system for all three courses was in full operation with

unlimited use of the no-penalty retaking of a randomly selected version

of each module test. The data accumulated during these six quarters

were used for this research. The basic design philosophy and outline

of the system prior to April, 1975 is herein called "the instructional


A detailed description of the instructional system was published

in the article "A Modular Instructional System For Introductory Courses

In Engineering Mechanics" by Martin A. Eisenberg, Ph.D., designer of

the system. This article was published in the December 1975 edition

of Engineering Education and is reproduced in part in Appendix A. A

brief description of the instructional system is presented in the fol-

lowing paragraphs.

The instructional system during the period of data collection

employed a flexible combination of modular curriculum packaging, unlim-

ited retaking of module tests, variable pacing, programmed learning

materials, and computer management of records.

There were 13 degree programs offered by the several departments

of the College of Engineering. The instructional system content was

designed to accommodate varying department objectives. Thus an elec-



Degree version Module number
program 301 302 303 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

Aero.Eng. 10 22 33 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3
Agric.Eng. 10 21 32 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3
Chem.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3
Civil.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3
C.I.S. 12 24 1 1 1 1 2 1 2 1 1 1 1 2 2 2 2 2 2 1 1 1 2 2 2 2 2
Elec.Eng. 11 20 31 1 1 1 1 1 1 3 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3
Eng.Sci. 10 22 33 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3
Env.Eng. 12 24 1 1 1 1 2 1 2 1 1 1 1 2 2 2 2 2 2 1 1 1 2 2 2 2 2
Ind.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3
Mech.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3
Matls.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3
Nuclr.Eng. 10 34 1 1 1 1 1 1 1 1 1 3 3 3 3 3 3 3 3 3 3
Syst.Eng. 10 20 30 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3

1 indicates module is part of ESM 301
2 indicates module is part os ESM 302
3 indicates module is part of ESM 303

Versions 10, 21, 22, 23, 31, 32, 33 and 34 are four credit hour courses
Versions 11, 20 and 30 are three quarter hour courses
Versions 12 and 24 are five quarter hour courses


trical engineering student and a civil engineering student did not take

the identical modules of ESM 301, Statics, nor did they receive the same

number of credit hours.

Table 1 shows the content of courses required for each of the under-

graduate degree programs. As far as the University registrar was con-

cerned, there were only three variable credit courses in which the stu-

dent could enroll. Within each of the courses, however, students could

be enrolled in one of three to five different subcourses whose existence

was of no concern to the registrar. A study of the content of subcourses

30, 31, 32, 33, and 34 of ESM 303 shows similarity. Subcourse 33 taken

by aerospace engineering students differed from subcourse 30 only in

the addition of shear center and column buckling modules to the curric-

ulum. Subcourse 34 taken by nuclear engineering students included

stress and deformations in thick-walled cylinders under thermal and

pressure loading, subcourse 31 taken by electrical engineering students

included damped vibrations of particles and rigid bodies. The differ-

ences among most of these subcourses was small.

Individualization by student major caused a variation in the choice

and number of modules to be included in the three basic course packages.

Table 1 previously shown provides the details. A summary of the indi-

vidualization by major is provided in Table 2. Versions 10, 20, and 30

were applicable to a greater number of students than were other versions.

Five professors taught the courses. A greater number of student

assistants staffed the learning laboratory and administered the tests.

The professors kept the records of individual student scores made on

the final examination of each course. These scores were the dependent

variable of this investigation. The student assistants recorded the



Course Version Number of Modules

301 10 9
11 8
12 12

302 20 9
21 11
22 11
24 11

303 30 9
31 9
32 11
33 11
34 10

results of individual module tests. Scores of 80% or above were

recorded numerically. Scores of below 80% were recorded as F for fail-

ure. The module test number, the result and the date were the three

kinds of information recorded. The student record sheet used is shown

in Appendix B. The recording was in handwriting. Scores of units

passed were transferred to a computer. Record of units attempted but

failed appeared at no other place than on the student record sheet,

which was stored with the module test and work papers in individual stu-

dent folders filed in the learning laboratory.

In addition to a standard textbook for the courses, students used


a set of programmed study guides. The study guide provided a learning

activity package. For each module there was a description of the con-

tent and rationale for study of the prescribed material, a statement

of prerequisites, a list of behavioral objectives, a commentary and

guide to the text, and a sample proficiency test. An example of a

sample proficiency test is shown in Appendix C.

The end-of-module test taken by the student was similar in level

of difficulty to the sample test. During the time period of interest,

there were five to ten separate end-of-module tests for each of the

34 modules. When ready to take an end-of-module test, a student went

to the learning laboratory and was given a test selected randomly from

the several for that module. If the student made below 80%, he failed

and could then take another of the tests when he so desired. Upon

obtaining a grade of 80% of better, the student proceeded to the next

module. Only the final grade was counted in determining the letter

grade for the course. The algorithm for computer calculation of this

letter grade was changed from time to time. The letter grade is not

relevant to this research. All students simultaneously took an end-of-

course final examination.

The self-paced feature of the instructional system was designed

to cater to the broad range of input student competencies. As origi-

nally implemented, the system employed a significant element of flex-

ible pacing strategy. Major traditional examinations were scheduled

during the fourth and eighth weeks of a ten-week quarter. Students

had the option, however, of demonstrating proficiency by passing the

related module tests prior to the dates of these two scheduled exam-

inations. Students demonstrating A or B proficiency as indicated by

grades of 80% on the module tests were excused. This feature was to

discourage procrastination. The details of this feature varied some-

what during the evolution period of the instructional system.

The learning laboratory staffed by student assistants provided

tutoring help in addition to administering the module tests.

The Test Population

The undergraduate engineering students who comprised the total

population had different types of educational institution backgrounds

prior to beginning the introductory courses of engineering mechanics.

The two main sources of students were the junior colleges and the

University College of the University of Florida. A few students came

from other universities. Thus the three general types of students

within the test population were native students and two types of

transfer students.

University College students begin as freshmen at the University

of Florida. At approximately the end of the sophomore years, students

transfer to the various upper division colleges such as the College of

Engineering. Some of these native students begin the junior level

engineering mechanics sequence of courses during their sophomore year.

Junior college graduates who transfer to the University as engi-

neering students begin as juniors. Such students have completed their

general educational requirements and have been awarded the Associate

of Arts degree. Most of these students come from the 28 Florida pub-

lic junior and community colleges. Some native and transfer students

do not initially register for the engineering mechanics course sequence

and therefore begin the sequence late.

Students from other universities who transfer to the University

of Florida as engineering students are sometimes well beyond comple-

tion of the sophomore year and occasionally have some type of bacca-

laurate degree.

The test population consisted of those students selected from

the total population. The total population consisted of all students

who took any of the three modularized engineering mechanics courses

from September 1973 through March 1976, in which unlimited no-penalty

retaking of module tests was permitted. Various types of native and

transfer students participated as described above. The following

paragraphs describe the selection of students who were included in the

data base of this investigation. Uniformity of situation and comple-

teness of data were the two general considerations used to select those

students who were included in the test population.

Students who took few other courses concurrent with the modu-

larized engineering mechanics courses accumulated too small a number

of credit hours to provide a reliable measure of their level of

achievement. The researcher set 20 usable hours as a minimum for pro-

viding a reliable average, defined as the modified junior year grad-

point average. Very few students were eliminated from the test popu-

lation by this criterion.

Students participated in part or all of the three course engi-

neering mechanics sequence during other times than their junior year.

Students who began in the last quarter of their sophomore year or who

completed by the end of the first quarter of their senior year were

considered by the researcher to have been in the standard time

sequence. Their data were combined with that of students who were

juniors at the time they took the courses. A considerable amount of

data would otherwise have been deleted. After review of the tran-

script of each student, the researcher deleted from the total popu-

lation those students whose matriculation pattern was further out of

time sequence or was otherwise nonstandard.

Further details about choices made in selecting students for

inclusion in the test population are given in Chapter IV.


Three engineering mechanics courses comprised the instructional

system. The system provided a modularized matriculation path indi-

vidualized to each student's engineering subdiscipline. It also pro-

vided a nonmodularized path utilizing periodic examinations. The modu-

larized versions of the three courses consisted of nine modules for

most students. Unlimited no-penalty retaking of module proficiency

tests was permitted. All students took the same end-of-course final

examination. Several factors influenced selecting a relatively homo-

geneous test population to provide a data base for analysis.




This chapter explains the investigative nature of the study, states

the null hypotheses, discusses the data collection decisions, and pre-

sents graphical representations of selected data for visual obser-

vation. An initial chi square analysis and a revised analysis are then

summarized and reasons are stated for discontinuing use of the chi

square method of analysis. An alternative analysis using linear

regression is then described and applied to 10 ESM 301, 302, and 303

sections, and the findings are stated. The chapter then presents

adjunct findings related to self-pacing and to student attitude. A

summary concludes the chapter.

Nature of Study

This was an associational study which sought to determine if a

measurable association existed between the number of module test retakes

and the final examination score. This was an ex post facto scientific

inquiry examining educational variables in a real life setting. Sys-

tematic controls of actions taken were lacking since no control group

was available for comparison. This study therefore sought to determine

an association and did not imply that a cause and effect relationship

necessarily existed. The "effect" or associational strength of a weak


variable was sought in the presence of many other variables that affect

student achievement.

The analysis proceeded heuristically, seeking to identify extra-

neous variables and control their influence by the choices of data

selection and the designs of statistical analysis.

This was an exploratory field study seeking to detect a possible

association between the variables that would provide usable information

to the faculty administering the innovative instructional system and

that would lay the groundwork for those who may conduct a systematic

test of the primary hypothesis in an experimental setting.

List of Hypotheses

Primary Null Hypothesis

The primary focus of this research related to the retaking of

module completion tests as to whether the amount of retaking was

associated with student achievement.

The primary null hypothesis was

Hg: The number of times that students utilized the

provision for unlimited no-penalty retaking of

module proficiency tests during the ESM 301, 302,

or 303 course was not related to student achievement

.as indicated by score on the end-of-course final

examination when other reference or causal vari-

ables were controlled.

Stated simply, this null hypothesis was

HO: The extent of retaking of module completion

tests was not related to student achievement.

Adjunct Null Hypotheses

Self-pacing and student attitude were two adjunct topics con-

sidered by this research. Analyses of these topics are presented

briefly at the end of this chapter. The adjunct null hypotheses were

H0: The completion date of student following

the modularized version of ESM 301, 302 or 303

was not related to student achievement.

H0: The attitude of students concerning the

modularization of ESM 301, 302, and 303 was

not related to student achievement.

Rationale for Using "Modified Junior Year GPA"

The dependent variable or criterion was student achievement,

which was multi-determined. The effects of ability and environment had

to be statistically controlled. High school GPA, high school rank,

freshman GPA, cumulative GPA, SAT scores, and the Florida battery of

12th Grade test scores were examples of ability variables that could

have been used to predict a student's performance. Class load, resi-

dence, car ownership, recreational and athletic participation, student

associations, family status, and part-time employment were examples of

environmental variables. Personality and work habit variables over-

lapped both of the above categories. A control was needed for the

effect of these various ability and environmental variables. A spe-

cially calculated grade-point average was selected as a means of con-

trolling for these variables.

Any grade-point average is to some extent a possible predictor of

a given student's expected performance in some other academic situation.

Chapter II discusses this point. In general, research suggested that

the more current the GPA selected, the better it would serve as a pre-

dictor. The plan of this research was, therefore, to use a specially

calculated concurrent GPA. The researcher proposed that a concurrent

GPA was a single measure reflecting the combined effort of the many

ability and environmental determinants summarized early in this section.

These variables affected a student's achievement in the modularized

courses. These same variables affected that student's achievement in

traditionally taught courses taken concurrently. These latter were

taken as a reference that reflected the effects of the many external


The researcher therefore established the term "modified junior

year GPA." Three quarters were included in order to correspond to the

three quarters in which a student took the three ESM junior level

courses. Many individual students were retained in the data pool even

though they began the series before their junior year, or completed

the series after their junior year, or did not take all three courses.

The researcher therefore had to select three representative quarters

for each student. The selected quarters were those in which the three

ESM courses were taken. For students not taking all three ESM courses,

junior year quarters were selected to provide enough data for a repre-

sentative average.

In order not to compare something to itself, the ESM courses were

deleted from the record of the three selected quarters. The modularized

course thermodynamics, ME 360 was also deleted. The remaining courses

were therefore independent of the modularized courses being investigated

but concurrent with them. These selected courses were taken in approx-


imately each student's junior year. The GPA of these courses was cal-

culated for each student and the term "modified junior year grade-point

average" was appropriate.

Analysis of the Uniformity of the Modules and Units

The several versions of the three courses of the modularized engi-

neering mechanics instructional system were described in Chapter III.

Each module had several end-of-module tests.

After several boxes of student learning laboratory folders were

obtained, an analysis was made of the relative difficulty of the 34

modules. A tabulation was made of how many students failed a given

module the first time those students attempted one of its tests. The

results are shown in Appendix D. The number of tabulated initial mod-

ule attempts varied because of course and version differences, drop-

outs, legibility of student assistant handwriting, and completeness

of records.

The variation shown within the percentage column showed clearly

that the difficulty of the modules was nonuniform. This variation in

itself was not serious for purposes of this research because all com-

parisons were made for students taking the same sequence of modules.

The total number of retakes was important, but where the retakes oc-

curred was not.

Next an analysis was made of the relative difficulty of the end-

of-module tests within each module. The number of different tests

available for random selection varied from module to module. A tab-

ulation was made of the number of times students failed each of these

tests on the first attempt. The detailed results are shown in Appendix

E. The large variation of difficulty of the end-of-module tests with-

in several of the modules constituted a problem for this research.

Students followed randomly occurring unequal paths as they progressed

through the sequence of modules. This condition was recognized and

accepted as a cause of dispersion that could not be statistically con-


Data Collection and Selection

Despite the researcher's intent to include all students of many

classes in the data base, the number of usable student histories

decreased considerably with the several types of preliminary tabulating

of the several components of the data. Many students were lost from

the test population because they did not follow or did not complete

the modular option. Students were also lost because they took the

courses unusually early or unusually late rather than in their junior

year, because they took an unusual version of a course, or because one

of the components of the data was missing. A discussion of the selec-

tion process is given in Appendix F.

The resulting data pool consisted of selected students from 10

usable sections of ESM 301, 302, and 303. There were numbered 1

through 10 and are identified as shown in Table 3. A further pro-

cessing of the data was made after the initial chi square analysis

described later in this chapter. Descriptive statistics concerning

the data are presented later in this chapter.




Section Designation Date Section Number
1 ESM 301 Fall '73 3743-V
2 ESM 301 Fall '73 3744-V
3 ESM 301 Winter '74 3548-V
4 ESM 301 Spring '74 3411-V
5 ESI 301 Fall '74 3762-V
6 ESM 301 Fall '74 3765-V
7 ESM 302 Summer '74 2136-V
8 ESM 302 Fall '74 3767-V
9 ESM 303 Spring '74 3414-V
10 ESM 303 Fall '74 3768-V

Characteristics of the Data

The data distribution was not bell-shaped. The final examination

score distribution had an upper limit at 100 and was skewed. The mod-

ified junior year GPA distribution has an upper limit at 4.00 and was

skewed. These deviations from the normal distribution were inherent

in the test instruments.

The number of retakes of module tests was a noncontinuous distri-

bution. This was necessarily so because this was a count function

rather than a mearurement.

Graphical Presentation of Selected Data

Three types of raw data were collected as described in the pre-

vious section. These were a GPA, the final examination score, and the


number of retakes of module tests. For purposes of chi square analysis,

the first two were combined into a measure of relative achievement.

Prior to that processing, this section presents selected graphs showing

the general nature of the selected data.

Graphs are presented in Appendix G for the data of Sections 1, 2,

3, and 9. Sections 1, 2, and 9 are the three sections having statis-

tical significance in the multiple regression analysis described later

in this chapter. Section 3 is typical of most of the other sections,

which have high dispersion of data and for which no statistically

significant conclusions could be found. These data were graphed to

permit visual inspection of whatever patterns might be discernible in

the selected data. These graphs did not control for the effects of

aptitude and environment upon each student's final examination score.

This is to say that none of the graphs showed relative achievement and

hence none directly related to the primary null hypothesis of this

investigation. The graphs show that considerable dispersion existed

within the data but did not show any clustering worthy of special


Summary of an Initial Chi Square Analysis

Three sections of ESM 301 were randomly selected for an initial

large scale evaluation of the method previously described as the

initial plan for chi square analysis. Ninety-nine student histories

were used to permit their being divided into three equal categories

for the contingency table. The chi square statistic indicated no

statistical significance. This lack of significance indicated that

the number of retakes was indeed a subtle influence. If it existed, a

more precise method of analysis would be needed to reveal it.

Summary of a Revised Chi Square Analysis

Procedural changes were made before proceeding with a full scale

data analysis using the chi square method. In the initial analysis

the dividing line between students typically fell with a group of stu-

dents all having the same number of retakes. Random numbers had been

used to determine how that group of students would be divided. The

use of equal size groups in the contingency table was discontinued to

increase the sensitivity of the method of analysis.

A chi square contingency table must have an expected values of at

least five in every cell. Otherwise the chi square statistic is not

valid because of the threat of instability. The problem of having

each individual cell expected frequency be equal to or greater than

five had originally appeared to be small because of apparent wealth of

large ESM sections available for analysis. The many deletions of stu-

dents to provide homogeneity and reduce dispersions substantially

reduced the data pool as was described previously.

The original data tabulation used only Version 10 of ESM 301, a

nine module version. Version 11 was identical except that it did not

include Module 7. Electrical engineering students took Version 11.

They were a substantial group. The researcher retabulated all retake

data, disregarding Module 7 information on the student record sheets.

This adjustment made Version 10 become an eight module package iden-

tical with Version 11. The total number of usable students per ESM

301 section was therefore increased.

Several chi square analyses were then made for different combi-

nations of the revised data for the six ESM 301 sections. The nine

cell contingency table was compressed into a six cell table by grouping

the number of retakes into the two categories of few retakes and many

retakes. This was to insure that the expected frequency of each cell

would be at least five. Some of these data groupings produced an appar-

ent statistical significance. Observation of the data points within the

contingency table led the researcher to conclude that the data pro-

cessing method probably introduced a bias that contributed to the

apparent significance. The method of controlling for aptitude and

environment was to subtract a student's expected score from his actual

score after standardizing both types of data. The expected score was

determined from the modified junior year grade-point average. This

process appeared to overcompensate and introduce a bias. For example

the researcher found that a high GPA student was almost precluded from

being categorized as achieving better than expected.

Because of threats to validity that appeared to have been intro-

duced in preparing the raw data for chi square analysis and because

of the lack of sensitivity to minor distractions between observed data

points, the researcher abandoned further use of contingency tables and

chi square analysis.

Linear regression analysis was the method chosen for a re-analysis

of the data.

Multiple Linear Regression Analysis for All Ten Sections

The variables associated in this investigation were (1) GPA,

(2) number of retakes, and (3) score on the final examination. A high


GPA student would reasonably be expected to do well on the final exam-

ination. A student requiring few module retake tests would seemingly

be having little difficulty and on this basis would be expected to do

well on the final examination. Stated conversely, a student requiring

many retakes would appear to be having difficulty in mastering the

subject matter and would reasonably be expected to have a resulting

low score on the final examination. These two expectations were com-

bined into a single mathematical equation as follows:

Score = B0 + B1 x GPA B2 x Retakes

This typical linear probabilistic model was appropriate as a deter-

ministic model for representing the best fit line through the scatter

diagram of the observed data. The method of least squares was used to

find this line of best fit to the empirical data. B1 and B2 were the

regression coefficients.

The above model was applicable only to students who chose to follow

the modular option to completion. Students who discontinued the modu-

lar path could have taken the intracourse examinations or they could

have become dropouts. Students not completing the modular option were

a sizable group. Research concerning them was conducted by Dr. Eisenberg

and was described in his article which appears herein as Appendix A.

The data for each of the 10 ESM sections was transmitted to the

Northeast Regional Data Center located at the University of Florida at

Gainesville for processing by the IBM 370 Model 165 computer. A

remote terminal at the University of North Florida in Jacksonville was

used for data transmission. The procedures used were part of the Sta-

tistical Analysis System (SAS) designed by Anthony James Barr and

James Howard Goodnight at the Department of Statistics at North Carolina

State University, Raleigh.

A confidence level of 95% was set as the criterion for judging

whether statistical significance existed. Stated in other terms, the

critical value chosen was at the .05 (5%) level of significance.

The appropriate Statistical Analysis System procedure was used to

determine the equation of the best fit line and analyze the dispersion

of the data points about this line to provide a numerical measure of

the statistical significance of the calculated B coefficients of the

multiple regression equations. Analysis of variance was the statistical

method used. Overall, if significance existed, at least one of the B1

and B2 coefficients was meaningful and further mention of these quan-

tities was warranted.

A summary showing overall probability for the 10 sections is

shown in Table 4, an overall analysis of variance summary.

The level of significance, which was the probability of chance

occurrence, is shown at the right in the column labeled P for proba-

bility. This was the probability of the occurrence being greater than

the F statistic calculated by analysis of variance. A probability of

.05 or less was significant.

The probability column shows values much larger and much smaller

than .05, the 5% value for adjudging significance. To seek an expla-

nation the researcher prepared Table 5, summarizing pertinent infor-


Table 5 shows the circumstances and the data that pertained to the

10 sections. The researcher looked for any clearcut explanation at the

great variation of probability value for the 10 sections. Variation in

difficulty of the final examination of each section was not an expla-

nation. This uncontrolled random external variable caused the numerical



Section Source of Var.

1 Regression
Corrected Total

2 Regression
Corrected Total

3 Regression
Corrected Total

4 Regression
Corrected Total

5 Regression
Corrected Total

6 Regression
Corrected Total

Sum of Squares







Mean Square







F Value








4.1 0.0239

1.1 0.3227

2.1 0.1371


TABLE 4 continued

Section Source of Var.

7 Regression
Corrected Total

8 Regression
Corrected Total

9 Regression
Corrected Total

10 Regression
Corrected Total

Sum of Squares





Mean Square





F Value




1.3 0.2642



6.4 0.0058

Source of Var. = Source of variation within the data
DF = Degrees of Freedom
P = Probability of chance occurrence. P = .05 or less indicates significance





Section ESM Quarter N GPA Score Retakes P
Number Designation 1' o a p

1 301 Fall '73 36 2.96 .68 62.9 18.1 3.58 3.13 .0003*
2 301 Fall '73 27 3.09 .59 68.2 16.6 3.62 3.32 .0002*
3 301 Winter '74 25 3.27 .54 69.2 14.7 4.20 3.62 .3195
4 301 Spring '74 38 2.94 .53 86.6 7.54 3.58 2.60 .0239*
5 301 Fall '74 49 3.22 .57 68.6 11.6 2.35 1.94 .3227
6 301 Fall '74 24 3.17 .62 66.1 15.6 3.08 2.02 .1371

7 302 Summer '74 9 3.30 .48 43.1 22.2 2.89 2.03 .0139*
8 302 Fall '74 30 3.06 .52 52.8 11.7 4.47 2.71 .2642

9 303 Spring '74 17 3.14 .58 56.8 12.6 2.35 2.12 .0001*
10 303 Fall '74 27 3.00 .63 48.0 17.4 1.70 1.66 .0058*

N = Number of students in selected data
P = Probability from analysis of variance summary
= Significance


value of the regression coefficients to vary from section to section but

did not effect their reliability. Three of the 10 sections were taught

by Dr. Eisenberg, the designer of the modularized instructional system.

These were Sections 1, 9, and 10. An interesting observation is that

all three of these sections showed significance. This suggested that

closer control of these sections may have existed with the result that

the influences of external variables was reduced. On the other hand,

sections taught by the other two professors involved showed cases of

both high and low value for the same professor. This suggests that

some external variable other than the professor introduced dispersion

into the data. Correlation between two independent variables increases

the likelihood that a more basic determinant can affect both variables.

Erratic results from separate investigations can result from this cause.

(That correlation existed was obvious from the graphs of Appendix G.

Quantatively the correlation coefficients between GPA and Retakes for

the 10 sections were as follows: (1) -.32; (2) -.50; (3) -.27; (4) -.09;

(5) -.34; (6) -.39; (7) -.08; (8) +.04; (9) -.29; (10) -.36.)

These sections having significance were selected for further exam-

ination. Section 10 was included as essentially being significant with

its 5.8% value of probability. Six sections were thus selected. Table

6 shows the regression coefficients and their associated P values for

these six sections. In all six cases it was found that GPA was a sig-

nificant predictor of the final examination score. The number of

retakes was not always significant at the 95% level. Table 6 was

therefore arranged into the two sections shown.

As shown in Table 6B, both GPA and Retakes were significant pre-

dictors of final examination scores in Sections 1, 2, and 9. Unexpect-



A. Summary data for sections in which only GPA was
individually significant

Section Source of Var. Regression Coef. P

4 Intercept 68.806 .0001
Retakes -0.110 .8050
GPA 6.178 .0076

7 Intercept -57.763 .1195
Retakes -4.360 .0923
GPA 34.420 .0061

10 Intercept 6.415 .6969
Retakes -1.438 .4448
GPA 14.684 .0061

B. Data for sections in which GPA and Retakes were
both individually significant

Section Sourve of Var. Regression Coef. Prob. Partial SS

1 Intercept
Regression equation

2 Intercept
Regression equation

9 Intercept
Regression equation

30.438 .021
-1.679 .047 866.38
13.000 .001 2464.59
is Score = 30.438+13.000(GPA)-1.679(Retakes)

38.495 .013
-1.722 .028 882.282
11.640 .009 1267.109
is Score = 38.495+11.640(GPA-1.722(Retakes)

-10.339 .313
1.963 .026 254.251
19.910 .0001 1954.818
is Score = -10.339+19.910(GPA)+1.963(Retakes)

A probability of .05 or less is significant
Partial SS = Partial sum of squares and indicates the amount of
variation attributed to one independent variable
when the effect of the other is controlled.

edly for Section 9 the retake coefficient was positive. This positive

coefficient indicated that students who had trouble in passing module

tests made higher than expected final examination scores. This finding

implied that students retaking the tests accomplished much learning and

surpassed the students initially appearing to be superior.

The Section 9 data were plotted to permit further investigation of

this unexpected result and are shown in Figure 10 of Appendix G. In-

spection showed one apparent wild point representing the moderately

high raw score grade of 62 for a student having the markedly high num-

ber of seven retakes. This one point in a leveraged position for a

small population appeared to be sufficient to cause the upward slope of

the data pattern and the corresponding positive value of the regression

coefficient. Removal of this one point would seemingly leave a data

pattern having a downward slope with increasing values of the retake


The wild point (Retake = 7, Score = 62) was removed from the data

base and the Section 9 data were reprocessed. The Retakes regression

coefficient (with GPA controlled) not only remained positive but decreased

only a trivial amount from 1.962 to 1.955. This indicated that when GPA

was controlled, the apparent wild point was "right on." The point was

very consistent with the regression equation when calculated from the

main body of data. The seemingly obvious explanation had failed to

explain the unexpected results for Section 9.

The researcher's next attempt to discredit or support the unex-

pected findings for Section 9 involved use of additional calculations

provided by the SAS regression analysis.

Table 6B also shows the partial sum of squares, which was variation

attributable to one independent variable when the effect of the other

had been removed. Table 6B therefore indicates the relative effect of

the two independent variables. Specifically it shows how the concurrent

GPA predominated over Retakes in predicting final examination scores.

The proceeding does not refer to the size of the regression coefficients,

which were measured by different scales.

The appropriate SAS procedure was also used to perform regression

analysis for predicting scores from GPA alone and from the number of

retakes alone. These two analyses were performed for each of the 10

sections. Table 7 shows the probability that these regression analysis

outcomes could have occurred by chance.



Section No. Probability for Chance Occurrence
Retakes GPA

1 .0063* .0002*
2 .0004* .0001*
3 .4185 .1442
4 .6521 .0063*
5 .1487 .2911
6 .1328 .0753
7 .2110 .0136*
8 .8673 .1006
9 .7954 .0001*
10 .0973 .0016*


Table 7 shows that except for Section 5 the probability for GPA

was much smaller than that for retakes. These smaller values for GPA

indicated that concurrent GPA was a more reliable predictor than was the


the number of retakes. This statement does not refer to the numerical

size of either regression coefficient. The size of the trend (the

slope of the regression line) was meaningless if there was little con-

fidence in its existence because it likely occurred by chance. The

probability figure related to this confidence. The smaller the prob-

ability figure, the less likely that the regression line, whether steep

or shallow, occurred by chance. It is worthy of note that only the six

sections showing significance of GPA used alone showed significance for

GPA and retakes used together.

The proceeding has shown that outcomes having statistical signi-

ficance were found, that other outcomes were far from significant, that

correlation between the independent variables and varying environmental

conditions provides a possible explanation as to why these large differ-

ences might have occurred among the sections, and that concurrent GPA

was a more reliable estimator (is more closely associated with) final

examination scores than was the number of retakes. For Sections 1 and 2

the final examination score appeared to decrease with increasing retakes.

For Section 9, the opposite appeared to be true.

At this point it could have reasonably been argued that the "effect"

of Retakes, if any, was so small that it was usually lost in the

residuals (the clutter remaining after the effect of GPA was removed)

and that occasional indications of significance which showed conflicting

results were random occurrences within the clutter. This reasonsing

would support a conclusion that any "effect" of the number of retakes

was too small to be detected by the methods of this investigation and

that the primary null hypothesis that no association existed could not

be rejected.

The researcher's final attempt to discredit or support the unex-

pected finding for Section 9 involved still another effort to identify

an external causal variable. The researcher again reviewed the

learning laboratory folders and discovered a satisfactory explanation

for the contrasting findings of Sections 1 and 2 as compared to Sec-

tion 9.

Sections 1 and 2 involved beginning engineering students who were

within the instructional system in Fall 1973. Section 9 involved stu-

dents approximately beginning their senior year and who were within the

instructional system in Spring 1974. More mature students were involved,

and some of the instructional system procedures had changed. Unlimited

retaking of module tests was still permitted but delays between tests

were encouraged by procedural changes that had been introduced. The

results was clearcut. The researcher found that for Sections 1 and 2

students typically took two, three or even four tests of a module on

the same day. The particular unit finally passed was often one of those

shown by Appendix E to be unduly easy. The students had manifestly

abused the system and eventually passed a module without properly

learning the subject matter. On the other hand the researcher found

that students in Section 9 predominantly had at least one day between

each retake.

The dissimilarity of the two situation was pronounced. A rea-

sonable explanation for the statistical results had been found and the

finding for Section 1, 2, and 9 can be supported rather than be attrib-

uted to random occurrences within the clutter that remained after the

predominant effect of GPA had been removed.

Effect of Date of Completion of the Modular Path

As an adjunct study, the "effect" of the rate at which students

progressed through the modularized version was sought. As an overall

indicator of whether a student raced through to completion or whether

he procrastinated, the researcher extracted from the student record

sheet the date of completing the final module. Two sections were ran-

domly selected. The date of the month was used as the numerical mea-

surement. This constituted a third independent variable in addition

to GPA and Retakes and was called Date. Table 8 shows the results of

a multiple linear regression analysis. The high value of P associated

with Date indicated high likelihood that the regression coefficient

for date of completion occurred by chance. This limited investigation

therefore did not indicate that date of completion had a reliable

association with final examination scores.



Section Overall Date

1 .0003* .7391
7 .0767 .3745

p = Probability of chance occurrence
* = Significant

Student Attitude

As a-second adjunct study, the association between student attitude

and final examination score was sought. Student attitude referred to

the feeling of students toward the modularizing of the engineering

mechanics courses and their taking of the courses in this form.

The researcher sought an attitudinal measuring instrument that

could be used at the time of the final examination. A quick response

instrument was needed in order to obtain the cooperation of professors

and students. Originally planned to suffice only to divide students

into broad categories for chi square analysis, a multiple choice ques-

tionnaire was prepared.

To validate this instrument, a trial run was conducted using a

group of electrical engineering students who had already experienced

the ESM modularized sequence. Many students checked the most favorable

of the three choices offered for several of the questions. An inade-

quate spread of results occurred. The researcher then prepared the

revised questionnaire of Appendix H, which had two choices of favorable

response such as "strongly prefer" and "prefer." This questionnaire

was presented to students in December 1974 during their final examination

period. The three sections of Fall ESM sections were involved. They

were Sections 5, 6, and 10. Table 9 shows the result of a multiple

linear regression analysis with attitude used as a third independent

variable in addition to GPA and Retakes. The high value of P associated

with attitude indicated that the regression coefficient for attitude

occurred by chance. This limited investigation did not indicate that

student attitude concerning the modularized ESM courses was reliably

associated with final examination scores.



Section Overall Attitude

5 .1025 .1948
6 .2606 .5999
10 .0594* .8398

P = Probability
* = Significant (borderline case)


This investigation was an associational study seeking to relate

student utilization of no-penalty retaking of module tests with final

examination scores when other variables were controlled. This ex post

facto investigation proceeded heuristically, seeking paths of inquiry

that offered promise.

Careful selection of data was used as a major statistical control.

Much effort was devoted to establishing a large data base in which

elements had been screened for uniformity of situation so as to reduce

dispersion. Ten sections of engineering mechanics classes were usable.

Much effort was devoted to establishing broad categories for chi square

analysis. Modifications to the initial procedures were necessary.

Much data processing was performed to produce a numerical measure of

a student's achievement relative to his expected achievement. A modified

junior year grade-point average was used as a reference for establishing

this relative achievement. The lack of a validated weighting factor

for applying this reference was the final reason found for questioning

the use of chi square analysis for the data of this investigation. A

finding of significance was disregarded because of threats to its


The investigation then proceeded by changing to regression anal-

ysis. Final examination score was the dependent variable to be pre-

dicted by concurrent grade-point average and number of retakes as two

independent variables. Significance occurred in several cases. GPA

was more associated with final examination score than was the number of

retakes. Regression equations were stated for the three sections found

to have significance for both GPA and Retakes. An observation of much

apparent relevance was that students taking more than one proficiency

test of a module on the same day showed a decreasing final examination

score with increasing total number of retakes but that students not

retaking on the same day of an intiial failure showed an increasing

final examination score with increasing total number of retakes. Adjunct

studies concerned whether final examination score was associated with

the date of student completion of the modules or with student attitude

about the modularized courses. Unsophisticated measuring instruments

were used. No significant association with final examination score

was found for either adjunct variable.




This chapter presents a summary of the preceding four chapters.

The topics have been rearranged and combined. Details of unproductive

methods of inquiry have been minimized.

The Instructional System

Three introductory junior level courses of engineering mechanics

at the University of Florida were offered in an innovative modularized

format. The courses were Statics (ESM 301), Dynamics (ESM 302), and

Mechanics of Materials (ESM 303). Various options were available.

One feature was the taking of end-of-module proficiency tests selected

at random with provision for retesting of students until a grade of A

or B was obtained. Initial failure carried no weight in determining

a student's letter grade for the course. This feature was called the

unlimited no-penalty retaking of module tests.

Students could initially or later choose the option of taking

two scheduled intracourse examinations in lieu of completing the mod-

ule tests. All students took the same final examination.

Individualization of the modular path was provided for students

of the several engineering specialties. This was accomplished by

having somewhat different versions of the three courses. The


prearranged versions consisted of slightly different grouping of the

total of 34 modules of the three course sequence. The different ver-

sions did not always have the same number of modules.

Student assistants provided tutoring in a learning laboratory

and administered the module proficiency testing feature of the instruc-

tional system.

Professors conducted regularly scheduled lectures for those stu-

dents wishing to attend.

All students participated in this instructional system although

they did not have to follow the modularized path. There were no sep-

arate traditional taught sections that could serve as control groups

for investigative comparisons.

Need for the Study

All students do not learn best in a given type of learning envi-

ronment. A trend in American education has been to provide less rigid

course structuring. When an innovative system with options is intro-

duced, a relevant question concerns how well the student population

masters the subject matter. This question concerns group attainment.

A control group would be needed to answer this question. Separate

questions could be asked concerning how various ability or personality

types achieve. Subgroup comparisons could be used to answer this ques-

tion. When the modularized instructional system of this investigation

was implemented in the real life circumstances of an engineering col-

lege, there was neither control groups nor knowledge of personality

parameters of individual students.

A remaining question concerned how student use of the unlimited


no-penalty retaking of module proficienty tests was related to student

performance on the final examination. The extent of retaking and also

the rate of progress along the modular path were observable character-

istics. They could be of assistance to the faculty provided that

reliable meanings could be associated with these two characteristics.

A major question was whether student reliance upon the unlimited re-

taking feature was associated with student achievement as measured by

the score on the final examination. (A meaningful answer required

that the effect of individual student ability and unique environmental

situation be eliminated.) Another question was whether a student's

rate of progress was associated with his achievement. Still another

question was whether an individual student's attitude toward the non-

traditional instructional system was associated with his achievement.

Nature of the Study

Concerning the above questions, this investigation predominately

treated the one concerning the retaking of module tests. The primary

null hypothesis stated that the extent of student use of the unlimited

no-penalty retaking of module tests was not associated with student

score on the final examination.

This was an ex post facto field study. It examined the recorded

records of a functioning college of engineering in which experimental

controls designed to facilitate this investigation were lacking.

Appropriate selection of the data that was included was the method

used to establish a somewhat uniform test population in which the

effect of extraneous variables was reduced.

This was an associational study. Because no randomly assigned

control group was available, this investigation sought associations

rather than causal relationships between the variables.

Student performance on the final examination of a course is multi-

determined. To provide a meaningful finding about the retaking vari-

able alone, the strong effects of individual student ability and unique

environment needed to be eliminated. The methods of data reduction and

statistical analyses were used to accomplish this. One of the tech-

niques used was to calculate a grade-point average of selected courses

taken at the same period of time that a student participated in the

modularized sequence of engineering mechanics courses. It was assumed

that the many aspects of ability and environment were compared to this

GPA. The raw score on the final examination was compared to this GPA

in a way appropriate to provide a measure of relative achievement,

achievement relative to the expected or par performance of a student.

With other variables thus partially controlled, the existence of an

association between performance and extent of retaking of module tests

was then sought.

Review of the Literature

Modularized courses at the college level and provisions for no-

penalty retaking of tests are seldom reported subjects. A computer

search of ERIC and of dissertation abstracts found few articles.

Manual searching was therefore used.

Examples of the extensive research in the use of grade-point

averages to predict academic performance were reported. These examples

related to the use in this investigation of a GPA to estimate the

expected performance of individual students, which was then compared

to observed performance. The literature indicates that GPA was often

found to be one of the best predictors. A recent GPA seemed to be

better than an older one.

Research finding concerning self-pacing and comparison of teaching

methods were included in the review. The gist of these articles

revealed that student ability was the primary determinant of student

performance. The teaching method used was less important.

Uniformity of Difficulty of the Module Tests

Analysis of student records showed that some modules were harder

to pass than others. This condition did not necessarily mean that

individual students followed unequal paths. More serious for the pur-

poses of this investigation was the finding that in some modules the

difficulty of different proficiency tests varied. The random selection

of unequal tests therefore created unequal paths for the students to

follow. This caused a dispersion within the observed data for which

there was no statistical control.

Data Collection and Selection

Data was collected from three sources. The special grade-point

average was calculated from the Registrar's records. The number of

retakes was determined from student folders from the learning labora-

tory. The final examination score was determined from the professors.

The data base used for statistical analysis consisted of carefully

selected histories of students taking and completing the same modular

path. The options and individualization that had existed caused the

records of many students to be unusable. Omissions of any part of

the data caused additional records to be unusable. The final data

base consisted of selected data from 10 sections that were taught

during approximately a two-year period.

Results of Chi Square Analysis

The independent variable for the chi square analysis was the

number of times students retook module tests. This was count data

and was used to establish categories representing the extent of stu-

dent utilization of the retake provision.

The dependent variable was the final examination score and was a

continuous variable. It was used to establish categories of achieve-

ment. This measure of a student's achievement controlled for ability

and environment was the difference between his standardized final

examination score and his standardized modified junior year GPA.

These standardized numbers were standard deviations measured from the

mean translated to zero. Subtracting these quantities produced a

number assumed to be indicative of how well a student did on his final

examination as compared to his expected achievement. These differences

were placed in rank order and were then divided into suitable categories.

A contingency table and chi square test was used to compare the

observed and expected frequencies. The chi square analysis tests for

statistical significance of the divergence of observed data from

expected data.

The grouping of the data of this investigation into categories

required the arbitrarily establishing of boundaries within the data

distribution. Several groupings were tested.

An analysis of the data of the six ESM 301 sections showed greater

dissimilarities than had been expected. The groupings of data to pro-

duce sufficiently large numbers was found to be less feasible than


One grouping involving compatible data from two sections produced

significance. Scrutiny of the steps of the data processing and of the

certain cell frequencies of the resulting contingency table suggested

that a threat to validity existed. The method of relating concurrent

GPA to raw score on the final examination to produce a quantitative

measure of a student's performance relative to his expected performance

appeared to overcorrect for GPA.

The several problems that had arisen in processing the raw data

into a form for use in the contingency table cast doubt on the validity

of whatever significance the chi square statistic might show. The

method of analysis using the chi square statistic was abandoned.

Results of Linear Regression Analysis

Ten ESM sections yielded usable data. This data was subjected to

linear regression analysis. The 95% confidence level was established

as the criterion for adjudging that statistical significance existed.

The dependent variable was raw score on the final examination.

The modified junior year grade-point average represented a con-

current GPA in other courses taken at the same time that the ESM

courses were taken. The concurrent GPA was the GPA used for all anal-

yses. It was considered to represent the effects of both ability and

environment. The number of retakes referred to how many times a stu-

dent attempted to pass a module proficiency test after failing an

initial attempt. Retakes was the other independent variable of interest.

Used alone as the single independent variable, GPA predicted final

examination score at a statistically significant level in six of the

10 sections. Used alone, retakes significantly predicted (was asso-

ciated with) score only a single time. Thus it was shown that grade-

point average was more reliable than retakes in predicting score.

The null hypothesis considered the effect of retakes when the

effect of GPA was controlled. A multiple regression analysis for the

10 sections was therefore performed using both GPA and retakes as inde-

pendent variables predicting score.

Overall significance was obtained for the same six sections for

which GPA alone produced significance. Including retakes on a second

variable to account for part of the observed variation did not increase

the number of sections showing overall significance. This is a further

indication that grade-point average is the dominant variable of the two.

The regression coefficients for GPA and score were calculated,

but the retake coefficient was not statistically significant for three

of the six sections showing overall significance. GPA alone had

accounted for enough of the variation to cause overall significance.

Thus, only three sections were found to have full statistical signifi-

cance. Conclusions were then drawn for the analysis results of these

three sections.

For these three sections, the portion of the variation attributed

to GPA was much greater than that attributed to retakes. This is a

specific indication that GPA is the more powerful variable in pre-

dicting the score on the final examination.

The existence of statistically significant nonzero regression

coefficients was the issue of vital interest to the hypothesis of this

investigation. The numerical values of these coefficients are


functions of the three scales used to measure the three variables (e.g.

Score was measured from 0 to 100) and are deliberately not repeated in

this summary. The details appear in Table 6 of Chapter IV.

The important finding concerns the nature of the regression coef-

ficient for the retake variable in the three sections where it had

been isolated from the random clutter (had been found to have statistical


Two of the sections were ESM 301. The regression coefficient for

retake was negative indicating that an increasing number of retakes

was associated with a decreasing score on the final examination. The

remaining section was a small ESM 303 section. A positive coefficient

was found indicating that an increasing number of retakes was asso-

ciated with an increasing score on the final examination. It was not

possible to attribute this unexpected finding to the effect of a "wild

point" in the data or to any other cause that would discredit the


Students of the two ESM 301 sections showed a strong tendency to

retake tests repeatedly on the same day that an initial attempt showed

that mastery of the subject matter had not been obtained. This was

not so for the ESM 303 section. This was the major environmental con-

dition found that offers an explanation for the different character-

istics of the regression coefficients for the retake variable.

Instances of statistical significance were found for an associ-

ation between the number of retakes and the score on the final exami-

nation. The lack of uniformity of the findings suggest that only

qualified conclusions should be drawn, although an explanation for the

lack of uniformity appeared to have been found. A more consistent


finding was that concurrent GPA is much more associated with final

examination score than is the extent of retaking of module tests.




This chapter presents the researcher's conclusions, recommendations

for the faculty administering the instructional system studied, and

recommendations for further research.


The innovative instructional system investigated by this explor-

atory field study was designed to provide options and alternatives to

the students. It was not designed for research purposes. Many causes

of dispersion could be only partially controlled.

A modified junior year grade-point average was calculated for

traditionally taught courses taken concurrently with courses of the

modularized instructional system. This concurrent GPA was felt to be

a composite representation of individual student innate ability, per-

sonality factors, and whatever external environmental conditions that

might have influenced individual students. Use of this concurrent

GPA as a predictor of student achievement was believed to control for

many of these basic influences.

The variation of difficulty from module to module was not felt to

cause a problem to this research. The variation of difficulty of pro-

ficiency tests within some of the modules may have been a cause of


dispersion that could not be statistically controlled. This variation

is believed to have been a contributing cause of the negative regres-

sion coefficient found for Sections 1 and 2 (ESM 301) for the retakes

variable. An easy test within a module allowed a student taking

repeat tests to have greater chance of receiving an easy test and

demonstrating an apparent proficiency that did not in fact exist.

The immediate retaking of a module test after an initial failure

did not occur in Section 9 (ESM 303). A further fact pertained. The

subject matter of ESM 301 was relatively easy. Students requiring

many retakes were likely to have been marginal students. Repeating

may not have produced genuine comprehension. The subject matter of

ESM 303 was difficult. The more mature students taking this course

may have profited from restudying and retaking the module tests.

Taking additional tests with restudying would constitute a greater

effort on the part of a student and probably provided increased

learning. The researcher believes that this was the cause contributing

to the positive regression coefficient for Section 9.

In summary, the extent of retaking of module tests was shown in

some statistically significant cases to have an association with stu-

dent scores on the final examination. This association was less

reliably shown and was of lesser magnitude than that of concurrent

GPA. The nature of this association rather clearly appeared to depend

upon the conditions. An immediate retaking of a proficiency test after

an initial failure probably did little to enhance comprehension and was

likely to lead to an erroneous appearance of mastery. Restudying prior

to retesting probably caused increased learning and comprehension.

The researcher concludes that while aptitude and environmental

factors were major determinants of academic achievement, the extent

of retaking of end-of-module tests in the modularized courses was a

second order effect that was occasionally detected by the investigation.

The researcher interprets the findings to conclude that there was an

association between the extent of retaking of modules tests and final

examination score. The researcher concludes that the extent of

retaking per se was not the entity that matters. The conditions under

which the retaking occurred influenced the outcome.

The researcher does not feel that the negative regression coeffi-

cient for Sections 1 and 2 conflicts with mastery learning theory.

Because of the several special features of the instructional

system investigated, the researcher suggests caution in generalizing

all observations and findings to other instructional systems. The

conclusion that the manner of retaking rather than the amount of

retaking is the entity that matters is felt to be a worthwhile con-

clusion of general applicability.


The researcher recommends that the end-of-module proficiency tests

be periodically monitored for uniformity of difficulty. This recom-

mendation is especially applicable if individual tests are replaced

from time to time to guard against the possibility of compromise of

the testing system. Concerning cheating, the researcher in reviewing

comments written in a space provided on the attitudinal questionnaire

did not find indications that the existence of cheating was a major

concern of the students.

There appears to be a discernible association between the extent


of module test retaking and the score on the course final examination.

The conclusion that the effect of retaking depends upon the conditions

is offered as a hypothesis to be investigated under controlled con-







In designing an instructional system for the introductory courses

in engineering, one must contend with a number of problems associated

with the heterogeneity of the backgrounds, interests and abilities of

the students served. Typically, a single department is assigned the

responsibility of teaching a course which must meet the curricular

demands of perhaps a dozen distinct degree programs. In large state

universities it is not uncommon to find that students have completed

their prerequisite studies at many different community colleges as well

as at the university in which the course is offered. Consequently, the

input competency of the students is highly variable. "Open door" and

"affirmative action" policies tend to contribute to this variability.

If the introductory courses are to be anything more than a passive

sorting or classification system, then special attention must be given

to the problems they present. The introductory mechanics courses in

statics, dynamics and mechanics of materials are archetypical of ser-

*The text of this article is reproduced herein without its acknowl-
edgements, references, and footnotes. Only Tables 4, 5, and 6 are
reproduced herein. The full article by Martin A. Eisenberg, University
of Florida, Gainesville, appeared in the December 1975 issue of Engi-
neering Education and is used herein with permission of the publisher.

vice courses which must meet the demands mentioned above.

At the University of Florida, an instructional system employing a

highly flexible combination of modular curriculum packaging, variable

pacing, programmed learning materials, computer managed instruction

and conventional lecture and examination methodologies has evolved.

In an earlier paper exploratory studies, the basic design philosophy,

and an outline of the system planned for the 1973-74 academic year

were described. In this paper the system as implemented and its evo-

lution in response to operational experience and increasingly severe

budgetary constraints are described. The effect of alternate staffing

strategies (graduate vs. undergraduate teaching assistants) upon cost

and effectiveness is analyzed, as is the overall cost of operation in

comparison with conventional systems. The results of preliminary

studies on the effectiveness of the prescribed learning materials and

activities are also presented.

The Basic System

The system of instruction consists of a sequence of courses in

the subjects of Statics (ESM 301), Dynamics (ESM 302) and Mechanics

of Materials (ESM 303) which are completed during the junior year by

most students in the College of Engineering.

In the design of the system the following criteria were adopted:

Course Criteria

1) The system must accommodate variable student
and departmental content objectives.

2) Variable input competencies must be recognized
and accommodated.

3) Continuity of program effectiveness should be

4) Variable faculty teaching styles must be

5) Provision for continuous improvement and
modification must be possible.

6) Instructional costs must be kept at eco-
nomical levels.

7) Sanctions sufficient to ensure reasonable
student progress should be imposed.

8) Objective measures of student performance
and program effectiveness should be provided.

9) A variety of learning modes should be
available to suit the student and subject

In response to the first criterion it became apparent that the sub-

ject matter of these disciplines would have to be broken down into mod-

ules, and a prerequisite hierarchy established. Figure 1 illustrates

a decomposition of the subject into the 34 modules upon which the sys-

tem is based. The prerequisite hierarchy is indicated by the arrows.

For example, the study of unit 16 on Plane Motion Dynamics requires

direct prerequisites of units, 4, 5, 8, 12 and 15, which in turn imply

the prerequisite study of units 1, 2, 3, 6, 10, 11 and 14. Thia a la

carte menu was presented to the curriculum committees of each of the

client departments, and a course of study tailored to the needs of their

students was negotiated.

Table 1 shows the content of courses required for each of the

undergraduage degree programs. As far as the University registrar is

concerned there are only three variable credit courses in which the

student may enroll. Within each of the courses, however, students may

be enrolled in one of three to five different subcourses whose exis-

tence is of no concern to the registrar. A careful study of the content

of subcourses 30, 31, 32, 33 and 34 of ESM 303 will reveal their

striking similarity. Subcourse 33 taken by aerospace engineering stu-

dents differs from the standard course (subcourse 30) only in the addi-

tion of shear center and column buckling units to the curriculum. The

nuclear engineers study the stress and deformations in thick-walled

cylinders under thermal and pressure loading, while electrical engineers,

for good but arcane reasons of a local nature, study units normally

part of other courses, e.g., the static analysis of beams and damped

vibrations of particles and rigid bodies.

The differences among most of the courses are small, but it is

these small differences that lead to major debates in college curric-

ulum committees, proliferation of courses in "Mechanics for XYZ Engi-

neering" and/or dissatisfied clients of service courses. By employing

the modular design it is possible to tailor the course to meet the dif-

ferent requirements of the clients and thereby provide better service.

The modular structure has the additional advantage of allowing

one to tailor a course of instruction to meet special requirements.

For example, a student who has completed a sophomore-junior level

physics course in mechanics probably has a good background in particle

dynamics, no experience in the static analysis of simple structural

elements, and a weak background in rigid body dynamics. In such a

case one'may create an ad hoc course, administrable under the aegis of

the standard dynamics course, which will allow the student to complete

his study of statics and dynamics without leaving major gaps and with-

out undue repetition.

Each of the courses employs a standard textbook, and a full

set of lectures is scheduled. The schedule of lectures is distributed

to the students at the beginning of the term so that they may use the

lecture as one of a number of learning resources. Our experience shows

that most students benefit from the lectures of most professors. In

any event, criterion no. 4 mandates the availability of a stage from

which the faculty member responsible for the course may deliver his

personal statement on the subject matter of the course.

In addition to the classical textbook a set of programmed study

guides has been developed for the course sequence. The study guides

are an adjunct and guide to the use of the standard texts. They are

designed to give the student sufficient guidance to permit him to

study the material independently. For each unit or module there cor-

responds a learning activity package as part of the study guide. Each

of the modules contains:

1) A description of content and rationale for
study of the indicated material.

2) A statement of prerequisites.

3) A list of behavioral objectives which the
student is encouraged to use as a check list
of tasks to be accomplished and abilities to
be developed.

4) Commentary and guide to the text. This, the
heart of the study guide, is written in a
colloquial style but is carefully structured
to require active participation of and fre-
quent feedback to the student. The student
is assigned tasks to read specific sections
of the text, to provide missing steps in
derivations, to identify implicit assump-
tions, to solve specific problems, to per-
form simple experiments, and to swear oaths
of allegiance to the use of free body dia-
grams. In addition, the study guide provides
remedial and advanced material. In this
manner the student is asked to abandon his
customarily passive role as reader. The
technique is similar to that of standard
programmed texts, but the programming steps

are considerably larger than is usual. This
"macroprogramming" approach has the advantage
of being considerable less time-consuming to
prepare than traditionally programmed mate-
rial. It also places greater responsibility
upon the student and hopefully, by example,
encourages the student to read technical
material from a critical perspective.

5) Sample proficiency test. After the student
has worked through the unit and reread the
list of objectives, he is instructed to
respond to a quiz printed at the end of the
unit. Upon completing the quiz under the
indicated conditions (duration, open book,
closed book), he is instructed to read the
solution provided in the package and to
grade his own paper. The actual quiz in
which he will be asked to demonstrate com-
petency will be of a similar level of dif-
ficulty and offered under similar conditions.

The study guides were prepared and the modular course system im-

plemented in stages during the academic year 1973-74. The statics,

dynamics and mechanics of materials guides were put into use in the

fall, winter and spring quarters, respectively. Since spring 1974,

modular courses have been offered to approximately 300 students each


To provide additional assistance to the student, a learning lab-

oratory has been established and staffed with graduate and undergrad-

uate teaching assistants whose duties include tutorial services. The

laboratory has typically been staffed about 40 to 50 hours a week and

is heavily utilized by students.

All of the above features of the instructional system have been

in effect since the fall of 1973 and are expected to be maintained in

the foreseeable future. In other respects the system is undergoing a

continuing evolution.

Modularity, Programming and Self-pacing

Modular instructional systems, programmed instructional systems

and self-paced instructional systems are not synonymous.

A modular system is one in which the complete system is substruc-

tured into discrete components with specified functions and specified

interreactions with other components. It is designed to meet special-

ized system criteria by selectively drawing upon a modular subset of

the total system. Modular instructional systems are individualized

systems in that they facilitate the construction of course syllabi to

meet individual needs.

Programmed instructional systems are characterized by: 1) the

explicit identification of changes (behavioral objectives) in demon-

strable competencies and/or attitudes expected of the user of the sys-

tem; and 2) by the adherence to psychological principles of behavior

modification through reinforcement (positive feedback), pioneered by

B. F. Skinner. These principles have been used to teach people to

assemble M-l rifles psychomotorr programming), to solve differential

equations (cognitive programming), and to develop value structures

(affective programming). Programmed instructional systems are indi-

vidualized systems in that they are addressed to individuals rather

than a public. They demand individual participation. They are based

upon dialogues rather than soliloquies.

Self-paced instructional systems are designed to cater to variable

student input competencies (criterion no. 2) and levels of effort.

They permit the system designer to demand minimum performance levels

of all students by permitting a variable time to meet these criteria.

The Keller or PSI methods are typical of such systems. Because self-

paced systems are inherently subject to abuse by procrastination

(criterion no. 7), they are rarely implemented in pure form. Most so-

called self-paced systems are actually flexibly-paced systems. Self-

paced systems are individualized systems of instruction in that they try

to accommodate individual differences in ability and are subject to a

significant extent to individual control.

Thus, modular, programmed and self-paced systems are, in different

senses, individualized systems. Any given individualized system may

assume to varying degrees the characteristics of all or some of these

system types. The instructional system for engineering mechanics in

use at the University of Florida since 1973 has been a consistently

modular and programmed system with variable elements of flexible


As originally implemented, the system employed a significant ele-

ment of flexible pacing strategy. Major traditional examinations were

scheduled during the fourth and eighth weeks of a ten week quarter.

Students had the option, however, of demonstrating proficiency (A-B

performance) on module quizzes which could be taken on demand and

repeated (different quizzes) as necessary without penalty. If a stu-

dent demonstrated reasonable progress by the dates of the scheduled

major exams, he was excused from the examinations. All students were

required to take a comprehensive, traditional final examination. For

students who completed nearly all the quizzes the final grade computa-

tion could be deferred until the end of the first week of the following

quarter, to permit completion of the remaining one or two modules at

the A-B mastery level.

With this system the student could proceed at his own pace and


walk into the final examination room with a minimum average of B, pro-

vided he made reasonable progress during the term in the completion of

his module requirements.

Although most students opted for this mode of completing the

course, not all students view such a system with favor. There are

those who would rather take a few rigidly scheduled exams on a tradi-

tional sink-or-swim basis. In accordance with criterion no. 9, such

provision was made. In fact, each student could elect a continuum of

options between the pure competency-demonstration self-pacing mode and

the traditional mode. With students registering for as many as five

different subcourse versions in the same class and with the possibility

of each student proceeding under some combination of self-pacing and

classical examination modes, a CMI routine was necessary. Such a com-

puter code has been developed by the author and reduces the record

keeping requirements to easily manageable proportions.

The cost-effectiveness of the system is analyzed in some detail

below. Such systems require moderate levels of funding for student

assistance. In particular, the updating and the maintenance of secu-

rity on extensive quiz files and the logistics of individually assigning,

monitoring, grading and recording quizzes places a heavy but not unrea-

sonable burden on such funds. In times of financial retrenchment, how-

ever, it is not always possible to maintain reasonable levels of

funding. Such is the case now. In response to increasingly severe

budgetary restrictions which forced a cutback in student assistant

staffing, it was necessary to first curtail and eventually eliminate

the flexible-paced aspects of the system. The modular unit quizzes

have been retained, but they are now scheduled on a weekly basis and


will not be repeatable. Thus, the system in adjusting to criteria no.

5 and no. 6 may lose some of its effectiveness in meeting criterion

no. 2.

Cost Effectiveness

The development of cost effectiveness data within a give institu-

tional setting is always subject to charges of prejudice. To transfer

such data to other institutions is still more difficult. The data

(table 2) on the per quarter cost of alternative systems of instruction

is reasonably accurate for the University of Florida. During academic

year 1972-73 two course sequences in engineering mechanics were

offered (ESM 304, 305 (10 hrs.) and ESM 301, 302, 303 (12 hrs.)).

The data on faculty contact hours represent typical course scheduling

patterns. If any error has been made, it has been in an underestimation

of the AY 1972-73 student assistant expenditures. Data for the modular

system are based on actual experience and budgeted expenses. No amor-

tization of development costs is shown for the modular system because

of the caprices to which such estimates are subject. If one assumes

that one-third of a faculty line item/quarter on a continuing basis

would be a reasonable allocation for the cost of development and mod-

ification, then that would add about $2000 to the indicated cost

figures. The data for the austere traditional system would represent

a severe cutback in service offered, since it would require all depart-

ments to take a common 12 hour course sequence. Such a development

would meet with severe resistance and lead to possible loss of cus-

tomers and the duplication of courses. It would offer no advantages.

Teaching Assistant Effectiveness

The one salient feature of the cost data of table 2 is the great

increase in expense incurred by substituting graduate student assis-

tants for undergraduate assistants.

At the end of the fall 1974 quarter the students were asked to

evaluate the effectiveness of each of the teaching assistants. This

was a quarter in which the students had extensive tutorial contact

with teaching assistants. They were asked to respond to the following


1) Had a good knowledge of the subject
matter for this course.
2) Graded quizzes fairly.
3) Helped you learn the material.
4) Treated students courteously.
5) Overall effectiveness.

The average scores of the three graduate assistants are compared with

those of the four undergraduate students in table 3. There is a con-

sistent pattern of superior performance by the undergraduate students

in comparison with the graduate students.

The differences may be attributable to a number of facts which are

widely applicable to other institution. First, the graduate students

were average graduate students in the department. They were not vol-

unteers, nor were they specially selected by the instructor. They were

available, were not working on reimbursable grants, and they were

assigned to the teaching lab as part of their assistantship respon-

sibilities. The undergraduate students were among the very best stu-

dents in the college. They were recruited and there were more appli-

cants than positions. Moreover, if called upon to work longer hours

on occasion, they were pleased to do so since they were being paid by


the hour. The graduate students are paid a monthly stipend in return

for which they may be called upon to perform some duties. It would

be understandable if they felt that to some extent they were over-

worked and underemployed.

The differences in performance are not large, and one should not

conclude that graduate students not be used for such assignments.

Rather, it is safe to say that no loss of effectiveness occurs if

undergraduate students are hired. It may be highly desirable to assign

graduate students to such duties as a means of supporting the graduate

program. Any increment in cost attributable to the use of graduate

students, however, should be considered as part of the cost of the

graduate program and not part of the cost of the undergraduate instruc-

tional system.

Effectiveness of the Instructional System

It would be felicitous to report evidence that the modular, flex-

ibly paced instructional system results in superior student mastery of

engineering mechanics in comparison with students educated by tradi-

tional methods. Although there is reason to believe this to be so,

evidence for such a sweeping conclusion is difficult to produce. One

cannot trust comparisons between performance on nominally comparable

tests or even comparisons in performance on the same test graded by

different people, or on the same test graded by the same person on

two different occasions.

There are data to report, however, which tend to corroborate the

improvement in student performance attendant to the use of the system

described above. Table 4 shows the average grades on the final exams

for six randomly selected courses taught under the modular system.



"A" "B" % Improvement
Term Course (0-7 quizzes (8 or more Sample ("B"-"A")xlOO
passed) quizzes size 100 "A"

F 73 ESM 301 47 63 52 30.2
W 74 ESM 302 54 68 23 30.4
S 74 ESM 303 53 56 49 6.38
F 74 ESM 301 84 91 34 43.8
F 74 ESM 302 48 47 27 -1.9
F 74 ESM 303 46 54 52 14.8

Total sample size = 237 Weighted average % Improvement = 20.2%

The students have been split into two groups. In column 'A' the final

examination scores of students who completed 0-7 units prior to taking

the exam are listed. In column 'B' similar results are shown for stu-

dents who completed 8 or more units. The average number of required

units for all courses is 9. The last column shows the percentage

improvement as 100 times the ratio of the difference in scores divided

by the maximum possible difference in scores. This figure of merit

provides a common basis on which examinations with different mean

scores may be compared.

During the spring 1975 quarter large numbers of students took the

major in-class examinations for the first time. Similar results for

these courses are tabulated in table 5.



No Some All % Improvement % Improvement
quizzes quizzes quizzes Sample Some quizzes All quizzes
Courses passed passed passed size passed passed

ESM 301 66 75 90 134 26.5% 70.6%
ESM 302 59 63 79 70 9.8% 48.8%
ESM 303 69 76 88 88 22.6% 61.3%

Total sample size = 292
Weighted average
% Improvement, some quizzes passed: 21.3%; all quizzes passed: 6.3%

Tables 4 and 5 show that participation in the competency demon-

stration quiz program results in significantly improved performance on

the comprehensive final exams and on the comprehensive fourth week

exams, although the data indicate that the effect is less pronounced

in the final examination, which covers a larger amount of material

and longer time span.

There is a possibility that the data are misleading, in that the

students who complete the units may be better students and that this

factor may account for the observed improvement. To test this hypoth-

esis the data for the large spring 1975 statics class were examined

in more detail. Table 6 shows the examination results for students

who passed varying numbers of quizzes prior to the examination. The

third column shows the average GPA for each of these groups of stu-

dents. The consistent GPA trend suggests the possibility that the



Number of Average University GPA corrected
quizzes grade on grade point examination
passed examination average grades

0 66 2.74/4.0 66
1 73 2.84/4.0 71
2 80 2.87/4.0 78
3 90 3.09/4.0 86

improvement in examination scores may be due to the fact that better

students pass more of the quizzes. However, when the GPA for each of

these groups was reduced to the base 2.74/4.0 level by randomly

deleting from the data set a sufficient number of students with high

GPAs, it was found that examination scores for the student with com-

parable GPAs, but different numbers of quizzes passed, were markedly

different. Participation in the competency quiz program accounted

for 20 points of the 24 point grade spread observed.

Thus, one may conclude that students with comparable histories of

academic achievement performed better by participating fully in the

use of the procedures and materials in this instructional system. To

this conclusion a skeptic might respond --what's new? All that has

been shown is that students who work hard do well. While to a degree

such criticism is justified, the data do indicate some useful informa-

tion, e.g., the tasks assigned to the student are relevant to the

objectives and effective in meeting them. While one may intuitively

anticipate such a conclusion, it is nonetheless not trivial. Without

experimental verification, an engineering professor may be no more

certain that completion of certain exercises will contribute to the

mastery of a given subject than may a swimming coach be certain that

completion of a regimen of calisthenics will increase the speed of his



As a result of experience with the modular system during academic

years 1973-74 and 1974-75, its feasibility has been demonstrated. It

is competitive on a cost basis with conventional systems, and prelim-

inary results indicate that student performance may be improved by the

use of the instructional materials and procedures of this system.

Course evaluation questionnaires and informal consultation with stu-

dents and teaching assistants indicate positive affective response to

the system and a belief by the students that they are learning more

than they would have in more conventional systems. Attempts to make

a more definitive evaluation of system effectiveness are in progress.

However, it is clear now that by using the modular structure, the

departments served by the courses have more detailed knowledge and

control of the content of the material for which their students are








Course: ESM 30 Credits

Major Department

U.F. G.P.A. Hours Registere

If working, how many hours

Lower Division Preparation at: UC


ESM 301



Subcourse Version

d This Quarter

College (which)


Math Completed PS 215 Grade

Grade When













University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs