Group Title: experimental study of two strategies for modifying impulsivity of children ages eight through eleven
Title: An Experimental study of two strategies for modifying impulsivity of children ages eight through eleven
Full Citation
Permanent Link:
 Material Information
Title: An Experimental study of two strategies for modifying impulsivity of children ages eight through eleven
Physical Description: vii, 110 leaves. : illus. ; 28 cm.
Language: English
Creator: Wen, Shih-Sung, 1936-
Publication Date: 1971
Copyright Date: 1971
Subject: School children   ( lcsh )
Child psychology   ( lcsh )
Foundations of Education thesis Ph. D   ( lcsh )
Dissertations, Academic -- Foundations of Education -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis--University of Florida, 1971.
Bibliography: Bibliography: leaves 106-109.
Additional Physical Form: Also available on World Wide Web
General Note: Manuscript copy.
General Note: Vita.
 Record Information
Bibliographic ID: UF00097695
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 - 000871581
notis - AEG8804
oclc - 014277450


This item has the following downloads:

PDF ( 5 MBs ) ( PDF )

Full Text

An Experimental Study of Two Strategies

for Modifying Impulsivity of Children

Ages Eight Through Eleven










The writer wishes to express his grateful appreciatior to the fol-

lowing people whose contributions have made his doctoral program a pro-

foundly enriching experience.

To Dr. William W. Purkey, Sr., Committee Chairman, for his constant,

warm, positive support and thoughtful guidance throughout the entire

program. To Dr. Hal G. Lewis, Chairman of Foundations of Education

Department, for his continuous concern and support. To Dr. Vynce A.

Hines, Committee Member, for his help, inspiration and suggestions re-

garding research techniques. To Dr. Hannelore L. Wass, Cornmittce Mem-

ber, for her understanding, encouragement, and advice. To Dr. Franz

Epting, Comn.ittee Member, for his interest and advice in regard to the

research approach employed in this study.

Special thanks are extended to those teachers and children of

P. K. Laboratory School at the University of Florida, Gainesville,

Florida, for their participation and cooperation. The writer is also

indebted to -Mr. William J. Bowers, for his assistance in testing, to

Miss Robin A. Mays and John Harrell, for their help in task training,

and to Dr. Charles E. Francis, principal, and pupils of Starke Elementary

School, Starke, Florida, for their participation in the pilot study of

this research. Special recognition is due Mrs. Annette Pizer for her

editorial assistance. Appreciation is also extended to Mrs. Earbara

Coaxum, for her preparation of this manuscript.

Finally, this writer wishes to express his gratitude to his lovely

wife, Alice (Tien-pao), and two-year old son, Stanley, for their pa-

tience, sacrifice, and support throughout his doctoral program. And to

his mother, Mrs. Chiu-yen Wen, living in Taiwan, China, he is deeply

grateful for her ceaseless concern and encouragement.


ACNOWLEDGMfENTS. . . . . . . . . . . . . ii

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

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

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


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

Statement of the Problem . . . . . . 2

II. REVIEW OF LITERATURE. . . . . . . . . 6

Modifiability of Reflection-Impulsivity Tempo. . 6
Relationship Between Response Time and Error . 14
Some Commentary in Regard to Direct Modeling
Method . ............... . 15
Language Jorms Affecting Cognitive Devclop r2nnt .17
A Possible Cause of Impulsive Children's Making
Errors . . . . . . . . .. .. 183
A New Indirect Method. . . . . . . ... 20

III. DESIGN OF THE STUDY . . . . . . . . . 22

Hypotheses . . . . . . . .... . 22
Sampling and Sample. . . . . . . ... 23
Instrumentation. ............... . 25
Experimental Tasks . . . . . . . . 29
Procedure. .............. . . . 31
Analysis of Data . . . . . . . ... 33
The Limitation of the Study. . . . . ... 35


Results of Hypothesis Testings . . . . . 39
Hypothesis 1. . . . . . . . .. 39
Hypothesis 2 . . . . . . . .. . '
Other Related Findings . . . . . ... 47

V. DISCUSSION. . . . . . . . . ... . . 54

Language and Forniing of a Cognitive Style. . . 54



Relationship Between Restricted Language and
Impulsivity. ..... . . . . . 55
Cognitive Mismatch by Mismatched Codes of
Communication. . . . . . . . . 56
Enriching Rather than Correcting . . ... 58
Enriching Through Hypothesis Testing . . .. 59
Evaluation of the Experimental Task. . . ... 60
Analysis of Training Effect on Each MFF Item . 61
Training Effect on Relationship Between Response
Time and Error . . . . . . . .. 64
Relative Efficiency of the Experimental Design 65
Conclusion . . . . . . . . ... 66


Summary. . . . . . . . .... . 67
Conclusion . . . . . . . . ... 69
Implications of the Study. . . . . .. 70

APPENDICES ............... . . . . 73



C. LONG SENTENCE TASK (LS TASK). . . . . ... .. 79


E. STATISTICAL DATA (TABLES 6, 7, 8, 9, 10, 11). . ... 101

BIBLIOGRAPHY . . . . . . . . . . . . . 106

BIOGRAPHICAL SKETCH. . . . . . . . ... .. 110
















Description of the Sample by Group. . . . . 24

MFF Error After Training . . . . ... . 40

Analysis of Variance of Error . . . . ..... 40

Multiple Comparisons: In Error . . . . . .. 41

Multiple Comparisons: In Error (Four Groups) . ... 41

Sex Differences in Error Change. . . . . . 102

Grade Differences in Error Change . . . . .. 102

Age Differences in Error Change ... . . .. . 103

TQ Differences in Error Change . . . . . ... 104

Extreme TQ Differences in Error Change . . . ... 104

Socioeconomic Difterences in Error Change . . .. 105

Summary of Mean Differences of Error Change by Sex,
Grade, Age, IQ, and Socioeconomic Index . . .. 43

Analysis of Covariance with CTMM Score as Covariate . 48

MFF Response Time . . . . . . . . . 48

Analysis of Variance of Response Tine . . . .. 50

Error Change by Extremely and Moderately Impulsive
Groups. . . . . . . . . . . 52



1. A Sample Etem of Matching Familiar Figures--
The Standard. . . . . . . . .... 27

2. A Sample Item of Matching Familiar Figures--
The Variations. ................. 28

3. Pretest Average Errors on Each MFF Item ..... 62

4. Posttest Average Errors on Each MFF Item. ..... 63

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



Shih-sung Wen

August, 1971

Chairman: Dr. William W. Purkey, Sr.
Major Department: Foundations of Education

For children with impulsive tempo, high errors on task performance,

among others, impede their optimal learning, informative measurement,

and healthy personality development.


The purpose of this study was to help solve the above problem by

modifying children's impulsivity in terms of error reduction as meas-

ured by Kagan's Matching Familiar Figures (bMFF). In this study, an

jssumpticn was made, relating language form with impulsive tempo.

Therefore, two strategies, enriching children's ability to deal with

redundancy as well as enriching children's ability to deal with vari-

ability, were introduced into this study to cope with the problem,

u.cthermore, two hypotheses representing the two strategies were for-

nulated for tasting. The two hypotheses are:

Hypothesis 1: Training in writing long sentences has a

significant effect on impulsivity in terms of reac.cing

impulsive children: errors as measured by Matching

Familiar Figures.

Hypothesis 2: Training in making multiple alternative

choices has a significant effect on impulsivity in Lt.rns

of reducing impulsive children's errors as measured by

Matching Familiar Figures.


Eighty-six boys and 89 girls in grades 3, 4, 5 of P. K. Yonge

Laboratory School at the University of Florida, Gainesville, Florida,

participated in this study. Their ages ranged from eight through eleven.

By using Kagan's ,MFF and the split-half method, 59 impulsive and 59 re-

flective subjects were identified. Fifty-seven impulsives were randomly

selected and assigned into three groups with block design in terms of

errors. Nineteen reflectives were randomly selected as :matched control


Two experimental tasks, Long Sentence Task (LS Task) and Multiple

Alternatives Task (MA Task), were designed by this researcher for test-

irg the two hypotheses. These two tasks were randomly assigned to two

impulsive groups with the third group as control group which received

no training. A male and a female graduate student were trained to con-

duct tne training. In LS Task training, subjects ,ere instructed to

write two long sentences for each task. Whn-le in VMA Task, subjects were

instructed t:o write at least three alternative choices for each part

of a taskl (each task includes two parts). Ten tasks for each group

w-:'re finished in Five weeks. After training, the same version of MFF

was irndividuill) d::ini .tered to all subjects in four groups.


The results of this study are sun-narized as follows: (_) LS Task

group and MA Task group significantly reduced their errors as compared

with control group. (2) LS Task group showed more error reduction

among boys, higher mental ability children, and those who were from

lower socioeconomic status. (3) MA Task group reduced more errors

aoli.g boys, lower mental ability children, and children from higher

'ocioeconomic families. (4) MA Task was effective in reducing errors

fur both extremely and moderately impulsive children. (3) No signifi-

car.: increase of response time was found in all four groups. (6) Be-

tween response time and error, the correlation of LS Task group was

significantly decreased, while the correlation of MA Task group was

increased significantly. (7) After training, the two experimental

gJro:ps pcduced equal or fewer errors than reflective group on three out

of twelve it'.i.s of MFF, while before training no impulsive groups pro-

duced fewer errors than reflective group on any items of MFF.

The results were interpreted as confirmation of the assumption

hat restricted language form is related to impulsive tempo. Therefore,

iinolications of the two strategies in terms of their possible contri-

buticns to learning, rmeasurerent, and personality development were dis-

c:-ssed. Also, suggestions f,~; further studies on impulsivity modifi-

cation w.?e proposed by this researcher.



Learning appears t be the result of interaction between the indi-

vidual and his environment. In the learning process, the individuals

cognitive style (here defined as the individual's relatively consistent

mode of conceptual process in dealing with his external environment)

seems to have important effects on perceiving or identifying stimulus

patterns, selecting perceived stimulus input, organizing input infor-

mation, interpreting information, storing organized information, and on

retrieving stored knowledge. A better understanding of the individual's

cognitive style should help us control learning situations and thus pre-

dict the result of lea-ning for the best benefit of the learner.

Numerous efforts, therefore, using different criterion measures

have been made by psychologists to identify or study various cognitive

styles (Kelly, 1955; Gardner, Holzman, Klein, Linton, and Spence, 1959;

Gardner, Jackson, and Messick, 1960; Rokeach, 1960; Witkin, Dyk, Faterson,

Goodenough, and Karp, 1962; Messick and Kogan, 1963; Kagan, Rosman, Day,

Albert, and Phillips, 1964, Schroder, Driver, and Streufert, 1967; Kipnis,

1971). These successes in ide-ntifying cognitive styles have triggered a

large number of studies on the relationship between various cognitive

styles and learning or instruction. Their findings appear: to be Eruit-

ful and may provide appropriate information about strategies for optimal

cognitive development in children.

Statement of the Problem

In terms of effective learning and optimally attainable achieve-

ment at school, undesirably high errors are frustrating, especially

when American educational standards emphasize accuracy and predetermined

correctness. When the cognitive style plays an important role both in

learning and in evaluation (Kagan and others, 1964; Messick, 1968;

Frederick and Klausmeier, 1970), the possibility of being misgraded,

misplaced, or mistreated due to the high incidence of errors on given

evaluative measures could be harmful to personality development.

The purpose of this study was to test the effectiveness of two

strategies designed by this researcher for modifying children's impul-

sivity in terms of error reduction. The two strategies .,ere: (1) en-

riching children's ability to deal with redundancy, and (2) enriching

children's ability to deal with variability. The two strategies were

assumed to have an effect on children's impulsivity in terms of reducing

errors of task performance on Kagan's Matching Familiar Figures (MFF).

Kagan and others (1964) have identified in children's cognitive

process a conceptual tempo which they call "reflection-impulsivity."

in order to measure children's reflection-impulsivity tempo, Kagan de-

vised Matching Familiar Figures (MFF) which is a set of visual analyti-

cdtl ta.sks r. lth relatively high stimulus uncertainty. There are 12

items in the task. Each item includes one standard picture and six

sailor stii7uli, only one of Twhich is identical to the standard. On the

basis of per--fcrm.ance cn iKagan's N!L'F, children whose average response

timCes (latc;,..y) o first se-lection across 12 items were below median

(faster), and total errors (accuracy) sur;.ed up from 12 items were above

median (more errors), for the saimle tested,were classified as impulsive.

Children were classified as reflective whose average response times

were above median (slower), and total errors were below median (fewer

errors) Those children who were either above medians on both variables

or below medians on both variables were classified as nonextreme (Yando

and Kagan, 1970).

Kipnis (1971) used the Impulsive Scale (IS) combined with the

Socialization Scale of the California Psychological Inventory (CPI) to

identify impulsive psychopaths. Each of these three categories comprised

about one-third of the total sample diagnosed as psychopathic. According

to Kipnis, this classificacion method can also be applied to normal pop-


The reflective-impulsive cognitive style appears to be stable with

age and may be generalized across situations for normal children (Kagan,

1965b; Yando and Kagan, 1970; Messer, 1970). However, the generali-

zation could also be applied with discretion to other specific situations

including preschool headstart children (Zucker, 1968) and educable men-

tally retarded children (Gozali, 1969) as well as those who were diag-

nosed psychopathic (Kipnis, 1971).

Results of certain studies indicate that normal children who are

classified as impulsive by using iFF show i"otor hyperactivity (Campbell,

1969; Meichenbaum and Goodman, 1969), and suffer from restlessness and

disitractibil ity (Kagan and others, 1964). Such children also use poor

problem-solvin-g strategies (Drake, 1970), jump at decisions before they

have carefully examined available alternatives (Kagan, 1965a, 1966;

Drake, 1970; Rherts, 1970), pile up errors as task complexity increases

(Rbetts, '970; Yarndc and Kagan, 1970), denirctrate poor verbal ability

(Schwebel, 1966), are usually pessimistic in the face of threatened frus-

tration (Campbell, 1969), and have higher percentages of school failure

(Mcsser, 1970).

In a study of one lower-class situation, Zucker (1968) also found

chaL preschool headstart children, compared with middle-cldss similar

aged children, were more impulsive and indulged in considerable guess-

work. On the other hand, Gozali (1969) observed that the educable men-

tally retarded children who were classified as impulsive tended to em-

ploy a position response set in responding to the test items.

In his character structure study of the impulsive psychopath, Kipnis

(1971) found that the impulsive psychopath experienced less anxiety or

shame and rejected conventional social values. In additiJn to becoming

easily bored, he also rejected advice and suggestions from others.

Besides, he tended to show inadequate psychomotor control, restlessness,

and poor interpersonal c.cping behavior. All of these characteristics

of impulsive psychopath certainly contributed to high percentages of

under-achievement in school. It might also be reasonable to infer that

the impulsive psychopath with some of the above personality character-

istics would have relatively high errors on a visual analytical task

iwth high ,unce Ltainty such as M'.

The studies referred to in the above section demonstrate that per-

sons with the impulsive cognitive style tend to show certain weaknesses

aind behavio,'al correlates which may impede successful and effective

learning, valid measurement, and adequate personality development.

Increasingly, such problems are concerning psychologists and educators

in fields of gene-al education as well as special education. With this

in. mind, then, this study was an attempt to contribute to one of t:hcse

vital fields. Reduction of performance errors through enriching impulsive


children's coin-unication codes was expected to achieve modification of

impulsivity to a significant degree.

Chapter II, Review of Literature, is a discussion of studies related

to the development of assumptions, hypotheses, and the approach to

achieving the purpose of this study.



The successful identification of reflection-impulsivity conceptual

tempo by Kagan and others (1964) has been followed by numerous studies

on the personality characteristics and other correlates of impulsivity.

As described in Chapter I, there are several undesirable behavior pat-

terns fouiid in children who are classified as impulsive either bv Kagan's

lMFF or by Kipnis' combinaticri of Impulsive Scale with Social ization Scale

of CPI. Such findings FaJe triggered interest among psychologists in the

inquiry into the modification of reflection-impulsivity tempo.

Modifiability of Reflection-Impulsivity Tempo

Since Kagan, Pearson, and Uelch (1966) found that impulsive tempo

cou-ld be modified, several researches have been conducted to further

k2.:wldge about modifiability of reflection-impulsivity conceptual tempo.

-'i.-ir:tlyv ist-ucting children to be reflective

In order to Inquire into the modifiability of impulsive tempo,

S;5n. Pearson and Welch (1966) tried instructing impulsive children

to be reflective. The first grade impulsive children were randomly

assigned to one of three groups. One training group (Group I-Id)

sha'rd some interests and crf-ributes with the trainer of the smae sex.

The ot:h"r L'aining group (Croup I-non id) shaced few interests and attri-

bli;-es wi-.h t(he cirainer of the same sex. The control group received no

training at all. For the Croup I-Id, th. subjects ware instructed to

try not to make a mistake on a haptic-visual matching task, a design-

matching task, and an inductive reasoning test. The trainer showed his

appreciation of the reflective way of doing things. For the Group I-non

Id, the trainer merely commented positively cn the subject:' c'hoiccs.

No delay of responses was required.

The training significantly lengthened the impulsive subject's re-

sponse time on PFF for both training groups as compared with the control

group and the matched reflectives. Unfortunately, as the report stated,

the error changes were not much affected by the training although the

direction of changes was in accordance with the expectation.

There were two kinds of manipulations involved in the research:

(1) trying not to make a mistake on the task, and (2) trying to inhibit

impulsive (rapid) answers. The resultsof the study showed that only the

second manipulation had its effect on MFF response time. The lack of

effect on MFF errors frum the first manipulation was not explained spe-


Manipulating reinforcement contingencies upon latency and accuracy

A Skl-ncerian method was introduced by Briggs (1966) to modification

of reflection-impulsivity tempo. He tried to manipulate reinforcement

contingencies upon response speed and errors on geometric figutal match-

to-sample tasks. Both impulsive and reflective subjects were trained to

respond in a way opposite to their own models. For impulsive subjects,

long latency and fewer errors -were positively reinforced by shoJing green

lights; otherwise, red lights were turned on with warning messages. The

two lights were separately operated to manipulate each variable by its

o-,.:n criterion. Similar training was given to reflective subjects except

that they had to respond quickly, or they were urged to guess.

Treatment results showed significant changes of MFF posttest scores

on response time and errors for both reflective and impulsive groups

with training. How well the result could be generalized to other sit-

uations was not ascertained. Modification of reflection-impulsivity

tempo was, however, tentatively p':oved to be possible.

Teacher as model for children

Yando and Kagan (1968) hypothesized that the teacher's conceptual

tempo had its effect on the student's conceptual tempo if both were in

the same classroom for a period of time. Eighty boys and eighty girls

in the first grade were randomly selected from 20 classrooms. Of

the 20 classrooms, 10 were impulsive teachers and the other 10 teachers

were reflective teachers according to the Adult Form of MFF. All had

differential experiences in terms of years taught. After about an eight

month period, all subjects were retested to see if any significant

changes in response time and errors took place.

Results indicated that the child with the highly experienced re-

flective teacher had significantly greater increase in response time

than all other children. Boys in the classroom of the experienced re-

flective teachers significantly increased response times. Girls in the

classrooms of the experienced reflective teachers had only sizable in-

creases in response times. Analysis of the main effect indicated that

the teacher temrpo had significant effect on the child's response time.

There was no significant main or interaction effect on errors, however.

In light of the above findings, then, Yando and Kagan argued that

the capacity to delay a response was more modifiable than was the ability

to perform pereptual discrimination. They further held that delay-

inhibitico vector could be altered without necessarily causing the change

of accuracy. They also suggested specific training in scanning strate-

gies for accuracy.

Directly instructing children to deploy their attention over tasks

Nelson (1968) assumed t1at attention might be related to the child's

response time and errors. He tried to manipulate the child's attention

in observing visual stimuli. Six sets of geometric figures as match-to-

sample task were used for the research. In the training, the trainer

presented the task to the fourth grade subjects. lie worked with the

subject in looking for the right figure which matched the standard.

While seeking a correct match, the subject was told to distribute his

attention over the standard and all available alternatives before making

his decision. The control subject was allowed to work on the same fig-

ures used in the training group, given equal time, and received the same

sorts of verbal reinforcements as the experimental subject.

It was found that there was a significant effect of attention de-

ployment strategy on response time. Analysis of the training effect

showed that the trained subject tended to give more overall observations,

to canvass the array of alternatives more evenly, and to process more

information before making a decision. The accuracy in this iiatching

task was also found affected by the training.

The effect of training was so confined within the training task and

the similar .ask that it could not be transferred to the tiTme and errors

of the WISC Ficture Arrangem:ent subtest. Lack of positive transfer of

training of attention deployment strategy was attributed to pretraining

group differences as well as differences in the format between the train-

ng tsk and WISC Picture Arrangenent subtcst.

Teaching children a scanning strategy of eliminating incorrect stimuli

Albert (1969) hypothesized that both the dalay of response and the

scanning strategy had effects on performance of the visual Tmatching task

such as the MFF. He assigned the second and the third grade impulsive

children into three groups. For the first group, the subject was told

to Lake his time and "think" for awhile before giving his answer to the

matching task. For the second treatment group, the subject was taught

to scan the alternatives by eliminating incorrect stimuli presented.

For the third group, no training was given.

Analysis of the data showed that the training for both groups had

a significantly greater effect on both response time and errors than the

group without training. It was found that subjects who learned the

strategy of eliminating incorrect stimuli had a significantly greater

increase in response time than the first or the third groups.

Concerning generalization of training effect, both Albert and

Nelson agreed that results were not to be generalized to other dis-

crimirnation tasks when those tasks differed from the training task.

Teaching children to learn rules of a model

Ayabe (1969) tried to analyze rules for different conceptual mod.,ls

and asked children 1:o learn the rules opposite to his own corlceptual

tempo. Forty fourth grade children were placed in four different groups:

reflective training group, reflective control group, impulsive training

group, and impulsive control group. For the reflective training group,

subjects were told to memorize the rules of impulsive tempo, to see the

demonstration of the rules by the trainer, and then to apply the rules

practically to the items provided. The impulsive training group followed

Lhe same staps of training for learning the rules of reflective tempo.

The rules learned by the control group were to be neutral. For both

training groups, the corrected the crrors aedd the tr' inccr pr-aised

,wlen subjects learned the rules correctly.

it was found that training in impulsive strategy induc. d L;'p;lsive

tempo for reflective subjects. However, training in reflect.ive strategy

failed to in.lduce reflective ten.po for impulsive subjects. Tl!e impli-

caticous of the research were not very clear in terms of modification of

cognitive styles.

Directly observing a film-mediated model

The possibility of using film-mediated models to modify conceptual

t';mpo w:as tested by Ridberg (1969). Ridberg asked fifty fouro-h grade

impulsive boys to view a film in which a peer was doing his match-to-

sample test in the reflective way. The other Eifty fourth grade re-

flective boys viewed a film in which a peer was doing the sime kind of

match-to-sarple test in the impulsive way. The viewing of the conceptual

models opposite to the subject's conceptual tempo was clarified by

verbalizing various cues of the models. In the film, the testing situ-

ation provided several response possibilities simultaneously. For

testing the effect of film-nediatad model nrg on conceptual te!n'po, all

subjects were tested immediately after viewing the film. They were

also retested a week later to see if the shift of conceptual tempo, if

any, was stable.

The studye i:pnilulsive subjtcl:t did per-

Lorm recflect'iely with significant increase in response timno aod signif.-

icant decrease in errors. Such a change was found stable over a week.

Tre reflective subi&jct,, icwever, show ?d inc r.ass in both crepo'se time

a.1d errors.

Both Ayabe and Ridberg tried to see if they could make the impulsive

child reflective and make the reflective child impulsive by providing

either experience or information of the model opposite to the subject's

conceptual tempo. They worked for the same purposes. But the slightly

different modeling methods yielded results contradictory to each other.

Ayabe made the reflective subject more impulsive but Ridberg did not.

On the contrary, Ridberg made the impulsive child reflective--unlike

Ayabe. It seems to indicate that the subject's reaction to the method

plays an important role in determining the effect of the training. Ayabe's

impulsive subjects might have been impatient with his step-by-step,

mmory-demonstration-practice approach. In Ridberg's stvdy, simply

viewing the impulsive model in the film lacked sufficient appeal to in-

duce the reflective subject to change his way of responding.

Observing patterns of ..odel behavior with associated reinforcement
contingency ies

Situations at home and in school often exist in ways which suggest

that a wide range of modeling is experienced by children. In view of

this, than, Debus (1970) asked third grade impulsive children to observe

sixth grade modelss in different behavior patterns with different rein-

forcenent contingencies. Four patterns were shown to fcur impulsive

groups with ona control grocp. The four patterns and reinforcement

contingencies were as follows: reflective model with positive rein-

forcement consequence, impulsive model -ich less positive reinforcement,

change model from impulsive ,odel to reflective model at halfway with

differing reinforcement, and dual model of two different models with

coni-jasting reinforcement consequences.

Observation of che reflective model showed significant effect for

both boys and girls of impulsive tempo in increasing latency. Change

model and dual model significantly changed girls' latency. Changes were

maintained over two and one-half weeks only for girls who observed change

model. Those subjects who increased latency made significantly fewer

errors than those whose latency was unchanged.

The studies reviewed above seem to contribute mn h to our knowledge

about modifiability of refiection-impulsivity conceptual tempo. The

following points are summarized from these researches: (1) It is not

impossible to modify reflection-impulsivity conceptual tempo. A new

conceptual tempo can be learned. (2) It seems more pertinent and

knowledgeable to modify impulsiviLy than reflection since these research-

ers seem to favor reflective behavior patterns. Impulsivity tends to be

related to some behavioral, patterns which seem to correlate with poor

strategies in learning and testing. (3) Increase in re-ponse time on

the figure-matching task does not necessarily bring about corresponding

decrease in errors. Accuracy is assumed to be related to the scanning

strategy. (4) The majority of those researches have been interested in

tackling this outstanding problem with strictly direct modeling methods.

(5) As far as the training methods and materials are concerned, the in-

tention of the training seems to be directed to "training for criterion

testing." Therefore, little significant effect can be generalized to

cther situations, even discrimination. (6) All samples are within the

range friom first grade to fourth grade. (7) Some controversial effects

of training r ay7 be related tu variations in sampling techniques, control

group situations, manipulation of experimental variables, subjects'

knowledge bout and reaction to the purposes of the research, duration

of training, reinforcement techniques, criterion of training, reaction

to measurements, and in the research design.

Relationship Between Response Time and Error

As indicated above, it is possible to modify impulsivity in children.

Theefcre, it seems reasonable to ask what is to be modified: response

time or errors? The combination of response time and errors for an op-

erational definition of reflection-impulsivity (Kagan, Pearson, and

Welch, 1966a, p. 591) indicates that impulsivity is not simply determined

by response time only. The combination of the two variables, response

time and errors, also avoids ambiguity caused by interchanging "quickness"

with impulsivityy" in the research.

The evidence shows thar sheer manipulation of the impulsive child's

response time has no significant effect on accuracy. The relationship

between response time and errors, furr.hermore, is not direct and simple.

Some other factors such as mental ability, attention, strategy and its

flexibility, interest, value, anxiety, incentive, motivation, other

personality characteristics, complexity and sequency of task presentation

may partly determine the degree of correlation between response time and

error. Yrndo and Kagan (1968) stated that delay-inhibition vector could

be altered without necessarily causing a change of accuracy. They sug-

gested specific training for scanning strategy to accomplish accuracy.

"e omingly, this indicates that the direct association between response

time and crror is rather weak.

In order to account for the differentiation of reflection and im-

pul.sivity, by means of combining response time and errors, Kagan (1966)

iistcd tnhruc possible causes: (1) constitutional predisposition, (2)

involvem.enc in the task, and (3) anxiety over task competence. The last

one, anxiety over task competence, was not supported by a study done by

Ward (1968) who partially replicated Kagan's study. After studying the

subject's eye movements by using Mackworth's eye-marker camera, Drake

(1970) argued that it was scanning strategy, not response time, that was

more powerful in differentiating the impulsive from the reflective.

From the analysis of past researches centered around modification

of conceptual tempo, one may assume that psychologists have been grad-

ually putting more weight on modifying scanning strategy than on in-

hibiting response, Thic may indicate a gradual increase in their con-

cern about accuracy in performance in areas of academic learning and

measurement. This trend has received much attention since evidence

shows that merely holding a response does not contribute much to error


Some Commentary in Regard to Direct Modeling Method

In summarizing past researches on modification of conceptual tempo,

it was also found that the majority of researchers were interested in

tackling this distinct problem by strictly direct modeling methods. For

example, one may ask a child to spend more time on the standard picture

and to keep his eyes on all alternatives before responding. If he does

not do so, he may be viewed as impulsive. The underlying assumption is

that the inpulsive behavioral rule or model brings about short response

time and more errors. On the other hand, the reflective behavioral rule

or model results in long response time and cwe. errors. Under such an

assumption, it should follow that if an impulsive child is taught to

follow the reflective model, he should henceforth act reflectively. This

has not been unequivocally supported by past researches. The problem

stems from the modeling method reversal of the operational definition of

rcflection-inmpulsivity concept.ial tempo.

In terms of sampling, those researches followed Kagan's operational

definition of reflection-impulsivity. That is,a child is classified as

reflective by having response time above median and errors below median

for the sample tested on the MFF. But, in training, those researches

introduced various "supposed' reflective models for the impulsive subject

to imitate. Such models may be oversimplified. Then, too, some rules

of the impulsive model may not be representative of the generalizable

impulsive behavior which can be tested out by a task such as MFF.

Furthermore, these models may have neglected several behavioral

correlates of impulsivity or reflectivity. Impulsivity, for example,

has been found to be statistically related to anxiety, restlessness and

distractibility, problem-solving strategies, verbal ability, acceptance

of traditional social values, and task complexity. The modeling tech-

nique which has been employed by some of past researchers may not take

into account the role of certain behavioral correlates in training and


In conclusion, only after seriously considering the possible causes

or behavioral correlates of the impulsive child's undesirably short re-

sponse time and his unreasonably high errors, tha appropriate method of

modification can be designed. Otherwise, it is not advantageous to in-

duce modification by modeling in order to achieve the testing goal. A

reasonable method should also diminish the subject's deliberately con-

forming behavior which has little to do with cognitive reconstruction.

Consequently, the present attempt to modify the child's impulsivity

emphasizes not only the reduction of errors but also the appropriate

pgroa.-.h to reducing errors. Thus, to accomplish this purpose, t.o

strategies have been developed with the following speculation in mind.

These strategies will be explained in detail in Chapter III.

Language Forms Affecting Cognitive Development

The present study originated from speculations about possible

causes of error-making by impulsive children. Bernstein (1965) stated

that children from the working-class (here defined as all members of a

semi-skilled and unskilled group) were accustomed to a restricted code

of communication which is contingent upon a specific form of social re-

lationship. The restricted code of communication, as Bernstein reported,

is characterized by comparatively short, simple, and incomplete sentences,

poor syntactical organizations, repetitive uses of conjunctions, rigid

and limited uses of adjectives and adverbs. Such individuals experience

relatively few alternatives. They are also restricted in their attempts

to express thenimse!ves in a verbally explicit form. The social relation-

ship which reinforces the use of restricted code of a.-immunication pro.-

duces cognitive process recognizing events as unconnected facts. Within

the social relationship of the working-class, children's cognitive dif-

ferentiation is comparatively less developed.

Bernstein's position which states that a relationship exists between

the use of restricted code of communication and poor development of

cognitive process (Bernstein, 1958) will be used in this study. It

seems reasonable to assume; that this relationship can account to a

marked degree for the high errors rnade by impulsive children on the

visual analytical task.

Based on this position, two assumptions have thus been derived and

formulated: (1) impulsive children are accustomed to reading or using

short sentences in communication, and (2) impulsive children are accustomed

to thinking or handling a ver-y limited number of alternatives in conmuri-


Furthermore, the relationship between thought and language is def-

initely a dynamic process (Vygotsky, 1962). Daily family language .usu-

ally symbolizes the family interaction system. Bernstein (1960) re-

ported that the middle-class families (here defined as family members

-who received high school education, some advanced education, or certif-

icated training in either skilled or non-manual occupations) used more

elaborate forms of language in contrast to certain working-class peo-

ple's less complex language forms. The elaborated form is characterized

by accurate grammatical order, complex sentence constructions, frequent

use of prepositions and impersonal pronouns, relatively wide range of

alternatives, verbal mediation of individual qualification L'nrough the

structure, and the relationship within and between sentences. Bernstein's

position was confirmed by Deutsch's study (1965). He reported that

lower-class was associated v'th poor language functioning. Furthermore,

Hess and Shipman (1968) found that family interaction system determined

the degree of provision of an array of alternatives for thought and ac-

tion. Used as an instrument of thinking (Bruner, 1966), language may

either facilitate or impede the individual's use of alternatives for

thought and action in interaction.

A Possible Cause of Impulsive Children's Making Errors

In impulsive children, an unduly shortened period of circumspection

bef'se maki-ig decisions (Kelly, 1955) may be related to the children's

bein,- accustomed to restricted codes of communication. High errors, in

this sense, may be caused by impulsive children's poor ability and low

desire to ccr.urnicate with reality in elaborated form which is the typi-

cal characteristic of today's learning materials and tests, This can be

interpreted qs a developmental phenomenon of inappropriate match between

the external requirement and the intrinsic ability (Hunt, 1961) of im-

pulsive children. When a restricted code is so familiar and accessible

to them, impulsive children who are confronted with elaborated code of

communication should be able to overcome the temptation of going back to

their habit of using the restricted code, On the other hand, they should

be able and willing to communicate in a seemingly redundant elaborated

code. Apparently, this is a matter of cognitive adaptation.

Pribram (1970) demonstrated his great concern for man's "human

state" in his analysis of the brain's function in the area of information

processing and retrieving. When a man encounters uncertainty, according

to Pribram, he can either seek real information for temporary relief or

reduce uncertainty by enhancing redundancy and varying patterns of a code.

To employ the first strategy of temporary relief by seeking real infor-

mation is very much similar to the way, described by Kelly (1955), used

by the impulsive to resolve confusion. To adopt the second strategy,

man is encoding and varying the patterns of redundancy in which infor-

mation is encoded. Enhancing redundancy and varying the patterns of

redundancy, as Pribram sees it, allows man to use stored information in

unexpected ways. This is the strategy, no matter how it is labeled,

which helps learning to be more meaningful for man.

At this point, it would seem appropriate to review Pribram's position

on the brain function and Bernstein's position in regard co differential

language forms. To facilitate meaningful learning, one should enable the

individual to live with redundancy in which information is coded as well

as to variate patterns of redundancy in order to permit using information

in various unexpected ways. It seems reasonable to infer that the elab-

orated code of conmrunication, compared with the restricted communication

code, is more redundant (long and complete) and variable (offering more


According to the two assumptions aforementioned, the individual,who

is unable and impatient in terms of communicating with reality in an

elaborated code, has no alternatives but to escape the uneasy situation.

Under such circumstances,he tends to seek a quick and temporary solution

without much regard for how many errors he has accumulated. Hence,

errors become inevitable results of impulsive children's performance

when they have to deal with elaborated codes.

Modification of children's impulsivity may thus be accomplished by

enriching impulsive children's cormuunication code. In so doing, chil-

dren may be more able and willing than before to communicate in elabo-

rated code. Once children are able and willing to work with an elabo-

rated code, they can tolerate some redundancy and various patterns of

redundancy which provide a powerful tool for reducing uncertainty.

A New Indirect Method

Reviewing past studies on modification of conceptual tempo has en-

abled this researcher to be more sensitive to the need of developing a

i-ore knowledgeable and workable approach for such modification. It is

a new, indirect method which seems more appropriate than other current

methods for modification of a conceptual tempo.

As modification of reflection-impulsivity has been shown possible

and in progress, it seems reasonable to see the modification in psycho-

logical terms rather than physical or physiological. By this it means

that some psychological strategies can be developed to achieve the goal

of modification. It also suggests that unless either some negative psy.-

cho!og'ical harriers are removed or unlearned, or some positive

psychological vectors are developed, there is little hope to obtain a

long term and generalizable effect of training.

The indirect approach was introduced in this study because of its

strength over direct modeling. The two strategies of this study did

not employ test-like match-to-sample tasks as training materials, n:or

did it follow procedures similar to criterion testing when traiinig.

The rationale for employing the indirect approach is more obvious when

the method not only discourages the trainee's temporary conformance to

the trainer's expectations, but also prevents the training from becoming

too parsimoniously attached to a specific task or criterion measure.

The immediate goals for the two strategies in this study, as a re-

sult, were two-fold in nature: (1) to help impulsive children develop

their ability to use a kind of communication code other than their own,

which exists and prevails around them, and (2) to help impulsive chil-

dren appreciate other cc.rnunication codes aside from their own. At

first glance this has nothing to do with children's performance on any

match-to-sample tasks. But as far as information is involved, the code

in which information is encoded and the cognitive style through which

information is assimilated should be relatively matched in order to

make optimal communication possible. The two strategies were used to

enable children to handle a code which, formerly, they had been unable

to deal with. Hence, if a match-to-sample task is representative of a

red~ndant type of code, those who are unable to make sense from that

code should be taught how to handle it. This exemplified the generic

method representative of the indirect approach used in this study.



An experimental research method was employed in carrying out this

study. The design included hypotheses, instrumentation, sampling,

training and control, data collection, and analyses of data.


To accomplish the purposes of this study, the following hypotheses

were formulated.

H"nothesis 1

Training in writing long sentences has a signifi-ant effect on im-

uulsivity in terms of reducing impulsive children's errors as measured

by Matching Fami.lar Figures.

Hyvothesis 2

Trasinig in making multiple alternative choices has a significant

effect on impulsivity in terms of reducing impulsive children's errors

:s measured by Matching Familiar Figures.

Thae length of a "long" sentence was arbitrarily determined by the

'i'i Lvi.dual subject according to his past ex:perM-nce. But it was antic-

ipated cLat a long sentence would be grammatically complete and rela-

tively complex in strucl.ure. It was also expected that the subjects

would d disco. er more than two alternatives.

Sampling and Sam-ple

Eighty-six boys and 89 girls, ages eight through eleven, in grades 3,

4, 5 of P, K. Yonge Laboratory School at the University of Florida,

Cainesville, Florida, participated in this study. At the ciime of the

pretest, 8. boys and 87 girls were individually administered Kagan's

Matching Familiar Figures. Seven children did not take the test mainly

because o. illness. A graduate student, working for his Specialist de-

gree in Educational Psychology, and this researcher administered MFF.

The major variables scored were the total number cf errors and average

response time (in seconds) to first selection across 12 items in the

MFF. To avoid the possible effects of grade and discrimination learn-

ing experience. medians of response time and errors of eacn grade level

*.-ere calco!latd separately. For the third grade, the medians of response

time and errors were 7.8 and 15. For the fourth grade, they were 9.4

and 11. Fifth grade medians were 3,8 and 9. These data supported this

researcher's prediction that response time would increase with grade,

:ad3 e-croi's- would decrease with grade. The only exception to this pre-

d icion was thl.rt response time did not increase as grade advanced from

Lhe f ~i- ., the fifth.

Aft.': tLhe tChre pairs of medians were calcula'ited, 59 impulsive sub-

j.La:m :'n6 39. rcfiElctive subjects ,ere classified. Impulsive subjects

i ulnuded those \ .o!se average respo,) e times were belo,-: median (faster) t~:.;al. irr or ere above r-edann (mocro errors) fcr each grade tested.

Pefic'iv-. s-ubjects included those whose average response times were

above median (slower) aiJ total errors were below median (fewer errors)

cO uaL 0

m 'n LC)

N cN- rN- -o

N'- N- i- N

0 '-. L O.0
.-4 .-1 -

0 C co < -" c

C0 )> >
> > *.-

e-i if. c' n C t) r-4 ,-1 r 0
P2 C? 3 0- : o. r--

o f c a o 0 *

(.3 *"* 0, s- C ~'-'3 *-'



x ^R
^ ,..j
0) I

O U-

4 (,-4
c. C-
cr 0
c3 0

(G '0

-o -

J -'-4 C
!- C)

4- I
0 C'

0 0

S0 ,
0 'Ud

- o

4- -A f)
) -) --0 4
I( 0

-. Z OX
'a OSJ

t u o c
'* 4J C .

03 0 (
0 - 0 U
:-4 0U

0 c0 1-

i-l- 01
*-0 C' CM
SC '.) 0 '&.

2- *
-'- 2

0 LO O M Lr

for each grade tested.

Fifty-seven impulsive subjects were randomly selected for the ex-

perinent. Ninteen reflective subjects were randomly selected as a

matched control group for later comparison. The completely randomized

block design was applied to eliminate individual differences in error

for the same blocks. All impulsive subjects were Lhen randomly assigned

irto three g-roups, 19 subjects in each, after they had been ramified in-

to blocks according to the number of their errors. The three groups were

randomly selected as two experimental groups and one control group.

Table I si..uiarized the main variables which included sex, grade,

age, response time and error of MFF, score on California Test of Mental

Maturity, and socioeconomic status, of the four groups--two experimental

and two control. Although subjects were assigned into blocks exclusively

by number of errors, grcup differences in response time and other main

vdrlables for three impulsive groups were small and therefore insignif-

icant. The reflective control group, however, deviated significantly

from the three impulsive groups in response time and error because of

the classification method. This reflective control group was selected

tor co;npa;-is-,ns. The table of original data for the four groups is

sh.,o in Appendix A.

The product-rCoiment correlation coefficients between response times

and errors for the four groups were: -0.498, -0.171, -0,470, -0.528.

the strikingly low correlation coefficient between response time and

error for the second group might have its effect on the late training.

The coeff icients of the other three groups, however, did not differ


In:r. trurenta tion

Cply cn, test, Matching Familiar Figures, ,as actually used by thi.

researcher for this study. The lest was given twice, a pretest and a


Matching Familiar Figures (MFF)

This test was developed by Jerome Kagan in 1965 to identify re-

flection-impulsivity conceptual tempo. Construction of the test was

based on the assumption that response time to visual analytical tasks

and the like, with high uncertainty, was a faithful reflection of de-

cision time for solution. The assumption also followed that speed of

making decisionswould affect the probability of producing a successful


The test consists of 12 items including: house, .:cssors, phone,

bear, tree, leaf, cat, dress, giraffe, lamp, boat, and cowboy. Each

item has one stand-rd picture on the top page. The other six pictures

are on the bottom page. Only one of these six pictures is identical to

the standard. Figures 1 and 2 show the standard and six variations of

a sample itcm.

According to testing procedure, the subject is shown the standard

picture and six similar pictures. The standard and the six variations

are always available to the subject. The subject is instructed to point

to a picture on the bottom that is exactly like the standard on the

topn AEter two practices, th tester begr'.s the actual test by cecord-

ing latency to first selections and Lotal errors across all 12 items.

T!he record form of MFF is slown in Appendix B.

Impu. I~s ive subjects are those w-hose average response times fall

below ::edin end total errors abo\va median for the pool being tested.

Reflective subjects are those whose average response tires are .above)

m:adian and total errors below: i-edian. Yanrio an1 l''agan l. abie otherc






-- 4c
-- -
c< \ ^

--< i \ _-~\ h '
c:~""^-^ \-4?

i-r_^? -____ F -\---- --- -) P




{" U'i > .' / .._ F_--- W L_ ----r-J -

/.---n,-<.-^i~.^-- / <^~
/' __ui /J' 1--- I--
F -,^ -- ,-. \ ^ ^ L -i I, r 'A:::^I;^ -) V \ 9?
( ~7 \- y^ *? f^J
~7i\ T^ L \\\'C^^-rr- --^ \ -
/ o ^)^J_.r^^ ^ ^- ~i

subjects as nonextreme.

The stability of MFF response time over ten weeks ranged from .46

to .92 (Yando and Kagan, 1970), over one year averaged .62 (Kagan,

1965a), over two and one-half years ranged from .25 to .43 (Messer,

1970). The stability of MFF errors over ten weeks ranged from .22 to

.87 (Yando and Kagan, 1970), over two and one-half years was .33 (Messer,

1970). The generality across MFF and Haptic Visual Matching (HVM)

ranged from .61 to .87 (Kagan, 1965a).

The correlation between MFF response time and error varied: -.347

(Gozali, 1969), -.51 and -.70 (Zucker, 1968), -.47 to -.66 (Kagan and

others, 1964), and -.46 to -.75 (Kagan, 1965a). All data showed sig-

nifiLant negative correlation between the two variables.

The decrease in stability of MFF response time and errors was neg-

atively related to the increase of time span between two administrations

of IF'-F to school chitdrei (Kagan, 1965a; Yando and Kagan, 1970; Messer,

1970). There were indications that growth and experience in discrimi-

nation learning, among others, affected MFF response time and errors in

teims of stability.

The test has been validated for children ranging from kindergarten

through Fifth grade. It can be used for children with normal mental

ability as well as e'lucable mental retardates (Gozali, 1969).

Experimental Tasks

To test the two hypotheses of this study, two corresponding ex-

pcr.ieni.a tf-sks were designed by this reewe.cher. The two experi-

mental tasks, using long sentences and making multiple alternative

choices, .we-e presented to the subject in the forpi of exercises.


It was assumed that to reinforce -sage of long sentences in train-

ing would increase the subject's subsequent frequency of using long sen-

tences. Therefore, it was expected that the process of becoming grad-

ually accustomed to using long sentences might eventually reduce re-

sistance against the redundancyy' of long sentences in which information

was coded. It was also expected that,with increased ability to deal

with redundancy of long sentences, the subject would assimilate more

information from stimulus situations.

In addition, it was assumed that to reinforce making multiple al-

ternative choices through training would increase the subject's subse-

quent frequency of making multiple alternative choices for a solution.

Therefore, it was expected that the increased ability to make alter-

native choices for solutions could also enhance ability to examine

existing alternatives more efficiently. Obviously, too, the subjects

would essimilate more information through examining more alternatives.

Long SentT ce Task (LS Task)

The task was designed to help impulsive children write a long,

elaborate and complete sentence. The arbitrary length of a long sen-

tence was based on both the grammatical completeness and the subjective

judgment of the subject according to his past experience.

To begin the task, the trainer reads the instruction aloud to the

group of subjects accordingly:

I am going to ask you to do something. We call it
a task. There are two parts in each task. irito a
long sentence fur each part. Write as long a sen-
tence as possible. Use as many words as you can to
make it a good long sentence.

If a sentence is not complete, the trainer should point outl the reason

rfo such gramsmatical incompleteness. If the sentence seems too short,

the trainer should suggest adding some words to make it longer. LS

Tasks vary in form, bu.t they are used for manipulating the same variable--

the formation of a long sentence. Samples of the LS Task are shown in

Appendix C.

Multiple Alternatives Task (MA Task)

The task was designed to help impulsive children make more than two

alternative choices. The main concern of the task is to increase chil-

dren's ability in making multiple alternative choices. In the beginning,

the trainer reads instruction aloud to the group of subjects accordingly:

I am: going to ask you to do something. We call it
a task. There -re two parts in each task. You will
have to change sj.e things. Or you will have to
think out different ways of doing something. I
would like you to put them together as many different
ways as you can.

It was suggested that the subject wr-ite at least three alternative choices

for each part of the task. MA Tasks also vary in form but they are de-

signed for manipulation of the same variable--making multiple altcrna-.

tive choices. Samples of MAu Task are shown in Appendix D.

A pilot task was used in Starke Elementary School, Starke, Florida,

before the task was revised and used in this experiment. This oppor-

tunity was helpful in rew,,ording instructions and adjusting complexity

for all tasks.


After this research proposal was approved by the research coordi-

nator at P. K. Yonge Laboratory School, the subjects and cther facilities

f.r testing were arranged. The pretest of 1F1F lasted six days, with ap-

proximately 30 subjects a day. During the testing period seven subjects

.ere absent, mainly because of illness. IT-mediately following classi-

fication and grouping of subjects, the names of the subjects in the two

experimental groups were sent to both the research section of the school

and the six classroom teachers. Concurrently, training times and other

schedules were settled. The two experimental tasks, LS Task and NA Task,

were randomly assigned to the two experimental groups, along with one

control group and one matched reflective control group. The two control

groups received no treatment.

Twice a week, each experimental group was given the experimental

task designed for that group. Each hand-out was a 8 1/2 x 11 xeroxed

worksheet which included the instruction and task. At assignment took

from 20 minutes to 35 minutes, depending on the subject's familiarity

with the task and the task complexity. For those who were absent from

a training session, an additional session was arranged by the subject,

the teacher, and the trainer.

A male and female graduate student, in the College of Education at

the University of Florida,vere trained to conduct the training sessions

of this study. At first, they read the proposal of this study and clari-

fied their questions, especially regarding the method. Later, they were

given some samples of the experimental task for practice. The two

trainers were then randomly assigned to the two training groups for five

weeks. Responsibilities of the trainers were defined as follows: (1)

reading the in:cruction of the task for the group in the beginning, (2)

distributing task sheets to sibje:ts, (3) reading instructions and ex-

amples on the task sheet, (4) asking subjects to see if they understand

how ta do their tasks, (5) using examples to illustrate instructions,

(6) using verbal reinforcemCnt such as "Very good," 'Yo- did a good jcb,"

"You wrote a good long sentence," and "You made many good ways of doing

things," to praise subjects' achievement, (7) helping subjects in

spelling, (8) urging subjects to write a sentence longer by saying, 'It

is a good sentence, but I know you can write it longer," and (9) check-

ing attendance and collecting finished assignments.

In order to avoid confounding experimental variables with other

nonexperimental variables, the trainers were told to refrain from using

the following admonitions: "Pay attention to...," "You have to follow

instructions carefully," "It is easier to do it like this...," "Think

of it before you write," "It is too short to me," and "Long sentences

are much better than short sentences."

The training took five weeks to finish ten tasks for each experi-

c.ental group. At the end of training, 89 per cent of the subjects in

LS Task group finished their assignments, while 95 per rent of the sub-

jects in MA Task group finished their assignment. The difference could

be attributed partially to the less desirable and inconsistent training

attitudes and techniques employed by the trainer of LS Task group. The

training in MA group, compared with LS Task group, was well conducted so

that many subjects made more alternatives than the space could accommo-

date. Enthusiasm was markedly higher in MA Task group than in LS Task


As the training ended, subjects in all four groups were retested

individually with the same version of MFF. The same testers administered

MFF to all subjects in the four groups.

Analysis of Data

The following data were collected for this study.





California Tert of Mental Matucity score

Socioeconomic index of Occupations

Pretest of MFF response time and errors

Posttest of MFF response time and errors

All data were coded and punched on IBM cards for calculation. The

data of 76 subjects either in experimental groups or in control groups

a:e shown in Appendix A. The main statistical computaLions were: analy-

sis of variance and analysis of covariance appropriate for this completely

randomized block design, association between the experimental variable

and the criterion variable, means and standard deviations for response

time and errors for each group, t test for mean differences, correlations

between response time and errors and between pretest and posttest of MFF,

and mean response times and errors for each item in MFF.

The two hypotheses were tested by using the same statistical methods.

Analyses of variance were followed by Dunnett's multiple comparisons be-

tween each experimental group and control group. (Hercafter, the impul-

sive control group will be labeled as control group, while the matched

reflective conutol grcup will be called reflective control group.) At

least .05 lcvel of significance was acceptable Cor F test and Dunnett's

test. The association between experimental variable and criterion vari-

able was calc lated to see how much variation in criterion measure could

L. accouni.e for by training, All tests between experimental groups and

control groups were one-tailed. All tests between the itwo experimental

groups or between the two control groups were two-tailed.

Further analysis of training effect on response time and errors

in various subgroups was done by t tests of mean differences. Signifi-

cance level of at least .10 was required for comparisons between sub-.

groups. Correlation coefficients of pretest and posttest between re-

sponse time and error were obtained to help understand( variation of

the relationship between response time and error.

Correlation method and other descriptive statistics were used to

help depict the nature of the population and the sample. This would

provide information about generalizability of the findings from this


The Limitation of the Study

The generalizability of the results of this study would be limited

by several factors: the size and the nature of the sample, the quality

and the duration of the training, the task complexity and validity, re-

peated measures and memory, and the control situation.

The sample

The major limitation of this study was the size of tne sample.

The .strategies employed by this study could be tested in either small

or large groups. In small group training, which has the advantage of

clinical observation, the trainer can better deal witl individual dif-

ferences. In large group trai ini- which resembles classroom instruc-

tion, the trainer can gain considerable insight into what a teacher can

no about. modification of a conceptual tempo in classroom teaching.

This research was originally designed for large group training.

It may have been advartageous to have asample of 25 to 30 subjects in

eacn group. Each group of this study consisted of 19 subjects.

The second limitation encountered by this research was the nature

of the sample. The population from which the sample was drawn' was

highly selected. Mean and standard deviation of CTIMM scores of the pop-

ulation were 111.1 and 13.5. Mean and standard deviation of socio-

economic indices were 70.6 and 21.9. Median of parents' occupational

hierarchy was 84. On the Socioeconomic Index, college professor was

ranked 84. Therefore, any generalization of the results of this study

to other populations should take these two factors into account. In re-

lation to this, Zucker (1968) found that correlations between response

time and error were difLerer.t between middle-class children and lower-

class children.

The training

The third limitation of generalizability of this study was the

quantity and quality of training. In LS Task training, the group fin-

ished 89 per cent of assigned tasks. Eleven subjects finished all 10

tasks, one finished 9, two finished 8, four finished 7, end one finished

6. While in MA Task training, the group finished 95 per cent of assigned

tasks. Fourteen subjects finished all 10 tasks, two finished 9, one

finished 8. and two finished 7. In addition, the trainers judgment cn

quality of training in terms of criterion might vary from person to per-

son. In this study, no instrument was designed and used to objectively

rmersure the subject's progre-s in ability to use long sentences or to

-make alternative choices.

The fourth irrmitetion came from i-he duration of training. The'; of five weeks sc-med not long enough to enable the subject to

adjust himself at ease with a communication code opposite to his ownm.

Besides, the modification of a cognitive style requires new coordination

and cooperation among cognitive factors which have been "harmoniously"

relating to that cognitive style. Therefore, the optimal length of

training time depends on the criterion and progress of the training.

The task

The fifth limitation of this study was the complexity of the task

designed by this researcher. Since general mental ability of this sample

was slightly higher than that of general population, the experimental

task for this sample might not fit a sample drawn from another population.

The sixth, a crucial, limitation to generalization, was the validity

of the task. The task could become invalid if it turned out to be an

exercise for remedial writing. LS Task, for example, emphasized the

length or redundancy of the sentence. But the remedial writing seemed

to work for acceptability or grammatical correctness of the sentence.

Similarly, MA Task stressed variation of something rather than prompting

ideas for "creativity" training. Fortunately, this research was effect-

ive in accomplishing the validity of the task in this instance. It is

hoped that: the result can be generalized across other situations.

The measurement

Kagan's Matching Fa-iliar Figures was the only crite-oion measure

of training effect. The two administrations, protest and posttest, of

lFF were spaced over seven weeks. It was not clear that if the "warm-

up" of pretest would have significant effect on MFF errors (RhetCts, 1970)

by favoring imipulsive children. Star.[c ically, no significant "'warm-up"

w:as found between the two control groups. The tiime in letw.een the two

administrations of the same test or the duplicate should be considered

before making any generalization. Since all subjects of this study took

the same version of MT'F, memory effect would be relatively similar for

each group.

The control

The last limitation of generalization was the control of nontreat-

ment groups. In this study, some subjects' reactions to and knowledge

about the training were "picked up" by some nontreatment subjects be-

cause of their curiosity. Some children who were not assigned to train-

ing groups did show their interest in training through different ways.

However, generally speaking, children in P. K. Yonge Laboratory

School seemed accustomed to being observed, being called for a test or

training, or not being called for a test or training. This greatly re-

duced any unnecessary psychological differences whicL were assumed to be

equivalent between all groups.

In conclusion, generalizability of the training effect of this

study was not tested against other task situations except MFF. Since

the assumptions of the two strategies employed in this study were generic

inr nature, it was expected that the training effect could be generalized

to some other task situations with moderate complexity and high uncer-




Results of this study were categorized into two parts: (1) hypoth-

esis testing, and (2) other related findings. The first category was

based strictly on error changes measured by MFF. In this part, two hy-

pothesis testing were discussed separately, although in all cases data

were shown in the same tables. The second category dealt with changes

of response time and other related findings.

Results of Hypothesis Testings

Means and standard deviations of both the posttest and the pretest-

postcest differences of MFF error were summarized in Table 2. The data

appeatring in the table provided a general picture of the training result

in tc~ms of error change.

tiypolthesis 1

Training in writing long sentences has a significant effect on im-

pulsivitv in term of reducing impulsive children's errors as measured

by M) %arching Familiar Figures.

The hypothesis was supported by this study at .05 (one-tailed) level

of si;inlficance. At first, ore-way analysis of variance in error was

'~ide for experimental groups and control group with completely randomized

block design. Data in Table 3 sho-:ed an F of 5.;90 (p<.01) which indi-

cated that at: least one significant variance could be found. Further

.xa:mination of the data in Table 4 by applying Dunnett's multiple



LS Task NA Task Control (reflective)
Mean S.D. Mean S.D. Mean S.D. Mean S.D.

Posttest 10.53 5.55 9.21 3.26 13.42 5.49 6.11 3.33

Pro- post-
test diff. -5.22 4.58 -6.53 3.32 -2.53 5.56 0.21 2.92



Source SS df MS F p

Treactmi en ts




Residual 527.405

2 78.965

18 33.624

36 14.650

Total 1290.562 56

F.01;2,36 = 5,
F.0532,36 = 3,28
F.0QZ18 36- 1-91



p <.01

p <.05

--------- --



(By Dunnett's Method)

LS Task

MA Task

LS Task

MA Task




*p <.05, one-tailed, d' 2.40



LS Task MA Task Control (reflective)

LS Task -1.31 2.69 5.43**

MA Task 4.00* 6.74**


*p <.01, one-tailed
**p<.005, one-tailed
IBy using simple t test

~I_ _~

_ ______ _~ _~ ~ILI_

_ ______________~____~____

comparisons (Kirk, 1968) indicated that LS Task group significantly

(p< .05) reduced MFF errors as compared with impulsive control group.

The mean difference between the two groups was 2.69. It exceeded the

critical value of 2.40 to be declared significant at .05 by the one-tailed


For all three impulsive groups, variances of pretest-posttest dif-

ferences in error, as shown in Table 2. did not violate the assumption

of homogeneity of variance as tested by Cochran's test (Kirk, 1968).

But violation would take place (C = .4327, p <.05) if reflective control

group was added to the three groups. In multiple comparisons between

experimental groups and reflective control group, pooled standard errors

of mean differences were used instead of using mean of square residual.

In Table 5, mean difference of errors between LS Task group and reflective

control group was 5.43 (t = 4.3579) which exceeded the critical value of

significance at .005.

The statistical analysis further looked into the training effect of

LS Task on errors ii various subgroups, Tables 6, 7, 8, 9, 10, and 11

in Appendix E show means and standard deviations of error changes in

different subgroups. Data in these tables were sumnari:ed in Table 12

for significance testing. Cochran and Cox's t statistic (Cochran and

Cox, 1957) was employed to overcome the effect of heterc .-i-:city of

variance between compared groups.

From data in Table 12, it appeared that LS Task training, in its

fffecr ,n reducing impulsive child rer 's errors, favored (p <.1b) 1ouys,

helped (p (.10) children whose ,ental ability was above Q3, benP'fitcd

(p <.0 1) those who were ten years old, and enabled (p<.10) children

froli low: r socioeccnroic status faT lies, as ccrpared to nontreatrLennt




(Summarized from Tables 6, 7, 8, 9, 10, 11 in Appendix E)

LS Task LS Task MA Task
vs. vs. vs.
Subgroups MA Task Control Control

Boy -1.21 3.46* 4.67**

Girl -1.47 1.40 2.8/

Grade 3 -2.80 1.20 4,60

Grade 4 -0.14 2.14 2.28

Grade 5 -1.43 3.00 5.43**

Age 8 -5.50 2.50 8.00

Age 9 -1.67 0.93 2.60

Age 10 2.53 5.13*** 2.60

Age 11 -3.17 2.50 5.67

IQ (CTMfl) above median 0.85 2.63 1.78

IQ (CTMM) below median -2.68 3.50 6.18*

IQ (CTMMI above 03 2.25 5.92* 3.17

IQ (CTMM) below Qi -4,33 2.25 6.58*

SocioeconQiic Fnlde, above median -2.99 2.93 5.92*

Socioeconomic Index below median 0.69 3.29* 2 .60

*p <.10, one-tail3d
**p <.025, one-tailed
;p <.01, one-tailed

situation. Although all other data showed positive training effect of

LS Task, significant level at .10 was not reached.

To sum up, training in using long sentence; had significant effect

(p <.05) on impulsivity in terms of reducing impulsive children's errors

as measured by Matching Familiar Figures. The effect was more apparent

(p -.005) when compared with reflective control group. It was also

found that impulsive children, who were boys, or with high mental ability,

or ten years old, or from lower socioeconomic families, were significantly

benefited from training with Long Sentence Task.

Hypo thesis 2

Training iL inaking multiple alternative choices has a significant

effect on impulsivity in terms of reducing impulsive children's errors

as measured by Matching Familiar Figures.

The hypothesis was supported by this study at .01 (one-tailed tes.)

level of significance. Data in Table 2 showed that the group having MA

Task training performed better than LS Task group in terms of accuracy.

Therefore, by one-way analysis of variance and Dunnett's multiple com-

parisons, data in Table 4 evidenced that training with A Task signifi-

cantly (p 4.01) reduced .FF errors as compared with control eloup. The

mean difference between the two groups was 4.00. It apparently exceeded

the critical v.;lue of 3.27 to be declared as significant at .01 by one-

tailed test.

A pooled ;tcqndard error of mean difference was used to compare the

error change of 1A Task group with reflective control group. Mean dif-

ference of the two groups was 6.74 in Table 5. The difference yielded

a t of 6.6456 which far exceeded the critical t value of significance

at .005 by one-tailed test.

Training effects on various subgroups were also examined through

comparing eiror changes between subgroups. Means and standard deviations

of error changes of various subgroups within MA Task group were com-

pared with means and standard deviations of corresponding subgroups with-

in control group. To overcome the effect of violation of homogeneous

variance assumption, Cochran and Cox's t statistic was again applied

to test significance. Data in Tables 6, 7, 8, 9, 10, and 11 were also

suTmmarized in Table 12.

From Table 12, it was evidenced that for impulsive children MA Task

training as compared with nontreatment situation engendered higher ca-

pacity to reduce errors for boys (p<.025), for fifth graders (p
for those whose CT{MM scores either below median (p<.10) or below Q1

(p-.10), and for those from higher socioeconomic status (p<.10).

All data demonstrated nos-tive effect of MA Task training on reducing

impulsive children's errors although some data did not reach the re-

quired significant level.

In sum, training in making multiple alternatives had significant

effect (p<.0l) on impulsivity in terms of reducing impulsive children's

errors as measured by Matching Familiar Figures. The effect appeared

to be more outstanding when compared with reflective control group.

Analysis of data also shc .ed that MA Task had significantly greater

effect on reducing impulsive children's errors among boys, fifi-h graders,

lower mental ability children, and children from high socioeconomic

families. Training effects of LA Task on all other subgroups were posi-

tive even though they failed to reach the required level of significance.

i' uilarity and difference between the two hypochffscs

The cwo hypotheses tried to suggest :to di erent :meanrs to arco' plisli

the same goal-to modify impulsivity in terms of reducing impulsive

children's errors. It was evidenced that the two tasks commonly pos-

sessed the same power to help achieve accuracy for subjects aged 8

through 11. To test the combined effect of LS Task training and MA Task

training, the two error changes (-5.22 and -6.53) were averaged and com-

pared with control group error change (-2.53). The standard error of

mean difference for the comparison was 1.52. A t was obtained by divid-

ing the mean difference of 3.34 by 1.52. The result, 2.197, exceeded

the critical t value of 2.015 to be significant at .025 by one-tailed

test. However, the difference between the two task groups was not sig-

nificant at .05 by two-tailed test.

Nonsignificant diffe:.e,.ce between the two training effects did not

provide much information about the relationship between the two tasks.

Tn Table 12, 11 out of 15 subgroups (approximately 73 per cent) showed more

improvement in error changes under HA Task training than under LS Task

training. Since none of differences were significant at the required

level, direct comparison between two task groups was by no means con-


However, if comparisons between the two tasks were done by separately

comparing LS Task group and MA Task group with control group, then the

effect of the two tasks could be differentiated. In Table 12, though

both tasks were effective for boys, MA Task showed more powerful than

LS Tas k in redcing errors in fifth grade children. But the direction

reversed when- LS Task showed its strength in training ten year old chil-


Another differentiable effect of the two tasks was demonstrated in

that high IQ impulsive children did better in accuracy through LS Task

training which was heavily related to verbal expression. On the other

hand, low IQ impulsive children tended to improve their accuracy through

MA Task training which was involved more in task variability.

The two tasks also functioned differently when each group was di-

vided into high and low socioeconomic subgroups. LS Task was found to

be related to achieving accuracy for low socioeconomic children, while

MA Task seemed to be associated with better error reduction for high

socioeconomic group.

To summarize, LS Task and MA Task had a similar function in reduc-

ing impulsive children's errors measured by 1MF. However, there was no

significant difference between the two tasks in degrees of error reduction.

Further examination showed that LS Task effect could be differentiated

from MA Task effect if separately compared with the control situation.

For error reaction, LA Task seemed more effective for i:gh IQ and low

socioeconomic status impulsive children, while MA Task tended to work

better for low IQ and high socioeconomic impulsive children.

Other Related Findings

Other findings related to testing the effectiveness of the two

strategies for modifying impulsivity were: (1) analysis of covariance,

(2) changes of response time, (3) changes of relationship between error

and response time. (4) association between the experimental variables

and the criterion variable, and (5) analysis of variance with selected


Analycis of covariance

It was considered that factors other than experimental variables

might confound the result of this study. From correlational analysis.



Source yy xy xx df SS MS F P

Treatments 157.929 302.605 581.625 2 173.102 86.551


5.915 p <.01

605.227 -205.200 3192.000 18

Residual 527.405 -334.621 7317.313 35 512.103 14.632

Total 1290.562

F.01-2,35 = 5.29

-237.035 11090.930 55



LS Task MA Task Control (reflective)
Mean S.D. Mean S.D. Mean S.D. Mean S.D.

Posttest 7,87 2.93 9.35 4.28 7.85 3.32 15.17 5.08

Pre- post-
Ltst diff. 2.31 2.83 3.C3 3.66 2.12 3.03 -0.94 5.71

- ----~-- ~ ---- ---------

--- --- ------~

the pretest of :MF error was negatively correlated with CTMM scores

(r = -.2413) for all 168 subjects tested. The correlation was signifi-

cant at .01. From analysis of task training effect on subgroups, LS

Task seemed to be more suitable for high IQ subjects, while MA Task

seemed to be more helpful to low IQ subjects. In order to see if any

differences io C TMM score beLween groups would have effect on treatment

effect, analysis of covariance was made.

Data sumr:arized in Table 13 evidenced that analysis of covariance

with an F ratio of 5.915 did not significantly change the result

through analysis of variance. Both were significant at .01, although

adjusted means of squares yielded a slightly higher F ratio. This sug-

gested that slight IQ differences among three groups did not affect

training effect significantly. The nonsignificant IQ differences among

three group was confirmed by checking F ratio of means of squares of

CTMI scores. The F ratio of 1.391 was below the critical value of 3.27

to be significant at .05.

Change of response time

Would change of error be accompanied by change of response time?

To answer the question, analysis of variance of response time was done

in this study. To begin with, means and standard deviations of posttest

and means and standard deviations of pretest-posttest differences of re-

sponse time ace summarized in Table 14. The result of analysis of vari-

ance of response time is Lhowni in Table 15.

The result (F 1.362) was not significant at .05. it seemed to

indicate that significant change of error did not necessarily bring about

significant change of response time. Therefore, reducing errors by

applying two strategies designed for this study supported an asumnpticn



Source SS df MS F p

Treatments 33.196 2 16.598 1.362 NS

Blocks 111.696 18 6.206 0.509 NS

Residual 438.588 36 12.183

Total 583.480 56

F.05;2,36 = 3.28

that to be reflective (lon-; latency and high accuracy) was not the only

alternative for modifying impulsivity. It also suggested that change in

accuracy did not necessarily induce change in response time.

Change of relationship be teen response time and error

LFF protest correlation coefficients Lbtween response time and

error for three impulsive groups were: -.498, -.171, and -.470 respect-

ively. They were lower than over-all correlation of -.534 from 168 sub-

jects. The second group, later assigned as MA Task group, showed sig-

nificantly low correlation between the two variables. Incidentally, this

group had slightly lower mean CTMI score and lower mean socioeconomic

status as compared with other two groups. The fi .st group, LS Task

group, had highest correlation of response and error, with both CTNM

and socioeccnomic status in between the second and the third groups.

Though somewhat confusing, an interesting phenomenon occurred, per-

haps, by training. The.posttest correlation coefficients between re-

sponse ti:e and error for all groups turned out to be: -.162, -.522,

and -.437. 'When control group remained stable, LS Task group and MA

Task group pretest correlation coefficients reversed direction remark-

ably, LS Task group reduced the correlation by .336, while HA Task

group increased the correlation by .351.

Association between experimental and criterion variables

Significant treatment effect of the two experimental tasks tested

by F ratio did not provide information concerning the size of its effect.

Indication of significant traiining effect of this study made further

exploration of association between the experimental variable and the

criterion variable necessary. By using the formula for calculating

association (Kirk, 1968, p. 134) W2 was found .10. The result showed

thEa task training was estimated to account for 10 per cent of variance

in error rate change. What caused the other 90 per cent of change was

unknown, statistically. This might be partially due to the character-

istics of indirect modification methods.

Analysis of variance for selected extremely impulsive subjects

Although the two experimental tasks were designed primarily for im-

puisive children to improve accuracy, it would be interesting to learn

how they worked when ''seriousness" of impulsivity varied. Subjects in

e-ch group were classified again into i.oderate impulsives and extreme

impulsves by combining response time and errors. Seven subjects in

each grcup xere classified as extremely impulsive, leaving 12 subjects

in each group as moderately impulsive.



LS Task MA Task Control
Mean S.D. Mean S.D. Mean S.D.

impulsive -6.86 5.40 -7.86 3.24 -4.86 3.62

impulsive -4.25 3.96 -5.75 3.25 1.17 6.16

Table 16 shows means and standard deviations of error changes by

extremely impulsive subjects. Simple t tests were used 0o compare LS

Task and MA Task groups with control group. MA Task tLaining had sig-

nificant effect (t = 1.976, p<.05, one-tailed) on error reduction as

compared with control group. Other differences were not significant.

For the moderately impulsive category, data in Table 16 were also tested

by t statistics. The result also favored MA Task significantly

(Q. = 2.278, p <.025, one-tailed) as compared with control group.

In general, extremely impulsive subjects tended to reduce more

errors on AIFr in both training groups and control groups than did mod-

erately impulsive subjects.

To conclude: (1) analysis of covariance did not show significant

group differences of IQ and their effect on training, (2) there was no

significant change in response time when errors did change significantly,


(3) different task training did alter the relationship betwPeen response

time and error, (4) the training was estimated to account for error

change by about 10 per cent, and (5) only MA Task training could have a

discernible effect on both extremely and moderately impulsive children

in achieving accuracy,



Results obtained in this study provided clear evidence to support

the hypotheses. Training in both writing long sentences and making

multiple alternatives was effective in reducing impulsive children's

errors as measured by Kagan's Matching Familiar Figures.

The two underlying assumptions proved to be sound. The indirect

approach to modifying impulsivity in terms of error was demonstrated

useful. By enhancing redundancy of a code and by enlarging variations

of codes, impulsive children learned to be more able and willing than

before to deal with reality in different codes. Although training tasks

in this study bore little resemblance to figural "match-to-sample" tasks

in terms of format and procedure, training effect provided generaliz-

ability in reducing error.

Language and Forming of a Cognitive Style

Tn the beginning section of this paper, cognitive style is defined

as the individual's relatively consistent mode of conceptual process in

dealing with his external reality. Furthermore, it seems reasonable

that no cognitive style is formed simply by a single factor. Formation

of a cognitive style thus is consistently operating in ea.h individual's

growth and development, environmental structure and dynamics, as well

as in international tools between the individual and his environment.

Once a style is for'Ted and proved to be workable and economical, it

will be -a'.rta'ned and ez'-anced by the individual thrc(igh his u-eferred


surroundings. Therefore reflective-impulsive tempo, without exception,

is formed, maintained,.and enhanced by several factors in a relatively

consistent way. Language is one factor consistently influencing re-

flective-impulsive tempo.

Relationship Between Restricted Language and Impulsivity

Language has been found to be a powerful tool for mental "operation,"

Bernstein's position (1958) that a relationship exists between the com-

munication code and cognitive development was considered for this study.

The relationship was extended by this researcher to assume that re-

stricted communication code is related to formation of impulsive tempo

which was identified by Kagan and others (1964). Short sentences and

limited alternatives were assured to be two major characteristics of a

restricted language form used by certain people. This researcher fur-

ther assumed, in light of certain people's using restricted language,

that impulsive children are accustomed to using short sentences and

handling very limited alternatives in communication. In other words,

impulsive children were relatively able and willing to communicate in a

less redundant and less variable form of language.

This assumption was confirmed by empirical observation in this

study. Subjects in both training groups did demonstrate their poor

ability and i;.patience in using long sentences and making alternatives

in the beginning. Boys, especially, made several efforts before making

a long sentence for the first few tasks. Certaai girls even put a few

short sentences together by using hands and called them a long sentence.

Bcth boys and girls experienced uneasiness when they were first; asked

to coimnunicaLe in a code different from that with which they were fa-

mi I iar.

Cognitive Mismatch by Mismatched Codes of Communication

The reason for the uneasiness described above seemed very clear. A

once "workable" and "economic" restricted code of communication cold

become "unworkable" and "uneconomic" if the situation changed. A learn-

ing, should it require or actually include an elaborated communication

code, may create a gap between information to be acquired and the cog-

nitive style which handles information assimilation. When information

is encoded in visually or auditorily redundant and variable ways, the

impulsive child is unable and consequently unwilling to information

acquisitable. Therefore, mismatch between a language form and the

child's own language form may result in mismatch between a cognitive

style which is called for by that language form and the child's own cog-

nitive style which is raLher akin to his language form.

When this is true, a child can show disgust at the appearance of

another form of language by avoidance or "living with it" without know-

ing what it is communicating. Some children would rather shun any books

in which sentences look redundant. Others may give answers without

understanding the questions

If an impulsive child is asked to act when encountering a language

form different from his own, he usually tries to "quit" the uneasy sit-

untLon by quickly doing something. He may guess or cheat if possible.

The correct answer is the one wh'ch signals that he is "allowed" to

':leave' that situation. When guessing, there is very little real conmmuni-

cation between che individual and the situation.

Such a'"guess-aA-d-see" response pattern was frequently deiu.onstrated

in this study, taking place in testing of MFF with impulsive children.

Subjects quickly chose an answer and glanced at the tester almost si-

multaneously. If the tester's feedback was positive, subjects would

usually guess again at the next item. If the first selection was indi-

cated incorrect, subjects tried to shift to the other answer as quickly

as possible. Inevitably errors were then accumulated progressively. In

this case, they "learned" the correct answer by being negatively rein-

forced. (The correct answer "took them away" from the uneasy situation

caused by the mismatch.) The answer was correct not because it should

be correct but because it signaled "passing."

Though frequent as they were, this was not the whole story of

impulsive children's response patterns in testing of MFF. For some im-

pulsive subjects, directly after the first unsuccessful pick, there was

a brief moment of hesitation'. Their fingers undecidedly hovering over

some other variations. Anxiety might partially account for that short

delay. But observation with scrutiny seemed to indicate that intrusion

of information took place before his second trial. Information, in-

volving either commonalities or differences between the standard and

variations, though mostly incomplete and inadequate for a solution,

intruded into subjects' guessing behavior. Subjects were tempted to

use intruded information because it was less redundant and less variable.

Ward (1968) found that impulsive subjects did appear more "reflective"

if the firsc choice failed.

It was also found that some impulsive subjects had little intention

to guess. But when confronted with redundancy and variations, these

subjects demonstrated little interest (or ability) in getting into

"details" oi the test ite-i.

They would neither concentrate on one single picture nor collect

elem:iets from all pictures. The only alternative they had seemed to be

accepting certain dominant features of the standard at first glance, and

using theii as criteria for scanning. They looked back at standards

for memory confirmation rather than seeking more information. Impulsive

subjects (they might be moderately impulsive) liked to have and to work

with a limited amount of information.

Consequences of mismatch in man-to-man communication were discussed

by Bernstein (1965) in an article on social class, speech systems and

psychotherapy. Bernstein reasoned that lack of clarity and presence of

ambiguity between people using different forms of language would likely

raise levels of tension for those who used restricted code. Those

tensions would be dissipated "quickly" through some "immediate channels"

other than verbal control.

Enriching Rather than Correcting

Psychological or educational intervention seems necessary to help

solve mismatches. Otherwise, communication of information will be

greatly handicapped by the existence of a gap created by mismatches.

In this respect, this researcher agrees with the suggestion made by

Cazden (1968) regarding language development that we forger about cor-

rv.ticin and concentrate on "adding," "enlarging," and "refining." With

regard to modification of cognitive styles, this researcher adopted the

idea that we should help children acquire and use alternative approaches

: .:ah'r -han restrict ourselves to specific approach. This idea was

also proposed by Frcderick and Klausmeier (1970).

Enriching Through Hypothesis Testing

The concept of enriching children's ability to encode information

into different forms was integrated into the two hypotheses of this

study. The first hypothesis represented a strategy to enrich impulsive

children's ability to communicate in seemingly redundant forms by in-

structing them to use long sentences. The second hypothesis represented

another strategy to enrich impulsive children's ability to communicate

in various ways by instructing them to find multiple alternatives.

The two hypotheses were substantially supported by s atistical

analyses of data collected from this experimental study with children

ages eight through eleven. In other words, impulsive children with

training in either using long sentences or making multiple alternatives

did cor municate better with visually redundant and various match-to-

sample tasks such as MFF. Furthermore, one may infer that significant

error reduction could be accounted for by increases in information as-

similation through better communication between subjects and MFF.

Therefore, the gap existing between subjects and MFF could be diminished

by the training. In addition, noticeable increases in response time

on MHFF could be partially interpreted as increased interest in coimmuni-


Results of this study, in tei-ms of an individual's ability to code

variability, were consistent with Munsinger and Kessen's findings (1964).

Preference for high variable stimulations was related to the individual's

ability to code variability. 1T other words, an individual will code

variable ferns of informatio-i only when he is able to do so. They further

pointed out that such ability could be changed by experience with vari-

able stimiulations or special training.

Evaluation of the Experimental Task

The success of task training in terms of its effect on error re-

duction for impulsive children should be attributed much to the contri-

bution of the experimental task. First, the task provided children with

consistent actual experience which was necessary for enriching a com-

munication code. There is no doubt that learning a specific communi-

cation code requires opportunity for repetition.

Second, the task was varied in form to enlarge the variety of stim-

ulat ons. Both LS Task ad ')A Task variations not only broadened chil-

dren's experience in communication but enhanced children's interest in

communication. En this respect, Fiske and Maddi (1961) found that varied

stimulation was related to enhancing subjects' attention arousal.

Third, the task served as stimulation for children to create their

own forms of redundancy and variations. It was believed that starting

from one's own experience, the training could be better integrated in

children's cognitive structures. The task avoided imposing any uniform

sentences or alternatives on children to learn or to imitate.

Finally, change of performance caused by the task training in this

study weas the best estimate of treatment effect. In the task training,

no comment, contrasts, valuation or devaluation of the two criterion

variables--response time and errors--were made. Therefore, the relationship

betvaocn the variation of the experimental variable and the change on

the criterion variables became highly interpretable without nuisance

effect. In other words, the task training induced error reduction while

it kept the criterion measure highly vdlid and sensitive. "'is

experimental task and task training demonstrated its superiority to

modeling in terms of explaining treatment effect. In modeling, the cri-

terion measure of the posttest becomes less valid or insensitive because

the variable involved in the criterion measure is already distorted or

desensitized by the training. Examples of desensitizing criterion mea-

sures can be seen in the original reports of those researches on modi-

fication of reflective-impulsive conceptual tempo using modeling tech-


Analysis of Training Effect on Each MFF Item

Significant effect of task training on error reduction for children

with impulsive tempo stimjaited further interest in analyzing changes of

errors on different items.

Figure 2 shows the average errors on each item of MFF pretest. The

average errors on each item of MIF posttest are given in Figure 3. On

the pretest, three impulsive groups were very similar in error-making

throughout 12 itens. They all encountered difficulties on items 1, 4,

7, 9, and 12. Similar to impulsive groups in trend, the reflective con-

trol group made relatively high errors on items 1, 3, 7, 12. Both im-

pulsive groups and reflective group "enjoyed" low errors on items 2, 5,

6, and 10. However, no single impulsive group made fewer errors than

reflective group on all 12 items.

Interestingly enough, the trend of MFF posttest for all four groups

changed very. little, But horizontally, error differences between three

impulsive groups increased, while error differences between training

groups and reflective group decreased. In other words, the task train-

uig iade three impulsive groups more discernible while it bought the

three groups closer to reflective groups. Among three impulsive groups,

O C. 3 o `,".: "4,

C i, ** -



S0 u ,- \-
-3 TJ *C 1^ I N.

^ cS ***. 0 I

0 0 E

('4 0
"'^ ^: *
< *: .

e*- ,- ^ H

101-13 aSc;a,~V

(I X C"D "

4j C) '--
- '"u- Li 4 1 -4
F .-i ... *- I

1.- Cj '- -
2 *, *

"; / bC

u \ riI

0 "i o il 0 O
2 ~ '

.o0j1t o .ua,,AV

LS Group did equally or better than control group on 9 out of 12 items,

v.while Ml, Group did equally or better than control group on 11 out of 12

items, Between the two experimental groups, il\ Group showed superior to

LS Group on 6 items while LA Group did better than IMA Group on 3 items.

Such results su -;:sted that the task training caused general improvement

(or overall error reduction) as well as specific improvement on sume


Surprisingly, in terms of accuracy on MFF posttest, LS Group did

equally or better than reflective group on items 3, 5, and 9. At the

same time, MA Group exceeded reflective group on items 5 and 9. These

three items, 5, 3, and 9, represented low to high task complexities.

This could be interpreted to mean that enrichment of ability to deal with

redundant and variable codes of conmunication also strengthened chil-

dren's ability to deal with less redundant and less variable codes.

Training Effect on Relationship Between Response Time and Error

In Chapter TV, data showed that after training the relationship

between response time and error for the two experimental groups changed

in different directions. LS Task training "pulled down" the correlation

coefficient from -,A98 to -.162, while MA Task training " up" the

correlation coeCficient from -.171 to -.522.

Thu unuxpN!-ected changes of correlation between the two variables

might be related to the two different task training when control group

corr.l-:atlo o:: rc:rse only slightly.

Since ave'rge response times and errors and average changes in re-

sponse times and errors did not provide any predictable rule for the

rcbaige of relationship between the two variables, a tentative interpre-

tation was proposed by this researcher. it was found that the two task

training had differentiable effects on test performance in terms of

response time and error. As data shown in Table 12, LS Task training

had more effect on error reduction for the subject with higher mental

ability or lower socioeconomic status. According to Zucker (1968),

correlation between latency and accuracy for: lower-class children's

task performance on MFF dropped from -.63 to -.49, when intelligence

.Jas partialed out. LS Task training in this study, therefore, might

function to "wash out" intelligence effect on MFF and "pull down" the


On the other hand, data in 'able 12 showed that MA Task training

.was more effective in error reduction for the subject of low mental

ability or high socioeconomic status. In the same study, Zucker found

that the middle-class subject showed almost no correlation between in-

telligence and response time or error. Therefore, MA Task training in

this study seemed to depend little on intelligence. The training, how-

ever, might enhance a certain scanning strategy which took longer latency

and produced lower errors, and the correlation would increase accord-


Relative Efficiency of the Experimental Design

Interpretation of research results has much to do with the ade-

quacy of the design of the study, especially an experimental study.

The adequacy of an experimental design can be sho,.-n by calculating the

relative efficiency of that design. For th's study, the relative ef-

ficirency of the completely randomized block design was the ratio of

mean of square within group (20.225) and mean of square of residual

(14.650). The ratio is 1,38 which is greater than 1.00, This ratio

suggested that the block design employed in this tu.idy demonstat-ed


its aJ' -.rl-,-.e in eliminating individual differences appearing in the

same b ocks i;- ter:s of error ch o ge on .

Such advantage was evidenced by a significant F ratio of block effect

shown in Table 3. The F ratio of the block effect (2.295) exceeded the

critical value of 1.91 to be declared significant at .05. If the block

design had not been used in this study, the block effect would then be

added to the error term. This would make the statistical analysis less

sensitive in exploring treatment effects.


This experimental study provided an opportunity to test the hypoth-

esis that training in using long sentences and making alternatives had

effects on error reduction for children with impulsive tempo. The re-

suit significantly supported the hypothesis.

Significant error reduction induced by the two tisk training was

mainly interpreted from results of children's enriched ability to deal

with redundancy and variability. The research demonstrated chat:(1)

the impulsive tempo was modifiable, (2) enriching impulsive children's

ability to deal with redundant and variable codes was one approach to

modification oE impulsivity, (3) to solve mismatch between the individ-

unl's cognitivP style and codes of communication was as important as

Teaching certainn scanning strategies, and (4) the correlation between

response tile and error could be altered in either directions by pro-

vilirg somg specific learning experiences.




This experimental study was to appraise the effect of two strate-

gies for modifying Jipulsivity in terms of reducing impulsive children's

errors measured by Kagan's Matching Familiar Figures. The assumptions

of this study related children's impulsivity with their being accustomed

to using restricted code of communication characterized partially by

shorc sentences and limited alternatives. To cope with impulsive chil-

dren's high errors when encountered with redundant and variable codes

of communication, two strategies were employed. These two strategies, which

were represented by the two hypotheses in this study, involved enrich-

ment of children's ability to deal with redundancy and variability.


The two hypotheses tested in this study were:

Hypothesis 1: Training in writing long sentences

has a significant effect on impulsivity in terms

of reducing impulsive children's errors as measured

by :arching Familiar Figures.

Hypothps's 2: Training in making multiple altErri-

ti'e choices has a significant effect on impulsivicy

1.: terms of reducing impulsive children's errors as

measuredd by Matching Familiar Figures.


Eighty-six boys and 89 girls in gr-ides 3, 4, 5, of P. K. Yonge

Laboratory School at the University of Florida, Gainesville, Florida,

participated in this study. Their ages ra.igcd from eight through eleven.

By using Kagan's Matching Familiar Figures (MFF), children whcse average

response ties fell below median and total errors above median ,were

classified as impulsive. Those whose average response time fell above

median and total errors below median were labeled as reflective. Among

59 impulsives and 39 reflectives, 57 impulsives were randomly assigned

into three groups with block design, and 19 reflective ;were randomly

selected as n-'tched control group. Data showed that the threp impulsive

groups were very similar in such variables as mental ability and socio-

economic status.

Two experimental tasks, Long Sentence Task (LS Task) and Multiple

alternativess Task (MA Task), were designed by this researcher to test

the hypoLheses. They were randomly assigned to two impulsive groups

:with the third group as control group which received no training. A

male and a female graduate student were trained to conduct the training.

In LS Task group, children were instructed to write two long sentences

for each task. While in MA Task group, children were instructed to write

at lIcst three alternative choices for each part of a task. Ten tasks

for c'uch group ver-e finished in five weeks. After training, the same

versi~r of MF' was individually adziinistered to children in four groups.

Results of the study significantly supported the two hypotheses at

.0'. Further c(,:parcicons of error changes by Duni:,tt's method showed

chat both LS Task group and IMA Task group significantly reduced errors

(at .05 a,-d .01 respectively) on :F as compared with control group.

Analysis of cuvanriance indicated that mental ability did not signifi-

cantly affect the discernible treatment effect. The LS Task training

was found to be more effective for boys, for children with high mental

ability, and for chose who .;ere from lower socioeconomic families. On

the other hand, the MA Task training was more effective for boys, for

children who had lower mental ability, and for, those who were from

higher socioc'-onor.ic status.

Other Eindings demonstrated that there was no significantt increase

of response time, and that the correlations between resporce time and

error were altered in different directions by diffreent training tasks.

The study also showed that .LM Task training affected both extremely and

moderately impulsive children in terms of error reduction.


The rrajor conclusion drawn from this stuy.v was that modification of

children's impulsivity in term.; of error reduction could be accomplished

by enhancing chl)Idren s ability to deal with redundancy and variability.

This conclusion was based upon the significant training effect of wJrting

]ong sentences end making alternatives on error reduction for impulsive

childreai ages eighC through eleven.

From the results of this study and subsequent discussion, the fol-

owing conclusions may be drawn: (1) The impulsive tempo can be modified

n tecmis of accuracy. (2) Eorichnog children's codes of communication

may enha:nc children's abii try in information assimilation. (3) Appro-

pr'.te meThod. for ra.oification of impilsivity can be developed if some

i.c.licaL1. behavioral corre:late; of imrpulsivity are taken into accccunt.

(4) The relationship between latency and accuracy can be affected by

certain experiences. (5) The indirect method for modification of im-

pulsive tempo can be as effective as direct modeling method. However,

the indirect ,method keeps the criterion -neasure as sensitive as possible.

It also eliminates influence from the experimenter's expectation. (6)

Modification of a tempo, either impulsive or reflective, should aim at

facilitating optimal learning, informative measurement, as well as a

healthy personality. These conclusions have not only emerged from the

study results and discussion but also suggested some implications for

consideration by educators and counselors.

Implications of the Study

From encountering a problem, seeking truth, to proposing keys to a

solution, research studies should contribute to making man more human.

With this ir' mind, the implications of this study will be considered ac-

cording to possible contributions to optimal school achievement, informa-

tive measurement, and healthy personality development.

First, the family and school may provide children, in their early

-.;,s, experiences in various language forms. Since language is one of

the powerful tools for cognitive development, a variety of experiences

in dJiffere .:t l:n '.ri-uaze Eorms, restricted and elaborated (or other for:ns).

may conseqsceiLy help children "internalize" a broader range of cogni-

tive skillsl, ith roreio available alternative cognitive skills, children

1y deal with realityy ,noie flexibly. Therefore, Icarning can be more


Secc: d, the curriculum planner may reorganize learning materials

in a ,.ay that such 'materials mn-v e:nich children's ceding and recoding

ability with different codes of communication. To vary language forms

used in learning materials, education could avoid stereotyped, o.notonic,

and fixed materials which are less effective in attention arousal.

Third; the school may encourage teachers to be more flexible in

verbal communication practices which not only allow children to use

their own language forms but also facilitate children's ability to use

other language forms. Obviously, this would presuppose no existence of

bias against any forms of language. Therefore, children can use language

forms more freely according to their judgment of the necessity.

Fourth, the test maker and test administrator may adjust measure-

m'nt techniques to assure child's comprehension of each test question.

They also should be sure that answers expressed in one form of language

are as valid as in the other form, providing the answer is correct, If

some children have difficulty understanding the question due to its

language form, appropriate adjustment should be made in order to make

optimal information communication possible.

Fifth, the school may give special training to those children who

cannot benefit from language enrichment through regular classroom teach-

ing. Such special training requires that a qualified trainer specialize

In language instruction and be sufficiently knowledgeable about cogni-

tive dcvel.opment. However, the training process should not evolve into

training language form for maintaining a cultural tradition.

Sixth, th!i counselor may improve interpersonal communication in

counseling. Tension increases if the counselor and the cc'nselee speak

"Jifferent" languages. Although the counselor can be trained to be fa-

iliar wt-h different forms of language used by children, it is more

advantageous to communicate with children who are not only proud of

their language, but also under,cand the counselor. This is essential

in establishing rapport betw:ecn the counselor and children.

Finally, the educator may enhance child en's healthy personality

development. If errors are caused by children's inability to conmmuni-

cate with stimulus situations, to remove the cause would reduce frus-

tration and probably enhance children's self-concepts. One possible way

to remove such a cause is to enrich children's ability to deal effective-

ly with redundant and variable forms of language.

These implications cannot become significantly relevant to childreris

effective learning and healthy personality development until more studies

are made in the area of impulsivity modification though error reduction.

lart!rermore, these results must be confirmed, and applied in terms of

increasing teacher's sensitivity to children's current language forms and

their cognitive styles.

The following suggestions in terms of research methodology for fur-

ther studies on modification of impulsivity through error reduction are

proposed by this researcher:

1. More detailed and professionally designed experimental tasks

ad standardized task training procedures are necessary for hypothesis

test ng.

2. An objective instrument should be designed to assess progress

iv task training.

3. In addition to the two characteristics cf restricted language

eectted for this study (shori. serternes and limited alternatives),

.ther s:uc.h characteristics can be isolated and added to the study.

These additions could provide increased depth in test ing the hypothesis

that language furrm is assclai:ed with conceptual terrmo.





CTMM3 SES NFF Pretest MFF Posttest
S Group 1 Sex2 Grade Age Score Status Time5 Error6 Time Error

9 2 2 5 11 116 84 5.4 9 10.6 2
11 4 2 5 10 109 62 11.3 3 8.9 11
12 2 1 3 8 105 72 5.3 20 12.8 9
13 I 1 5 11 118 84 4.5 10 8.0 15
16 2 1 4 10 125 44 6.0 12 10.5 10
17 4 2 5 11 118 12 22.0 6 19.7 5
18 4 2 4 10 117 65 11.6 11 24.3 10
19 1 2 4 11 104 44 4.2 12 8.3 7
21 2 1 4 9 102 61 6.5 19 12.9 10
23 3 1 5 11 120 92 8.8 14 16.3 8
24 1 1 4 10 102 84 5.6 21 10.2 13
25 4 2 5 10 120 96 31.5 1 22.5 4
29 1 2 3 8 86 44 3.4 18 6.0 21
30 3 1 5 11 86 14 7.3 10 5.2 15
32 1 2 5 10 117 92 8.0 9 5.6 4
38 2 1 5 11 111ii 66 7.7 15 21.6 4
40 2 2 5 11 115 92 6.7 10 10.6 8
41 2 1 5 11 121 59 7.4 13 7.5 7
42 2 2 5 10 114 36 3.0 11 6.3 9
43 4 2 5 11 130 84 10.2 8 10.2 5
46 4 2 5 10 108 49 9.9 4 16.0 1
48 1 1 5 11 121 84 6.8 11 8.5 3
50 3 2 5 10 100 85 6.2 16 5.8 11
52 2 1 5 11 94 84 6.5 13 4.7 9
53 3 1 5 10 95 92 4.1 20 6.2 18
54 3 1 5 10 126 84 5.5 11 2.5 16
62 3 1 4 9 115 84 3.3 14 10.4 13
63 3 1 4 10 127 92 7.0 11 8.1 8
64 2 2 4 10 112 61 6.3 14 7.9 12
63 2 2 4 10 96 21 7.0 18 11.7 9
70 2 2 4 10 131 92 4.5 15 10.0 7
71 1 2 4 10 101 72 5.3 13 5.2 8
7? 4 2 4 10 115 84 11.0 5 10.8 4
72 3 2 4 9 116 93 4.5 17 6.2 9
75 4 2 4 10 127 84 16.2 2 19.0 1
78 3 1 4 10 119 68 5.8 19 8.0 12
81 3 1 4 10 133 85 3.6 13 8.1 10
82 4 2 4 10 82 19 14.5 10 15.9 9
b4 3 1 4 10 108 39 5.8 17 10.0 7
85 1 1 4 10 123 67 7.9 12 9.5 4
87 1 2 4 10 69 44 7.3 14 10.6 9
88 3 1 4 10 122 84 9.5 12 9.2 13
89 4 1 4 9 93 62 11.5 6 11.5 12
91 4 2 3 8 127 84 14.8 8 21.1 7
99 3 1 3 9 117 92 4,0 23 5.5 16


Grade Age Score


S Gro pU

T im-5












etest MFF Posttest
Error6 Tin:e Error


One for LS Group,
Reflective Control

2 for MA Group, 3 for Control Group, 4 for Matched


20;e for Boy, 2 for Girl.
"California Test of Mental Maturity.
4Duncan's Socioeconomic Index by Occupations.
5Average Time in Second.
6Total number of error.








Matching Familiar Figures
(record sheet)

Na ------Sex.__ Birth date / /

!tem # Time to first choice (sec.) Number of errors














Datep of recording / /

No e




LS Task 1

Na. Write one long sentence descibin how you look.

W. rite one long sentence describing how you look.

2. Write one long sentence to tell what you do in school.

LS Task 2


1. Write one long sentence to tell how you spend your money.

2. Wrice one long sentence to tell what you like to read.

LS Task 3


Write a long sentence which tells more about each of the following.

1. Happy look.

2. Smiling face.

LS Task 4

rite a long sentence which tells more about each of the following.

L. Go play.

2. Don t move.

--~--- --

- -- ----I----

-- I-----

LS Task 5

Name :

Write a long sentence which tells more about each of the following.

1. Not me,

2. Find him.




LS Task 6


Write a long sentence which tell more about each of the following.

1. She got gift.

2. lHe caught fish.

LS Task 7


1. Write a long sentence to show the reason you should keep your hands

2. Write a long sentence to show the reason you like to have fresh air.


LS Task 8

Name :

1. Write one long sentence to tell why you have a house.

2. Write one long sentence to tell why you have (or you do not have)

LS Task 9


The words of following sentences are all mixed up. They do not make any
sense. W 'ite the sentences in the right orders so that they make good

1. my father hardworking man is a and a lot of he had energy

2. a lot of she had Itrouble but give up still did not she

~_~_~I_ ~I~~ ~I_ I

_1~1_ _I~ __ ~

--~-`------------ --

LS Task 10

The words of following sentences are all mixed up.
sense. Write the sentences in the right orders so

1, and blue eyes
feel make me

They do not make any
that they make good

blond hair good looking ears I have which pretty

2. cornflakes my
h is

I eat when lies he

in bed with juice and toast

~ _~ ~

_ ~~I~ ~C~

_ __ __~I_ ___

___ ~_I_ _^I__ ~~_~I_~



(~r TASK)

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