Group Title: descriptive study of three methods of programming flash cards
Title: A Descriptive study of three methods of programming flash cards
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00097690/00001
 Material Information
Title: A Descriptive study of three methods of programming flash cards
Physical Description: ix, 98 leaves : illus. ; 28 cm.
Language: English
Creator: Skypek, Genevieve, 1945-
Publication Date: 1971
Copyright Date: 1971
 Subjects
Subject: Reading -- Programmed instruction   ( lcsh )
Psychology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Psychology -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
 Notes
Thesis: Thesis - University of Florida.
Bibliography: Bibliography: leaves 95-97.
Additional Physical Form: Also available on World Wide Web
General Note: Manuscript copy.
General Note: Vita.
 Record Information
Bibliographic ID: UF00097690
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 - 000565856
oclc - 13588393
notis - ACZ2278

Downloads

This item has the following downloads:

PDF ( 3 MBs ) ( PDF )


Full Text














A DL crlpnti e Study of Three et-hods of
Programmining Flash Cards













By

GENEVIEVE SKYPEK


A DiSSERTATlON PRESENTED TO TIET GRADUATE COUNCIL OF
THE UN\'VESITv OF FLTOR'A TN PAPRTTAL
IFULFILLM'ENT OF liiE 'REOUTIRrTE MTS FOR THE DEGREE OF
DOCTOR OF P'I LOSCOPHY


UNIVERSITY OF FLORIDA

















ACKNOWLEDGEMENTS


The author wishes to express her appreciation to both Dr.

William D. Wolking, her committee chairman, and Dr. H.S. Pennypacker,

without whom this paper would not have been. In addition, both men

contributed considerably to her development of a meaningful and

coherent view of the complexities of human behavior, without which

she could be of no value whatsoever as a practicing clinician.

The author wishes to thank her other committee members, Dr.

Harry Grater, Dr. Donald Avila, and Dr. E.F. Malagodi, for their

support and patience.
















TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ................................................. ii

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

LIST OF FIGURES ................................................. vi

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

CHAPTER 1. Introduction.......................................... 1

The Problem........................................... 1

Relevant Research..................................... 3

Research on the teaching of reading................ 5

Research on verbal learning ......................... 6

Research on programmed instruction................. 8

Research on matching-to-sample discrimination
training. ........................................... 13

Purpose of the Study................................... 16

CHAPTER 2. Method................................................ 18

Subjects .............................................. 18

Curriculum Materials .................................. 18

Procedure............................................. 19

CHAPTER 3. Results............................................... 24

Celeration Ratios During Acquisition.................. 27

Retention Measures ..................................... 27

Step Size........................... ........ ......... 33

Repetition. .......................................... 37




iii















TABLE OF CONTENTS (CONT.)

Page

Celeration Ratios Diring Method IV........................ 40

Accuracy at the Median Performance During Method IV....... 42

Summary of Results................................. ....... 42

CHAPTER 4. Discussion................................................. 48

CHAPTER 5. Summary................................................... 59

APPENDIX A. Individual Graphed Data................................... 62

APPENDIX B. Notational System Proposed by Findley (1968).............. 75

APPENDIX C. Intra-day Performance.................................... 83

LIST OF REFERENCES.................................................... 95

BIOGRAPHICAL SKETCH.................................................. 98















LIST OF TABLES


Tables Page

1 MEAN CELEBRATION RATIOS DURING ACQUISITION.................... 28

2 TREATMENT X SUBJECTS ANOVA ON MEAN CELEBRATION RATIOS
OBTAINED DURING ACQUISITION .................................. 29

3 RETENTION SCORES............................................. 34

4 TOTAL NUMBER OF NEW WORDS PRESENTED IN EACH METHOD ........... 35

5 MEAN NUMBER OF REPETITIONS PER WORD.......................... 39

6 CELERATION RATIOS DURING METHOD IV........................... 41

7 ACCURACY SCORES FOR MEDIAN PERFORMANCE DURING METHOD IV...... 43

8 SUMMARY OF THE RESULTS ........................................ 44















LIST OF FIGURES


Figure Page

1 Celeration coefficients and celebration ratios.............. 25

2 Robin, age 16, grade 10 -- Method I........................ 30

3 Robin, age 16, grade 10 -- Method II....................... 31

4 Robin, age 16, grade 10 -- Method III...................... 32

5 Mean frequency multipliers per number of new words
added daily ................................................ 38

6 Michael, age 9, grade 3 -- Method I........................ 53

7 Michael, age 9, grade 3 -- Method II....................... 54

8 Michael, age 9, grade 3 -- Method III...................... 55

9 Jesse, age 6, grade 1 -- Method I.......................... 63

10 Jesse, age 6, grade 1 -- Method II......................... 64

11 Jesse, age 6, grade 1 -- Method III........................ 65

12 Sharon, age 6, grade 1 -- Method I......................... 66

13 Sharon, age 6, grade 1 -- Method II........................ 67

14 Sharon, age 6, grade 1 -- Method III........................ 68

15 Aubrey, age 9, grade 3 -- Method I......................... 69

16 Aubrey, age 9, grade 3 -- Method II........................ 70

17 Aubrey, age 9, grade 3 -- Method III....................... 71

18 Debbe, age 13, grade 8 -- Method I......................... 72

19 Debbe, age 13, grade 8 -- Method II........................ 73

20 Debbe, age 13, grade 8 -- Method III........................ 74
















LIST OF FIGURES (CONT.)


Figure Page

21 Serial arrangement of the single matching-to-sample
discrimination tasks (word reading), represented as
options, in each method of flash card programming........... 78

22 Movement through the chains and/or trees during each
daily experimental period ................................... 79

23 Scheduling of delivery of the conditioned reinforcer,
"being correct on all words", under each method............. 82

24 Aubrey, age 9, grade 3 -- Method I: Intra-day
performance. ............................................ .... 85

25 Aubrey, age 9, grade 3 -- Method II: Intra-day
performance......... .. .................................... 87

26 Aubrey, age 9, grade 3 -- Method III: Intra-day
performance. ................................................. 89

27 Jesse, age 6, grade 1 -- Method I: Intra-day
performance ................................................. 91

28 Jesse, age 6, grade 1 -- Method II: Intra-day
performance................................................. 92

29 Jesse, age 6, grade 1 -- Method III: Intra-day
performance................................................. 93














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


A DESCRIPTIVE STUDY OF THREE METHODS OF
PROGRAMMING FLASH CARDS

By

Genevieve Skypek

August, 1971

Chairman: William D. Walking, Ph.D.
Major Department: Psychology


The purpose of the present study was to describe the continuous,

on-going performance of students as they progressed through a pool of

discrete stimulus items under three different methods of programming

flash cards containing these stimuli and to compare the methods.

Six Ss were studied. They.differed in age, sex, and adequacy of

reading performance. Curriculum differed for each S. Young Ss re-

ceived basic English reading words, while older Ss received Spanish

vocabulary words.

A multiple baseline procedure was used in which each S received

each of the three different methods of programming flash cards.

Methods I and III required Ss to perform sets of ten new words and

sets of eight old and two new words, respectively, to criterion (each

set entirely correct for three consecutive days) before receiving new

sets of words. Method II required criterion attainment on each separate

word before receiving a new word.


viii







Frequency correct and frequency incorrect were obtained daily for

each S under each method. New measures describing continuous changes in

frequencies over time were used for describing performance differences

during acquisition. Discontinuous accuracy measures were used to describe

the effect of the methods following acquisition.

A reacquisition phase, M'Iethod IV, followed the acquisition phase.

For each method, all originally new words were presented together for

four days. Behavior measures over time and accuracy measures were obtained.

The general conclusions were as follows: (1) Method III produced

highest retention. Day to day behavior change was low. The reacquisition

phase was not necessary to increase either accuracy or speed of reading,

which was at ceiling level following acquisition. (2) Method I produced

an Jntermediate amount of retention. Day to day behavior change was

highest. The reacquisition phase produced increases in Method I perfor-

mance to equal that of Method III. (3) Method II produced lowest reten-

tion following acquisition. Little sign of over-all behavior change

appeared during acquisition. However, striking improvement occurred

during reacquisition, bringing median accuracy up to equal that of Method

III with speed of reading continuing to improve beyond median performance.

Since Method II produced acquisition of the greatest number cf

words, it appears that Method II incombination with Method IV programming

procedures produced maximum efficiency, in terms of speed and accuracy,

of progression through a pool of discrete stimulus items.

The study further demonstrated that the use of the newly developed

continuous measures of behavior change in combination with the more

traditional measures of discontinuous change provides a fuller descrip-

tion of the effects of procedures on performance.














CHAPTER 1

Introduction



Work with special education teachers and others at the Children's

Learning Laboratory of the University of Florida led to the observation

that there were two most frequently used methods for teaching reading to

children referred to the laboratory because of reading difficulties. On

one hand, some sort of programmed reading series, such as the Sullivan

Reading Program, was used from the very beginning of training. On the

other hand, flash cards, such as the Dolch Basic Reading Words, were fre-

quently used to establish a basic sight vocabulary before going on to the

use of some programmed reading series. Other more specific Lechniques,

such as teaching letter discrimination (b, d, p, q) or phonetic sounds,

have been used where necessary to teach specific discrimination needed

to enhance efficient progress through the reading series or the flash

card task. Training for specific discrimination may differ from child

to child and may not be used except where the behavioral definition of

the reading problem indicates that it is needed in order for the child to

progress successfully through whatever program has been outlined for him.

Generally, the basic program outlined for a child includes one or both of

the two methods mentioned above, i.e., a basal or programmed reading series

and flash cards.

The Problem

It would seem that all classroom teachers, both remedial and regular,


-1-









are faced with having to train students to respond appropriately to a

number of discrete stimulus items making up a portion, at least, of

some subject matter area. Examples of pools of such items include the

alphabet, names of numbers, basic reading words, phonetic sounds which

accompany letters and letter combinations, multiplication tables, and

foreign language vocabularies. Given a pool of such stimulus items,

e.g., basic reading words, the question one might ask is how should the

items be presented (programmed) to achieve an optimally efficient use of

instructional time. Actually, there are any number of questions one might

want answered: (a) Given some preset criterion which the student must

achieve, one might wish to know the quickest way to move a student through

a given pool of items. In this instance, time would be of primary impor-

tance, with accuracy acquisition somewhat secondary to time. (b) One

might want to know- how to maintain accuracy at its highest level, regard-

less of the time taken to achieve it. (c) Another question might be how

one maintains a consistently high level of success for the student.

If one had decided to use flash cards as a method of teaching a

basic-sight vocabulary, the question which might arise is what is the best

way to program the flash cards in order to answer the questions asked

above. This is a practical question concerning the technology of a method

in use. It is not a statement of the author's preference for the use of

flash carJs over a programmed reading text. Rather, it is a statement of

the author's belief that, if a given method is going to be used for some

purpose, the effects of that method, in its different forms, should be

carefully described.

There are a number of ways in which flash cards could be sequentially

programmed. For example, given that a preset performance criterion must










be met in each case, one might present a set of items from a pool of items

and require that the performance criterion be met for the entire set be-

fore presenting the student with a new set (composed of the same number)

of items from that pool. Each time criterion -as met, a new set of items

could be presented to the student until all items in the pool had been

presented and mastered to criterion.

Another method of programming might be to "embed" new items from the

pool among old or already-known items from that pool. Once the performance

criterion had been met by the student on this set of old and new items,

he would receive some new items "embedded" among the old ones he had just

performed. Some of the old items could be removed to maintain the same

number of items.

A third way of programming items in a pool might be to keep the num-

ber of items presented at any one time constant, but to require that the

performance criterion be met on each word separately rather than on a

whole set of items. Once performance on a single item had reached cri-

terion, the item would be dropped and a new item added. The student

would progress through the items in a given pool separately rather than

through sets of items from a pool.

The present study was an attempt to describe the on-going performance

changes resulting from the presentation of flash cards in the three dif-

ferent ways presented above, using basic reading words as curriculum ma-

terials.


Relevant Research

Much of the research on the acquisition of verbal kinds of behavior,




-4-


such as reading, has involved discontinuous measures of behavior change,

i.e., the comparison of posttest scores with pretest scores. Frequently,

any differences obtained in the before and after scores are attributed

to rather vaguely defined treatment procedures. However, where treat-

ment procedures are fairly precisely defined, the continued use of dis-

continuous measures of behavior change tends to obscure or disregard

much valuable information about the behavior engaged in while under treat-

ment. Much of the controversy existing within and between different

areas of research might be resolved if, instead of arguing over differ-

ences in before and after scores, researchers would stop to look at pat-

terns of behavior change occurring during their treatment procedures.

Such a change in experimental procedure might enable them to define more

precisely the variables resulting in the differences in their respective

findings. For example, a finding of no differences between posttest

scores obtained from differing treatment procedures does not mean that

the treatment procedures do not produce differences in patterns of beha-

vior change. Although two procedures do not differ in their effects on

final performance, they may differ with regard to the amount of success or

strain the individual experiences while acquiring the final perfoiimance.

One treatment procedure might result in continuously high frequencies of

correct responding while another procedure might result in an initially

high frequency of incorrect responding and low frequency of correct re-

sponding followed by accelerations in frequency of correct responding and

decelerations in frequency of incorrect responding. The student would

experience considerable differences in total amount of success under the

two different procedures. However, if discontinuous posttest scores of











acquisition did not differ, one would never know from such traditional

measures alone that such treatment differences existed.

The first three areas of research to be discussed suffer from the

experimental inadequacies, among many others, just mentioned. The abso-

lute relevance of their findings to any results based on discontinuous

measures used in the present research project is left to the reader's

judgment.


Research on the teaching of reading

Practical research directly relevant to the programming of flash

cards in teaching a basic sight vocabulary is scarce among educational

researchers. Those studying the teaching of reading seem not to have

addressed themselves to the task of describing immediate on-going per-

formance differences resulting from the different ways of programming ma-

terials within the different methodologies for teaching reading which

they propose. They have, instead, attempted to assess long-term reading

differences resulting from rather vaguely defined in-classroom teaching

procedures. For example, a major controversy seemed to exist within the

field of education between proponents of a "whole-word" or "look-say"

approach to teaching reading and those of a "phonetics" or "sounding"

approach. The controversy still exists, although to a lesser extent,

today. Seldom are the specifics of either approach precisely outlined.

Chall's (1967) summary of the literature on the "whole-word" versus

"phonetics"controversy suggests that a "phonetics" approach to reading

results in better reading skills as measured by conventional reading

achievement tests. However, in reviewing articles on reading failures or




-6-


children who read poorly, Chall seems to suggest that there is value in

teaching the service words and frequently used nouns and verbs by a

"whole-word" approach or even through the use of flash cards. Dolch

(1959) makes the same suggestion for all beginning readers. Such a

method is presumably easier, provides for quicker attainment of reading

success, and thus helps to instill confidence in the child. Following

this, the teacher could then teach phonetics to achieve lasting effects

in the form of successful reading. However, no article by Chall or

Dolch describes the best method for presenting basic words using either

flash cards or some other "whole-word" technique.

Even if -- or perhaps because -- flash cards or the "whole-word"

method are defined by these educators as useful only with those children

who are hardest to teach, an appropriate technology of flash card use

should be developed and researched.


Research on verbal learning

Acquisition of a sight vocabulary through the use of a flash card

methodology is most like a paired-associates learning paradigm used by

verbal learning researchers. The stimulus term is the written word and

the response term is the spoken word corresponding to the written symbol.

The typical paired-associates paradigm begins with one study trial in

which the stimulus term is presented alone (2 seconds), followed by pre-

sentation of the response term. The subject's task is to give the re-

sponse term to the stimulus term before it (the response term) appears.

This is called the anticipation method because the subject's response must

anticipate the response term presented on the memory drum. Beginning with

the second trial, the E can measure any learning that took place on the











first trial and so on. Trials are continued until the subject reaches

some performance criterion, such as one perfect trial or some number of

trials. The order of presentation of the stimuli is varied unsystematically

so as to avoid serial learning (Jung, 1968). Once the subject has reached

the performance criterion set by F, tests of immediate and long-term re-

tention are generally made. According to Jung, amount of learning is

the single most effective variable during learning on retention. One way

of defining amount of learning might be number of repetitions per stimu-

lus-response pair. The author knows of no research on total number of

repetitions per stimulus-response pair; instead, research on number of re-

petitions seems to lie primarily in the area of overlearning, which is de-

fined as number of repetitions beyond initial mastery. De Cecco (1968)

states that overlearning produces striking gains in retention up to a

point. To be more specific, 50% overlearning (half as many repetitions

beyond initial mastery as was required to reach initial mastery) produces

impressive gains in retention, while 100% overlearning, although it pro-

duces some gain, does not produce much gain beyond that produced by 50%

overlearning.

With regard to the three methods of flash card programming studied,

the total number of repetitions per stimulus-response pair provided by

each method might relate to retention of the words programmed by each

method. This suggestion involves an extension of the overlearning find-

ings that may or may not be justifiable. Assuming that each method does

not provide repetitions beyond the 50% overlearning level, such an ex-

tension is reasonable.




-8-


Research on programmed instruction

Another area of research that might be relevant to the question of

how best to program flash cards is that of programmed instruction. More

specifically, research on two of the structural or independent variables

of which a program's efficiency might be a function, step size and/or re-

petition or density, provides another way of describing discrete differ-

ences in the three flash card programs studied.

Skinner (1961), in an attempt to maximize success and minimize failure,

recommends that the steps between items be small so that the student is

seldom wrong. With fewer errors, the student receives more reinforcement.

The acquisition of a complex behavior, such as a skill or knowledge, is

best achieved through a series of progressive approximations to the final

complex behavior desired. Research from the operant animal laboratory

has generally shown that acquisition is best achieved under conditions of

continuous reinforcement. For this reason, the ideal program should have

steps small enough to produce no errors, thus placing the student under

conditions of continuous reinforcement (assuming in many cases that a

correct response is a reinforcer). Skinner is not necessarily referring

to repetitions when he recommends small step sizes. Rather he is saying

that changes in behavior required for reinforcement should be so small

that the organism can easily make them without error.

A consistent quantitative definition of step size has proven difficult

to achieve. Lumsdaine (1960) says that step size has typically been de-

fined in one of three ways: (1) difficulty, measured as item error rate,

i.e., the percentage of students missing a given item; (2) size of re-

sponse required, measured as number of letters, number of words, number of











sentences, etc.; and (3) size of frame, measured as number of informational

items or number of phrases.

Briggs et al. (1962) compared programs with two step sizes, defined

as size of frame. The small step program followed Skinner's recommenda-

tions and contained a phrase, sentence, or two sentences at most in each

frame. The large step program contained frames as long as eight pages

and was much like conventional textbook reading procedures. Briggs'

posttest of learning contained items to which a response was required by

both programs and items to which no response had been required by the

programs but which were based on the additional information contained in

the frames. He found no differences in total test scores between the

two groups of Ss using the different programs. However, those Ss using

the small step program performed better on items to which a response

had been required than those Ss using the large step program, but poorer

on items to which no response had been required by the programs. Briggs

et al. interpret these results as favorable for large step programs.

However, one could interpret the results in favor of small step programs

since the one they used taught what it was intended to teach better than

the large step program taught that same material (contained in the small

step program). The results seem to suggest that a small step program

including all the material that was additional in the large step program

would produce better results on that same material than would the large

step program.

Furukawa (1970), using a program to teach Hawaiian words, defined

step size as number of associations or chunks per frame. A chunk con-

sisted of one sentence, one question, and one answer. The differences in




-10-


step size consisted basically of how many sentences the subject had to

read in a frame before he began answering the questions (requiring a one-

word response). This definition differs from that of Briggs because

every bit of information in the sentences was necessary for correct re-

sponse to the questions. The large step program of Briggs contained

much information irrelevant to the correct response. Performance through

the program (measured by number of errors) was better in the small step

program, but retention on an immediate posttest was better for programs

containing seven to fourteen chunks (intermediate step size). A confound-

ing factor, making the results difficult to interpret, was the existence

of additional stimulus cues in the larger step programs. Every seven

items formed a meaningful whole, i.e., seven items were about flowers,

seven were about food, etc., thus, perhaps giving the subjects with large

step programs an advantage.

Shay (1961) defined step size as difficulty of making a correct re-

sponse. The programs taught the construction of roman numerals. A small

step program had a higher probability of producing correct responding in

subjects. In addition to the small step size program, he compared a medi-

um and a large step size program and found that as step size decreased,

posttest scores, requiring translation from arabic to roman numerals

and vice versa, increased.

Another way in which step size has been defined, not mentioned by

Lumsdaine, is the number of frames for a given program. Smith and Moore

(1962) defined step size in a spelling program as the number of frames per

word. They found no significant differences in achievement related to

step size on a posttest. However, the number of frames per word for a




-11-


small step program varied from four to seven and for a large step program

from three to six. These differences in step size may not have been

large enough to produce a difference. Evans, Glaser, and Homme (1960)

also defined step size as number of frames in the program. According to

them, a large step program was constructed by leaving out repetitious and

redundant items contained in the small step program. They found that

smaller step programs were associated with significantly fewer errors on

immediate and delayed performance tests and also led to fewer errors in

progressing through the program. Coulson and Silberman (1960) defined

step size in a manner similar to Evans et al. and obtained similar re-

sults. Small step size programs produced better performance on criterion

tests than did large step size programs.

In the studies defining step size as number of frames per program

or item, it is difficult to tell whether or not step size is defined as

Skinner meant it. It seems more likely that these studies confuse step

size with repetition of items. There have been relatively few studies

addressing the problem of item repetition in programs; however, those

that have generally find that more repetition increases criterion test

scores. For example, Hassinger (1965) used as his independent variable

a density ratio, i.e., the ratio of the number of different responses in

a program to the total number of responses required by the program.

The denser the program (maximum density = 1.00), the more different re-

sponses required. Using a program to teach simple statistical concepts,

Hassinger found that the groups using the less dense program obtained

higher gain scores from pretest to posttest.

A programming procedure directly related to the repetition vari-




-12-


able is one called the drop-out procedure (Ferster and Sapon, 1960;

Coulson and Silberman, 1960). In this procedure, the student progresses

through the program, marking in a scoring column items correct and items

incorrect. Once he finishes a given unit of the program, he goes back to

the beginning of that unit and repeats every missed item. He continues

doing this until each item has been performed correctly to some criterion,

e.g., two consecutive correct responses in a German program (Ferster and

Sapon, 1960). Fry (1960), in a study designed to test constructed re-

sponse versus multiple choice response procedures, had three conditions

under which subjects progressed through the program. In the first condi-

tion, the subject worked toward a criterion of mastery (as in the drop-

out procedure). Total working time and number of repetitions were allowed

to vary with each individual. In condition II, total working time for

all subjects was controlled, leaving number of repetitions to vary depen-

dent upon the individual variations in speed of progression. The last

condition controlled both working time and number of repetitions, holding

these the same for all subjects. Although Fry did not test the differ-

ences between these conditions statistically, the mean posttest scores

for condition III, which presumably resulted in less repetition than

either conditions I and II, were lower on both an immediate and a delayed

posttest. Holland and Porter (1960) compared a drop-out procedure with

a procedure requiring subjects to perform on a single trial only, regard-

less of accuracy. The program used was Holland and Skinner's psychology

program. The drop-out procedure, involving repetition to a criterion of

each item answered correctly at least once, produced lower error percent-

ages on three tests administered at various points throughout the program.




-13-


Step size and item repetition seem to have been confused in the

research on step size. They may or may not be related, but the data

are not conclusive. Sidman and Stoddard (1966), in their article on

developing a program for teaching a circle-ellipse discrimination to

the mentally retarded, suggest that too much practice can result in

satiation and boredom. 'According to them, if the steps are empirically

determined so that they are just small enough to keep errors down, re-

petition is not needed. Ideally, this should be the case. However, it

is unlikely that the economic contingencies are such that the amount of

work required to effect such precision in the training of an academic

skill or knowledge will ever be done. In that case, it appears that

some degree of repetition will be necessary to correct for the lack of

correct step size of any program.

With regard to the three flash card programs studied, one might de-

fine step size as the number of new words added after criterion on pre-

vious words had been reached, while repetition could be defined as the

number of times each word was performed until mastery to criterion was

achieved. The data seem to suggest that small step sizes and/or greater

repetition result in higher posttest scores. It might be interesting to

see how patterns of behavior change as the flash card methods relate to

step size.


Research on matching-to-sample discrimination training

Researchers in the area of operant conditioning or the experimental

analysis of behavior have defined the task of reading a word as a match-

ing-to-sample discrimination task for people with a pre-existing vocal

repertoire. More specifically, reading a word has been defined as a com-




-14-


plex discrimination task in which the subject matches his already exist-

ing vocal repertoire to a textual stimulus, the written word. Frequently,

a fading procedure is used in which the textual stimulus is presented,

a verbal prompt is given by the teacher, the child then imitates the

prompt and emits the appropriate vocal response. The prompt can be

gradually faded out as the child acquires the appropriate matching re-

sponse (Staats and Staats, 1968; Ferster and Perott, 1968). It appears

that much of the work of Staats and Staats has focused on the necessity

of an appropriate response-reinforcement contingency in keeping the

child at the reading or matching task and in producing acquisition of the

reading behavior. They show that when reinforcement is unavailable, the

child will show a decrease in rate of responding, in rate of correct re-

sponding, and in rate of occurrence of prerequisite, orienting behaviors.

Matching-to-sample is a discrimination task that has most often been

studied as one response to one set of stimuli to be matched to a sample.

The matching-to-sample task may be extended to the shaping of abstractions,

such as same shape, larger than, smaller than, to the left of, etc.

(Ferster and Perott, 1968). However, this typically involves the presenta-

tion of one sample and one set of to-be-matchad-stimuli until the appro-

priate matching response occurs at some given frequency. Then different

sets of stimuli involving the same abstraction are presented individually

until the abstraction can be demonstrated on presentation of new sets of

stimuli.

Reading a word is a relatively simple task, given that the student

has a pre-existing vocal repertoire. Reading does not in and of itself

require complex abstraction training procedures. The variables necessary

for producing acquisition of a reading or matching-to-sample response




-15-


have been sufficiently demonstrated in the laboratory that a teacher,

using this technology, could easily teach single words to students. How-

ever, teaching 100 words one at a time to a criterion with a daily time

base could become an arduous task for both teacher and student. The

question most likely to be asked by a teacher would be how to present

single items in a set so as to speed progression through the pool of

items to be learned with maximum retention. How many new words or single

matching-to-sample discrimination tasks can be presented at one time?

Should new words or new matching-to-sample tasks be presented with al-

ready acquired words so as to maximize total reinforcement or success?

Must the words be treated arsa single unit or set or can the make-up of

the set be changed as individual words are acquired before others? Re-

search on how best to program the acquisition of a number of different

matching-to-sonple discrimination tasks does not seem to be available.

It does not seem that the matching-to-sample paradigm is complex enough

to describe in detail the three methods of flash card programming studied.

In light of the fact that operant methodology is the only experimen-

tal methodology to make use of continuous measures of behavior change

and to focus on patterns of behavior change, it would be most valuable

to describe the procedural differences in the three flash card methods

as precisely as possible in terms that relate to operant methodology.

It may be that the different arrangements of matching-to-sample stimuli

provided by the different methods result in different frequencies of rein-

forcement and different response requirements for obtaining reinforcement.

Findley (1962) proposed a notational system to characterize complex pro-

cedures as special combinations and arrangements of simpler components.

In his system, the different methods can be seen as chains of simple





-16-


matching-to-sample discrimination tasks. The different stimulus arrange-

ments provided by the different methods seem to set the occasion for

different frequencies of reinforcement and different response require-

ments. Findley has demonstrated that complex operants composed of simpler

operants are controlled in a similar manner by the same variables that

control simpler operants. A more detailed application of his system

to the three methods of flash card programming will be presented later

in the paper.


Purpose of the Study

As one can readily see, aside from the operant studies on matching-

to-sample behavior, the majority of studies presented used pre-and post-

test criterion scores and/or error rates (measured as percentage or ab-

solute number) to assess and describe the effects of any variable under

study on performance. However, the performance under observation in

most studies was not typically acquisition of verbal or reading behavior,

but rather some behavior occurring before and/or after the acquisition

of behavior had been completed. The procedures did not allow for a point

to point analysis of the relationship between whatever kind of program

was used and the subject's behavior. Edinger (1969), however, has demon-

strated the extreme sensitivity of direct and continuous recording of

student performance in his analysis of the Sullivan Reading Program.

Using rate (or frequency) correct and rate incorrect as his behavioral

measure, Edinger was able to assess the diagnostic value of the Sullivan

Diagnostic Test in placing students in the Programmed Readers. Further-

more, he was able to assess the efficiency of various contingency ar-

rangements on the correct and incorrect rates. Finally, he demonstrated




-17-


that correct and incorrect rates of responding are concurrent operant

behaviors, i.e., they are not complementary with respect to their con-

tingencies. It is doubtful that such findings could have been obtained

with the more traditional before and after criterion tests and/or per-

centage measures. More traditional measures seem to focus on absolute

differences in quantitative scores, not on changes in behavior processes

over time as a result of environmental influences.

The purpose of the present study was to describe the continuous,

on-going performance of students as they progressed through a pool of

discrete stimulus items under three different methods of programming

flash cards containing these stimuli and to compare the three methods.

The emphasis was on the direct and continuous recording and description

of behavior changes in frequency (rate) correct and frequency incorrect

associated with the different programming procedures. The author hoped

to demonstrate that the patterns of behavior change observed under each

method were the same for each subject individually, a much better demon-

stration of both reliability and generality than that provided by the

more traditional group evaluative statistical procedures (Sidman, 1960).














CHAPTER 2

Method


Subjects

A total of six subjects were studied. Two boys, aged nine years,

were obtained from a remedial reading class at Sidney Lanier School

in Gainesville, Florida. Both boys were in the third grade, and both

had repeated a grade previously. They were classified as retarded

readers by the reading teacher. Two girls, aged six years, were

obtained from E's neighborhood. Both were in the first grade and

had not repeated this grade previously. Furthermore, these two Ss

received grades of Satisfactory in Reading and were promoted to the

second grade level. These two "normal" readers were studied for the

purpose of providing a systematic replication of the different

flash card program effects on Ss with different regular classroom

reading behaviors than the two boys. An additional systematic

replication was conducted on Ss of a different age using different

curriculum materials. Two females, aged 13 and 16 years, were also

obtained from E's neighborhood. They were in the eighth and tenth

grades, respectively. Neither had repeated a previous grade.


Curriculum Materials

The two boys were presented with the Dolch Basic Reading Words,

considered to be the equivalent of a first-grade reading level. It

was initially intended that the two six-year-old girls would also

receive the Dolch words. However, they knew them all during initial


-18-




-19-


testing. Therefore, the Graded Picture Words Cards by Mills, non-

picture side, were presented to these girls. The Mills cards were

grouped according to first-, second-, and third-grade levels. Both

girls began at the second-grade level and moved into some of the

third-grade level words. The two older Ss were presented with

Spanish vocabulary cards written by E. Words were randomly selected

from a Spanish-English dictionary. The Spanish word was written on

one side of a 3-X-5 index card, with the English equivalent on the

other side.


Procedure

It was initially planned that each S would be seen once a day

during the five school days of the week. Exceptions to this plan

occurred when Ss were ill, absent from school, or absent from the

neighborhood. An additional exception occurred with one of the

female first-graders who chose to come every day of the week.

A multiple baseline procedure (Baer, 1968) was used, in which

each S received each of the three different methods of programming

flash cards. Such a procedure minimizes intersubject variability,

thus allowing one to attribute differences observed in performance

to the program differences rather than to intersubject variability.

In order to find out which words were known and which were

unknown to each S prior to presentation, each S was required to

read words from his item pool. These were then separated into

correct (or known) and incorrect (or unknown) words. The items were

then programmed according to the conditions specified below. No




-20-


frequency measures were taken during this pretest. They were not

considered necessary since gain could be measured from zero correct

responses.

Items from the item pools were programmed in three different

ways. Method I used a set of ten unknown words. Each S was required

to perform on these words until he reached criterion for the entire

set. He was then presented with a new set of ten unknown words.

Following criterion attainment on the second set, he was presented

with a third set of ten unknown words. Attainment of criterion on

the third set of words marked the end of the acquisition stage of the

study.

Method II used ten unknown words on initial presentation. Whenever

criterion was reached on any single word, it was dropped from the set

and a new word was added. The number dropped on any given day could

conceivably range from 0 to 10, depending on the S's performance on

each separate word. Presentation of Method II words stopped at the

same time as presentation of Method I words stopped.

Method III used eight known words and two unknown words. Upon

reaching criterion performance on this set of ten old and new words,

the S was presented with two new words, and two of the originally

known words were dropped. The S continued in this manner, receiving

two new words at a time, until he finished the third set of words in

Method I and stopped. The particular set he was on in Method III was

then presented to criterion and no more were presented.

The performance criterion required for a change in words or set

of words was three successive days of correct responses or 100% accuracy.

This performance criterion was selected on the basis of previous experi-





-21-


ence with children in remedial reading classes. It seems that the

performance of such Ss shows a high amount of inconsistency from day

to day. Three successive days correct was considered minimum in

order to assume that the stimulus had acquired sufficient control to

maintain correct responding from day to day.

Each day the S was seen by E he was presented with three groups

of flash cards, with item changes occurring according to the conditions

presented above for each method. Each group of words under the three

different programming methods was presented three times each daily ex-

perimental period to allow for practice and observation of intra-day

changes in performance. (See Appendix D for intra-day performances.)

The measures used in recording daily observations were frequency

correct, obtained by dividing total minutes for all three presentations

of each method's group of words into total number of words correct in

those three presentations, and frequency incorrect, obtained similarly

by dividing total minutes for all three presentations into total num-

ber of words incorrect.

Specific external reinforcers were not given for correct respond-

ing. Again experience with remedial reading Ss suggests that the important

requirement is that Ss receive reinforcement for responding, whe-

ther correct or not. The goal was to keep them working at the reading

task. This procedure may require the assumption that "being correct"

is a conditioned reinforcer which controls the acquisition of the dis-

crimination task itself. The four young Ss received M & M's for res-

ponding, regardless of the correctness of the response. After each

presentation of all three method's groups of words and S response, S




-22-


received three to four M & M's. Thus, Ss were on a fixed ratio

schedule of 30 responses per reinforcement delivery. The total num-

ber of responses required per day was 90. The total number of rein-

forcement deliveries was three. The two Ss performing on Spanish

vocabulary cards chose not to receive an external reinforcer, stating

that learning the language for class was sufficient incentive.

Following the completion of the acquisition stage of the study,

each S had two to four days "vacation", depending on when the week-

end intervened. A posttest or retention test was then given. All

originally new words presented during acquisition in all methods were

presented again in one large group. Percent correct on each set of

previously presented words was the accuracy measure used to assess re-

tention. This accuracy measure was obtained only for the first presenta-

tion of all previously presented words.

For four days following the "vacation", a new programming method

was used in presenting all previously acquired words. This procedure

was called Method IV and was similar to the procedure used in Method I.

The total set of words acquired during the acquisition stage under each

method condition was presented during Method IV. Each of the total

sets was presented three times in each daily experimental period for

four successive days. Reinforcement delivery occurred after completion

of each of the three total sets of words. Thus, frequency of reinforce-

ment delivery remained the same, i.e., three times per day. However,

the response requirement for reinforcement delivery was raised consid-





-23-


erably and varied from S to S. The lowest ratio requirement under

Method IV conditions was 68 responses per reinforcement; the highest

was 102. There was no performance criterion to be met under condi-

tions of Method IV. Method IV results provided observations on the

re-acquisition of previously presented words, if such re-acquisition

was needed.

Under all methods, E presented the stimulus item, the word

written on a flash card, to each of the young Ss and then waited for

approximately ten seconds until the response, the vocal equivalent,

was made by S. If such response was not forthcoming, E then provided

the correct response. S was required to look at the word and repeat

E's prompt. The next item was then presented. If a response was

given but was incorrect, E gave the correct prompt and S was required

to repeat it. Items responded to correctly were placed in a pile

separate from items responded to incorrectly. Correctness was con-

firmed by placing the item in the appropriate pile.

The Ss reading Spanish vocabulary words were given each set of

ten words to present to themselves, while E watched and timed. These

Ss were required to read the Spanish word (pronunciation did not mat-

ter) and then to give the English equivalent. Following this, they

were to turn the card over and, if the response they gave was incorrect,

they were to read aloud the correct English word.

A movement-based recording plan was used in which E measured the

time taken to complete responding to a fixed number of stimulus items.















CHAPTER 3

Results


The measures used in this study, frequency of correct responding

and frequency of incorrect responding, were plotted on six-cycle semi-

log paper (Lindsley, Behavior Research Co.). Plotting linear data on

a log scale provides one with a picture of proportional changes in be-

havior or frequency of responding rather than absolute changes. Infor-

mation that the frequency of occurrence of a given behavior has doubled

or tripled is considerably more valuable than information that the fre-

quency of occurrence of that behavior has changed by one arbitrarily de-

fined unit.

Few statistical measures are available for describing continuous

changes in behavior over time. Therefore, researchers interested in con-

tinuous observation and recording of behavior have developed several new

measures for this purpose. One such measure is the celebration coeffi-

cient (CC), functionally related to the slope of the line of best fit,

obtained by using the least squares method of regression. The celebration

coefficient tells by how much one must multiply the predicted frequency

of responding on day (n) to obtain the predicted frequency of responding

on day (n+7). It is a measure of change occurring in frequency of re-

sponding over a week's period of time and can be used to describe pro-

portional changes in behavior having either accelerating or decelerating

slopes. For examples of this measure applied to different lines, see


-24-































































cj



C.;,
-1


S I--


0 o
*o


~---~-I------


21inrJIN )133a S1N3113AON


-25-


o0-


0 0
o0
0 LO


I I


0






0
,O









0
.0






oo
41
0


o. 0 0
0


r- i




Li


-.4
> 0
.,-
co

Q )






r-i
U

f >




o^ a
uCo


----' .




-26-


Figure 1. The top line in the first set of lines in this figure has a

celebration coefficient of 2.00. This number means that the behavior

changed from a frequency of one movement per minute on day 1 to two

movements per minute on day 8. The frequency of occurrence of the beha-

vior doubled within a week's time. The bottom line in the first set

has a celebration coefficient of .50. This number indicates that the fre-

quency of occurrence of the behavior decreased by half over a week's

time from a frequency of one movement per minute on day 1 to a frequency

of one movement every two minutes by day 8. The second set of lines

in Figure 1 represents different frequencies of behavior change. The

celebration coefficients for the top line and bottom line respectively

are 1.1 and .90. The change in behavior from day 1 to day 8 in the top

line reflects a 10 percent increase in frequency. Similarly, the reduc-

tion in frequency of responding shown in the bottom line reflects a 10

percent decrease in frequency from day 1 to day 8.

In the event that one wishes to relate observed changes between two

frequencies of responding, such as frequency correct and frequency incor-

rect, to the same set of stimuli, one can use a celebration ratio (CR).

The celebration ratio tells by how much one must multiply the celebration

coefficient of frequency incorrect to obtain the celebration coefficient

of frequency correct. This measure is a function of the angle of separa-

tion between the acceleration and deceleration of the two frequencies of

responding observed. Or perhaps another way of stating the function of

the measure is to say that the celebration ratio describes the rate at

which the two lines separate from each other. The first set of lines

depicted in Figure 1 has a celebration ratio of 4.0. This means that the





-27-


celebration of the top line is four times that of the bottom line. A

similar interpretation may be made of the second set of lines.


Celeration Ratios During Acquisition

The mean celebration ratios obtained during acquisition for the

three methods of flash card programming are presented in Table 1.

(The individual celebration coefficients and celebration ratios for

each set of ten words are presented on the Ss' graphs in Appendix A.)

A treatment X subjects analysis of variance, presented in Table 2,

was conducted on these mean celebration ratios. The results (F=7.83;

df=2, 10; p .01) indicate that the treatments differed significantly

in their effects on the Ss' performance. For all six Ss, Method I pro-

duced a higher rate of separation between frequency correct and frequency

incorrect than did either of the other two methods, suggesting that Method

I produced more behavioral change from day to day than either Method II

or III. Similarly, for all six Ss, Method III produced more behavioral

change than did Method II. In fact, Method II appeared to produce little

over-all change in separation between frequency correct and frequency in-

correct. In general, one might say that accuracy developed faster or

simply that behavior changed faster using Method I than using Method II

or III and faster using Method III than Method II. Figures 2, 3, and 4

present a typical S's data under the three different methods.


Retention Measures

The retention measure was percent correct obtained when all previ-

ously presented new words were presented together two to four days fol-

lowing the end of the acquisition stage. The retention scores are present-






-28-


Table

MEAN CELEBRATION RATIOS


Subjects


Jesse

Sharon

Robin

Debbe

Aubrey

Michael


I

49.37

25.54

5.83

38.68

7.02

9.57


1

DURING ACQUISITION


Methods

II

.95

.92

.94

.94

1.08

1.15


III

3.07

4.10

3.76

2.24

2.81

3.11


~I~ ~





-29-


Table 2

TREATMENT X SUBJECTS ANOVA ON MEAN CELEBRATION
RATIOS OBTAINED DURING ACQUISITION


Source df SS MS F p

Treatment 2 1708.33 854.17 7.83 .01

Subjects 5 587.65 117.53

Treatment
X Subjects 10 1091.29 109.13

Total 17 3387.27






-30-


0u


O


O
q-


os o -> o ^!
ui CM w
0 u ,
uO 0
00 a- 0 0 r _



rd 0
wO
0-
u 0 u 0 0 0







0 I' O
o 0


.6-4











\ __^^.: ___ .0







3l"^(dS133O





-31-


\ 0


N -.
Cl.
O

Sus a

0 UO
oD ~-H





( i0
C 0

















ft *
o< o 2


P0
II 0
2, : 0



o )

oCo



Ot


o .,









-- -
I !I r


















3 0".fNI N Hd SIN3 WA ON .
\o .' O- W Q
*\, ,,*4-J <_. .- j o
s ~ s. 8
0 L' -* &
-- 3nNI'\' ~d Si'













O


O


O
1-



0





O


O
0














-o










0
-0 o
A M

-d
C.'


o

o 0


31Ar:'V 'd3d SIN3JV3AOI'


-32-


CC



'p,


0 0
O o
O "


0 o


-


o





-33-


ed in Table 3. A Friedman's X2 test was conducted on the retention
r

scores. Results (X =10.3; df=2; p < .01) were significant, indicating

that, for all six Ss, Method III produced reliably greater retention.

In addition, Method I produced greater retention than Method II. Thus,

it appears that, although Method I produced greater behavioral change

during acquisition, Method III was conducive to greater retention.


Step Size

There were several possible definitions applicable to this study.

One definition of step size was the total number of new words acquired

or presented per method. Data on the total number of new words pre-

sented to each S in each method are presented in Table 4. A Friedman's

Xr test, conducted on these data, indicated that the number of new words

presented under each method differed significantly (X2=ll.1; df=2; p -<.01).

The smallest step size or number of new words was produced by Method III;

the largest by Method II, with Method I intermediate.

Another way of defining step size was in terms of the number of new

words added each time criterion was attained. Since only two new words

were added per phase in Method III and ten new words were added per

phase change in Method I, one could say that Method III involved smaller

steps in progressing through a pool of items than Method I.

The retention data indicated that Method III produced the highest

amount of retention, Method I produced the second highest amount of reten-

tion, while Method II produced the lowest amount of retention. Programmed

instruction researchers have suggested that smaller step sizes result

in higher amounts of retention. The findings of the present study with




-34-


Table 3

RETENTION SCORES


Subjects


Jesse

Sharon

Robin

Debbe

Aubrey

Michael


I

90%

90%

93%

90%

85%

73%


Methods

II

92%

76%

88%

75%

61%

58%


III

100%

100%

100%

100%

100%

87%


- -


I


- ---- --------




-35-


Table 4

TOTAL NUMBER OF NEW WORDS PRESENTED IN EACH METHOD


Subjects Methods Total

I II III

Jesse 30 35 6 71

Sharon 30 51 10 91

Robin 30 37 8 75

Debbe 30 50 8 88

Aubrey 30 60 12 102

Michael 30 30 8 68




-36-


regard to retention and step size appear consistent with the previous

research in that area.

Support for programmed instruction findings on the positive rela-

tionship between step size and frequency of responding can be found in

the step size changes in Method II and their relationship to frequency

correct and incorrect. Step size in Method II was defined as the number

of new words added per day. Step size could range from zero to ten;

however, the actual range was from zero to seven new words per day.

Frequency multipliers were used to describe the changes in frequency

from one day to the next as a result of the addition of a given number

of new words. The frequency multiplier is a measure which indicates by

what multiple the frequency changes (increases or decreases) from one

presentation to the next. For example, if the frequency correct on the

last presentation of day (n) was 3 movements per minute and the frequency

correct on the first presentation of day (n+l) was 6 movements per minute,

the frequency multiplier would be 2.0, indicating that the frequency had

doubled. Similarly, if the frequency incorrect on the last presentation

on day (n) was 3 movements per minute and the frequency incorrect on the

first presentation of day (n+l) was 1.5 movements per minute, the frequen-

cy multiplier would be .50, indicating that the frequency had decreased

by half from one day to the next. The frequency multiplier differs from

the celebration coefficient in that it measures a single-day difference

while the celebration coefficient provides a means of describing continu-

ous differences over longer periods of time.

Frequency multipliers for frequency correct, frequency incorrect,




-37-


and total frequency of responding were obtained by dividing the appropri-

ate frequencies of the first presentation of Method II set of words on

day (n+l) by the frequency of the last presentation of this set of words

on day (n). The mean frequency multipliers over all Ss for frequency

correct, frequency incorrect, and total frequency of responding plotted

as a function of the number of new words added per day are presented in

Figure 5. These data indicate that, as the number of new words added

per day increased, the changes in frequency from the last presentation

on day (n) to the first presentation on day (n+l) decreased, suggesting

that the total frequency of responding itself was decreasing as a func-

tion of an increased number of new words. Similarly, as the number of

new words added increased, the frequency of correct responding decreased.

The frequency of incorrect responding appeared to increase as the number

of new words added increased. These results might be supportive of pre-

vious findings that small step sizes produced higher frequencies of total

responding and correct responding and lower frequencies of incorrect

responding.


Repetition

The mean number of repetitions per word under each method are pre-

sented in Table 5. A Friedman's X2 test, conducted on these data, indi-

cated that the methods differed significantly (X2=12; df=2; p < .01) in
r
the mean number of repetitions per word provided. Method III produced

the greatest amount of repetition per word. Method I produced the

second highest amount of repetition per word, and Method II produced

the lowest amount of repetition per word. In light of the fact that





-38-


I .


0 0 0


> $ >.


0; 0



,-4 $-4 4, .
0 0 C. 0O
-i -i Q )0 l

S0 H 0 $H



C CZEOC
0.00.0 00C0
,..4 -. $ -4







C 0i l
r1 I
i-


8 .4


I I

0


31ntllnl Z3d SL-TM33AOH


0 O
On


0 0
o O
O r


:0


,I


I 1


I


-





-39-


Table 5

MEAN NUMBER OF REPETITIONS PER WORD


Subiects


Jesse

Sharon

Robin

Debbe

Aubrey

Michael


Methods

II

13.1

14.6

15.2

16.7

17.3

25.8


III

34.0

46.5

64.5

57.0

56.5

55.5


--


-----





-40-


Method III produced the highest amount of retention, Method I the second

highest amount of retention, and Method II the lowest amount of reten-

tion, the data on number of repetitions per word support previous

findings that increased amounts of repetition resulted in higher reten-

tion. The repetition data under Method II also suggested that, when not

limited by methodological factors, all the Ss took about the same num-

ber of repetitions to "learn" a new word.


Celeration Ratios During Method IV

Celeration ratios, presented in Table 6, were computed for fre-

quencies correct and frequencies incorrect during Method IV. A Fried-

man's X2 test was conducted on these celebration ratios. The results

(X2=10.3; df=2; p < .01) indicated that the different methods differed

significantly in their effects on the Ss' performances during Method IV.

Performance on Method II words showed the greatest amount of behavior

change during Method IV. Performance on Method I words showed the

second highest amount of behavior change during Method IV. And, finally,

performance on Method III words showed little behavior change during

Method IV presentation of all previously presented new words.

Accuracy on Method III words was at a ceiling following acquisition,

meaning that retention was 100 percent or that the Ss were reading

Method III words as accurately and quickly as they could. Therefore,

correct performance on these words could not have improved. The celera-

tion ratios also indicated that speed of performing the reading task

was at a ceiling or could show no further improvement on Method III words,

something which the single retention measure could not have possibly

shown.





-41-


Table 6

CELEBRATION RATIOS DURING METHOD IV


Subjects


Jesse

Sharon

Robin

Debbe

Aubrey

Michael


I

6.36

48.36

4.89

137.67

20.00

15.84


Methods

II

2.01

176.33

136.20

2783.33

21.33

94.38


III

.77

.75

1.10

1.17

.36

1.69


I~





-42-


Accuracy at the Median Performance During Method IV

An accuracy measure, percent correct, was computed for the median

performance during Method IV. These accuracy measures are presented

2
in Table 7. A Friedman's X test, conducted on the median accuracy
r
measures, indicated that accuracy on the different sets of words did

2
not differ significantly (X =1.3; df=2) at the median performance

during Method IV. Thus, accuracy of performance on the different sets

of words was attained rapidly during Method IV conditions. However,

the celebration ratios indicated that improvement was continuing beyond

the median performance. Such improvement probably can be explained by

the speed of reading the sets of words. Reading is a behavior involv-

ing more than just accuracy. As with all behaviors, it also involves

speed or time. A highly accurate reader who takes ten minutes to

read ten words is less likely to be called an accomplished reader by

a teacher than one who reads those ten words accurately in less than

one minute. The more traditional measures of accuracy, such as single

quantitative posttest scores, do not assess reading speed or improvement

in it. The procedures and measures used in this study provided a

means of independently assessing both parameters, accuracy and speed,

of accomplished reading behavior.


Summary of Results

Table 8 presents a summary of the findings on each variable and/or

measure used for the description of the three programming methods. For

purposes of simplification, the methods are ranked 1, 2, and 3, respec-

tively, in terms of which had the greatest, next greatest, and lowest





-43-


Table 7

ACCURACY SCORES FOR MEDIAN PERFORMANCE
DURING METHOD IV


Subjects


Jesse

Sharon

Robin

Debbe

Aubrey

Michael


I

97%

99%

99%

98%

91%

96%


Methods

II

99%

96%

99%

99%

88%

91%


III

95%

96%

97%

97%

94%

96%


_ _~___~ __




-44-


Table 8

SUMMARY OF THE RESULTS


Variables


Step size (number of new
words covered)

Repetition

Retention


I II III


2 1 3

2 3 1

2 3 1


Celeration
change)
sition

Celeration
change)
IV


(behavioral
during acqui-


(behavioral
during Method


1 3 2




2 1 3


Accuracy at median in
Method IV


Methods


i_~ ~


Methods


--~-=L-~ __ ___




-45-


effectiveness with respect to each measure. The rankings were based

on the size of the measures of the particular effects. Where no

differences were found statistically, dashes instead of number rank-

ings were substituted.

One can readily see that, using traditional discontinuous posttest

scores, Method III produced the highest retention scores. Programmed

instruction and verbal learning researchers might say that the high

amount of retention produced by Method III was a result of the fact

that the smallest step sizes and/or greatest number of repetitions

per word occurred under Method III conditions. The smallest number

of new words were learned, however, using Method III. It certainly

seems logical that one must reduce the total amount of material one

covers if one is to increase the number of repetitions made on that

material. Method III conditions did not produce the greatest change

in reading behavior during acquisition. It appears that, on the basis

of Method IV data, the Method III procedure utilized during acquisition

was sufficient to produce performance, measured in ter.s of both speed

and accuracy, at ceiling level.

Method I produced the highest rates of behavior change during

acquisition. In addition, Method I produced higher retention scores

than Method II. The higher retention scores may be related to the fact

that Method I produced smaller step sizes and greater repetitions per

word than Method II. Although accuracy showed improvement during

Method IV conditions, performance on Method I words was fairly close

to ceiling level following acquisition. It may be that the primary




-46-


gain during Method IV in performance was in the speed of reading

Method I words.

Method II produced little sign of over-all behavior change during

acquisition and a low amount of retention following acquisition relative

to the other two methods. However, the largest number of new words

were learned to criterion in this method. Step sizes in Method II

varied considerably from day to day. As the number of added per

day increased, frequency of correct responding decreased and frequency

of incorrect responding increased. Performance on Method II words showed

a striking improvement during Method IV conditions. Accuracy of

reading improved quickly; it was equal to the accuracy of the other

two methods at median performance. However, speed of reading continued

to improve dramatically beyond median performance. In fact, it may be

that the primary effect of Method II was to retard speed of reading or

frequency of responding as a result of the constant addition of new

words. When the retarding conditions were removed in Method IV,

frequency of responding was allowed to increase rapidly.

It is interesting to note that the measures of continuous change

in behavior do not show much relationship or similarity in rankings to

the measures of discontinuous change. Method IV discontinuous measures

show no differences between the different sets of words in median

accuracy of performance, but continuous measures (CR) certainly indicate

that differences in performance on the three sets of words exist under Method











IV conditions. Similarly, continuous measures of behavior during

acquisition indicate that more behavior change was occurring under

Method I conditions than under Method III conditions. The discontinu-

ous measures of retention alone might not have led to such a conclu-

sion. The importance of using measures of continuous change as well

as measures of discontinuous change in fully describing different

procedural effects should be adequately demonstrated by these data.

One type of data is independent of the other in its descriptive

ability. Both are necessary for complete description.
















Chapter 4

Discussion


The notational system proposed by Findley (1962) for describing

complex operants was used in the present paper to conceptualize the

three methods of flash card programming and associated behavioral differ-

ences within the framework of operant conditioning research. This nota-

tional system was not applied to the methods and the findings until af-

ter the data had been gathered and analyzed. As a result, the following

discussion of the data within that framework is highly speculative. It

is, however, also quite suggestive of further studies which might help

in isolating or defining the variables responsible for the behavioral

differences existing between the methods of flash card programming.

A complete description of the methods of flash card programming using

Findley's system is presented in Appendix C. For purposes of discussion

here, some important points must be made. In Findley's system, an option

is a parallel arrangement of operants. A single word to be read provides

the sample stimulus in a simple matching-to-sample discrimination task.

There are two operants which the subject or student might emit in the pres-

ence of this stimulus: (1) the appropriate matching vocal response or (2)

an inappropriate matching vocal response. These two possible operants

compose a single option. Each method of flash card programming can be

described as a random serial arrangement of ten options (ten words) or a

chain of options.


-48-





-49-


The problems rendering speculative the use of his system here arise

from the following two assumptions: (1) "being correct" on a single

word was a conditioned reinforcer and (2) "being correct on all words"

in each set of ten words was a conditioned reinforcer. Finally, the

definition of response requirement was limited somewhat in the present

application of his system since response requirement was defined as the

number of new discrimination to be acquired rather than the total

number of words to be read.

The value in using Findley's notational system lies in the fact that

it provided a means by which to describe the different programming methods

as different stimulus conditions occasioning different schedules of rein-

forcement delivery. The discussion will focus primarily on this aspect

or this conceptualization of the behavioral effects of the different

flash card programming methods.

The total chain of options composing Method I was on an FR 10 schedule

of reinforcement. In other words, the response requirement for delivery

of the conditioned reinforcer, "being correct on all words," was the ac-

quisition of ten new matching-to-sample discrimination tasks. Such a

schedule should produce a fairly high rate or frequency of response and

a fairly stable frequency of response, which it did, in fact, do. The

celebration coefficients and the celebration ratios for Method I phases in-

dicate a high degree of behavior change over time. In fact, both sets

of measures were higher for Method I than for either of the other two

methods. In addition, it appears that the ratio requirement was not too

high since "strain" in the form of decreased response frequency did not

occur (Morse, 1966; Reynolds, 1968).




-50-


The total chain of options composing Method III was on an FR 2 schedule

of reinforcement. The response requirement for delivery of the conditioned

reinforcer, "being correct on all words," was the acquisition of two new

matching-to-sample discrimination. The response requirement for rein-

forcement in Method III was lower than the response requirement for rein-

forcement in Method I. Frequency of reinforcement would, therefore, be

higher in Method III conditions. The frequency of response required to

obtain the same frequency of reinforcement in Method I as in Method III

was higher. The celebration coefficients and celebration ratios for Method

III phases indicate a lower rate of responding, both in accelerating and

decelerating directions, and a lower rate of behavior change over time

in Method III than in Method I. It appears that the response require-

ment differences and related frequency of reinforcement differences be-

tween the two methods occasioned by the differences in stimulus conditions

could account for the differences in frequency of responding between the

two methods.

The total chain of options composing Method II was on an FR 10 schedule

of reinforcement that could never be achieved. Approaching a response

frequency sufficient for reinforcement occasioned changes in the stimulus

conditions such that a state of non-reinforcement was reinstated or main-

tained. In essence, a high frequency occasioned conditions of non-rein-

forcement. This state of affairs is similar to those existing under FR

schedules with a very high response requirement. Such schedules produce

"strain" and/or long pauses immediately after reinforcement, resulting

presumably because reinforcement is a stimulus for conditions of non-

reinforcement. The celebration coefficients and celebration ratios ob-




-51-


trained for Method II did not adequately describe the cyclic changes

occurring in response frequency under Method II conditions. If one

will look at Figure 6 and the Method II graphs in Appendix A, one will

notice that high response frequencies are followed by low response fre-

quencies. Perhaps the low response frequencies indicate "strain" or

pausing occurring as a result of high response frequencies which occasion

conditions of non-reinforcement. If this is the case, then the low over-

all rate of response and rate of behavior change, indicated by the celera-

tion coefficients and the celebration ratios, could be accounted for in

terms of the periods of responding and the periods of "strain" or pausing.

Method IV conditions apparently do away with pausing, thus allowing re-

sponse frequency to increase over-all. Perhaps Method IV conditions pro-

duce such high frequencies of responding on Method II words because the

response requirement was, in essence, lowered.

Much of the presentation of methodology and discussion in Findley's

analytical terminology is speculative. A study designed to test these

speculations would be most valuable. One could, for example, leave

Method I as it was in an experimental test of these hypotheses and change

Method III so that only two words were presented at a time, thus making

the response requirement definition more in line with present usage.

Method II could be designed so that an extremely large number of words

had to be performed accurately before the conditioned reinforcer, "being

correct on all words"or some other reinforcer, was delivered. To correct

for the assumptions about "correctness" as a conditioned reinforcer, pri-

mary external reinforcers could be used as reward for responding correctly.

Then, under what are here termed Method IV conditions, raise the response




-52-


requirement for reinforcement in Methods I and III, while lowering it

in Method II. The results of such a study would be quite interesting.

Apparently raising the response requirement in Method III words during

Method IV conditions had little affect on frequency of response in the

present data. However, raising the response requirement for Method I

words during Method IV conditions did produce an increase in response

frequency. It appears, from the above discussion, that beginning ac-

quisition of a new set of words under conditions requiring a high rate

of response, followed by a lowering of the response requirement for that

set of words, might in the end, produce reading behavior acquisition

faster, i.e., more words read at a high rate accurately, than other

methods.

With regard to questions that might be ashed by teachers about which

method to use to achieve accuracy or speed or some combination of both,

it appears that Method I will produce the fastest change in reading be-

havior during acquisition of that behavior. Retention will be adequate.

If accuracy at ceiling level is the goal, without regard to the speed

with which a student moves through a pool of stimulus words, then

Method III will suffice. However, if one wants to move quickly through

a pool of stimulus items and maintain a high level of accuracy, then

Method II in combination with Method IV might be the best procedure to use.

These data should not be taken as universal without regard to the

student. For example, consider Michael's data (Figures 6, 7, and 8).

His performances in Methods I and III were fairly typical. However, his

Method II performance was atypical. He only performed on 30 words in a





-53-


31nNIN 123d SLN34W3AOVi


O
N












`(I-






0-N
C







0









O
V,


0
-

0
-c

0


0
o
--



C

W Q


0



o
-o)
SQ)













rl






-54-


" L) o co
0 a0
0 u
u 0
U'

O0 0* -
c u)
so ai c 1


ri-P
Z3 ) C )

II II n

U o


c6


o n


o 0
0


31nNi'N u3d SL1N3V3A0/i


(NI










or


0


O .l L ,_ I t I ,
O --L


o

O o


O O























be
.14
ua
0









.-0
-oe
^ T

p> c
0 ^

^ 01


0

o
o o


___ _


I -I ..


S ,





-55-


- ,

















0 o
4 o

0


6n
*0*O


3.0nNIN 'd3d SIN3V3AO/4


.0
oa a
cc
O

~O Q O
<


-J


0 .
c,1 ,







Cb
0 a





-O
C3 a)



bC






&

"-.4

3
&C





-56-


period of time that allowed other Ss to perform on 50 or more. He did

not show the half log cycle separation between frequency correct and

frequency incorrect that the other Ss showed. Instead, his frequencies

correct and incorrect fluctuated back and forth with sometimes one and

sometimes the other being higher within the same cycle area on the graph.

Michael was the only black student used. He had failed one grade

and was in a remedial reading class of primarily black students in a

middle class white school that was integrating. It is possible that

"being correct on one word" was not a sufficient reinforcer for Michael'a

frequency of correct responding to be maintained at a level similar to

that shown by the other Ss. This conditioned reinforcer may have lost

any strength it might have had initially in the face of the high require-

ments for "being correct on all words." Perhaps Michael needed a higher

frequency of reinforcement than the other Ss to maintain other than per-

functory attentive behavior.

"Being correct" is a conditioned social reinforcer among middle

class white families through its association with praise, approval, and

more primary reinforcers. However, social approval and praise are not

as frequently used as reinforcers among black family members (Risley,

1969). It may be that Michael has not had sufficient experience with

"being correct" as a reinforcing event for it to affect his behavior.

One means of testing this idea might be to make a primary reinforced con-

tingent on Michael's emitting the appropriate response to each single

word and to all of the words in a set presented under Method II condi-

tions. Perhaps his performance would more closely approximate that of

other Ss under conditions of primary reinforcement.





-57-


It should be noted that, without the use of direct and continuous

recording procedures, such speculative and hopefully testable analyses

of the variables controlling the behavior under observation would not

have been possible. It is true that, on the basis of the discontinuous

retention scores following the acquisition stage, median accuracy

scores in Method IV, and total number of words covered under each method,

one could predict that Method II followed by Method IV conditions

would afford the quickest movement through a pool of stimulus items with

accuracy at completion being equal to any other method. However, one

could probably not have predicted or known that little apparent behavior

change would occur during initial acquisition using that procedure with-

out the use of direct and continuous recording. Similarly, since the

above discontinuous measures do not account for speed of reading, one

would not have been able to differentiate among the three flash card

programming methods with regard to this particular aspect of reading be-

havior. One would never have known that Method III conditions alone pro-

duced the highest accuracy and speed of reading, while Method IV condi-

tions had to be added to the other two methods in order to produce both

aspects of reading behavior at ceiling level of performance. In essence,

without a full description of acquisition behavior during a given pro-

gram, one cannot adequately utilize the knowledge of the final product

produced by discontinuous measures of behavior.

Direct and continuous recording of behavior generally requires the

observation of individual subjects. In this study, statistical signi-

ficance meant that all Ss showed the same relative differences in their

performances under the three different methods of programming. Group





-58-


statistics and procedures would probably not have done violence to the

data in this case, i.e., group summary procedures would not have masked

the relative, gross differences in performance between the three methods

because these differences did, in fact, exist. However, without the

use of direct recording procedures and the multiple baseline procedure,

one would never have known whether group procedures were masking differ-

ences or not. It certainly lends power in the form of reliability and

generality to the findings in this study to know that all of the Ss

showed the same relative differences in performance.

Where performance under a given method differs from that of other

students, as did Michael's in Method II, direct recording would enable

a teacher to first see the difference and then begin manipulating possible

variables to change the performance. One certainly cannot change any-

thing until one first sees that there is something happening to be changed.

One must be able to see a behavior before one can do anything rational

and systematic about it. Knowing retention scores tells little, if any-

thing, about the necessary preliminary behaviors engaged in to produce

that retention score. Knowing Michael's retention score would not

have given a teacher any idea that possibly "being correct" was not a

strong enough reinforcer for him under the conditions imposed by Method

II. It is the variables controlling the behavior producing the reten-

tion score which must be changed if one wants the retention score changed.

And to do this, one must observe the behavior to see if different vari-

ables do, in fact, produce the desired change.
















CHAPTER 5

Summary


The purpose of the present study was to describe the continuous,

on-going performance of students as they progressed through a pool of

discrete stimulus items under three different methods of programming

flash cards containing these stimuli and to compare these methods.

Six Ss were studied. They differed in age, sex, and adequacy of

reading performance. Curriculum differed for each S. Young Ss re-

ceived basic English reading words, while older Ss received Spanish

vocabulary words.

A multiple baseline procedure was used in which each S received

each of the three different methods of programming flash cards.

Methods I and III required Ss to perform sets of ten new words and

sets of eight old and two new words, respectively, to criterion (each

set entirely correct for three consecutive days) before receiving new

sets of words. Method II required criterion attainment on each

separate word before receiving a new word.

Frequency correct and frequency incorrect were obtained daily for

each S under each method. New measures describing continuous changes

in frequencies over time were used for describing performance differ-

ences during acquisition. Discontinuous accuracy measures were used to

describe the effect of the methods following acquisition.

A re-acquisition phase, Method IV, followed the acquisition phase.


-59-





-60-


For each method, all originally new words were presented together for

four days. Behavior measures over time and accuracy measures were ob-

tained for this procedure.

The general conclusions were as follows: (1) Method III produced

highest retention. Day to day behavior change was low. The re-acquisi-

tion phase was not necessary to increase either accuracy or speed of

reading, which was at ceiling level following acquisition. (2) Method

I produced an intermediate amount of retention. Day to day behavior

change was highest. The re-acquisition phase produced increases in

Method I performance to equal that of Method III performance. (3)

Method II produced lowest retention following acquisition. Little sign

of over-all behavior change appeared during acquisition. However,

striking improvement occurred during re-acquisition, bringing median

accuracy up to equal that of Method III with speed of reading continuing

to improve beyond the median performance.

Since Method II produced acquisition of the greatest number of words,

it appears that Method II in combination with Method IV programming pro-

cedures produced maximum efficiency, in terms of speed and accuracy, of

progression through a pool of discrete stimulus items.

The study further demonstrated that the use of the newly developed

continuous measures of behavior change in combination with the more

traditional measures of discontinuous change provides a fuller descrip-

tion of the effects of procedures on performance.






































APPENDICES

































APPENDIX B

Notational System Proposed by Findley 119621





-63-


uL
x r- o -
-\ -

'-. z 0












0 0 -
a)
r- 0N









44
-. uC r- ,

c-0
Co 4


00 0

























O 4o- No o 3
N- ooo
-0














\NVs '! 3 0






-64-


0
CA


0




u- 14
O



0 0
a -o
03W- U x


S 0U
m u c





\ c- o

3)A

l -
a 3 -





< o



0CO23 > w
O-


o0

cO








4 ) 0 T o 4
LSU)











o o O
*- .0



-N IS
^- o <"Y
\ l~b *~ CL3':~V Fr







-65-


.- o r-
Ir.-. 0 r,.- .


0
oN O
H M



C *


0 en c -
u 0


C 0






9- a;0
S0 0
m O


SC-
O




,o I 0
M lu a u




















0
003



uo

3:


-a



-C-i z







0 -y









31iNIEJ U3d SIN3W3AON FzI






-66--


C) O



UN C 9 O -
r o C0
1 ' 01
U
u u





LU U








a)a
LC U




o_
O 00 Cc'
o 0o 0 o O a-










S.


























031NIN o SIN3-3AO-
o 0








0 w









0 o

U) U








o r-
o 0 N O, "






























































ce


31nNI Ul3d Si3N,3AOYi


































































0o 0 o 6-'
- o O
*0 0S
3J1flNI- U3d SIN3.W3'AOV4


-O
o


.0


-o
0
.4



0

r
*r-


0u


co


0 \
0 o
0
O





-69-


U U a
r? n O
QCN Q) -0 Cl
0

("- CN U *





1-44
U U
c
Sr0 -u o o

V- O




S0
6 L O


00
















O23
3'k~ o3 S-
\ \ I















0t\, 0 0




o ) '-
"3 L3NIV' U3d SiN3'/<3ArOY





-70-


C a0
C r 00 CO-


0

ol
O



\ -O




SII I Q




-,4









/ L O
\n i










O
0 O
\0 0






< o 0
ii > ^^ \ ^













f > -

<\O I


31iNIlN d3d SI3NW3AOV4






-71-


0o
('4




















(i-










-~0)













Ai~


u (
U 0
> 0j N C


o >S
C U


0 *w


0 0 c^
U0 0


R a
0 0 1
0 0 0




C I
f. ' *-----------_-














_ \_____ _____


1.


0
0 LD


:J

-0 0
o
-O 0




4

)U









^ g


31NflCln1, J3 SIN3V,3AOV


cc n cc


-7 -
uu f r-' 1C
- 00
0 ~ U
S-. c
C O
C C
0 .-4 C. 00

I-)
N \0 CO


O 0
0
0


I i

o 0 o
Go
*


I I 1 I I I I 1


- - --


I
0 A


~ L


,I







-72-


OJ I I I I O



O
o0 0





Vt0
QQ-




O 1
S: 0
o u ,




o o u- m e
3lucl u N ---
) D 0 o Uc 0



: O m Q oo Co
I- c ;j co 0
\ C' U >o
A 4 C Qe u












> (
oo 0 W






uN u

r w







ftIA I \ d S aI








00 00 oul -J w






-73-


O




oo o
(0






u) c0
Pa r j 0' O


0 o -C --. op







0 0




x 0o

T4 5 . .








CL


o- -I t o



-4C
------^ ~---'------ 5








0 f I n O
\ -*





c 0
o l "I


DinNIN 83d SIN3W3AOW


""



















' OI. I 0 I I O
NH










U in r-9
0 C O0

"'0
\ L) CM 00


UU

o 0 O



C 0




co -
1)0 0
















S ---O v
4 ca



















\ o OO O < -- -- m --
O0 3- -0o O0
- 0 .-..









31NIN 3d SI343AO
-ac
03











31NI Y
^--04?-



































APPENDIX A

Individual Graphed Data












In Findley's system, an option is a parallel arrangement of operants

and is a unit of analysis in its own right. A single word to be read

provides the sample stimulus in a simple matching-to-sample discrimina-

tion task. There are two operants which the organism might emit in the

presence of this stimulus: (1) the appropriate matching vocal response

or (2) an inappropriate matching vocal response. In the study conducted

here, producing the appropriate matching vocal response resulted in the

delivery of an assumed conditioned reinforcer, i.e., placing the card in

the correct pile or "being correct". Producing an inappropriate matching

vocal response resulted in an assumed conditioned punisher, i.e., prompt

by E, imitation of prompt by S, and placement of the card in the incor-

rect pile or "being incorrect". Prompts were included as part of the

punishment because occurrence of prompts provided a signal that "being

incorrect" was at hand. The option described above is a nonreversible

option since the first emission of a response of either class removes

the conditions for the other operant, i.e., if an S was correct, he could

not then be incorrect and vice versa. The following diagram represents

the option involved in making a reading matching-to-sample discrimination

to a single word:

op 1
(word correct) Sr ("being correct")


FR 1 3&4


op 1
(word incorrect) Sr- ("being incorrect")


FR 1 3 & 4


-76-




-77-






The open arrows on the vertical line indicate nonreversibility. The ar-

rows on the horizontal lines indicate that the alternative behaviors

of operants 1 and 2 both produce the conditions for operants 3 and 4.

More specifically, either a correct or an incorrect response to a stimu-

lus word produces first the appropriate conditioned consequence and

second the presentation of a new stimulus word. The procedures of each

method of programming flash cards conceptualized as the serial arrange-

ments of single matching-to-sample discrimination tasks can be diagrammed

in the same manner. Such a diagram is presented in Figure 21. It can be

seen that the S moves through ten options with twenty possible operant

behaviors in each method. The completion of the tenth option within

each method sets the condition for presentation of the ten options in

another method or delivery of primary reinforcement which in turn sets

other conditions.

Findley does not provide a label for the serial arrangement of op-

tions. A chain in his system refers to the serial arrangement of single

operants; however, there appears to be no reason why the label, chain,

cannot also be applied to the serial arrangement of options. Therefore,

it can be said that each method is composed of a chain of options.

Three chains must be completed before presentation of M & M's (primary

reinforcement). The third primary reinforcement delivery sets the con-

ditions for ending chain presentation that day and beginning it again

the next day. Figure 22 presents a diagram of movement through the chains.

The chains are presented in random order to the Ss by E.

Each chain may be conceived of as an option itself, i.e., all of





-78-


0






o 0


a 0 0
u w o

0 3: 0 4 \c




0a
E


Mv
















UU
),C









04 CO)


0 1H 4-4
o o
CJ
4-a
1 a
00

U4 -H 1C)
A





























o 'uH
-4uc

o U C-4
CU 0 u n




0 0 0
004-1
00 0 oJ





S0 a




aQ a,

4-- ca cO




$ -I 0 Q(
o u y
CL u cz &. \ Zo
0 0 4O ;4

0 1 o4
0 00
1 4






-79-


U) *



0) 0
u au



o. P-1 4-W










^-z
a)















pU
cn


SA A
V) W V) ) V)c




c cs c co l clN




UU U


UI l
Ad d d d
a c c co ct c










C1 C C CC C4



ca \ c c \





CO C CO








0-<-


0
4o



o E























A 0
0)
a)

























0)






W,




-80-


the words are read correctly or not all of the words are read correctly

(some may be read incorrectly). A second conditioned reinforcer might

be assumed to operate at the single chain level, i.e., "being correct

on all words". Similarly, a second conditioned punisher might be as-

sumed to operate at the single chain level, i.e., "being incorrect on

some of the words". It is at this level that the differences in methods

can be described.

All methods, seen as chains of ten separate options, produce the

conditioned reinforcer of "being correct on all words" on an FR 10 sche-

dule. However, the response requirement for delivery of this conditioned

reinforcer differs from method to method. If one defines response re-

quirement as the number of new discrimination to be acquired, one can

see that Method I has a response requirement of ten words per conditioned

reinforcement delivery, while Method III has a response requirement of

only two words per reinforcement delivery. Therefore, it might be said

that Method I provides reinforcement delivery on an FR 10 schedule,

while Method III provides reinforcement delivery on an FR 2 schedule.

Method II also provides reinforcement on an FR 10 schedule or has a re-

sponse requirement of ten words for reinforcement delivery. However, pro-

cedural conditions are such that as the individual approaches satis-

faction of the response requirement and reinforcement delivery, stimulus

conditions change (new words are added) to reset the response require-

ment back. In other words, approaching reinforcement delivery becomes

a stimulus for conditions of nonreinforcement. In a sense the response

requirement is constantly being raised immediately prior to attainment

of conditioned reinforcement, thus making reinforcement delivery an im-




-81-


probability. A diagram of the above descriptions is presented in

Figure 23.

In general, one might say that responding or movement through the

options making up the chains is maintained by the primary reinforced,

M & M's. Acquisition of the single matching-to-sample discrimination

tasks is controlled by the first conditioned reinforcer of "being cor-

rect" on a single word. Any differences in rate of acquisition or fre-

quency of responding might be attributed to the differences in response

requirements for the second conditioned reinforcer, i.e., "being correct

on all of the words".





-82-


, A L A



0

' -





4--4- -



,$-4
0






o




u U

-"

0 0
)2
41
o














ca ca
00)










0
-M *H -U 4H



-e O w






a0














-H
0 r4
0 g U













1-1 a





































APPENDIX C

Intra-day Performance












Frequency correct and frequency incorrect were obtained for each

separate presentation of each set of ten words, allowing for observa-

tion of intra-day changes in performance. There were three sets of

frequency correct and frequency incorrect per day for each method. The

following graphs provide examples of these data. Each set of three

connected data points were obtained on a single day.

Although these data were not analyzed, the technological addition

of this procedure to the continuous analysis of behavior could be valu-

able in doing more refined analyses of practice effects, for example.

Or perhaps one could discover how much practice (with and without rein-

forcement) to require of a student daily to get him to acquire some beha-

vior within a given period of days. An interesting experiment might be

discovering some daily criterion performance for each student which

would enable him to progress through a given set of words within five

days. Further, one could assess how the addition of some specific rein-

forcer might change the daily requirement, if at all. In essence, such

a technology provides a description of changes in behavior without large

amounts of time intervening and allows a comparison of intra-day changes

with daily changes.

In the figures presented, note that Aubrey, in Method I conditions,

takes longer to get a set of words "all correct" the first time than

does Jesse. Once he finally achieves a set "all correct" he continues

to have difficulty maintaining the performance after a long lapse of

time. However, after the first presentation of the day, he quickly re-

turns to near perfect performance. Jesse more quickly reaches the point


-84-





-85-


* ii
4s-


K---


I


H '
-*?.
"ri
~~1<


to
uC

5"


< i






J.
13 4


I I
0 0
0 L0


__^- _


II I I '
0 o t -- Uo -O
_ O Oo
3 S
3 10IN: b3d SVL13qW'3AO'


1 ,


0 0
0


5


I I


0
-1

0
o)

0

o

Oe
=

Owi





O4
o -

o 5






-0





Lr)
:o C

0 -
o
O












o







.,-4
rM



r


-


*




-86-


0





o


0
O




o i



-CO
V)





o U
I- I

ow
n)
o7
o "
Q 0








00 o






00 Q3 010 d 30 ,3AO"
L 0
31InIN13 933 0 JLN3W33AON





-87-


L I ,. I I a9









Sl 0 0
S0-














00
0 W
-o


.- 0O





.^ <
^L-C










J--.^JW4 JnO
8g ^ .-- -g 03 g
-~~~~~~ 3i"l 3( LNMAO





-88-


00 0o
0 -


u L -t r -
-- 9 O o O
0 O
31tri(ION 13d SiLM3'33AOVi


~O
co

0
)o

.0
(D T
Z I
O O "-
r\ 0 c
*? ,


o







-0
-a
o_ do
u


-2





-89-


o
1_ I II I 13 I I I -- -I i -



f I! -



o\
occ
** ---- fi~---------- -^

0)
lz ..





-0

":






CD
) 0
I I .O <
o 0 --, 0 0 o













S10"o -0 .O O
0 0 A"
c. " ^'10 ^ <
<. <~ ,
^-^ > -0
__----_ _------ c,
^ ^
hl i^K,
*' ^ 0 >
^
^ ^ -
*~ -~ -. r ^
.C^
8 g S -" 8 5 j g
3nMN ~dSi3WAO '




-90-


0



(D 4
0
-(1

_O
-or



-o
.0



o


O )
o >







o o
O

















OO Q)
vi 1



10 u O
j s




o 2







3 IN 83d s





-91-


I I I!


0
J


1 a~


t-. '. -"-




I I I
0 o 0 0 to O tO -
S o O 8o
) 0


0
-4"

0



ci





o
O0








o -


o
(J





-O







O

o "
-0
0 *



V I

c?




0













I4


XI4


3inrIN 3d SLN3W~3AOM


C
0


f




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