Citation
An exploration of character emission rates in composition

Material Information

Title:
An exploration of character emission rates in composition
Creator:
Merbitz, Charles Timothy, 1946-
Publication Date:
Copyright Date:
1975
Language:
English
Physical Description:
vi, 57 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Adjectives ( jstor )
Alphabets ( jstor )
Creative writing ( jstor )
Literary characters ( jstor )
Nouns ( jstor )
Philosophical psychology ( jstor )
Pollutant emissions ( jstor )
Topography ( jstor )
Writing exercises ( jstor )
Written composition ( jstor )
Dissertations, Academic -- Psychology -- UF ( lcsh )
English language -- Composition and exercises -- Research ( lcsh )
Psychology thesis Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis--University of Florida.
Bibliography:
Bibliography: leaf 56.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Charles Timothy Merbitz.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
025487163 ( AlephBibNum )
02972994 ( OCLC )
AAT8954 ( NOTIS )

Downloads

This item has the following downloads:


Full Text
















AN EXPLORATION OF CHARACTER EMISSION RATES IN COMPOSITION


By

CHARLES TIMOTHY MERBITZ





















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




UNIVERSITY OF FLORIDA

1975

















ACKNOWLEDGEMENTS


I am especially indebted to Dr. H. S. Pennypacker,

the Chairman of my Supervisory Committee, and to all of

the other members of my committee, for their unflagging dif-

ferential positive reinforcement of successive approxi-

mations of productive professional behavior, both with

respect to this paper and to other work. Dr. M. K. Gold-

stein, Dr. John Newell, and Dr. Frank Sieka were especially

instrumental in my progression here.

I am also deeply grateful for the support and en-

couragement given me by my wife, Gloria, without whom even

the beginning of this undertaking would have been doubtful,

and who gave crucial support to the whole enterprise.

Special thanks go to David M. Smolen, who wrote

the computer programs used and gave help and valuable ad-

vice on data analyzed.

I wish also to thank the people who volunteered to

serve as subjects, often giving time and effort well beyond

any possible formal reward. I cannot express my apprecia-

tion deeply enough for the efforts of all these people.

ii











Physical support was provided by several organi-

zations. The Personalized Learning Center helped greatly

with some office space, a stopwatch, and some computer

terminal time. Dr. M. Branch lent the Running Time Meter.

Finally, the University of Florida College of Arts and

Sciences and the Northeast Regional Data Center jointly

provided the computer funds used.


















TABLE OF CONTENTS


Page
ACKNOWLEDGEMENTS . . . . . ... .... ii

ABSTRACT . . . . . . . . . . . v

INTRODUCTION . . . . . ... . . . . 1

EXPERIMENT 1 . . . .. . . . . . . 8

EXPERIMENT 2 . . . . . . . . . 19

EXPERIMENT 3 . ... . . . . . . . . . 26

EXPERIMENT 4 . ... . . . . . . . 30

EXPERIMENT 5 ... . . . . . . . . 41

EXPERIMENT 6 . . . . . . . ... . 48

GENERAL DISCUSSION .... . . . . . . .51

REFERENCES . .... . . . . . . ... .56

BIOGRAPHICAL SKETCH . . . .. ....... .57

















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



AN EXPLORATION OF CHARACTER EMISSION RATES IN COMPOSITION

By

Charles Timothy Merbitz

December, 1975

Chairman: H. S. Pennypacker
Major Department: Psychology

Six experiments explored character and word emission

rates under controlled conditions with a variety of subjects.

Day to day stability was seen in both character and word

emission rates, but first session rates were not reliable

indicators of general rate. Specific experimental tasks

tended to control distinct rates. Both self-timing and re-

lease from tasks changed character rates, while self-timing

was also seen with longer durations of writing. The control

of character topography by consequences was also demonstrated,

and a large effect was noted. Thus, both operant antecedent

and consequent control of both response rate and topography

was demonstrated for written composition.

Both hand written compositions and compositions

written on a typewriter (computer keyboard) on various topics

v











and with various constraints were assessed. Suggestions

for easier and cheaper measurements were made. Results

were related to measurement problems in the teaching of

essays, and suggestions for teacher measurement of compo-

sition activity in the classroom were made.

















INTRODUCTION


If an experimental analysis of written composition

is to proceed, a unit of response must be defined. Find-

ley's (1962) work on multi-operant repertoires clearly

showed the possibilities of a unit that itself was com-

posed of smaller units of response. If language behavior

in general and written composition in particular are anal-

ogous to Findley's multi-operant repertoires, we may be

able to proceed with units defined at several levels. Bas-

ically, written materials are defined by combinations of

curved and straight lines and dots. Gonzalez and Waller

(1974) reported records generated by two subjects writing

on an apparatus that responded to variations in pen pres-

sure, and defined one class of responses by microswitch

closures of .5 sec minimum duration between minimally

specified IRTs. They found handwriting rates while writing

were stable despite differences in schedule of reinforce-

ment in effect but the schedule did control the pattern of

pausing such that interval and ratio components of a mult

FR FI LH are clearly distinguishable.

A larger response unit that has been considered

is the word. Brigham, Graubard, and Stans (1972) reported

1










the results of several operations on mean word counts in

a 5th grade "adjustment" class. In "Baselind'all students

got points for sitting and writing during the writing

period. Points were exchangeable for generally unspeci-

fied privileges and activities in the room. In the "Number

of Words" condition, one point was given for each word in

the stories written. In the next condition, "Different

Words," two additional points were given for each initial

use of a word in that story. Finally, in the "New Words"

phase, the children were given an additional three points

for each word that had never appeared in a previous story.

An alphabetized list of all words used in earlier stories

was provided for each child. Groups A (n=5) and B (n=3)

experienced all conditions, while Group C (n=5) was ex-

posed only to baseline and "Number of Words" conditions.

The authors reported mean numbers of words written per

group per session in each condition, mean numbers of "Dif-

ferent Words," and mean numbers of "New Words," as well as

the mean times. In addition, the stories were rated on

a 25-point scale and the ratings of each story were given.

Inspection of the figures revealed a marked similarity in

the shape of the functions for each measure within each

group. The similarity of the time function to the others

suggested that conversion of the counts to rates may re-

veal otherwise obscured relationships.










Figures 1 through 4 of Brigham et al. were con-

verted to tables and the rates of words emitted per minute

were calculated and charted (Figures 1 and 2, see data dis-

play section below for graphing conventions). Since

Brigham et al. report sessions "twice a week" on the aver-

age, T Th lines were used for plotting. It was apparent

that the major effect of the first contingency imposed

(Number of Words) was to simply increase the time the

children spent writing. Rates of word production were

quite variable, and no consistent effect was obvious. The

authors report rising quality point ratings during the ex-

perimental period, and suggest that it was possible that

the rise in ratings associated with the "New Words" condition

may have been the result of direct attempts to use new words, as

.they worked harder and spent more
time on the sentences, thereby producing
more and better ideas while actually
writing less. (p. 429)

More time was indicated on the graphs, but the word

production rates did not drop, as this interpretation

requires. Also, the mean ratings the stories received

were charted as quality units per minute emitted by each

group. Little change appeared in Group A's quality unit

production rates throughout the experiment, except for a

drop in sessions 9 and 10. Group B shows more variability,

and also seems to share a small dip for those two sessions

(among others). Group C shows the clear experimental ef-

fect of the "Number of Words" contingency-while the word









I I I


1000

500



100
50




*10 -

S5

-i
Q)
0 1


0




.05
.1




.005



.001


F rl""l""l( 1-1~r-I7


Phases:
1 Baseline
2 Number of
words
3 New words
4 Different
words


Group B


- m I- -" I


Figure 1. Chart of


Successive Calendar Days-
longer marks indicate Sundays
data from Brigham, Graubard and Stans.


- -reciprocal of. mean
minutes spent writing
*-mean words/min.
A-mean different words/min.
mean new words/min.
x.-mean quality units/min.


Group A


, . I ,


-


I I A


S "- .. .. |





























- -reciprocal of mean minutes spent writing
* -mean words per minute
A -mean different words per minute
a -mean new words per minute
X -mean quality units per minute





Phases:
1 Baseline
2 Number of words


1.TT1rr;,.1I.rJr. I II**I*11 I~ ~lI..*,~S5J*I...I*II*I~*.II


! .... ''.I. "1 ...... I" '. " ".| "
Successive Calendar Days. Longer marks indicate Sundays.

Figure 2. Chart of data from Brigham, Graubard and Stans. Group C.


I i I II 9 I I


1000

500



100 -

50



10 -
5


1

.5



.1

.05


.01

.005


. UUI


' 0 *
0










rates remain roughly stable, much more time was spent, and

a very marked decrease in quality units per minute emitted

is seen, denoting less efficiency. Plots of quality unit

and word rates for the same sessions seem to suggest co-

variation in trend, although the effect was not reliably

seen. Difficulties in further interpretation lie in

the problems inherent in converting graphed points to rate

data and the report of group means rather than individual

performances (Sidman, 1960).

In another study involving composition writing,

Maloney and Hopkins (1973) reported the use of a permuta-

tion of the "Good Behavior Game" to change the number of

different adjectives, action verbs, and sentence beginnings

in 17 ten-sentence stories written by 14 fourth through

sixth graders. They found no change in character or word

emission rates throughout the study, but did report that

selective reinforcement with extra recess and candy did

change subjective ratings of the creativity of the stories.

They did not report the rates of behavior.

Van Houten, Morrison, Jarvis, and McDonald (1974)

reported rates of word emission for children in one second

and two fifth grade classes. They discovered that ex-

plicit timing and feedback from the children counting

their own words roughly doubled the response rate, from

three to six words/min.,in ten sessions. Stories were

also rated on five dimensions, and the ratings were higher











during each of the timing and feedback phases. Feedback

was provided by letting the children count their own

words, which they did with 80-100% accuracy.

Of the three studies, only Van Iiouten et al.

showed a rate increase, while all three reported "quality"

increases during the experimental period.

While assessment of composition quality remains

an important issue, measurement of composition rates is

also needed. The studies cited above support the notion

that counting the words emitted per minute by subjects

writing essays may reveal interesting relationships be-

tween essay rates and environmental events. However, a

gap existed in units of analysis at the level of the

character between the behaviors recorded by Gonzalez and

Waller and words counted by the others. The goals of the

present work were,therefore,completely exploratory to

measure some writing rates, to begin definition of a

dependent measure of writing rates, and to examine rela-

tive sensitivity of rates of character and word emission.

The research was not structured initially; rather as in-

teresting phenomena came to light, new directions were

taken.

















FXPERTMENT T


To begin the investigation of compositional rea-

sponse rates, subjects were timed while they performed var-

ious writing tasks. One subject was seen in three pre-

liminary sessions to establish timing procedures and ex-

perimental tasks.



Method


Subjects and Apparatus


Subjects S-802, S-803, S-506, and S-805 were fe-

male college students enrolled in Psy. 201 and serving for

experimental credit. The first three were 18-year-old third

quarter freshmen; S-805 was 19. Three subjects were black;

S-506 was white. Session 1 for S-802 was timed with a

Westclox (General Time) stopwatch; all other sessions were

timed withaRunning Time Meter (Cramer Controls). Experi-

mental sessions were held in an office approximately 4 m

x 5 m x 2.5 m containing two desks, files, chairs, and

common office furnishings. Subjects were run individually,

generally during normal school hours. Paper and a vari-

ety of ballpoint pens were provided by the E. Before
8











beginning, the purposes and methods of the experiment were

explained to each subject and a standard informed consent

form was signed by each subject.


Procedure


While sitting at the E's desk, subjects were given

an experimental task number from the E's data sheet to

write at the top of a fresh sheet of paper. They were then

given specifications for the writing task to be'performed.

When ready, the timer was started and they wrote until

told to stop. They then counted the words (or characters)

written, the count was recorded, and rates calculated and

displayed on six cycle semilog Standard Behavior Charts.

The pattern was repeated until termination of the session.

Subjects were asked to indicate any excessive hand fatigue.

An informal and relaxed atmosphere was maintained, and all

questions were answered fully. Words were counted by S-802

and during S-803's first two sessions. All subsequent counts

were of characters.


Experimental Tasks


Experimental tasks included copying a text which

had been typed with six words per line on a duplication




Since tasks were standardized for Exp. 2, a more
detailed description of procedure is presented in the
method section of Exp. 2.










master. Triple spacing permitted subjects to write between

the lines and thus keep eye shifts to a minimum. Subjects

were also asked to write short essays on topics with which

they felt some emotional involvement. A free association task

(FA) was defined by instructing subjects to simply write

whatever came to their mind. Limited association (LA) was

simply all that they could think of to write relevant to a

given topic. Other tasks included listing all of the pos-

sible uses for a specified common object ("Uses for . ."),

signing one's name, writing the alphabet, and writing a list

of all the objects in the room. Subjects were also asked to

generate lists of nouns, and were later asked to list ad-

jectives or verbs appropriate to the nouns listed.


Data Display


Following a common practice in graphing dynamic

phenomena in other natural sciences, semilogarithmic charts

were used to keep and display frequency of response data.

Two general types of charts were used: In the first,

the count per minute of behavior was displayed on the

line corresponding to the calendar day the data were col-

lected. The second displayed the count per minute of be-

havior, but successive lines represent successive experi-

mental tasks within one session. All data were labelled

with true calendar days to permit unambiguous indication

of the time relationships between various experimental










events. The log scale allows accurate predictions and both

absolute and relative comparison of data from different

subjects. The dots and x's indicate frequencies of be-

havior on the indicated day, and the dashes show the recipro-

cal of the time spent observing. Thus, the lower the dot,

the lower the rate of behavior, while the lower the dash,

the longer the time spent observing.


Reliability ', r


Subjects were informed that spot checks were made

of their counts. Limited resources did not permit all

counts to be verified, but 50% of the experimental tasks

were recounted. The accuracy of match between subjects'

counts and reliability counts ranged from 88% to 100%,

with 13 % of the subjects' counts inaccurate. Letters,

numbers, and punctuation marks were all counted as characters.

Of -the errors, 55% involved miscounts by two or less characters.



Results


Frequencies of words written per minute by S-802

are displayed as x's on Figure 3, while frequencies of char-

acters written per minute (counted post hoc) are indicated

by dots. Note the parallel relationship between the two

frequencies in the copying task, in which the average word

length was fixed by the original author. In the FA tasks,








1000
500.




50 -
100



50


10







.1

o .05 -
o .05

Copying
Rates FA Rate Essay
.01- Rates

.005 -



.001
23 Apr 24 Apr23 Apr 22 May 22 May
S24 Apr

Figure 3. Chart of character (*) and word (x) emission rates for S-802 on indicated days.











some deviation from a fixed relationship was seen, partially

due to a comparatively low word rate in the first FA task.

Also, average word length changes from approximately 4

characters per word in the FA tasks to 3.5 for the essay.

Figure 4 shows frequencies of word and character

production for S-803. During sessions on 5/21 and 5/22,

only word counts were made. Note the steady rise in the

FA and LA word frequencies for those days. Post hoc char-

acter counts indicated that the rise in word count was a

result of shortening word length. (A similar effect may

be seen in the 4/23 FA rates of S-802.) The effects of

the subsequent shift to character frequencies over the next

four sessions were to slightly raise the character frequen-

cies while word frequencies gradually declined as word

length increased. Frequencies of word and character pro-

duction when the S was asked to list nouns were also shown,

as were the frequencies of production when the S was asked

to write adjectives and verbs appropriate to the nouns on

the list. Note the much lower frequencies seen with verbs.

Two lists were used, in the order ABA AB BA. Note that

list B always controlled lower frequncies. List B was gen-

erated on 5/22, and list A was the second noun list generated

on 6/5. The last two pairs of rates were generated under

the instruction to simply produce lists of adjectives. The

first performance of the "Uses for common objects"

tasks was not a good indicator of usual frequencies
































3 4' 5 '11

June
Letter and word
count


22 5
May June
B A
List
Nouns


All FA and LA tasks


S'12 June

1 adjectives
2 verbs
3 adjectives


June
"Uses for"


4 adjectives (free)


I.I.,..~r


June
Essays


Calendar days


Figure 4. Chart of character (.) and word frequencies (x) on successive experimental
tasks done on indicated days by S-803. Signature rates are plotted the
calendar days emitted.


--f I-- -I- I


50-
1000-
500


100 -
50


May
Word counts
only


52 .I


-d


1
June
Signature











displayed by S-803. Also shown were her essay rates and

her rate of character writing when signing her name.

Data from S-805 are shown in Figure 5. On two

successive days, the same object was presented (a brick)

in the "Uses for . ." task. Note the jump in frequency

of response. Essay and signature rates were unexceptional.

Four tasks (list nouns, list adjectives appropriate to

the nouns, list more nouns, and list verbs appropriate

to the nouns) involving listing and relationships of nouns,

verbs, and adjectives were presented on 6/8. Note the

slower adjective rate-the first list of nouns generated

contained all proper nouns. Also shown on Figure 5 were

.5 min. and five 2 min. FA rates which were collected on

6/3. Each half minute during the longer samples was

announced, and the S drew a line and continued writing,

thus permitting sectioning of the sample into half-minute

segments. Note the slight drop in frequency as the samples

become longer.

Frequencies of character writing from S-506 were

plotted on Figure 6. Her signature and alphabet writing

rates seemed stable, as did her essay writing rates. Her

description of the room tasks was segmented into quarter

minutes, and the total overall rates as well as the first.

quarter and half minute rates were shown. The first sample

rate was taken for .25 min., and was lower than comparable

first quarter rates,.although close to the overall rates
















I


8 June


w- '.


3 June 11 June


Successive Days


Signature "Uses for
Brick"


ouns Total
Ijectives


Essays


0)

O








.5 1.0 1.5
rates for successive
cumulative recording
time
(June 3)


Figure 5. Chart of frequencies of character emission rates by S-805. Signature and "Uses
for a brick" rates are plotted on the calendar days the tasks were performed.
Longer lines indicate Sundays. Other tasks are plotted as successive perform-
ances on the days listed.


25 May






































1 June

Signature


1 t h... i


,1 I*^ ^ "'''" "''" I. .


























1 June 1 June 1 June 1 June

Alphabet Describe room quarter time half time

Successive Calendar Days within Task Groups.


1 June

Essays


Figure 6. Chart of character rates on indicated tasks performed by S-506. Panels 5 and 6
show .25 and .5 minute recording times for the performances shown in Panel 4.


. 1 t 1 0

8 June

"Uses for"


4P

H

t-w

4J

0
U
1LL 0
c-)











rates of the longer timings. The essay and "uses for

rates were unexceptional.



Discussion


In general, character emission rates seem to be

stable and predictable. Word rates seem less desirable

as indicators, as S-802 and S-803 show a change in word

rate without parallel movement in character rates. Further

work involving fewer but more standardized tasks and

shorter session lengths but more repeated measures was

undertaken.

















EXPERIMENT 2


Two subjects served to establish the daily sta-

bility of character emission rates, and one of those sub-

jects demonstrated a change in rates. Procedures were

designed to shorten experimental sessions and tasks were

standardized.



Method


Subjects and Apparatus


Subject S-380 was a 38-year-old female college

graduate. Subject S-922 was a 27-year-old female teacher

with a master's degree. Both of these Ss consented to

serve "to help the E find out about writing." Experi-

mental sessions were held in the E's kitchen in the even-

ings after both Ss completed working days. A variety of

ballpoint pens was again provided and unlined paper

approximately 19 x 28 cm. was obtained by cutting used

computer output paper in half with a bandsaw. Subjects

wrote on the blank side. Timing was again done by manually

operating the running time meter to signal the beginning

and end of writing activities. The timer was checked against

19










the internal clock of the NERDC IBM 370/165 computer and

found to be accurate.


Procedure


Since tasks were much more standardized in Exp. 2,

procedures will be described in more detail.

A subject was seated at a table with pens and paper.

Each task was numbered, and the S was asked to write her

subject number and the task number (supplied by the E) on a

sheet of paper. At a signal from the E the subject signed

her name as many times as she could until the E signalled

her to stop. She then counted all of the characters written

and announced the count to the E, who recorded the details

on her data sheet and plotted the performance on a standard

behavior chart. The alphabet was next written under sim-

ilar conditions, followed by either a short essay concerning

a picture the E presented, or a task in which all the dif-

ferent possible uses for a common object named by the E

were listed. The final task was typically a "free-associ-

ation," in which the S wrote anything (comparable to Gonzalez

and Waller's task).


Contingency


After 6 sessions, S-922 asked when she could quit,

and at the next session she was told "we can drop any task

on which you exceed a goal rate for 5 out of 6 days." At











her next session S-380 was also presented with this state-

ment.


Stimuli


Pictures were provided to write about for the

"Picture Essay" (PE) experimental task. Illustrations were

cut from magazine advertisements or feature articles (typi-

cally color photos) and identifying text was cropped or

obliterated. An effort was made to locate provocative

pictures with an obvious point, such as those often seen in

Psychology Today.

Another experimental task ("Uses for .") in-

volved listing as many uses as possible for common objects

named by the E. Objects included rubber bands, a pencil,

paperclips, a brick, a bicycle, a safety pin, Scotch tape,

a house, a car, a park, a barn, money, a rope, and a coin.

Objects were named by the E in varying orders to each subject.

Daily charts and data sheets were kept on each S,

and were shown whenever requested. Usually, subjects merely

glanced at both.



Results


Figures 7 and 8 show the rate of character emission

per minute by S-922 and S-380 respectively. The heavy dark

line on the figures indicate the presentation of the statement








..... .... .... I

- e. l +


22 .June
22 June


A *0 e


22 June


22 June


. a .9 -*,


.1 Imrhmr



a- rI...


ti t


22 June


Signature
Mean rate 186.3 211.3
Stability
index .119 .175


Alphabet
158.6 187.6

.063 .065


Picture essay. Uses for . .


168.5 193.6

.20 .113


152.1 178

.071 .28


FA
155.8 186.8

.082 .23


Figure 7. Chart of characters emission rates of S-922 before and after contingency
arrangement (vertical line). ++ indicate repeated presentation of the same
stimuli. Performances are plotted in successive calendar days within each
panel. Mean pre and post contingency rates are given with mean stability
indexes. Stabilities were calculated by dividing the ranges of the last 5
persons by the median.


1000-
500



100-
50.


.... .. ....


., N


S, % A


22 June








1000-
500


22 June


*.*. .


22 June


S...T1 ......


22 June


Alphabet


Uses for


Picture essays


Figure 8. Chart of character emission rate for S-380 on indicated tasks. Successive
calendar days plotted within each panel. ++ indicate repeated presentation
of same stimuli. Vertical lines indicate contingency announcement.


- Itrrr3mr
.


100 -

50


10

5 -


* * *


Signature


22 June


*


.











on quitting to the subjects. Note the immediate effect on

S-922. She was writing her signature at an average rate of

183.6 characters per minute prior to the announcement, and

after it her rate averaged 211.3 characters per minute.

Similar increases may be seen in the other tasks. Note

also the increase in variability after the contingency an-

nouncement. Daily and average pre- and post-announcement

rates are given on Figure 7.

No such effect was noted for S-380. For'this S,

replication points were provided by presenting the same

stimulus object in the "Uses for . ." task on two dif-

ferent occasions, as indicated. Note also the "banding"

exhibited by this subject. Each task controls a different

set of rates of character emission.



Discussion


These two subjects provide evidence of the stability

of the rate of character emission in daily experimental

sessions. The effect of the contingency arrangement on

S-922's rate indicated that perhaps the operant conditioning

model would indeed be appropriate for written composition.

In addition, S-922 started at a very high baseline rate, com-

pared to the subjects seen in Experiment 1. The failure of

the announcement to move S-380's rates simply suggest that

it was inappropriate for her. In view of the weakness of





25




the E's nominal control over the total environments of both

of these subjects, the experimental control shown by the

stability for the rates and the movement of S-922 suggest

that character rates may well be an excellent experimental

measure and that the tasks selected produce behavior that

is readily ameniable to experimental analysis.

















EXPERIMENT 3


Since the tasks used in Experiment 2 never lasted

more than 2 min each, longer sessions and longer task

lengths were necessary to begin exploration of the effects

of these conditions on the writing rates.



Method


Subject and Apparatus


A 22-year-old male student (S-061) was promised

"experimental credit" slips toward Psychology 201 extra

credit on an hour-for-hour basis. He had received an

incomplete in PSY 201 the previous term and was allowed

to turn in credit slips as well as completing other re-

quired course work. The credit slips were due July 16,

but he agreed to keep serving after that time even though

all credit usable had been earned by that date. Typically,

20-30 min of conversation with the E followed each session.

Apparatus and procedures were the same as in Experi-

ment 2, except that essays on "emotional" topics were sub-

stituted for FAs, S-016 timed himself and sessions were held

in the E's office on campus, with S-016 seated at the E's desk. No

26











graphs or charts were made or shown to S-016 during this

period. However, his data sheet was not concealed and he

occasionally inspected it.



Results


Figure 9 shows the daily rates for S-016 on each

task. Note that with free essays as S-016 was permitted

.to time himself, the length of time he spent writing appre-

ciably increased. However, in spite of the increase, his

rates remained remarkably stable, even when he wrote for

as long as 18 minutes. After that session his time de-

creases, and then he showed a fluctuation in rate. His

signature and alphabet rates seem stable. (The first four

alphabet rates were written in capitals; he spontaneously

changed to lower case letters at the fifth session.) Four

objects were presented twice, and only one showed very close

rates on the two occasions. Picture essay rates increased

also after the procedure was changed so that S-016 time

himself.



Discussion


The data from the first 6 free essays of S-016 indi-

cate that character counts can be quite stable even for

relatively long essays (18 min.). However, later sessions

show unexplained jumps in the rates. It was planned that








1000
500


100 .'f A *

50 "


10 11 11 D
5 AB IC BA C
SN N N D N


a 1
6 July Alphabet



o Signature ...... I..i "" | 'i .
o 500
6 July S-Timed 6 July

100 Uses for
50
1 6 July Free essays

N Figure 9. Charts of character emission rates for S-016
10 on indicated tasks. (N) indicates different
5 locations for session. Letters indicate re-
peated presentation of stimuli. Successive
calendar days are plotted within panels.


6 July





29




S-016's alphabet and picture essay performances would be

charted in order to observe the effect of presentation of

the chart on both a steady and a fluctuating rate. However,

S-016 decided to terminate and the operation was impossible.

The effects of permitting the S to time himself was, in

general, to raise the rates, and to increase the time spent

writing.



















EXPERIMENT 4


One of the problems associated with investigation

of character emission rates was the tremendous tedium of

counting characters. Accordingly, a computer program was

obtained to count, time and calculate rates on material

typed in as text. Comparison of typed rates of character,

word, and sentence production seemed appropriate, and the

use of the machine permitted more efficient use of subject

time, since counting was not required. In addition an

attempt was made to investigate the effects of constrain-

ing the experimental task in different ways. Rates of

character emission of both hand written and typed materials

were collected.



Method


Subject and Apparatus


An 18-year-old female currently enrolled in PSY 201

(S-157) served under the promise of experimental credit and

help with study skills. Se accumulated approximately 21











hours of experimental credit, of which 5.5 was the maximum

useful. Her efforts to find a job were discussed with the

E, as well as other personal issues, although study skills

were not. She reported preexisting typing skills and agreed

to compose essays on a typewriter. For the first 11 sessions,

S-157 followed the same general procedures and used the same

apparatus as S-016, except that she always used and brought

the same pen to experimental sessions, and her first ten

sessions were charted as were S-922's and S-380's. In addition,

sessions 12 through 19 included work on the computer terminal,

and some tasks were dropped. (See Figures 10 and 11.)

Two sessions (12 and 13) were held in a room as-

signed for student computer terminal use, containing ap-

proximately 20 assorted computer terminals and keypunches.

Since S-157's session time was approximately noon, the ter-

minal work room was full and noisy. Accordingly, arrange-

ments were made to use another machine during the lunch

hour of a terminal operator in the central administration

building, since no access was permitted to the psychology

department's terminal. All terminals used were IBM 2741's,

which resemble Selectric typewriters. Basically, the com-

puter timed each line typed on the terminal, and counted

the characters, words, and blanks. Upon command, the count

and rate of word, character, and blank production were

printed out for each line, as well as the time and its

reciprocal, and the overall averages for the whole perform-

ance. Sentence rates were determined by post hoc hand count.










Sequence of Operations


For her first 11 sessions, S-157 performed the same

tasks as S-922 and S-380, except that at her 9th, 10th, and

llth sessions she ended the "Uses for . ," Picture Essay,

and Free Assoc. tasks "when she was done," as did S-106.

After session 9, her charts were no longer shown. After

session 7 she was asked to raise her rates, and after ses-

sion 10 she was told that if she could raise her alphabet

rates to 175 (from approximately 156) characters per min.,

the alphabet task could be dropped. At session 9 an essay

on an "emotional" topic was added. At session 12, all tasks

except the picture essays and the essays on emotional top-

ics were dropped, and two or three of each were done every

session. For sessions 15 through 29 two PE's were done by

hand, and two were done on the terminal as well as one or

two essays on the terminal. Hand tasks were done first.

One in each mode was to contain a question as every other

sentence, and the orders were varied. At session 21, in-

stead of questions, each sentence was to start a new line

on half the PE's. At session 24, errorless performances

were attempted and three texts were copied, as S-157 was

going for a job interview and she wished to practice her

typing skills. Before each performance during all subse-

quent sessions, word counts and rates done on the previous

performance were announced with a request that she raise

word rates, regardless of the effect on character rates.









After each performance, the rates were checked and high

rates occasioned praise. For all terminal work, a clock was

started when each performance began and after two to three

minutes the S was told she could stop writing whenever she

wished.


Drugs


After the final experimental session, S-157 revealed

that she had been taking amphetamine (Desoxyn, 15 mg.) every

Monday, Wednesday, and Friday throughout the experimental

period and for months before it. She had selected the alter-

nating days regimen to avoid habituation. At the same time,

she also admitted to regular (daily) use of marijuana.


Mechanical Difficulties.


Several problems are associated with use of the com-

puter. First, some sessions were terminated early due to

general temporary failure of the whole computing system.

Secondly, overstrikes resulted in transmission of unrecogniz-

able characters, and the computer responded by reprinting the

line typed in up to the point of the overstruck character.

Finally, during the return of the typing ball from the right

side of the keyboard to the left, the keyboard locked, and

remained locked momentarily before more characters could be

typed. Timing was only of the actual time the keyboard was

unlocked, but the slowness of the return was noted by sub-

jects.










Results


Figure 10 displayed S-157's character emission rates

for Free Association, "Uses for ..," alphabet, and sig-

nature tasks for her first 11 sessions. Variability in the

data for the week of July 6 for the signature and "Uses for

task suggests drug activity. The effects of self-

timing on both FA and "Uses for ." rates seems to have

been to drop the rates slightly, but recovery and rise was

rapid. The interruption of the experimental session by a

telephone call to the E (point a) had a powerful effect on

character rates. Adaptation to the experimental data tasks

was especially marked in the alphabet task.

Figure 11 showed rates of behavior for the first 11

Picture Essays (PEs) and experimental PEs with declarative

statements and questions alternating. Both hand and computer

PEs were performed in both free and "alternating question"

style, and the order of free vs. question was balanced. No

great effect on rates was noted.

Figure 12 displays overall rates of word and char-

acter production for all computer "Emotional Essays" (EEs).

Several days of "Uses for . ." and copying tasks were also

shown. Rates appear steady.

Character, word, and sentence rates for all PEs

done on the terminals were shown on Figure 13. Notations

of X with performances from 7/27 to 7/30 indicate questions
0









1000
500



100 -
50


10 -
5



1-


22 June

Signature


Alphabet


I I .


S-V


29 June

Uses for


1














29 June

FA


Charts of character emission rates for S-157. Repeated symbols with
"Uses for .. ." indicated repeated presentation of same stimuli.
Point "a" indicates phone call to E. Successive calendar days are plotted
within task groups. The change seen in FA rates seems to be the result
of an increase in writing time.


Figure 10.


~~T~rCmrrrCrrrrrr(rrr~r(


-- -1- .


































29 June


iooo .....I ......"1 ............I ...
500 + .


Hand written PE rates


Panel 1
S-Timed
Stopped showing chart
Rates not shown


Panel 2
PE ratesX indicates
past perfect tense
(. indicate every
other sentence
a question


Computer written PE rates

Panel 3 Panel 4
Computer PE rates Computer PE rates
1 Every other
sentence a ques-
tion
2 Start each sentence
at right margin


Figure 11. PE rates for S-157. All panels are of successive calendar days.
Repeated symbols indicate repeated presentation of stimuli.


1. 1.


100-

50 '



10-
5 .


..... .. .... I .. .... 1
c b


13 July
WC '

















13 July


20 July


20 July


27 July









1000
500


100-
50



10-
5


S


k


F
7--- -


Dates 18 July 21 23 24 25


M Tu Th


28 29 31


Emotional Essays


IW Th


6 Aug 7

Copy
Text


_I- I


10
11 12 15 Aug

Emotional essays


Successive tasks within experimental sessions for S-157. Character (.)
and word (X) frequencies are indicated for the days and dates listed.
"Uses for. ." rates are plotted on successive calendar days with the
amount of money presented as the simulus to write about.


Figure 12.


I 1$1000
$10,000
$10
Uses for


** **


W Th F


I I








V- h ---


4 Aug 6 7


11 Aug 12 14 15


Figure 13. Character (4) word (x) and sentence (A) emissions. Rates of S-157
working on the computer terminal are plotted for successive picture
essays during experimental sessions on listed days. "R" indicates
normal instructions while "0" indicates essays in which every other
sentence was to be a question.


1000
500-


F
RR
18 July


... pr...


-rr


A dV


Uw


I M T ITh
'*,*** **-.









1000
500



100-
50



10"
5



1


--


11~


August


V


~-Y


"Yvvry


12 August


Successive lines of print









Figure 14. Charts of line by line character (.) and word (x) rates for picture
essay written 8/11 and 8/12 by S-157.


nThmv--







)VVfi











alternating with declarative sentences, while R indicates

no special instructions. Beginning on 8/4, a carriage re-

turn was keyed with every period, question mark, or ex-

clamation point. Subsequently, character and word rates

rose slightly. The imposition of positive statements after

higher rates and statements of previous word rates on 8/11

was associated with increased variability in character and

word rates, and a slight rise in rate. Line by line rates

of word and character production from all PEs written on

8/11 and 8/12 were displayed on Figure 14. Note the dif-

ferences in bounce and patterning of the word rates and

character rates.



Discussion


Stable character rates were seen within task

groupings with S-157 and an effect was noted for self-

timing. Inspection of the PE records on Figure 14 re-

vealed changes in rate trends between each level of re-

sponse recorded, which may be due to the fact that while

character sizes were fairly standard, word and sentence

sizes varied.

The effect of the drugs taken by S-157 seemed to

be minimal, as the most marked effect easily attributable

to the drug was not seen across all behaviors. A larger

dose may have shown greater, more easily identified effects.
















EXPERIMENT 5


The general measurement tactics and procedures de-

scribed previously were also applied to other subjects,

thus forming a set of replications. One of the subjects

serving in this phase was not a native English speaker.



Method


Subjects and Apparatus


Four subjects served in the replications of experi-

ments 1-3. Subject S-517 was an 18-year-old male first term

freshman, who served in the study in an effort to learn more

about his writing. Subject S-001 was a 19-year-old fourth

term female college student with a history of academic dif-

ficulty. She was serving in return for study skills help.

Subjects S-99 (24 years old) and S-202 (20 years old) were

both female college students serving for PSY 201 experimental

credit. All subjects except S-999 were black, native English

speakers; S-999 was an Iranian citizen who learned English

in school. Apparatus, tasks, and procedures were the same

as those used in Experiment 2, except that S-202 wrote let-

ters to friends instead of essays,andcharts were not drawn

41










and shown to the subjects. On essay and letter tasks, sub-

jects were generally permitted to time themselves.

Subject S-314 was a 21-year-old female serving for

PSY 201 experimental credit. She generated stories, picture

essays, and letters using a computer terminal and the same

program as was used by S-157.



Results


Figures 15, 16, 17, 18, and 19 display the rates of

character permission per minute for each task performed by

each subject. The data resemble charts previously shown for

other subjects, except that attendance was generally more

sporadic and fewer sessions were run. Note that the first

data point was quite representative of frequencies of be-

havior for S-001, and was moderately representative for

S-517, while S-202's first letter rate was not a good indi-

cator. For three out of the five tasks, S-999's first rate

was also not a good predictor. The data for S-314 are simi-

larly more sparse than those shown for S-157, and S-314's

rates are lower.



Discussion


In general, these data conformed to the data gener-

ated by the other subjects. First session measurement of

rates of behavior was not seen to yield a reliable baseline.









1000 .....
500



100
50


10
5




5



.1
13 July 13 July 13 July 13 July

Signature Alphabet Picture essays Essay





Figure 15. Charts of character emission rates for S-001 on tasks listed. Successive
calendar days are indicated for each panel.









..... ..... ...


1000 -
500



100
S50



10
5


1


.5 "


/ / r


r -













* .-







6 July

Signature




Figure 16.


6 July

FA


- .~


Charts of character emission rates for S-517 on tasks listed. Suc-
cessive calendar days are indicated for each panel.


6 July

Uses for


6 July

Picture
essay


6 July

Essays


6 July

Alphabet









1000
500


* *

T



















20 July

Signature


Charts of character emission rates for S-202 on tasks listed. .Suc-
cessive calendar days are indicated for each panel, except the last
one, in which three performances were emitted in the same day.


Successive
letters
written on
8/12


100
50



10
5


20 July

Letters


Figure 17.


TW


20 July

Picture
essays


20 July

Alphabet


M.









1000
500


100
50


-:















13 July

Signature


. **


".... t i i ... I '"
13 July 13 July

Picture essay Uses for


13 July

FA Essay


Figure 18. Charts of character emission rates for S-999 on tasks listed. Successive
calendar days are indicated.


13 July

Alphabet


10
5



1
.5









.I........rhm


20 July


17 July


29 July 31 July 12 Aug


13 July 27 July

Picture essays FA Letters to friends
Successive calendar days Successive experimental tasks

Figure 19. A. Character (-) and word (K) emission rates
on indicated tasks by S-314 using computer.


1000
500 +


1000-
500 -



100-
50


1
.5



.1
.05



.01
12 Sept



Successive tasks


B. Correct (*) and in-
correct (X) character
emission rates for S-
922 on successive trials.


100-
50



10-
5 -


.1-
.05



.01

















EXPERIMENT 6


Experiments 1-5 dealt with the frequency of emission

of character and words to the relative exclusion of inves-

tigation of the topography of the responses. Experiment 6

was designed to demonstrate control of the topography of

characters emitted by the consequences of emission of

selected characters.



Method


Subject and Apparatus


A subject who had participated in an earlier phase

of the investigation (S-922) was asked to serve again. One

experimental session was held under conditions approximating

those of Exp. 2, except that a yellow lined 8 1/2 x 11"

legal tablet was used to record the responses, and timing

was done by reference to a Westclox electric clock.


Procedure


The subject was instructed to write characters "to

guess a simple word,"and the E would respond "no" after in-

correct characters and "yes" after correct ones, and circle
48











the correct ones. The same word, "CARDBOARD," was used

for both trials.


Topography Measurement


The "correct" topography was defined as the next

letter in the stimulus word, while an "incorrect"topography

was defined as any other character. Hence, a direct measure

of topography was possible by simply counting the correct

and incorrect letters.


Preliminary Sessions


One preliminary session was held with S-922, in

which the phrase "white dogs are quieter" was selected. The

following difficulties were found: after the first two

words, S-922 adopted the same procedure (writing the alpha-

bet) as a system for locating the correct response, and she

announced a desire to stop the session as the task was bor-

ing. Hence, CARDBOARD was chosen for the next task, as the

first letters appear early in the alphabet, as does the B

in the last part of the word. The "B" was important be-

cause few cues exist to signal the next letter after the

first syllable.


Experimental Error


After the first trial, the S pointed out that "CARD-

BOARD" was a compound word, not a simple one. The instructions










were amended for the second trial by the substitution of

"easy" for "simple," and the S was encouraged to try again.



Results


Figure 19 part B displays the correct and incorrect

character emission rates for trials 1 and 2. Note the con-

trol of the topography of the response by the consequation,

as the incorrect characters decelerate and the correct char-

acters accelerate. The change in topography of the response

went from 9 of 47 responses correct to 9 of 23, and the rates

show an even more dramatic effect, from 2.0 per minute cor-

rect and 8 per minute incorrect to 3.18 per minute and only

4.87 per minute incorrect.



Discussion


The change in topography of response under the experi-

mental conditions supports the notion that the consequences

of emission of a given character (and hence a given word)

control the topography of emission as well as the frequency.
















GENERAL DISCUSSION


Since issues are raised that fall in several diverse

categories, the discussion is arranged topically.


Unit of Response


The findings with respect to the unit of measure-

ment appropriate to compositional behavior may seem unclear.

On a FA task in Exp. 1, S-803 demonstrated a clear rise in

word emission rates while simultaneously lowering her char-

acter emission rates by shortening her average word length.

However, S-157, when writing texts, showed a general parallel

movement of word and character frequencies, as did S-314.

More research is needed to clarify the differences

between these two rates, but one possible explanation for

the lack of separation with S-157 may be the textual con-

straints imposed by the requirements of the essay task. In

Exp. 1, S-803 was under no contingency to produce words

with any specific denotation, but S-157 was required to emit

connected discourse, which may have prevented her from short-

ening her words as drastically as did S-803. Thus, until

more research is undertaken, it would seem most prudent to

record both frequencies, but, in situations in which lack










of resources makes that difficult, word frequencies may

be easiest to measure and explore.


Machine Writing


The ease of measurement of character and word

emission rates with the aid of the high speed computer in-

creases experimental efficiency tremendously. A simpler

way of obtaining similar measures may be to instruct sub-

jects to type on a standard electric (or manual) typewriter

and to type all the way to a specified right margin and no

further before operating the carriage return, thus typing

lines of a specific fixed length. An external timer and

a count of the words or blanks in each line would provide

the basis for word rates, and subtracting the number of

blanks from the line length would give the information for

the character rate. Thus, the advantages of mechanical

recording may be available without the expenses and incon-

venience associated with computer use.


Environmental Control


Writing, like other behaviors, has two basic char-

acteristics: some frequency of occurrence in time and some

topography or form. Several examples of control of fre-

quency of character and word rates have been prescribed.

An announcement of a contingency was sufficient to dramatically











move S-922's character emission rates in all measured be-

haviors. In general, announcement of the task controlled

differing rates for most subjects. A more subtle example

is shown in the data from S-803 (Figure 4): the part of

speech (adjective or verb) to be generated to match one of

the two lists of nouns had a dramatic effect on the fre-

quency, and indeed the noun lists themselves also differ-

entially controlled the rates.

Generally, control of the topography of the response

was done simply by announcement. No subject made an ex-

tended topographically inappropriate response; all produced

written materials that met the specification of the experi-

mental situation. For example, no subject responded with

an alphabet to the signature task, or vice versa. Overall,

specification of the acceptable response was sufficient to

produce that response, and a change in specification pro-

duced an immediate change in topography of response.

An effect of the topography on the response rate

may be seen in the alphabet rates of S-016. One topography

(all capital letters) was emitted at a response rate dif-

ferent from the other (small letter) rate. Another example

may be seen in the data for S-803 cited above. Finally,

Exp. 6 unequivocally demonstrated control over the topography

of a written response by the consequences of that response.










Pedagogy


Perhaps the greatest value of the present work

lays in its relationship to instruction in written compo-

sition. The orderliness seen in rates of character and

word emission under experimental conditions justify the

diversion of scarce classroom time and resources to pro-

grams of daily measurement of word emission rates during

composition classes by teachers and experimental psychol-

ogists.

Zoellner (1969) requested a measurement system for

classroom use that would allow teachers to display incre-

ments of improvement in writing. The present work pro-

vides a basis for attack of the problem of measuring im-

provement. Since writing frequencies seem to be stable,

changes in the quality of a student's writing may be shown

by marking all of the words in a composition that are in-

appropriate, and plotting the rates of appropriate and

inappropriate words written per minute by the students.

Clearly, any improvement will show up as a relative drop

in the frequency of incorrect words emitted or a rise in

the frequency correct. Specific types of errors may be

handled in a similar fashion. One type of choppiness,

for example, may be seen as a frequency of emission of

terminal marks (.?!) that is too high relative to the

frequency of words emitted. After initial measurement, any










change in sentence length must show up as a change in the

frequency of terminal marks relative to word frequency.

Hence, one may assess the effect-of an announcement or

instruction on choppiness by examining the frequency of

terminal marks relative to the word frequency. Thus, pre-

cise measurement of the change in student performance due

to instruction is possible, and with such measurement the

efficiency and effectiveness of different instructional

strategies may be assessed. Accountability in terms of

cost for a given unit of student improvement is also

possible.

Specific pedagogical tactics remain to be explored,

but some implications may be drawn from the present study.

First, the results seen with S-016 suggest that self-timing

by students may be a useful tactic for increasing the time

spent writing. Secondly, the data with respect to measure-

ment of word versus character emission frequencies equivo-

cally suggest that word emission rates may be sufficiently

stable, but further investigation of this point seems in-

dicated. Thirdly, both the frequencyof emission and the

topography of the characters emitted can be controlled by

manipulation of the environment, so that teachers can arrange

conditions under which students will write well. Thus, the

extension of the findings of the experimental analysis of

behavior to written composition as aids to teachers and

students may be slightly closer.

















REFERENCES


Brigham, T. H., Graubard, P. S. and Stans, A. Analysis of
the effects of sequential reinforcement contingencies.
Journal of Applied Behavior Analysis, 1972, 5, 421-
429.

Findley, J. D. An experimental outline for building and
exploring multi-operant behavior repertoires.
Journal of the Experimental Analysis of Behavior,
1962, 5, 113-166.

Gonzalez, F. A. and Waller, M. B. Handwriting as an operant,
Journal of the Experimental Analysis of Behavior,
1974, 21, 165-175.

Maloney, K. D. and Hopkins, B. L. The modification of sentence
structure and its relationship to subjective judgments
of creativity in writing. Journal of Applied Behavior
Analysis, 1973, 5, 425-433.

Sidman, Murray. Tactics of Scientific Research. New York,
Basic Books, Inc., 1960.

Van Houten, R., Morrison, E., Jarvis, R. and McDonald, M.
The effects of explicit timing and feedback on
compositional response rate in elementary school
children. Journal of Applied Behavior Analysis,
1974, 7, 547-555.

Zoellner, R. A behavioral approach to writing. College
English, Jan. 1969, 30, No. 4.

















BIOGRAPHICAL SKETCH


Charles T. Merbitz was born on March 30, 1946, in

Chicago, Illinois. He graduated from a private high school

in Chicago in 1964.

In 1970, he married Gloria Golec and also received

a B. S. from the University of Illinois at Chicago Circle.

In 1971, after a year of work in business, Mr.

Merbitz began graduate studies at the University of Florida

in Gainesville. He received his M. A. in Psychology in

1973. While at the University of Florida, Mr. Merbitz was

employed as a research assistant in an animal operant behav-

ior lab; served as Academic Advisor in University College to

Minority Students (EEOP/SSS program): was Tutorial Coordin-

ator for the Personalized Learning Center and taught in the

Department of Behavioral Studies.

Mr. Merbitz presently resides in Lake Wales, Florida,

with his wife, and is employed as a school psychologist by

the Winter Haven Hospital.


















I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Do pr of Philosophy.




i y. ennyp cker, Chairman
Professor of Psychology


I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor o Ph osophy.




M .K. Goldstein
Asst. Professor f Psychology


I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor f Philosophy.




M. Ps. yeyer co
Professor, Psychology















I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.




J. M. Newell 1
professor of Education


I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.




F. S. Sieka, Ph. D.


This dissertation was submitted to the Graduate Faculty of
the Department of Psychology in the College of Arts and
Sciences and to the Graduate Council, and was accepted as
partial fulfillment of the requirements for the degree of
Doctor of Philosophy.

December, 1975




"4A De Graduate School




Full Text

PAGE 1

AN EXPLORATION OF CHARACTER EMISSION RATES IN COMPOSITION By CHARLES TIMOTHY MERBITZ A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS OF THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1975

PAGE 2

ACKNOWLEDGEMENTS I am especially indebted to Dr. H. S. Pennypacker, the Chairman of my Supervisory Committee, and to all of the other members of my committee, for their unflagging differential positive reinforcement of successive approximations of productive professional behavior, both with respect to this paper and to other work. Dr. M. K. Goldstein, Dr. John Newell, and Dr. Frank Sieka were expecially instrumental in my progression here. I am also deeply grateful for the support and encouragement given me by my wife, Gloria, without whom even the beginning of this undertaking would have been doubtful, and who gave crucial support to the whole enterprise. Special thanks go to David M. Smolen, who wrote the computer programs used and gave help and valuable advice on data analyzed. I wish also to thank the people who volunteered to serve as subjects, often giving time and effort well beyond any possible formal reward. I cannot express my appreciation deeply enough for the efforts of all these people. ii

PAGE 3

Physical support was provided by several organizations. The Personalized Learning Center helped greatly with some office space, a stopwatch, and some computer terminal time. Dr. M. Branch lent the Running Time Meter Finally, the University of Florida College of Arts and Sciences and the Northeast Regional Data Center jointly provided the computer funds used.

PAGE 4

TABLE OF CONTENTS Page ACKNOWLEDGEMENTS ii ABSTRACT v INTRODUCTION 1 EXPERIMENT 1 8 EXPERIMENT 2 19 EXPERIMENT 3 ' 26 EXPERIMENT 4 30 EXPERIMENT 5 41 EXPERIMENT 6 48 GENERAL DISCUSSION 51 REFERENCES 56 BIOGRAPHICAL SKETCH 57

PAGE 5

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 AN EXPLORATION OF CHARACTER EMISSION RATES IN COMPOSITION By Charles Timothy Merbitz December, 1975 Chairman: H. S. Pennypacker Major Department: Psychology Six experiments explored character and word emission rates under controlled conditions with a variety of subjects. Day to day stability was seen in both character and word emission rates, but first session rates were not reliable indicators of general rate. Specific experimental tasks tended to control distinct rates. Both self-timing and release from tasks changed character rates, while self -timing was also seen with longer durations of writing. The control of character topography by consequences was also demonstrated and a large effect was noted. Thus, both operant antecedent and consequent control of both response rate and topography was demonstrated for written composition. Both hand written compositions and compositions written on a typewriter (computer keyboard) on various topics v

PAGE 6

and with various constraints were assessed. Suggestions for easier and cheaper measurements were made. Results were related to measurement problems in the teaching of essays, and suggestions for teacher measurement of composition activity in the classroom were made. VI

PAGE 7

INTRODUCTION If an experimental analysis of written composition is to proceed, a unit of response must be defined. Findley's (1962) work on multi-operant repertoires clearly showed the possibilities of a unit that itself was composed of smaller units of response. If language behavior in general and written composition in particular are analogous to Findley's multi-operant repertoires, we may be able to proceed with units defined at several levels. Basically, written materials are defined by combinations of curved and straight lines and dots. Gonzalez and Waller (1974) reported records generated by two subjects writing on an apparatus that responded to variations in pen pressure, and defined one class of responses by microswitch closures of .5 sec minimum duration between minimally specified IRTs. They found handwriting rates while writing were stable despite differences in schedule of reinforcement in effect but the schedule did control the pattern of pausing such that interval and ratio components of a mult FR FI LH are clearly distinguishable. A larger response unit that has been considered is the word. Brigham, Graubard, and Stans (1972) reported 1

PAGE 8

the results of several operations on mean word counts in a 5th grade "adjustment" class. In "Baseline" all students got points for sitting and writing during the writing period. Points were exchangeable for generally unspecified privileges and activities in the room. In the "Number of Words" condition, one point was given for each word in the stories written. In the next condition, "Different Words," two additional points were given for each initial use of a word in that story. Finally, in the "New Words" phase, the children were given an additional three points for each word that had never appeared in a previous story. An alphabetized list of all words used in earlier stories was provided for each child. Groups A (n=5) and B (n=3) experienced all conditions, while Group C (n=5) was exposed only to baseline and "Number of Words" conditions. The authors reported mean numbers of words written per group per session in each condition, mean numbers of "Different Words," and mean numbers of "New Words," as well as the mean times. In addition, the stories were rated on a 25-point scale and the ratings of each story were given . Inspection of the figures revealed a marked similarity in the shape of the functions for each measure within each group. The similarity of the time function to the others suggested that conversion of the counts to rates may reveal otherwise obscured relationships.

PAGE 9

Figures 1 through 4 of Brigham et al. were converted to tables and the rates of words emitted per minute were calculated and charted (Figures 1 and 2, see data display section below for graphing conventions). Since Brigham et al. report sessions "twice a week" on the average, T Th lines were used for plotting. It was apparent that the major effect of the first contingency imposed (Number of Words) was to simply increase the time the children spent writing. Rates of word production were quite variable, and no consistent effect was obvious. The authors report rising quality point ratings during the experimental period, and suggest that it was possible that the rise in ratings associated with the "New Words" condition may have been the result of direct attempts to use new words, as . . . they worked harder and spent more time on the sentences, thereby producing more and better ideas while actually writing less. (p. 429) More time was indicated on the graphs, but the word production rates did not drop, as this interpretation requires. Also, the mean ratings the stories received were charted as quality units per minute emitted by each group. Little change appeared in Group A's quality unit production rates throughout the experiment, except for a drop in sessions 9 and 10. Group B shows more variability, and also seems to share a small dip for those two sessions (among others). Group C shows the clear experimental effect of the "Number of Words" contingency — while the word

PAGE 10


PAGE 11

bfl

PAGE 12

rates remain roughly stable, much more time was spent, and a very marked decrease in quality units per minute emitted is seen, denoting less efficiency. Plots of quality unit . and word rates for the same sessions seem to suggest covariation in trend, although the effect was not reliably seen. Difficulties in further interpretation lie in the problems inherent in converting graphed points to rate data and the report of group means rather than individual performances (Sidman, I960). In another study involving composition writing, Maloney and Hopkins (1973) reported the use of a permutation of the "Good Behavior Game" to change the number of different adjectives, action verbs, and sentence beginnings in 17 ten-sentence stories written by 14 fourth through sixth graders. They found no change in character or word emission rates throughout the study, but did report that selective reinforcement with extra recess and candy did change subjective ratings of the creativity of the stories. They did not report the rates of behavior. Van Houten, Morrison, Jarvis, and McDonald (1974) reported rates of word emission for children in one second and two fifth grade classes. They discovered that explicit timing and feedback from the children counting their own words roughly doubled the response rate, from three to six words/min., in ten sessions. Stories were also rated on five dimensions, and the ratings were higher

PAGE 13

during each of the timing and feedback phases. Feedback was provided by letting the children count their own words, which they did with 80-100% accuracy. Of the three studies, only Van Ilouten et al . showed a rate increase, while all three reported "quality' increases during the experimental period. While assessment of composition quality remains an important issue, measurement of composition rates is also needed. The studies cited above support the notion that counting the words emitted per minute by subjects writing essays may reveal interesting relationships between essay rates and environmental events. However, a gap existed in units of analysis at the level of the character between the behaviors recorded by Gonzalez and Waller and words counted by the others. The goals of the present work were, therefore, completely exploratory to measure some writing rates, to begin definition of a dependent measure of writing rates, and to examine relative sensitivity of rates of character and word emission. The research was not structured initially; rather as interesting phenomena came to light, new directions were taken .

PAGE 14

EXPERIMENT I To begin the investigation of compositional reasponse rates, subjects were timed while they performed various writing tasks. One subject was seen in three preliminary sessions to establish timing procedures and experimental tasks. Method Subjects and Apparatus Subjects S-802, S-803, S-506, and S-805 were female college students enrolled in Psy . 201 and serving for experimental credit. The first three were 18-year-old third quarter freshmen; S-805 was 19. Three subjects were black; S-506 was white. Session 1 for S-802 was timed with a Westclox (General Time) stopwatch; all other sessions were timed withaRunning Time Meter (Cramer Controls). Experimental sessions were held in an office approximately 4 m x 5 m x 2.5 m containing two desks, files, chairs, and common office furnishings. Subjects were run individually, generally during normal school hours. Paper and a variety of ballpoint pens were provided by the E. Before

PAGE 15

beginning, the purposes and methods of the experiment were explained to each subject and a standard informed consent form was signed by each subject. Procedure While sitting at the E's desk, subjects were given an experimental task number from the E's data sheet to write at the top of a fresh sheet of paper. They were then given specifications for the writing task to be' performed . When ready, the timer was started and they wrote until told to stop. They then counted the words (or characters) written, the count was recorded, and rates calculated and displayed on six cycle semilog Standard Behavior Charts. The pattern was repeated until terminat ion of the session. Subjects were asked to indicate any excessive hand fatigue. An informal and relaxed atmosphere was maintained, and all questions were answered fully. Words were counted by S-802 and during S-803 ' s first two sessions. All subsequent counts were of characters. Experimental Tasks Experimental tasks included copying a text which had been typed with six words per line on a duplication Since tasks were standardized for Exp. 2, a more detailed description of procedure is presented in the method section of Exp. 2.

PAGE 16

10 master. Triple spacing permitted subjects to write between the lines and thus keep eye shifts to a minimum. Subjects were also asked to write short essays on topics with which they felt some emotional involvement. A free association task (FA) was defined by instructing subjects to simply write whatever came to their mind. Limited association (LA) was simply all that they could think of to write relevant to a given topic. Other tasks included listing all of the possible uses for a specified common object ("Uses for . . ."), signing one's name, writing the alphabet, and writing a list of all the objects in the room. Subjects were also asked to generate lists of nouns, and were later asked to list adjectives or verbs appropriate to the nouns listed. Data Display Following a common practice in graphing dynamic phenomena in other natural sciences, semilogarithmic charts were used to keep and display frequency of response data. Two general types of charts were used: In the first, the count per minute of behavior was displayed on the line corresponding to the calendar day the data were collected. The second displayed the count per minute of behavior, but successive lines represent successive experimental tasks within one session. All data were labelled with true calendar days to permit unambiguous indication of the time relationships between various experimental

PAGE 17

11 events. The log scale allows accurate predictions and both absolute and relative comparison of data from different subjects. The dots and x's indicate frequencies of behavior on the indicated day, and the dashes show the reciprocal of the time spent observing. Thus, the lower the dot, the lower the rate of behavior, while the lower the dash, the longer the time spent observing. Reliability Subjects were informed that spot checks were made of their counts. Limited resources did not permit all counts to be verified, but 50% of the experimental tasks were recounted. The accuracy of match between subjects' counts and reliability counts ranged from 88% to 100%, with 13% of the subjects' counts inaccurate. Letters, numbers, and punctuation marks were all counted as characters. Qf the errors, 55% involved miscounts by two or less characters. Results Frequencies of words written per minute by S-802 are displayed as x's on Figure 3, while frequencies of characters written per minute (counted post hoc) are indicated by dots. Note the parallel relationship between the two frequencies in the copying task, in which the average word length was fixed by the original author. In the FA tasks,

PAGE 18

12 M >> CD -P oj O •H -d c CO -P CD CS W OS >> S CM CM j£_ .!_«.* *« ini|i i

PAGE 19

13 some deviation from a fixed relationship was seen, partially due to a comparatively low word rate in the first FA task. Also, average word length changes from approximately 4 characters per word in the FA tasks to 3.5 for the essay. Figure 4 shows frequencies of word and character production for S-803. During sessions on 5/21 and 5/22, only word counts were made. Note the steady rise in the FA and LA word frequencies for those days. Post hoc character counts indicated that the rise in word count was a result of shortening word length. (A similar effect may be seen in the 4/23 FA rates of S-802.) The effects of the subsequent shift to character frequencies over the next four sessions were to slightly raise the character frequencies while word frequencies gradually declined as word length increased. Frequencies of word and character production when the S was asked to list nouns were also shown, as were the frequencies of production when the S was asked to write adjectives and verbs appropriate to the nouns on the list. Note the much lower frequencies seen with verbs. Two lists were used, in the order ABA AB BA. Note that list B always controlled lower frequncies. List B was generated on 5/22, and list A was the second noun list generated on 6/5. The last two pairs of rates were generated under the instruction to simply produce lists of adjectives. The first performance of the "Uses for . . . common objects" tasks was not a good indicator of usual frequencies

PAGE 20

14 aq.nuTui Jtad ^.unoo CD CD >> •-a CO CO >> ea 73 Jh sd X) c cd rH cd a i ** ;

PAGE 21

15 displayed by S-803. Also shown were her essay rates and her rate of character writing when signing her name. Data from S-805 are shown in Figure 5. On two successive days, the same object was presented (a brick) in the "Uses for ..." task. Note the jump in frequency of response. Essay and signature rates were unexceptional. Four tasks (list nouns, list adjectives appropriate to the nouns, list more nouns, and list verbs appropriate to the nouns) involving listing and relationships of nouns, verbs, and adjectives were presented on 6/8. Note the slower adjective rate — the first list of nouns generated contained all proper nouns. Also shown on Figure 5 were .5 min. and five 2 min. FA rates which were collected on 6/3. Each half minute during the longer samples was announced, and the S drew a line and continued writing, thus permitting sectioning of the sample into half-minute segments. Note the slight drop in frequency as the samples become longer . Frequencies of character writing from S-506 were plotted on Figure 6. Her signature and alphabet writing rates seemed stable, as did her essay writing rates. Her description of the room tasks was segmented into quarter minutes, and the total overall rates as well as the firstquarter and half minute rates were shown. The first sample rate was taken for .25 min., and was lower than comparable first quarter rates, although close to the overall rates

PAGE 22

16 aq.nuTiu .iad ^.unoo [ 5 J I 03 Fh Cfl O 0) O O O fn m co-0 > H h'H O P 0»< d fa d xi o d 0) en cc 3 1 03 CO d -h ^P J 03 > •H 03 co 03 O o 3 ' CO CO >» d ta en w 03 >> d Q CD. > H 03 03 0) o o CO O :
PAGE 23

17 aq.nuxui xdd q.unoo < » I r CO >> si ra CO w I to a o a CO >> Q !h T3 CD -a • a m cd a w d r-< a* a; c e d -H ft c • o CO X) O CO m I CD CO CD O >> fi S 0) O £• =H Sh ^ O CD CO ^! U W O rf 4-1 -P w T3 CU en s 73 c C -H •H TJ +-> CD d +-> c ^ -H CD S -P o m d • aS T3 -in n 1XU+. 1UJ+ urn i.i — l. ^ -US' o in d iri 7TS" 1

PAGE 24

18 rates of the longer timings. The essay and "uses for . . ." rates were unexceptional. Discussion In general, character emission rates seem to be stable and predictable. Word rates seem less desirable as indicators, as S-802 and S-803 show a change in word rate without parallel movement in character rates. Further work involving fewer but more standardized tasks and shorter session lengths but more repeated measures was undertaken.

PAGE 25

EXPERIMENT 2 Two subjects served to establish the daily stability of character emission rates, and one of those subjects demonstrated a change in rates. Procedures were designed to shorten experimental sessions and tasks were standardized . Method Subjects and Apparatus Subject S-380 was a 38-year-old female college graduate. Subject S-922 was a 27-year-old female teacher with a master's degree. Both of these Ss consented to serve "to help the E find out about writing." Experimental sessions were held in the E's kitchen in the evenings after both Ss completed working days. A variety of ballpoint pens was again provided and unlined paper approximately 19 x 28 cm. was obtained by cutting used computer output paper in half with a bandsaw. Subjects wrote on the blank side. Timing was again done by manually operating the running time meter to signal the beginning and end of writing activities. The timer was checked against 19

PAGE 26

20 the internal clock of the NERDC IBM 370/165 computer and found to be accurate. Procedure Since tasks were much more standardized in Exp. 2, procedures will be described in more detail. A subject was seated at a table with pens and paper. Each task was numbered, and the S was asked to write her subject number and the task number (supplied by the E) on a sheet of paper. At a signal from the E the subject signed her name as many times as she could until the E signalled her to stop. She then counted all of the characters written and announced the count to the E, who recorded the details on her data sheet and plotted the performance on a standard behavior chart. The alphabet was next written under similar conditions, followed by either a short essay concerning a picture the E presented, or a task in which all the different possible uses for a common object named by the E were listed. The final task was typically a "free-association," in which the S wrote anything (comparable to Gonzalez and Waller's task). Contingency After 6 sessions, S-922 asked when she could quit, and at the next session she was told "we can drop any task on which you exceed a goal rate for 5 out of 6 days." At

PAGE 27

21 her next session S-380 was. also presented with this statement . Stimuli Pictures were provided to write about for the "Picture Essay" (PE) experimental task. Illustrations were cut from magazine advertisements or feature articles (typically color photos) and identifying text was cropped or obliterated. An effort was made to locate provocative pictures with an obvious point, such as those often seen in Psychology Today . Another experimental task ("Uses for . . .") involved listing as many uses as possible for common objects named by the E. Objects included rubber bands, a pencil, paperclips, a brick, a bicycle, a safety pin, Scotch tape, a house, a car, a park, a barn, money, a rope, and a coin. Objects were named by the E in varying orders to each subject Daily charts and data sheets were kept on each S, and were shown whenever requested. Usually, subjects merely glanced at both. Results Figures 7 and 3 show the rate of character emission per minute by S-922 and S-380 respectively. The heavy dark line on the figures indicate the presentation of the statement

PAGE 28

22

PAGE 29

23

PAGE 30

24 on quitting to the subjects. Note the immediate effect on S-922. She was writing her signature at an average rate of 183.6 characters per minute prior to the announcement, and after it her rate averaged 211.3 characters per minute. Similar increases may be seen in the other tasks. Note also the increase in variability after the contingency announcement. Daily and average preand post-announcement rates are given on Figure 7. No such effect was noted for S-380. For 'this S, replication points were provided by presenting the same stimulus object in the "Uses for ..." task on two different occasions, as indicated. Note also the "banding" exhibited by this subject. Each task controls a different set of rates of character emission. Discussion These two subjects provide evidence of the stability of the rate of character emission in daily experimental sessions. The effect of the contingency arrangement on S-922 's rate indicated that perhaps the operant conditioning model would indeed be appropriate for written composition. In addition, S-922 started at a very high baseline rate, compared to the subjects seen in Experiment 1. The failure of the announcement to move S-380 's rates simply suggest that it was inappropriate for her. In view of the weakness of

PAGE 31

25 the E's nominal control over the total environments of both of these subjects, the experimental control shown by the stability for the rates and the movement of S-922 suggest that character rates may well be an excellent experimental measure and that the tasks selected produce behavior that is readily ameniable to experimental analysis.

PAGE 32

EXPERIMENT 3 Since the tasks used in Experiment 2 never lasted more than 2 min each, longer sessions and longer task lengths were necessary to begin exploration of the effects of these conditions on the writing rates. Method Subject and Apparatu s A 22-year-old male student (S-061) was promised "experimental credit" slips toward Psychology 201 extra credit on an hour-f or-hour basis. He had received an incomplete in PSY 201 the previous term and was allowed to turn in credit slips as well as completing other required course work. The credit slips were due July 16, but he agreed to keep serving after that time even though all credit usable had been earned by that date. Typically, 20-30 min of conversation with the E followed each session. Apparatus and procedures were the same as in Experiment 2, except that essays on "emotional" topics were substituted for FAs, S-016 timed himself and sessions were held in the E's office on campus, with S-016 seated at the E's desk. No 26

PAGE 33

27 graphs or charts were made or shown to S-016 during this period. However, his data sheet was not concealed and he occasionally inspected it. Results Figure 9 shows the daily rates for S-016 on each task. Note that with free essays as S-016 was permitted to time himself, the length of time he spent writing appreciably increased. However, in spite of the increase, his rates remained remarkably stable, even when he wrote for as long as 18 minutes. After that session his time decreases, and then he showed a fluctuation in rate. His signature and alphabet rates seem stable. (The first four alphabet rates were written in capitals; he spontaneously changed to lower case letters at the fifth session.) Four objects were presented twice, and only one showed very close rates on the two occasions. Picture essay rates increased also after the procedure was changed so that S-016 time himself . Discussion The data from the first 6 free essays of S-016 indicate that character counts can be quite stable even for relatively long essays (18 min.). However, later sessions show unexplained jumps in the rates. It was planned that

PAGE 35

29 S-016's alphabet and picture essay performances would be charted in order to observe the effect of presentation of the chart on both a steady and a fluctuating rate. However S-016 decided to terminate and the operation was impossible The effects of permitting the S to time himself was, in general, to raise the rates, and to increase the time spent writing.

PAGE 36

EXPERIMENT 4 One of the problems associated with investigation of character emission rates was the tremendous tedium of counting characters. Accordingly, a computer program was obtained to count, time and calculate rates on material typed in as text. Comparison of typed rates of character, word, and sentence production seemed appropriate, and the use of the machine permitted more efficient use of subject time, since counting was not required. In addition an attempt was made to investigate the effects of constraining the experimental task in different ways. Rates of character emission of both hand written and typed materials were collected. Method Subject and Apparatus An 18-year-old female currently enrolled in PSY 201 (S-157) served under the promise of experimental credit and help with study skills. Se accumulated approximately 21 30

PAGE 37

31 hours of experimental credit, of which 5.5 was the maximum useful. Her efforts to find a job were discussed with the E, as well as other personal issues, although study skills were not. She reported preexisting typing skills and agreed to compose essays on a typewriter. For the first 11 sessions, S-157 followed the same general procedures and used the same apparatus as S-016, except that she always used and brought the same pen to experimental sessions, and her first ten sessions were charted as were S-922 ' s and S-380's. In addition, sessions 12 through 19 included work on the computer terminal, and some tasks were dropped. (See Figures 10 and 11.) Two sessions (12 and 13) were held in a room assigned for student computer terminal use, containing approximately 20 assorted computer terminals and keypunches. Since S-157' s session time was approximately noon, the terminal work room was full and noisy. Accordingly, arrangements were made to use another machine during the lunch hour of a terminal operator in the central administration building, since no access was permitted to the psychology department's terminal. All terminals used were IBM 2741 's, which resemble Selectric typewriters. Basically, the computer timed each line typed on the terminal, and counted the characters, words, and blanks. Upon command, the count and rate of word, character, and blank production were printed out for each line, as well as the time and its reciprocal, and the overall averages for the whole performance. Sentence rates were determined by post hoc hand count.

PAGE 38

32 Sequence of Operations For her first 11 sessions, S-157 performed the same tasks as S-922 and S-380, except that at her 9th, 10th, and 11th sessions she ended the "Uses for . . . , " Picture Essay and Free Assoc, tasks "when she was done," as did S-106. After session 9, her charts were no longer shown. After session 7 she was asked to raise her rates, and after session 10 she was told that if she could raise her alphabet rates to 175 (from approximately 156) characters per min., the alphabet task could be dropped. At session 9 an essay on an "emotional" topic was added. At session 12, all tasks except the picture essays and the essays on emotional topics were dropped, and two or three of each were done every session. For sessions 15 through 29 two PE ' s were done by hand, and two were done on the terminal as well as one or two essays on the terminal. Hand tasks were done first. One in each mode was to contain a question as every other sentence, and the orders were varied. At session 21, instead of questions, each sentence was to start a new line on half the PE's. At session 24, errorless performances were attempted and three texts were copied, as S-157 was going for a job interview and she wished to practice her typing skills. Before each performance during all subsequent sessions, word counts and rates done on the previous performance were announced with a request that she raise word rates, regardless of the effect on character rates.

PAGE 39

33 After each performance, the rates were checked and high rates occasioned praise. For all terminal work, a clock was started when each performance began and after two to three minutes the S was told she could stop writing whenever she wished. Drugs After the final experimental session, S-157 revealed that she had been taking amphetamine (Desoxyn, 15 mg . ) every Monday, Wednesday, and Friday throughout the experimental period and for months before it. She had selected the alternating days regimen to avoid habituation. At the same time, she also admitted to regular (daily) use of marijuana. Mechanical Difficulties . Several problems are associated with use of the computer. First, some sessions were terminated early due to general temporary failure of the whole computing system. Secondly, overstrikes resulted in transmission of unrecognizable characters, and the computer responded by reprinting the line typed in up to the point of the overstruck character. Finally, during the return of the typing ball from the right side of the keyboard to the left, the keyboard locked, and remained locked momentarily before more characters could be typed. Timing was only of the actual time the keyboard was unlocked, but the slowness of the return was noted by subjects.

PAGE 40

34 Results Figure 10 displayed S-157's character emission rates for Free Association, "Uses for . . .," alphabet, and signature tasks for her first 11 sessions. Variability in the data for the week of July 6 for the signature and "Uses for . . ." task suggests drug activity. The effects of selftiming on both FA and "Uses for ..." rates seems to have been to drop the rates slightly, but recovery and rise was rapid. The interruption of the experimental session by a telephone call to the E (point a) had a powerful effect on character rates. Adaptation to the experimental data tasks was especially marked in the alphabet task. Figure 11 showed rates of behavior for the first 11 Picture Essays (PEs) and experimental PEs with declarative statements and questions alternating. Both hand and computer PEs were performed in both free and "alternating question" style, and the order of free vs. question was balanced. No great effect on rates was noted. Figure 12 displays overall rates of word and character production for all computer "Emotional Essays" (EEs). Several days of "Uses for ..." and copying tasks were also shown. Rates appear steady. Character, word, and sentence rates for all PEs done on the terminals were shown on Figure 13. Notations of X with performances from 7/27 to 7/30 indicate questions (V \

PAGE 41

35 I " 1 r

PAGE 42

36 • x > * 1 « m T > X

PAGE 43

37 X T^ ^ ^F^ I T r ' 1 -j— i r i — ' r i — r — — i — r ! o

PAGE 44

38 U i L i > 4 j. _i ^...i 1 1 i3e i < < > -s I in CM Hfc"> JJo iQ 01 _f ffi
PAGE 45

39 r~ r > i -i 1 — i r *v ^ i > ^ "> r . O X5 !h (N . crj rH o oo G C •H crj t»rH 00 cu C C •H CD H 4-> P
PAGE 46

40 alternating with declarative sentences, while R indicates no special instructions. Beginning on 8/4, a carriage return was keyed with every period, question mark, or exclamation point. Subsequently, character and word rates rose slightly. The imposition of positive statements after highger rates and statements of previous word rates on 8/11 was associated with increased variability in character and word rates, and a slight rise in rate. Line by line rates of word and character production from all PEs written on 8/11 and 8/12 were displayed on Figure 14. Note the differences in bounce and patterning of the word rates and character rates. Discussion Stable character rates were seen within task groupings with S-157 and an effect was noted for selftiming. Inspection of the PE records on Figure 14 revealed changes in rate trends between each level of response recorded, which may be due to the fact that while character sizes were fairly standard, word and sentence sizes varied. The effect of the drugs taken by S-157 seemed to be minimal, as the most marked effect easily attributable to the drug, was not seen across all behaviors. A larger dose may have shown greater, more easily identified effects

PAGE 47

EXPERIMENT 5 The general measurement tactics and procedures described previously were also applied to other subjects, thus forming a set of replications. One of the subjects serving in this phase was not a native English speaker. Method Subjects and Apparatus Four subjects served in the replications of experiments 1-3. Subject S-517 was an 18-year-old male first term freshman, who served in the study in an effort to learn more about his writing. Subject S-001 was a 19-year-old fourth term female college student with a history of academic difficulty. She was serving in return for study skills help. Subjects S-99 (24 years old) and S-202 (20 years old) were both female college students serving for PSY 201 experimental credit. All subjects except S-999 were black, native English speakers; S-999 was an Iranian citizen who learned English in school. Apparatus, tasks, and procedures were the same as those used in Experiment 2, except that S-202 wrote letters to friends instead of essays , andcharts were not drawn 41

PAGE 48

42 and shown to the subjects. On essay and letter tasks, subjects were generally permitted to time themselves. Subject S-314 was a 21-year-old female serving for PSY 201 experimental credit. She generated stories, picture essays, and letters using a computer terminal and the same program as was used by S-157. Results Figures 15, 16, 17, 18, and 19 display the rates of character permission per minute for each task performed by each subject. The data resemble charts previously shown for other subjects, except that attendance was generally more sporadic and fewer sessions were run. Note that the first data point was quite representative of frequencies of behavior for S-001, and was moderately representative for S-517, while S-202's first letter rate was not a good indicator. For three out of the five tasks, S-999's first rate was also not a good predictor. The data for S-314 are similarly more sparse than those shown for S-157, and S-314 ' s rates are lower. Discussion In general, these data conformed to the data generated by the other subjects. First session measurement of rates of behavior was not seen to yield a reliable baseline.

PAGE 49

43 I— 1

PAGE 50

44 » r t s J i — » r -i — > 1 — r i . i — • r iiii} i 1 i jiii 1 1 ' i i f iin[ i i i juu^, mil i I i iiui I i i — l. o o o o o o o m o m r-t

PAGE 51

45 1 s 1 1 1 — «" 1 •H CO CO CD CD P O o £ 00

PAGE 52

46 T " r O O o o o o o m O in rH r

PAGE 53

47 +-> a 03 CO ,»-? s. 1 1 ' 1 o o o o o m o o o m CO w aJ p 0) > H 03 co cd o o CO

PAGE 54

EXPERIMENT 6 Experiments 1-5 dealt with the frequency of emission of character and words to the relative exclusion of investigation of the topography of the responses. Experiment 6 was designed to demonstrate control of the topography of characters emitted by the consequences of emission of selected characters. Method Subject and Apparatus A subject who had participated in an earlier phase of the investigation (S-922) was asked to serve again. One experimental session was held under conditions approximating those of Exp. 2, except that a yellow lined 8 1/2 x 11" legal tablet was used to record the responses, and timing was done by reference to a Westclox electric clock. Procedure The subject was instructed to write characters "to guess a simple word," and the E would respond "no" after incorrect characters and "yes" after correct ones, and circle 48

PAGE 55

49 the correct ones. The same word, "CARDBOARD," was used for both trials. Topography Measurement The "correct" topography was defined as the next letter in the stimulus word, while an "incorrect" topography was defined as any other character. Hence, a direct measure of topography was possible by simply counting the correct and incorrect letters. Preliminary Sessions One preliminary session was held with S-922, in which the phrase "white dogs are quieter" was selected. The following difficulties were found: after the first two words, S-922 adopted the same procedure (writing the alphabet) as a system for locating the correct response, and she announced a desire to stop the session as the task was boring. Hence, CARDBOARD was chosen for the next task, as the first letters appear early in the alphabet, as does the B in the last part of the word. The "B" was important because few cues exist to signal the next letter after the first syllable. Experimental Error After the first trial, the S pointed out that "CARDBOARD" was a compound word, not a simple one. The instructions

PAGE 56

50 were amended for the second trial by the substitution of "easy" for "simple," and the S was encouraged to try again. Results Figure 19 part B displays the correct and incorrect character emission rates for trials 1 and 2. Note the control of the topography of the response by the consequation , as the incorrect characters decelerate and the correct characters accelerate. The change in topography of the response went from 9 of 47 responses correct to 9 of 23, and the rates show an even more dramatic effect, from 2.0 per minute correct and 8 per minute incorrect to 3.18 per minute and only 4.87 per minute incorrect. Discussion The change in topography of response under the experimental conditions supports the notion that the consequences of emission of a given character (and hence a given word) control the topography of emission as well as the frequency.

PAGE 57

GENERAL DISCUSSION Since issues are raised that fall in several diverse categories, the discussion is arranged topically. Unit of Response The findings with respect to the unit of measurement appropriate to compositional behavior may seem unclear. On a FA task in Exp. 1, S-803 demonstrated a clear rise in word emission rates while simultaneously lowering her character emission rates by shortening her average word length. However, S-157, when writing texts, showed a general parallel movement of word and character frequencies, as did S-314. More research is needed to clarify the differences between these two rates, but one possible explanation for the lack of separation with S-157 may be the textual constraints imposed by the requirements of the essay task. In Exp. 1, S-803 was under no contingency to produce words with any specific denotation, but S-157 was required to emit connected discourse, which may have prevented her from shortening her words as drastically as did S-803. Thus, until more research is undertaken, it would seem most prudent to record both frequencies, but, in situations in which lack 51

PAGE 58

52 of resources makes that difficult, word frequencies may be easiest to measure and explore. Machine Writing The ease of measurement of character and word emission rates with the aid of the high speed computer increases experimental efficiency tremendously. A simpler way of obtaining similar measures may be to instruct subjects to type on a standard electric (or manual) typewriter and to type all the way to a specified right margin and no further before operating the carriage return, thus typing lines of a specific fixed length. An external timer and a count of the words or blanks in each line would provide the basis for word rates, and subtracting the number of blanks from the line length would give the information for the character rate. Thus, the advantages of mechanical recording may be available without the expenses and inconvenience associated with computer use. Environmental Control Writing, like other behaviors, has two basic characteristics: some frequency of occurrence in time and some topography or form. Several examples of control of frequency of character and word rates have been prescribed. An announcement of a contingency was sufficient to dramatically

PAGE 59

53 move S-922's character emission rates in all measured behaviors. In general, announcement of the task controlled differing rates for most subjects. A more subtle example is shown in the data from S-803 ' (Figure 4): the part of speech (adjective or verb) to be generated to match one of the two lists of nouns had a dramatic effect on the frequency, and indeed the noun lists themselves also differentially controlled the rates. Generally, control of the topography of the response was done simply by announcement. No subject made an extended topographically inappropriate response; all produced written materials that met the specification of the experimental situation. For example, no subject responded with an alphabet to the signature task, or vice versa. Overall, specification of the acceptable response was sufficient to produce that response, and a change in specification produced an immediate change in topography of response. An effect of the topography on the response rate may be seen in the alphabet rates of S-016. One topography (all capital letters) was emitted at a response rate different from the other (small letter) rate. Another example may be seen in the data for S-803 cited above. Finally, Exp. 6 unequivocally demonstrated control over the topography of a written response by the consequences of that response.

PAGE 60

54 Pedagogy Perhaps the greatest value of the present work lays in its relationship to instruction in written composition. The orderliness seen in rates of character and word emission under experimental conditions justify the diversion of scarce classroom time and resources to programs of daily measurement of word emission rates during composition classes by teachers and experimental psychologists . Zoellner (1969) requested a measurement system for classroom use that would allow teachers to display increments of improvement in writing. The present work provides a basis for attack of the problem of measuring improvement. Since writing frequencies seem to be stable, changes in the quality of a student's writing may be shown by marking all of the words in a composition that are inappropriate, and plotting the rates of appropriate and inappropriate words written per minute by the students. Clearly, any improvement will show up as a relative drop in the frequency of incorrect words emitted or a rise in the frequency correct. Specific types of errors may be handled in a similar fashion. One type of choppiness, for example, may be seen as a frequency of emission of terminal marks (.?!) that is too high relative to the frequency of words emitted. After initial measurement, any

PAGE 61

55 change in sentence length must show up as a change in the frequency of terminal marks relative to word frequency. Hence, one may assess the effectof an announcement or instruction on choppiness by examining the frequency of terminal marks relative to the word frequency. Thus, precise measurement of the change in student performance due to instruction is possible, and with such measurement the efficiency and effectiveness of different instructional strategies may be assessed. Accountability in terms of cost for a given unit of student improvement is also possible. Specific pedagogical tactics remain to be explored, but some implications may be drawn from the present study. First, the results seen with S-016 suggest that self-timing by students may be a useful tactic for increasing the time spent writing. Secondly, the data with respect to measurement of word versus character emission frequencies equivocally suggest that word emission rates may be sufficiently stable, but further investigation of this point seems indicated. Thirdly, both the frequency of emission and the topography of the characters emitted can be controlled by manipulation of the environment, so that teachers can arrange conditions under which students will write well. Thus, the extension of the findings of the experimental analysis of behavior to written composition as aids to teachers and students may be slightly closer.

PAGE 62

REFERENCES Brigham, T. H. , Graubard, P. S. and Stans, A. Analysis of the effects of sequential reinforcement contingencies. Journal of Applied Behavior Analysis, 1972, 5, 421429. Findley, J. D. An experimental outline for building and exploring mult i-operant behavior repertoires. Journal of the Experimental Analysis of Behavior , 1962, 5, 113-166. Gonzalez, F. A. and Waller, M. B. Handwriting as an operant, Journal of the Experimental Analysis of Behavior , 1974, 21, 165-175. Maloney, K. B. and Hopkins, B. L. The modification of sentence structure and its relationship to subjective judgments of creativity in writing. Journal of Applied Behavior Analysis , 1973, 5, 425-433. Sidman, Murray. Tactics of Scientific Research . New York, Basic Books, Inc., 1960. Van Houten, R., Morrison, E., Jarvis, R. and McDonald, M. The effects of explicit timing and feedback on compositional response rate in elementary school children. Journal of Applied Behavior Analysis , 1974 , 7 , 547-555 . Zoellner, R. A behavioral approach to writing. College English, Jan. 1969, 30, No. 4. 56

PAGE 63

BIOGRAPHICAL SKETCH Charles T. Merbitz was born on March 30, 1946, in Chicago, Illinois. He graduated from a private high school in Chicago in 1964. In 1970, he married Gloria Golec and also' received a B. S. from the University of Illinois at Chicago Circle. In 1971, after a year of work in business, Mr. Merbitz began graduate studies at the University of Florida in Gainesville. He received his M. A. in Psychology in 1973. While at the University of Florida, Mr. Merbitz was employed as a research assistant in an animal operant behavior lab; served as Academic Advisor in University College to Minority Students (EEOP/SSS program): was Tutorial Coordinator for the Personalized Learning Center and taught in the Department of Behavioral Studies. Mr. Merbitz presently resides in Lake Wales, Florida with his wife, and is employed as a school psychologist by the Winter Haven Hospital. 57

PAGE 64

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Do^tTpr of Philosophy. Pehnypycker , Chairman Professor of Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of^ Philosophy Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor gf Philosophy M. ty. l^eyer Professor, Psycho l6gy

PAGE 65

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ;Vv\\JUvJL^ VJ. M. Newell 'rofessor of Education I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. F. S. Sieka, Ph This dissertation was submitted to the Graduate Faculty of the Department of Psychology in the College of Arts and Sciences and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December, 1975 Cwt l -v^\^ouJ^ Graduate School

PAGE 66

UNIVERSITY OF FLORIDA 3 1262 08285 169 1


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EWTXO28UP_TJ6HTI INGEST_TIME 2017-07-14T21:47:19Z PACKAGE UF00099533_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES