The Influence of fixe-ratio schedules of reinforcement on acquisition and resistance to extinction of the instrumentally..

PDF ( 3 MBs ) ( .pdf )

THE INFLUENCE OF FIXED-RATIO SCHEDULES
OF REINFORCEMENT ON ACQUISITION AND
RESISTANCE TO EXTINCTION OF THE
INSTR UMnrENTALLY C ON EDITION ED GSR









W~ILLIA111 A.1. GREENE


A4 DI55ERTATIONN PRESENTED TO THE GRADUATE COUNCIL OF
THE: UNP.ER?[TY OF FLORIDA
IN PARTLAL FULFILbLMENT OF THE REQUI~tlREENT5 FOR. THE
DEGREE: OF DOCTOR Of' PHILOSOPHY










UNIVERSITY OF FLORIDA
August, 1964











ACKNOWLEDGMEfffTS


The author wishes to express his appreciation to the ~memers of

his supervisory cormmittee for their aid and critical perusal; to the VA

Hospital at Coral Gables, Florida for providing time and facilities for

data analysis and writing; and especially to my committee Chairman,

Professor. H. D. Kimmel. Acknowledgement is also given to the U. S.

Public Health Service for their financial support (MH 06060-2).

















TABLE Of CONT~ENTS


ii





villa




1


ACKNOWLEDGMENTS .........................


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


LIST OF FIG RS .... .... ....


ChaJpter

I ~ INTRODUCTIGN ...... ......


The evidence. .. .. .. .. . ..
Recent studies. ...........
The effect of partial reinforcemelnt .
Procedural difference of the present


. . .


. .
. .


study
. .


Hypot heses .


. . .


11 RETHOD. .. .... ..


. *


Subjects. . ..
Instructions. .
Apparatus .. .
Design. . .. .
Procedure ....
Interviewr .. .


. .
. .
. .
.
. .
. .


11 RESULTS .. .. .. .. . .. * * * * *


Acquisition .. .. . *. * * * * *
Extinction~. . .. . * * * *
0 percent control group . . .. . *
Small and large emnitted GSRs. .. .. .. .. *



IV DISCUSSION. .. .. . .. ... * * * * *


Hypothesis 1. . .. ... .
Hypothesis 2. .. .. .. .
Hypothesis 3. .. .. .. *
Hypothesis 4. .. .. *











TABLE OF CONTiENTS (Continued)


Page


Incomplete adaptation and responding
below the 100 percomt base rate ... .. .
Hypothesis 5. . . . . ......
Dercrased responding anrd t:he s.nall
responses ...................
:Jature of thz large response. ..........
Conditioning the small and large responses. ..


Adaptation of small
Conclusion. ....


and largj respo~nse; .


V SUMMARY .. .. . ... . . . . . . .

REFERENCES. .. .. ... ... .. .. .. .. . .

APPENDICES. .. .. .. .. .. .. ... .. .. ... .

Appendix A. .. .. . . . . . . . . . .
Appendix B. ... .. .... .. .. .. .. .. .
Appendix C. ... ... .. ... .. .. .. ..
Appendix D. .. .. .. .. .. .. .. . ... .. ..
Appendix E. ... .. .. .. . ... ... .. ..
Appendix F. ... .. .. .. .. .. ... ... .
Appendix G. .. .. ... .. ... . ... .. ..
Appendix H. .. .. .. .... .. .. .. .. ..
Appen~dix I. . . . . . . . . . . . .
Appendix J. ... .. ... .. .. .. .. .. .. .











LIST OF TABLES


Table Page

1 MEANS AND STANDARD DEVIATIONS OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY
ANDRATIO OF: REINFORCEMENTS GROUP DUR.ING ADAPTATIONI
(IN 2-MIlN. BLOCKS). . ... . ..... .. .. 20

2 MtEANS ANID STANDARD DEVIATIONS OF ~tE F WHERE X.
IS THE AVERAGE OF THE LAST TWO 2-MIN. CLOCKS OF
ADAPTATION, FOR EACH SUBJECT. . . . . . . 22

3 MEANS A\NO STANDARD DEVIATIONS OF TRIANISFORMED
PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY
AND R~ATIO OF REINFORCEHErTT GROUP DURING ACQUIS~ITON
(IN 2-MIN. BLOCKS). . .. .. .. .. ... ... 26

4 SUMMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES DURING THE 16 MINI.
OF ACQUISITION (IN 2-MIN. BLOCKS) .. .. .. ... 27

5i MEANS AND STANDARD DEVIATIONS OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY
AND RATIO OF REINFORCEMIENT GROUP DURING EXTINCTION
(IN 2-MIN. BLOCKS). ... .. ... .. .. .. 30

6 SUMMIARY OF ANALYSIS OF VARIANCE OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES DURING THE FIRST 8
MIN. OF EXTINCTION (IN 2-MIN. BLOCKS) .. . .. ... 32

7 SUMMAURY OF ANALYSIS OF VARIANCE OF LINEAR TRENO OF
TRANSFORMED PERCENT RESPONSE FREOUENCIES DURING THE
FIRST 8 MIN. OF EXTINCTION (IN 2-MIlN. BLOCKS) .. .. 34

8 MEANS AND STANDARD DEVIATIONS OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES FOR THIE 0 PERCENT
CONTROL GROUP DURING THE 42-MIN. SESSION (IN
2-MIN. BLOCKS). . .. .. .. . * * * 36

g SUMMARY OF ANALYSIS OF VARIANCE ON LARGE RESPONSES
FOR THE LAST 8 MIN. OF ACQUISITION AND THE ENTIRE
16 MIN. OF EXflNCTION FOR THE CONTINGENT AND) NON-
CONTINGENT GROUPS RECEIVING 100 PERCENT REINFORCE-
MENFT (IN 2-MIN. BLOCKS) ......... ** 42









LIST OF TABLES (Co~ntinued)


Page


Table

10 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION
FOR 100 PERCENT CONTINGENT SUBJECTS . . .. .

II NUMBOER OF RESPONSES PERi 2-MIN. BLOCK FOR REINFORCEMENT
FOR 100 PERCENT CONTINGENT SUBJECTS. .. .. ..

12 NUMBER OF RESPONSES PER: 2-MINH. BLOCK FOR EXTI1NCTION
FOR 100 PERCENT COMTINGE~NT SUBJECTS ..........

13 RU~MBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION
FOR 66 PERCENT CONTINGENT SUBJECTS. . .. .....

14 NUIMBER OF RESPONSES PER 2-MINJ. BLOCK FOR REINFORCEMENrT


.76


.77


.78


. 8


FOR 66 PERCENT CONTIINGENT SUBJECTS.


. .


15 NUMBER OF RESPONSES PERr 2-flifl. BLOCK FOR EXTINJCTION
FOR 66 PERCErT CONTCINGENT SUBJECTS. .. .. .. ..

16 NUMBER OF RESPON~SES PER 2-MIN~. BLOCK FOR ADAPTATION
FOR 33 PERCEKT CONTI1NGENT SUBJECTS. .. .. .. .

17 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMEITT
FOR 3? PERCE~rT CONT~INGENT SUBJECTS. .. . ... .

18 NUMBER OF RESPONSES PER 2-nlit. BLOCK FOR ENTINCT:0Nl
FOR 3? ;r'CC*ET CONCTINGEPTT SUBJECTS. . . .. . .

19 NUMBER OF RESPONSES PER 2-MIN. LOCK FOR ADAPTATION
FOR 100 PERCENT NO)NCONTINBGENTI SUBJECTS. .. .. .. ,

20 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEM~ENT
FOiR 100 PERCENT NDONCOETTINGENCT SUBJECTS. .. .. .

21 NUMBER OF RESPONSES PER 2-MNl~. BLOCK FOR EXTINCTION
FOR 100 PERCEET NON~CONTINGENT SUBJECTS. .. .. .

22 NUMBER OF RESPONSES FOR 2-MIN. BLOCK FOR ADAPTATION
FOR 66 PERCEeT NONCOMTINGEN~T SUBJECTS ........


.. 85


. 86





.. 89


.. 90


.


23 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMENT
FOR 66 PERCENT NONC~ONTINGENT SUBJECTS . . ....

24 NUMER OF RESPONSES PER 2-MINi. BLOCK FOR EXTINCTION
FOR 66 PERCENT IONCONTINGENT SUBJECT ... ...


. 93


. 94










LIST OF TA\DLES Continue


Table

25 NUMBER OF RESP~ONSES PER 2-MIN. BLOCK FOR ADmAPTAION
FOR 33 PERCENT NONCONTINGENT SUBJECTS ,........

26 NUMBER OF: REsPONISES PER 2-MIN. BLOCK FOR REINFORCEMENT


Page


. 96


FOR 33 PERCENT NONJCONTINGENT~ SUBJECTS . .


. ... 97


27 EnUMBER OF RESPONSES ."ER 2-MIN. BLOCK FOR EKTINICT~ION
FOR 33 PERCENT NONCONTINGENT~ SUBJECTS .......

28 NUMBER OF RESPONISES PER 2-MIN. BLOCK FOR 1-10 ftlN.
FOR 0 PERCENT CONTROL SUBJECTS. . .......

29 F3UMBER OF RESPONSES PER 2-MlIN. BLOCK FOR 11-26 HIN.
FOR 0 PERCENT CONTROL SUBJECTS. ..........

30 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR 27-42 MIN.
FOR 0 PERCENT CONTROL SUBJECTS. . . . ..

31 MEANS AND STANDARD DEVIATIONS OF THE EAURBER OF
RESPONSES FOR EACH GROUP DURING ADAPTATinN (IN
2-MIN. BLOCKS). ... .. . . . **

32 MEANG r"� STANDARD DEVIATION-S OF THE NUMBER OF
RESPONSES FOR EACH GROUP OURING ACQUISITION (IN
2-MIN. DLOCKS). .. .. .. . .. * **

33 MEANS AND STANDARD DEVIATIONS OF THE NUMBER OF
RESPONSES FOR EACH GROUP DURING EXKT~CINGIO (IN
2-MIN. BLOCKS). .. .. .. . . ***


100


101


102



104



105



106


. .


. .




. ,


* *


* *


* ,












LIST OF FIGURES


Figure Page

1 Percentage of transformed response frequencies
during the 10-min. adaptation period in 2-min.
blocks for the Contingent and ENoncontingent
groups. . . . . . . . . * * * 19

2 Percentage of transformed response frequencies
during the 16 min. of reinforcement and the 16
min. of extinction for the combined Contingent
and combined N~oncontingent groups .. .. .. .. 23

3 Percentage of transformed response frequencies
during the 16 min. of reinforcement and the 16
min. of Extinction for the Contingent groups. ... 24

4 Percentage of transformed response frequencies
during the 16 min. of reinforcomeont and 16 min.
of extinction for the Ijoncontingent groups. .. .. 25

5 Percentage of transformed response frequencies
during the 16 min. of acquisition and the 16
min. of extinction in 2-miin. blocks for the
laisa responses in the 100 percent groups . .. .. 39

6 Percentage of transformed response frequencies
during the 16 min. of reinforcemjent and the 16
min. of extinction in 2-min. blocks for the
saggl) responses of the 100 percent groups . .. .. 40

7 Percentage of transformed response frequencies
during the 10-min. adaptation period in 2-min.
blocks for the sgl. and l31101 responses of the
Contingent and Noncontingent groups. .. .. .. 61

8 Diagram of the system controlling reinforcemeant .. 74


viii













CHAPTER I


INTRODUCTION


The purpose of this study wJas to determine the inifluence of psr-

tiol reinforcement upon the rate of emnissioni of the unelicitedi GSR, a

response madiated by the autonomic nervous system. Only recently has

any evidence been reported to support the contention that autonomically

mediated responses may be influenced by the presentation of reinforce-

ment followJing their emission, The present study, thus, was Intended to

provide further empirical evidence on operant conditioning of the GSR

and, hopefully, to relate operant conditioning of autonomic responses

more fully to conventional operent conditlioning.

K~imble (1961) distinguishes between tw~o types of conditioning--

classical and instrumental, He supports the distinction by indicating

the possible differences b~etween the two types of conditioning. Among

the differences marshalled is the autoniomi c-classical and somatic-

i nstrumental hypothes i s, imble concluded, from the evidence he pre-

sents, that this distinction cannot be completely maintained on empiri-

cal grounds as some nonautonomic responses have been classically condi-

tioned. His strongest statement on the matter, however, indicates that

even though some skeletal responses have been classically modified, au-

tonomically mediated behavior is not modifiable instrumentally: "Thus,

for autonomically mediated behavior the evidence points unequivocally to










the conclusion that such responses can be modified by classical, but not

i instrumental train ni ng methods.'* (p. 100)

Furthermore, the efficacy of this technique, wi th autonomic re-

sponses, was questioned by Skinner,for he wrote, ''Glands and smooth mus-

cles do not naturally produce the kinds of consequences involved in op-

erant reinforcement, and when we arrange such consequences experimen-

tally, operant conditioning does not take place. We may reinforce a man

with food whenever he 'turns red,' but we cannot in this way condition

him to blush. . .'' (1953, p. 114) Skinner's belief, however, was

based upon meager factual evidence.


The evidence

Mowrer (1938) attempted to make the GSR Instrumeintal in a shock

avoidance situation. His attempt ended in failure. In another experi-

ment Skinner and Delabarre (Skinner, 1938) tried to condition vasocon-

striction instrumentally. In this experiment a positive reinforcement

was made contingent upon the response, but the authors reported no con-

clusive results. The Pavlovians (Kimble, 1961) simply have indicated

that glandular responses cannot be conditioned instrumentally, without

citing any data.


Recent studies

More recently, Mandler et al. (1962) attempted to influence the

rate of occurrence of unelicited GSRs by rewarding the response with

money, the earning of which was signalled by a light. They reported in-

creased emission of GSRs for some of their Ss. The results reported by

Shearn (1960) are also somewhat equivocal. He attempted to condition









heart rate acceleration by making the criterion response instrumental In

avoiding a shock. His results Indicated that while the experimental

group differed significantly from the control group in number of accelor-

ations, there was no significant reduction in the number of shocks re-

ceived. Harwood (1959, 1962) tried to condition heart rate deceleration

by rewarding the criterion response. Although the results reported are

not conclusive his research Is continuing with improved instrumentation.

In contrast to the negative findings of Skinner and Delabarre

(Skinner, 1938) reported above, Razran (1961) has reported a Russian

study which showed instrumental conditioning of blood volume changes.

In this study five human Ss wrere stimulated with painful electrical

shock. This type of stimulation usually produces vasoconstriction, howr-

ever, if vasodilation occurred the stimurlus was terminated. When the S~s

were allowed to observe their own photo~pl ethyso~igraphs (in real time)

they learned to escape the shock.

In another study using an avoidance conditioning paradigm,

Kmmlrl and Baxter (1964) found that an avoidance group gave superior

performance whecn compared to partially reintorced classical controls

with trace conditioning and the GSR.

In a series of experiments, Kimmerl and his associates (Fowler

and Kimsrel, 1962; Kimmetc and Hill, 1960; and Kimmrel and KImmael, 1963)

have reported positive results with operant conditioning of the GSR. In

the first of these studies, Kfimsel and Hill (1960) found that the rate

of the unelicited reinforced GSRs waJs influenced by operant training al-

though the effect was not apparent untli the reinforcem;ent period wars

term nated. Using a dim white light as a reinforcer Fowler and Kimmel









(1962) attempted to determine whether the unelicited GSR could be condi -

tioned instrumentally. Threse investigators found that the ex~perime~ntal

groups receiving 1C min. of response-contingent training differed from

the control group during acquisition and the early minutes of extinc-

tion. In addition, the relative response frequency trends during ex-

tinction between the two groups were opposite in direction, although

this effect was not statistically significant. These results were in-

terpreted as providing further support for the conclusion that the un-

oilcited GSR can be conditioned operantly.

The essentioI aspects of the Fowler and Kirmml (1962) study were

replicated by Kimsecl and Kimsel (1963). In this study S~s were divided

into two groups, Contingent and Noncontingent. Contingent S~s received a

dim white light following each emitted GSR while S~s in the Noncontingent

group received the light onlL when no GSRs we~re being emitted. The dim

white light wJhich served as the reinforcer was the sole source of visual

st imulIat ion. Each S, i n the Noncontingent group was "yoked" to an S, in

the Contingent group to receive the same number of visual stimusli per

minute during the 16-mnin. training period. Following training, all S~s

received 10 min. of extinction, during which time no lights were pre-

sented. Before the actual training period was begun, 5s were allowed to

adapt to the experimental room for 10 main. This allowed the 5s to

achieve a mnore stable response rate than was obtained in the earlier

studies using shorter adaptation periods. Due to this longer adaptation

period, considerable reduction in the ~amplitude of the emitted GSR was

noted. Therefore, the criterion for counting responses and for present-

ing or not presenting the reinforcer was changed to include all responses










detectable by the naked eye. The grace perriodl following delivery of a

reinforcer was changed from 5 sec. to 3 sec. to allowr more responses to

be counted. Kimmel and Kinmel's results Indicated that the Contingent

groups responded more frequently during acquisition than during the last

5 min. of adaptation (base lovel) while the Noncontingent group responded

less frequently during the reinforceme3nt period than they hiad previously.

The large obtained differences in response frequencies between the twoK

groups were statistically significant. During extinction the trends of

the two groups were converging toward the level which had obtained durr-

ing adaptation, thus Indicating opposite trends. Both the Contingency

and the Contingency X Minutes Interaction effects w~ere statistically

significant during extinction.


1151 effect of partial reinforcemn

The previous studies (Fowrler and Kimmel, 1962; Kiimmel and Hill,

1950; Kimrsel and Kimmel, 1963; and Mandler c~t~., 1962), using the op-

erant conditioning technique with the GSR, have reinforced every re-

sponse. Other criteria for the delivery of reinforcement may be fol-

lowed. For example, every other response may be reinforced, every tanth

response may be reinforced, or almost any number of responses may be re-

qui red of the organism before reinforcement i s del ivered. In addition

to the procedure of requiring a set number of responses to occur before

reinforcement is delivered the experimental arrangement may be designed

so that the organism obtains a reinforcement after a certain time


1A response elicited by the stimulus light was neither counted
nor reinforced. Thus, any response occurring from I to 4 sec. following
the light was in the grace period.










interval has passed, provided the criterion response occurs after the

specified time period. The influence of such schedules of reinforcoment

has been studied in detail (Ferster and Skinner, 1957).

A fewr of the basic schedules of reinforcement are: (a) Fixed-

ratio schedules, in which the reinforcoemrent is delivered after the spac-

ified number of responses have occurred, (b) Variable-ratio schedules,

in wrhich the number of responses required for reinforcement varies fromn

reinforcemnrct to reinforcementn about some mean value, (c) Fixed-Interval

schedules, in which reinforcemecnt is delivered following the first re-

sponse occurring at the end of some fixed-interval, (d) Variable-interval

schedules, In which the period of time during which no reinforcemen~t is

available varies between reinforcements around some mean value. Each of

these schedules results in performance curves peculiar to the schedule.

In the typical free operant instrumental conditioning paradigm

(e.g., bar pressing) high rates of responding can be maintained w~hen the

number of responses required to obtain a reinforcement is very large.

Skinner (1957) has reported high rates of responding even when the num-

ber of responses required for one reinforcement was as high as 900.

Such schedules of reinforcement (in this case a fixed-ratio schedule)

have been found to result in high stable response rates during training

with much greater resistance to extinction than can be obtained with

conti nuous reinforcamnent. For example, Ebowrer and Jones (1945) found

that rats trained to press a bar for food reward differed in resistance

to extinction as a function of the schedule of reinforcement during ac-

quisition. The greater the ratio of nonreinforced responses to reln-

forced responses, the greater the resistance to extinction.









Recently, 80ren (1951) trained rats to bar press on several

fixed-ratio schedules and found increasing response rates with the in-

creasing fixed-ratios used. More simply, the greater the number of re-

sponses required for a reinrforcemecnt, the higher the rate of responding

obtained. In addition, it was found that the number of responses

emitted during extinction was approximately a linear increasing function

of the fixed-ratio used during acquisition.

Both the Fowler and Kimmetl (1962) and the Kimmlrn, and Kirmmel

(1963) studies reported decreasing response rates during the reinforce-

ment period for the Mloncontingent Ss. Klrmmel (1962) hypothesized that

the Ss who w~ere receiving the light when they were not responding (Mon.-

contingent Srs) may have been learning not to respond. That is, those S~s

were being reinforced for not mraking GSRs. Analysis of the stimulus

conditions during the reinforcement period (by the present writer) of

the Noncontingent group Indicated that the Ss were being reinforced on

a m~odified variable-interval schedule (mnodified because the mean inter-

val changed from minute to minute depending upon the number of rein-

forcem~ents delivered). Assuming that the hypothesized process of not-

responding (not emitting GSRs) involves active inhibition, then rein-

forcing the Inhibition would tend to increase the probability that inhl-

bition would occur as reflected by decreased responding.

In ther prl'vious studies the number of lights delivered to a N~on-

contingent S for any given minute was determined by the number o: li ghts

received by the matched Contingent S for the particular minute in ques-

t ion. The E_ th~en del iverred the light to the Njonconti ngent S unsystema-

atically (varied the interval), but only when the S was not responding.










Assuming that this analysis is correct and thiat thle Honcontingent groups

were actually being reinforced for inhibiting the GSR\ on a modified

variable-interval schedule, then predictions can be ru~de regarding the

behavior of the Dioncontingent S~s in the present study.

Each N~oncontingent S in the present study will be yoked ;o a

Conti ngent S. The number and pattern of lights received by the Noncon-

tingecnt S~s will be ill~iaediately detenmined by the Contingent is. Si nce

there are three fixed-ratio schedules of reinforcement for the Contin-

gent groups (100 percent, 66 2/3 percent, and 33 1/3 percent) the actual

num',er of lights received by a Noncontinge~nt S w~ill vary according to

the fixed-ratio schedule of the yoked-Contingent S. Thus, group 10)NC

receives the most lights, followed by group 6GtJC, and, then group 33NC.

The number of lighrts received will determine the modified variable-

Interval schedule a particular Noncontingent S. recolves; the greater the

number of lights the shorter the variable-interval.

Holland and Skinner (1961) report that a short variable-interval

schedule resulted in higher response rates than a long variable-interval

schedule. For example, when the variable-interval during which no rain-

forcement was available is long the organism tended to emit fewetr re-

sponses during the Interval than when the interval is shorter.

Assuming that group 100NC recolves the shortest variable-interval

schedule thsn this group should be reinforced more for Inhibiting GSRs,

anid, thus respond the least (see Hypothesis 3).









Procedural di fferenos gi gg est ug

The present study employjed procedures similar to those described

by Kimsel (1962) with a few major changes. These differences wJere as

follows:

1. All of the previous recent GSR studies were done utilizing a

partially soundproofed room and recording and amplifying equipment dif

ferent from that used in the present study. Although the over-all dif-

ference was quantitatively unknown it should be noted that the present

recording and amplifying equipment was more sensitive, and was able to

monitor and record from Ss with higher base resistance levels. In addi-

tion a double-walled Industrial Acoustic Chamber (IAC) was used in the

present study, in place of the partial soundproofing of the previous

laboratory.

2. The responses to be reinforced were? automatically determined

by a differential amplifier. The amplifier was adjusted by the E during

the adaptation period so that it would reinforce all noticeable deflec-

tions of the GSR recorder pen (all responses greater than or equal to a

10 ohm ilrop activated the circuit). The higher reliability of the rein-

forcaement apparatus coupled with the greater sensitivity of the equipment

resulted in the reinforcement of a larger number of unelicited GSRs than

in any previous study in the series.

3. A "true" yoked control procedure was utilized. In the pre-

vious studies eachl S wes run separately, and the PNoncontingent S~s re-

ceived a reinforcement aQlv when no GS~s were occurring. In the present

study two S~s were run simultaneously. Under these conditions the yoked-

control S received the same number and pattern of reinforcements ag th









same time as the Contingent Ss. This important modification resulted in

the delivery of a reinforcement to the Noncontingent S regardless of his

ongoi ng GSR act ivi ty. Thus, Ss in thle No~ncontingent groups might re-

ceive the reinforcer whien they wer'e making a GSR or not, since detlivery

or nondolivery was comp;letely determnined by th- behavior of the Ss in

the Cont ingent group. This yoking procedure is essentially the same as

that used by Moore and Gormezano (1961).

4. The period of extinction was increased to 16 min. Since it

was expected that the resistance to extinction would be increased after

partial reinforcement training, this increase was introduced to enhance

the likelihood of detecting the effect.

5. The previous studies all utilized a 100 percent schedule of

reinforcoment. The present study used several different ratios of par-

tial and continuous reinforcement to determine their effects.

6. Since the effects of prolonged absence of stimulation on the

unelicited GSR were unknown for the present laboratory a special 0 per-

cent group was run. The Ss in this group merely sat in the experimental

chamber but received no stimulation.


~ L~otim..t.1e i

The major hypotheses concerned the effects of Contingency of re-

Inforcemetnt (Contingent vs. Noncontingent) and Schedules of reinforce-

meant (100 percent, 66 2/3 percent, 33 1/3 percent, and 0 percent) upon

the rate of emi ssion of GSRs during the experimental sess i ons. The

specific hypotheses were as follows:









i. The rate of occurrence of emitted GSRs will vary as a func-

tion of Contingency of reinforcemrent. The Contingent groups will emnit a

greater number of responses duringJ acquisition and extinction.

2. Within the Contingent groups, thre rate oF unelicited GSRs

will vary ;?vorsely with the percentaya of reinforcemepnt, both during

acquisition and extinction. Thus, thre highest rate of responding will

be obtained in the 33 1/3 percent group, followed by the 66 2/3 percent

group and 100 percent group, respectively. in addition, all of the Con-

tingent groups w~ll respond above the 100 percent base level during ac-

quisition and extinction.

3. Within the Noncontingent groups, the rate of emission of un-

elicited GSRs will vary inversely wJith the schedule of reinforcement re-

ceived by the yok~ed-Contingent is. Thus, the ordering of groups will be

the same as in hypothesis 2, but by Hypothesis 1, the over-all rates of

the Noncontingent groups wJill be lower then the Contingent groups.

4. During the extinction period the rate of GSR emission will

tend to return to the 100 per cent level (base level). Thus, the Con-

tingent groups will decrease responding, and the Noncontingent groups

will increase responding during extinction.

S. The rate of unelicited GSns in the D percent group w~ill tend

to stabilize and remain constant for the experimental session.












CHAPTER II


METHOD


Subjects

One hundred and five undergraduate students at the University

of Florida, 63 men and 42 women, volunteered as Ss. Each 5 who served

in the experiment was paid one dollar for participation.2 Ss of the

same sex were run in pairs and the proportion of males to females in

each treatment was equal.


Instructions3

The Ss were Instructed to relax, to remain as motionless as pos-

sible, but to remain aw~ake and alert. An interview was conducted at the

end of the experimental session to determine whether these instructions

had been followed. If It was determined that an S had not followed in-

structions he/she and the yokted partner were rejected. Ss would be re-

jected due to experimenter error, equipment malfunction, or failure to

follow Instructions (sleeping, moving, closing the eyes, etc.).


Appa rat us

A small, dim patch of white light (0.7 ft. candles), located in

the direct line of sight approximately 30 in, in front of 5, served as


2Fhis support was provided by grant HM 05050-2.

3See Appendlx A for complete instructions.










the reinforcing stimulus. The experimental rooms were double-walled

soundproofed IACs equipped with a padded chair, a small minlbox housing

the reinforcing light, and a speaker which served as an intercom and for

commiiunication with the S. The GSR wa~s picked up as a DC resistance

change from the palm and back of the I's left hand by 3/lr In. zinc elec-

trodes. The electrodsr- were covered with small gauze patches saturated

withl zinc sulfate solution and wJere housed in lucite cups filled with

saline electrode paste (Grings, 1954). The constant current through the

S was 20 microamps. The control panel in an adjoining room contained

two Biophysical GSR amplifiers, and Texas instruments Co. Rocti -Riters.

The recorders wrere equipped with signal-megnet pens for recording the

occurrence of the reinforcing stimulus. The circuit controlling the re-

Inforcing stimuilus contained a differential amplifier and a switching

systemn for automatic delivery and programming of the stimulus. in addi-

tion a hold circuit prevented the delivery of the reinforcing stimulus

more than one time In any i, sec. period (the grace period). A block di-

agram of the reinforcing system can be found in Appendix 8.


Delsisn

The basic experimental plan consisted of a 2 x 3 factorial de-

si n with 15 is assigned at random to each cell, w~ith the restriction

that each pair' of Ss be of the same sex and the male/female ratio be

constant between cells. Ss were run in pairs, one member being assigned

to the CqgatitREL~ group and one member to the Noncontingrent group. The

independent variables were percentage of reinforcemeent (100 percent,

66 2/3 percent, and 33 1/2 percent) and Contingency of reinforcemrent










(Contingent and Noncontingent). The reinforcing circuit was automati-

cally activated by any response (= 10 ohms) of the Contingent S and de-

livered to both members of a pair at: the same time.

In addition, another group of 15 S~s wJas run under conditions of:

no stimulation (0 percent group), to determine the course of GSR omis-

sion without stimulation.


Procedure

The two> S~s of a pair entered the laboratory via different doors

and were seated in separate experimental chambers. The palm and back of

each S~as left hand were then cleaned by acetone and the GSR electrodes

were attached and held in position by a rubber band. The overhead light

wras extinguished and the experimental chamber doors were closed placing

each S in total darkness. The experimenter (~ then read the instruc-

tions over the Intercom. The S.'s questions wJere answered by rereading

the appropriate part of the instructions. During the entire experiman-

tal period the ~s w~ere monitored auditorily to detect any noises due to

movements (shuffling, changing position in the chair, etc.), sneezes,

coughs, etc. whstich might produce a GSR. All GSRs so identified were not

scored as responses.

After a brief interval, during which the instruments were checked

and adjusted a 10-min. adaptation period was begun. At the conclusion of

the 10-min. adaptation period, the 16-min. reinforcement period was be-

gun. Responses made by Contingent S~s were passed through the differen-

tial amp~lifier which closed the circuit delivering reinforcements; the

reinforcements wrere delivered to both Ss of a pair simultaneously. The









programming apparatus delivered rainforcoments on a fixed-ratio schedule

(I.e., 1:1, 2:3, and 1:3). The circuit was designed to deliver a rein-

forcement at the point at whlichh the particular GSR had reached its

maximum. The duration of the reinforcement was 1/10 sec.

FollowinUI the reinforcement period the S~s were given a 16-min.

period of extinction during which no stimulation was given. At the con-

clusion of the extinction period, the electrodes were removed and the

InterviewJ was conducted separately for each S.


interview

Each 5 was asked, (a) "Did you have the impression at any time,

during the experiment thiat the occurrence of the light depended upon

something you might have done?"4 If the answer to question (a) wars neg-

ative, the S was told to make the hypothetical assumption that: his be-

havior controlled the light, and then he was asked, (b) "Can you now

think of any possible connection between your behavior and the presence

or absence of the II ht?" On~ly the data of Ss whose answers indicated

that they had followed the instructions (particularly regarding move-

ment) completely were used. Five Ss were rejected for fai lure to follow

the instructions regarding mo~vemewnt.5 The Ss were then thanked, given a


three is, one from each Contingent group,indicated that they
thought the light came on whecn they expected it. Exami nation of their
records Indicated that these is were near the mean of their respective
groups.

SThree contingent Ss in the 100 percent group were rejected for
moving. TwJo Contingent is in the 33 1/3 percent group we~re rejected,
one for moving, thp other .~or deep and irregular breathing. The rejec-
tion of these Contingent Ss necessitated the rejection of their yoked
controls, therefore, a total of 10 5s were rejected.






16


chit exchangeable for one dollar, and instructed not to discuss the ex-

periment with anyone. The Ss of a pair left the laboratory by the doors

through which they had entered.












CCHAPTER III


RESULTS


The primary performance measure was the number of responses

emitted per minute during adaptation, acquisition, and extinction. A

response was defined as any deflection of the GSR recorder pen which In-

dicated at least a 10 ohm drop in resistance. The reliability of this

measure was determined separately for the Contingent and No~ncontingent

groups. Th~is was accomplished by having two judges (one of whom was the

E) read an unsystematically selected minute from the record of' each of

the Ss in the main experiment. The second judge read the records

"blindly," in that she did not know to which group a record belonged.

The cotrreation (Pearson) between the readings of the two judges was

calculated for each group and both indicated high reliability (Contin-

gent, r r .96, N~oncontingent, r a .97). Both of these correlation coef-

ficients wrere significantly different from zero (p< .001) but not from

each ot her. Only unelicited GSRs were counted, thus all responses be-

ginning between i and 4 sec. after onset of the light were not scored.

These criteria applied to all records regardless of the group to which

they belonged.

The basic response frequency mePasures6 were first. grouped into

blocks of 2-min. Intervals to increase their stability. Then they w~ere


bAppendices C-1 present the untransformed frequency scores for
each of the 105 Ss in 2-nin. blocks. Appendix J presents the Means and
Standard Deviations of the untransformed frequency scores in 2-min.
blocks for all groups.









transformed to'\ K E to overcome the sktewness of their frequency dis-

tributions (Snedecor, 1956). The transformed measures were then ex-

pressed as percentages of the transform of the average of the last two

2-min. blocks of the 10-min. adaptation period. These transformations

were done separately for each 2-min. block for each S.

The group performance curves during adaptation were expected to

indicate declining relative response frequencies as adaptation progressed,

as well as no differences in average frequency at the end of adaptation.

The relative response frequency curves, shown in Fig. 1, indicate that

the general trend of the curves is doiwnwrd, as expected. Table 1 pre-

sents the means and standard deviations for the adaptation period for

t~he transformed percent response frequencies in 2-min. blocks. An anal-

ysis of variance of these data indicated that only the M4inutes effect

was5 significant (F r 7.02, df = 4/334, p
effect, in conjunction with the shape of the curves in Fig. 1, indicated

a significant adaptation trend.

The response frequencies during the last 4 min. of adaptation

wrere examined for each of the seven groups. It was necessary that these

pretraining frequenclos be comparable between groups so that any differ-

ences found after training could be attributed to the experimental treat-

mient and not to sampling errors. The average response frequencies (and

SDs) during the last two 2-miln. blocks of adaptation transformed to


7For one pair of Ss the recording apparatus was not started un-
til the beginning of the third minute of adaptation. Thus, 2 missing
data wcere replaced and 2 df were subtracted from the error term.








19










rcI
cC

Z Z ZJ cr
wu. u cc

~.. o
u J w u J 15 ,~
OOO3CZZZ 1,'
ZZZO OO u-a

ZZZZZZ f,
000000 !'t o i
uou09 z z.o. -
o-o mo m r/ C,? Cn

oa a ," o.

I~~Q : 'sy?
o a *e * nr

Ij I o ro.
if E

I .I v O M
i : : c-






rUOE

I~ aQ.
r I

I I I Iu -





CO













TABLE 1

MEANIS AND STA~RDRD DEVIATIONS OF TRANJSFORM:ED PERCEIiT& RESPONSE
FREQUENCY IES OF EACH C~TW I NGENCY AND RAT IO OF
REINFORGCEMEK GROUP DURING ADAPlTATON:


Adaptation
Group 1 2 3 4 5


Contingent


100
SD


66
SD


33



Noncontingent


100
SD


66
SD


33
SD


105

33.6

105

23.2

121

25.2


101

26.3

96

13.7

100

13.8


102

27.2

92

17.9

103

19.3


399

11.0

96



37

11.8


100

11.9

103

6.7

102

10.9


114

28.7

103

20.0

Ill

23.9


101

11.5

103

12.0

105

18.3


110

23.6

399

14.5

102

13.2


101

7.2

97

5.9

98

9.6


99

7.7

102

5.4

101

9.3









l6 E for all groups are shown in Table 2. Analysis of vari once i ndi -

cated no significant differences amonng the groups (F
To compare the effect of the reinforcer on the different Contin-

gent and Noncontingent groups during acquisition and extinction, the

performance curves for the different groups were plotted in 2-miin.

blocks for the 16 min. of acquisition and extinction. Firrs. 2, 3, and

4 present these curves.


Acquisition

The first section of Fig. 2 shows the combined performance curve

of the three Contingent groups and of the combined yoked Noncontingent

controls during the 16 min. of acquisition. Inspection of these curves

reveals that the over-all mean of the Contingent (95.83 percent) was

higher than that of the Noncontingent groups (88.33 percent). Further

examination of the acquisition curves indicates that the over-all dif-

ference between the Contingent and Noncontingent groups became larger

during reinforcemnent.

To evaluate the statistical significance of these differences

the data obtained during acquisition were subjected to an analysis of

variance. Table 3 presents the means and standard deviations for the 90

Ss of these groups for th~e 16 min. of reinforcemnt, and Table 4 summaJ-

ri zes the analysis of variance. The analysis indicated that the differ-

ence between Contiingent and Noncontingent groups were statistically sig-

nificant (F = 6.23, df f 1/84, p
found to be significant.




















Group Hean Standard Deviation

Cont intent

I00 3.33 0.82

66 3.29 0.72

33 3.06 0.76


Non cont Ingent

100 3.31 1.04

66 3.30 06

33 3.36 0.68


01 control 3.I5 0.76


TABLE 2

MEANS AND STANDARD DEVIATIONS OF jj- WHERE X IS THE
AVERAGE OF THE LAST TW~O 2-MIN. BLOCKS OF
ADAFFAT10N, FOR EACH SUBJECT





I I I _




( L+ISa~'. l+rSa11\00L)

X~uanbaij asuodsa~ paiuiojsuoil jo rua~iad


r
.C


CC



in







CC



LC-

4- c

O --


E
CLO

.cE



CCD
re


OU





~C


SE


I






o a





o'

o'


O


I r
O
n V


o


oO


a

a I


C

* V



O

O r


O
"
3
SC























1,


z z z ..C9 -- -5E
ZZZ C"E



v v. v -






,~~Q c
o' LC
,'~ t O--0

.d 0I -n N )

DO




o 0 --



OO



o *C


o
*C bE


O L-








25

















i O -- C

Z Z Z .E O
ooo c moJ

o Z o .; .- U
~~s I- .a 0




-CO
~~oo -- '




I OO
, vi *



.- O -


oac







COX
.- L..












TABLE 3

MEAllS AND STANDARD DEVIATIONS OF TRANSFORM1ED PERCENT RESPONSE
FREQUENCIES OF EACH CONTINGENCY AND RATIO OF
REHI~FORCEM3EIIT GROUP DURING ACQUISITION
(IN~ 2-MIN~. CLOCKS)

Acquisition
Group 1 2 34 5 G 7 8


_ __ __


Contingent


100
SD



SD


33
SD


22.4

93

11.0

106

16.0


96

14.6

96

15.9



13.9




82

14.4



20.4

93

16.9


12.1

97

20.0

103

24.9




88

13.6

92



89

20.7


94

14.2

39



97

21.9




81

24.0

90

17.8

88

24.8


89

16.1

101

16.0

9;3

23.5



30

20.3

89

18.4

87

22.1


92

25.5

95



97

15.7




95

31,.7

83

16.1

92

9.5


95

20.2

90

14.1

99

19.4




86

26.0

82

19.9

93

14.6


95

13.7

94

21.8

97

20.9




79

27.9



17.7



19.9


Moncont ingent


100
SD



SD


33
50


98

37.9

91

15.3

94

17.4














TABLE 4

SUMMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCEIJF RESPONSE
FREQUENCIES DURING THE 16 MIN. OF ACQUISITION
(IN 2-MIN. BLOCKS)

Source df MS F

(Betweeon Ss) (89) 1,700.33

Contingency I 10,132.50 6.23*"

Rat io 2 2,164.r46 1

C xR 2 151.83

Error (b) 84 1,625.76


(Within Es) (630) 21,7.14-

Minutes 7 214.29

Mx C 7 233.55-

M x R 14 279.79 11

M x C x R 14, 260.02 1.05

Error (w~) 588 246.78


*p <.025.









Fig. 3 indicates that the differences an~ong the Contingent

groups were greatest during the first 2-mrin. block of reinforcement.
These differences oscilIlated and became smaller as the number of rein-

forcements increased, until, at the end of reinforcement, there wras al-

most no difference among these three groups.

To evaluate this trend, the data for the Contingent groups were

subjected to an additional separate analysis. Although this analysis

indicated no significant over-all F ratios, f tests on the first block

of 2 min. showed the response rate of the 330 group to be significantly

greater than that of the 6GC and 1000 groups (_t r 1.93, p
&L = 2.68, p
These comparisons tend to support the observation that the larg-

est differences among the Contingent groups occurred during the first 2-

min. block of acquisition.

Several facts about the data suggested that the linear component

of the acquisition trand in the Contingent groups be examined. The 330

group's mean performance during the first 2-min. block was about 100) per-

cent, while both the GGC and 100C were below~ 100 percent. Secondly, the

over-all curves of the two groups that were below 100 percent at the be-

ginning of acquisition appeared to increase slightly as training pro-

gressed while the 33C group appeared to decrease. It was clear that the

slopes of the 100C and 66C groups were positive wrhile that of the 33C

group wras negate ve. The di fferences among the liner components of

these trends were significant (F = 8.91, df = 2/42, p

&The error term used in this analysis Is the partitioned error
term suggested by Grant (1956) and more recently by Galto and Turner
(1963).










establishing the statistical reliability of the apparent convergence of

the Contingent groups during training.

Contrasted to the behavior of the Contingent groups, were the

performance curves-of the Noncontingent groups. Fig. 4 indicates that

the tendency of these three groups was toward fewer responses as train-

ing progressed. No statistically significant differences in i~ncar

trend w~ere present In these data nor was their over-all trend signifi-

cantly negative.


Extinction

The second half of Fig. 3 and the second half of Fig. 4 show the

transformed response frequency curves for each Contingent and NSoncontin-

gent group, respectively, for the extinction period, in 2-mln. blocks.

The means and standard deviations of these data are presented in Table 5.

The first thing to be observed In the extinction data is the

general convergence of the groups. By the end of the first 8 min, of

extinction, the differences that were present at the end of training had

almost entirely disappeared, with the exception of group 100NC, which

fluctuated greatly. Furthermore, the only group that remained consist-

ently above the base level of 100 percent was group 33C. Of special in-

terest was the behavior of the Contingent groups as compared to the Nlon-

contingent groups during the first 2 min. of extinction. The Cont ingent

groups separated considerably from each other,w~hile the differences

present during the last 2 min. of acquisition for the N-oncontingent groups

almost entirely disappeared during the first 2-mi~n. block of extinction.

These changes werre not statistically significant.













TABLE 5

MEANS AND STFANDARD DEVIATIONS OF TRANSFORMED PERCEFIT RESPONlSE
FREQUENCIES OF EACH CONTI7NGENCY AND RArTIO OF
REINFORCEMENT GROUP DURING EXTINCTION
(IN 2-MI~N. BLOCKS)

Extinction
Group 1 2 3 5 6 7 8


Contingent


100
SD



SD


33
SD


103

20.6

107

3'.6





17.9

96

19.9

92

25.5


101

17.5

103

21.3

102

25;.6




87

17.4

102

23.6

101

22.3


97

8.4

97

17.5

107

32.4




92

26.8

98

15.8

95

25.2


93

18.7

102

22.1

105

26.8




93

26.1



19.8

100

22.0


36.0

108

23.6

100

22.3




92

22.1

93



399

20.9


93

17.5

93

25.5

106

28.8




89

22.7

91

23.8

103

18.7


95

13.7

103

25.0

112

32.7


97

10.2

95

26.5

105

28.0




97

20.0

99

21.2

92

20.3


Noncontingent

x
100
SD

X

SD

x


96

24.7

92

19.3

93









An analysis of variance wa3s carried out on the data from the

first 8 min. of extinction and Is presented in Table 6. The analysis

showed several things. First (see Table 5), extinction9 occurred to a

sufficient extent that the over-all difference due to Contingency of: re-

Inforcement was not significant. Secondly, the Minutes X1 Contingency

interaction was not significant, indicating again the extent to which

extinction mitigated against significant group differences (see: the sig-

nificant mondified Minutes X Contingency interaction below). To evaluate

possible differences which occurred early in the extinction period but

soon disappeared, analysis of variance was carried out on the data of

the first 2-min. blosk of extinction. In thils analysis the Contingency

effect resulted in an F ratio of 3.81 (df X 1/81,, p
breakcdown of the data during the first 2-min. block was performed by

ranking the measns anid carrying out Duncan'G Range test (1955).10 Thi s

test indicated that the mean of group 33C was significantly different

from the mean of the G6NC group (p <( .05). Analysis of the data of the

second 2-min. block~ of extinction indicated that the 33C group was sig-

nificantly greater than the 100NC group.

Since the differences betw-een the means of groups 1000, and 100100

fluctuated widely during the first 8 min. of extinction, these two

groups were set aside and another analysis was performed on the remawin-

ing four groups: 660, 66NC, 33C, and 33NC. Analysis of variance on the

9Oy "extinction" Is meant "return to pretraining response
level." For the Noncontingenrt groups this means increasing, while for
the Contingent groups It means decreasing.

10For the rationale behind such a test after obtaining a nonsig-
nificant F see Edwards (1960).















TADILE 6

SUMMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCENT RESPONSE
FREQUENCIES DURINtG TH~E FIRST 8 MIN~~. OF EXTINCTION
(IN 2-MIN. B)LOCK:S)

Source df MS F

(Detwreen S~s) (89) 1,551.75

Contingency I 4.319.47 2.80

Ratio 2 1,304.72

C xR 2 694.03

Error (b) 84 1,545.10


(Within Ss) (270) 199.46

Minutes 3 153.74

M x C 3 261.46 1.35

Mx R 6 181.77

1x ;(C x R 6 459.07 2.37~

Error (ur) 252 193.48


*p <, 05.









data from these four groups during the first 8-min. of extinction re-

suited In a significant Contingency x Minutes interaction (F a 3.05,

df = 3/168, p <1 .05). This significant Interaction indicated the con-

vergence of these groups over minutes. Both Contingent groups decreased

responding, and both Noncontingent groups increased responding.
Table 6 showrs that the Minutes X Contingency X Ratio effect was

significant (F a 2.37, df : 6/252, p clarified by examination of Figs. 3 and 4. The performance curves indi-

cate that the differences between groups 33C, 33NC, and the differences

between groups 66C, 66NC appear to have become smaller while the differ-

ences between groups 1000 and 100NC fluctuated. Further analysis of

this significant interaction was accomplished by analyzing the trend of

the data for the first 8 min. of extinction. Table 7 presents the re-

sults of this analysis. A significant Contingency X Ratio linear com-

ponent was found (F 1 3.1,1, df : 2/84, p < .05). This effect resulted

largely from the fact that the difference between the linear extinction
trends of the 33C and 33NC groups was significant (F = 4.32, df a 1/84,

p <1.05). The triple interaction of Table 6 was probably significant
because of the convergence of groups 33C and 33NC, the partial conver-

gence of groups 66C and 66NC, and the considerable fluctuation between

groups 1000 and 100NIC.


O percent control aroup_
It was hypothecsizesd that thle Ss comprising the no stimulation

group, who ererly sat in the experimeantal chamber, would adapt to the
situation and then stabilize (indicated by an initial decrease in


















TABLE 7

SUMMA~tRY OF ANALYSIS OF VARIANCE OF LIN0\AR TRENID OF TRANSFORMED
PERCENT RESPONSE FREQUENCIES DURING THE FIRST
8 MIN. OF En`TINCTION
(IN 2-MIN. BLOCKS)

Source df MS F

(G roups) (5) 499.90

Contingency I W41.10 1.G2

Rat to 2 98.61

C xR 2 930.56 34

Error (wr, lincor) 84 272.60


*p < .05.









relative response frequency followed by fairly constant responding).

Table 8 presents the means and standard deviations of the transformed

response percentages in blocks of 2 min. for the 15 Ss in this group.

The data of the 0 percent group are presented in the same timeL divisions

as the earlier data (10-min. adaptation, 16-min. acquisition, 16-min ex-

tinction). Table 8 indicates an immediate decrease in responding dur-

ing the first 8 min. followed by a slight Increase during the next 2-

min. block. During the next 16-min. period (comlparable to the acquisi-

tion period) the mean percentage responding remained approximately con-

stant at a little above the 100 percent base line and then decreased be-

low~ this base line toward the and of the period. The last 16 min. (com-

parable to the extinction period) were marked by a general decrease

after a slight Initial increase in responding.

Statistical analysis of the first 10-min. period (adaptation)

indicated that the variation of the means of the five 2-min. blocks wa~s

statistically significant (F = 2.64, df r 4/56, p <( .05). This find-

ing, In conjunction with the means shown in Table 8, Indicated that

adaptation didloccur during the first 10 min. Analysis of the variation

of the 2-min. means from minutes 11-26 and minutes 27-42 showed no sig-

nificant differences, indicating that the subjects Ihad more or less

stabilized.

Another indication of the effect of the reinforcing stimulus was

obtained by comparing the 0 percent control group to all other groups in

the main experiment during acquisition and extinction. Examination of

Figs. 2, 3, 4, and Table 8 shows that, after the first 2-min. block of

acquisition, the 0 percent control group responded at a higher level of















TABLE 8

ItEArlS ANID LTANIDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE
FREQUCNCI ES FOR TH1E 0 PERCEFTT COtWROL GROUP DURING
THE~ 42-MIN. SESSION
(IN 2-MIN, BLOCKS)

Minutes 1 2 3 14 5 6 7 8

Ill1 103 100 96 103 --
1-10
sD 17.4 16.6 11.4 9.3 8.G --

10 106iO 104 106 107 102 96 100

so 16.4 16.8 20.4 17.5 17.4 20.7 17.6 1.

10 104I~r 103 98 95 99 100 93
27-42
SD 23.9 23.3 17.7 23.6 23.3 28.4 29.1 20.8









relative response frequency than all the other groups until the 14th

minute. However, the 0 percent Control group gradually gave fewer and

fewer responses while the Contingent groups we~re either gradually giving

more responses or remaining about the samne; all of the N~oncontingent

groups were gradually decreasing.

Analyses of variance carried out on these data indicated that

the 0 percent Control group gave significantly more responses during ac-

quisition than: (a) 100C group (F 5.58, df a 1/28, p <1 .05), (b)

100NC group (F = 7.61, df = 1/28, p <( .025), (c) 66NC group (F' r 10.20,

df = 1/28, p
p ( .025).

During extinction thre 0 percent control continued to decrease re-

spondi ng gradually. For thre first 6 min. of extinction the 0 percent

control group's response level wa~s equal to or above oil the groups ex-

capt group 33C. (Group 330 responded consistently above all other

groups throughout extinction.) For the remainder of the extinction pe-

rIod the 0 percent control responded consistently below or equal to all

groups except 100NC.


Smell and large emitted GSRr

It was expected that the relative response frequencies for the

Contingent groups would be above the 100 percent (pretraining) level

during the reinforcem~ent period. Fig. 3 shows that the opposite oc-

curred. In general, all of the Contingent groups responded below their

owsn preliminary levels.









There was the possibility that the sensitivity of the equipment

used in the present study resulted in recording and reinforcing of very

small responses whlich mauy not halve been conditionable.11 Lacey and

Lacey (1958) have pointed out that some GSR activity is fairly stable

over time and is not correlated with environmental stimulation. If

these responses are actually not conditionable, then their "reinforce-

ment'' and inclusion in the frequency count for determination of the

present dependent variable would contaminate the 'true" conditioning ef-

fects. To examine th~is possibility the total response frequency was

dichotom~ized into those responses greater than or equal to 1 percent of

the base resistance of the S at theQ time a response occurred (large re-

sponses), and those responses less than 1 percent of the base resistance

(smnall responses).12 Base resistance in this case was defined as the

resistance (in Kilohm~s) of the S. Each Sis base resistance was obtained

for each\ minute of the experimental session and samgg and lgarg responses

wrere measured and tabulated accordingly. These were then transformed

and expressed as percentages of the number of responses during the last

4 min. of the adaptation period. Those transformations were carried out

on the data for all Ss separately and the results for the 1000 and 100NG

groups are presented in Fig. 5 (large) and Fig. 6 (smell) In blocks of 2

mi n. The first half of Fig. 5 (large responses) Indicatles that the 1000

group responded consistently above the 100 percent base level for the


ilThe reason these s~mal responses may not be subject to condi-
tioning by instrumental methods will be dealt wilth in the discussion
section.


12Kimmael (SEPA, 1964).


























vZ Z



0 - u .-




~ 1"
n 0 O 3 c

o* OC


ccoi





OOo


EO



-- C
C In.

0 CE
o .*s~ LIa1\0 -

X~a\ai -soda LOu~u i ola~a





o o o
o ,z cc,



(L*l~a~'.L+lra1~00L )

X~uanbai~ asuodsatl paiuiolsuD11 10 iua~iad


LJ

Ccn



O


S.uc
OYE
SI



- -
gc
aO

OnC
*U -
L J

y 0 *



S-c
TEL)
I- *
-- 0





SC'D


I u




*C C
crl *- O
-- *U a.

oLL


r~ I





z, c




.-**


o


o





o


o
/
o


O


N U

co O



m O
v
-


















P Q


- C


O

0 a


o *I~~C






o


o

,::::::----- ~,' ;









entire 16 min. of acquisition although no apparent trend is evident.

During this same period the 100NC group responded consistently belowJ the

100 percent base level and below~ the 1000 group (except for one data

point). An analysis of variance over the last 8 min. of acquisition wJas

carried out on the transformed relative response frequencies grouped

into 2-min. blocks and is summnarized in Table '3. The analysis indicated

that the obvious difference (for the last 8 min. of acquisition) bztweecn

the two, groups was significant (F 4.40, df = 1/28, p <1 .05). In ad-

dition, the apparent increasing separation between the two groups over

time from the 10th through the 16th min. was reflected in a significant

Contingency X Minutes interaction (F r 2.91, df 3/84, p
The second half of Fig. 5 shows the large response performance

during the 16 mln. of extinction. The figure shows that the 1000 group

initially increased responding, reaching a peak; at 8 min., and then re-

sponded below this peak for the remaining 8 min. The 1000 group re-

sponded above the 100 percent base loval throughout the extinction po-

riod. The second half of Table 9 presents the analysis of variance for

these data over the entire 16 min. of extinction. The analysis Indi-

cated that the difference between the Contingent and Noncontingent

groups was significant (F = 4.98, df = 1/28, p
gency X Minutes interaction was not significant, Indicating that the

differences between the two groups did not change greatly during the ex-

tinction period.

Fig. 6 presents the curves of the mean transformed response fre-

quencies of the 1000 and 100N~C groups for responses less than 1 percent

of the base resistance. The figure shows the performance curves in
















00
c I



.$* ~r
Zb~I
c~L1
V~hCI







U1 CP
v m l





-C 0

-f




I
~C~uI
r




hI










5 O




CIULI



a a ni
- UO -


a a o



mt an -
****cL













N










CO 01







LD m m C

**** f
- LA O



O N L
If" c O













*.ol n 3
0 0
r u
a L
.c" O
4.8 .- L.
*- I 1
3 w


2










uO
0
2 2
..O
U


CCD 0 -
22 ~


20
U.1 &
up a

oa e-h
LU. V3



LI1 ce" uJ
..I O. O G.
(I J) b. 2

0-"- I

w-ez E
.. > s
EU.1
Ome U
W.M
WLLL
a Us
z ev.
2Z"

\o -
<0 %
>0
10c
U. -
Iw
vtI*
--
..J


k.




vi


VV\
L. O
*IC


..










blocks of 2 min. for the 16 min. of acquisition anid 16 min. of: extlinc-

t ion. Since this class of response was tentatively identified as non-

c~onditionable, the course of change for these responses was expected to

be uncorrelated with reinforceme~nt contingency. The first half of Fig.

6 indicates unsystematic changes. Analysis of variance of these data,

however, resulted in a significant Contingency X Minutes Interaction

(F r 2.97, df = 7/136, p
this significant interaction effect was due to the large shifts in the

magnitude of the differences between the two~ groups and the shifts in

their relative positions. The extinction curves for this class of small

responses are presented in the second half of Fig. 6. ImmeRdiaJtely obvi -

ous from the figure is the apparent hiigher response rate and large fluc-

tuations of the 100NC group. Thr~oughout this period the 100NC group re-

sponded at a consistently higher rate than the 1000 group, a result

which is just the opposite of what wJould be expected were this response

class being conditioned. Also apparent from Fig. 6 is t).s large incre-

ment in relative response frequency of the 100NC group from the last: 2

min. of acquisition to the first = min. of extinction. An analysis of

variance was carried out on the data for this 16-min. period of extinc-

tion. The apparent suprelority of the 100NC group in relation to the

1000 group was statistically supported (F 1 4.77, df = 1/28, p < .05).

Using the large and small responses, previously defined, as the

dependent variables of 66C, 66NG, 33C, and 33NC groups were also further

examined. Since no systematic effects were found, the data from these

groups are not presented. The purpose of the presentation and analysis

of thre small and large responses will be considered in the next chapter.













CHAPTER IV


DISCUSSION


The purpose of the present study was to determine whether the

course of conditioning and extinction of the uneliclted GSR could be

differentially Influenced by different schedules of reinforcement. The

results indicate that a differential effect was obtained. One group of

Ss received a response-conti ngent, dim, white light as a positive rein-

forcement. This Contingent group was subdivided into three equal s~ub-

groups. Each subgroup received the reinforcer on a different fixed-

ratio reinforcement schedule, viz., 100 percent, C6 2/3 percent, 33 1/3

pe rcont Yokced to each S In the Contingent group was an S wrho recei ved

the same number and pattern of reinforcements, but the delivery of the

reinforcement was completely determined by the Contingent S. Thus, Ss

in the Noncontingent groups received the reinforcer independent of their

behavior.

Several facts led to the assumption that the dim white light used

in the present study would act as a positive reinforcer. The 55, sart in

an experimental noom which was sound- and light-proof. Befor-e actually

beginning the delivery of the light, a 10-min. adaptation period was

given during which time the 5 m~erel:y sat In the dark room. Thus, it

would seem to follow that any mild enanges in this sensorily deprived

environment light serve as a reward. Several studies on animal Ss con-

firm this hypothesis. Barnes and Baron (1961), Barnes and Kish (19583),









and Kish (1955) found that when mice were confined to a dark chamber

significant increases in number of bar contacts were obtained when light

onset was made contingent upon bar contacts. Fox (Solomon, 1961) wJas

able to train monkeys to press a lever to turn on a light w~hen they were

kept in a dark chamber. When the m~onk~eys wero next deprived of light

for a period of hours and then given access to the bar, a large and sig-

nificant increase in responding occurred. These studies help define the

brief light used in the present study as a positive reinforcer. Perhaps

even more important for the definition of the i~ght as a positive rein-

forcer are thec investigations by Fow~ler and Kimserl (1962) and Kimsel and

Kimmel (1963). Both of these studies were designed to condition the GSR

instrumentally with a brief flash of dim wJhite light as the positive re-

inforcer. Both studies reported positive results.


Nlvpothopis J.

The first: hypothesis of concern in the present study stated that

the Contingent groups would emit a greater number of responses during

acquisition and extinction than the Noncontingent groups. Analysis of

the data indicated the confirmation of this hypothesis for the acquisi-

tion period. The overall average relative response frequency for the

Contingent groups was 35.8 percent while for the Noncontingent groups it

was 88.3 percent. This superiority of the Contingent groups was statis-

tically significant (F 6.23, df 1/84, p
studies (Fowler and Kimmel, 1962; and Kimsel1 and Kimseel, 1963) on condi-

tioning of the unelicited GSR also reported such differences during the

acquisition period. Thus, the present result confirms the previous









findingsand is in agreement with th~em. However, an important procedural

change introduced in the present study should be emphasized at this

point. The earlier studies using this technique delivered noncontingent

reinforcement only at times when the S was not making a response. That

restriction precluded the delivery of the reinforcer fortuitously whein a

response was being emitted. HowJever, in the present study an S, in the

Moncontingent group could receive a reinforcement at any time, whecnever

his Contingent mate received one. It had been anticipated that this

would result in smaller absolute differences between the Contingent and

ilonconti ngent condi tions. This, in fact, did occur. Tj :etermine the

extent to which a Noncontingenlt S was reinforced for emitting a GSR\, the

number of reinforcing lights (a response was considered reinforced if

the light occurred contiguously with the response or within I sec. after)

wa~s divided by the total number of lights received by an S. This score

w~as then expressed as a percentage for each S. The means and standard

deviations of the percentages of the reinforcing lights for the three

Nonconti ngent groups were: 100NG, Mean r 18.93 percent, SD = 10.25 por-

cent; 65NC,, Mean 20.07 percent, SD = 9.26 percent; 33EC, Mean = 22.33

parcunt, SU r 12.05 percent. Wlhen the present study is compared to the

Kimmel and Kimme~l (1963) study, the absolute effects were smaller. How-il

ever, despite this procedural change, reliable differences wsere obtained

in the present study.

During the extinction the over-all difference between Contingent

and Ioncontingent groups only partially supports the hypothesis. The

mean of the Contingent group was 101 percent while that of the Noncon-

tingent group was 95 percent. However, this difference was not










statistically significant. Analysis of the data of the Noncontingent

groups during the first 8 min. of extinction showed that the 66NC and

33NJC groups gradually increased in response frequency until, at the end

of 8 min., their relative percent response was approximately equal to

their base rates as determined during the last 4 min. of adaptation.

These: tw~o groups showed extinction curves comparable to the extinction

curve obtained in the Kimme~l and KimmelR~ (1963) study for a 100EC group,

although the extinction was faster in the present two groups. This more

rapid extinction was probably due to the di fference in the procedure as-

sociated with the delivery of thle noncontingent reinforce~r~ers (cdescribed

above), and to th~e fact that the SEs In the 66NC3 and 33N~C groups of the

present study received considerably fewer Ilghtsl3 than the Ss of the

100NC group of the other study. The main reason the hypothesis under

considerationl was only partially supported was due to thle relative per-

formances of the 100C a~nd 100NC groups. Of considerable interest was

the fact that th~e 100NC group demonstrated what can best be termed "Im-

med iate ex i nct ion. This group shifted fromn having the lowest relative

response rate in the Notncontingent groups during the last 2 min. of ac-

quiiton to having the hrighest relative response rate during the first

2 min. of extinction. Furthermore, the 1O0lNC group was actually re-

sponding higher than1 the 1000J group during the first 2 min. of extinc-

tion. This finding was not in agreement with other the Fowler and

Kimmnel (1962) or KImmrel and Kimseil (1963) study. In the Fowler and


ITThe mean number of lights received by the iS in each Noncon-
tingent group was: group IDONC, 67.73; group 66NC~, 47.53; group 33NC.
22.40.










Kimmel study the difference between the tw~o groups (1000 and IDONC) was

found to be statistically significant in extinction during the first

minute. In the Klmmel and Kimmel study large and significant diffor-

ences occurred during thle entire 10-miin. period of extinction for the

two) groups (1000 and 100N1C) comprising that study. The contrast between

these and the present results cannot be easily dismissed and will be

dealt with below. It murst be concluded, nonetheless, that the noncon-

tingent reinforcement procedure used in the present study was not as of-

fective in maintaining the difference between Contingent and Noncontin-

gent groups during extinction as was the procedure of delivering the re-

inforcoment to N~oncontingent Ss only when an S was noL making a response.


Hypothqpis Z,

The second hypothesis tested in the present study was concerned

with the influence of the three schedules of partial reinforcement upon

the performance of the Contingent groups. The predict ion speci fi cal1ly

stated that the highest rate of responding would occur in the 33C group

both during acquisition and extinction. The next highest level of: re-

spondinn was expected to occur in group 66C followed by group 10DC. It

was expected that this ranked relationship would hold during acquisition

and ex i nct ion.

The results Indicated that during both acquisition and extinc-

tion the highest rate of responding was given by group 33C followed by

group 66C and then group 100C, thus confirming the hypothesis. It was

noted that the, maximum difference between the three Contingent groups

occurred during the first 2-min. blocks of acquisition and extinction.









The over-all differences amonng th~e Contingent groups were not statisti-

cally significant, but analysis of the first 2-min. block~ of acquisition

showed tbo 33C group to be significantly more responsive than the 65C

group (1 1.93, p < .05) and also greater than the 1000C group (t =

2.68, p
verged during acquisition. Thus, only during the early part of training

did the higher ratio of responses required for reinforcement result in

higher response rates.

In the Kimmel and K~inmel study (1963) it was noted thrat thle

strong conditioning effect apparent during the early part of training

decreased as the acquisition period continued. The present results par-

tially confirmed their findings. One possible explanation of this ef-

foct may be that postulated by Kimmel (1962), namely, that the "need'

for the light gets satiated, or it macy be that some kind of inhibition

grows as a function of the number of reinforcements. A high initial re-

sponse rate followed by decreased responding has also been found in

rats, when light served as the reinforcer (Barnes and baron, 1961); and,

in the present study, In group 33C. It will be of interest to determine

in future research if the obtained acquisition curves for conditioning

of the unelicited G5R are a function of the reinforcer used or thle re-

sponse systemi itself.

During extinction the ranked means of the three Contingent

groups weare in the hypothesized order, but statistical analysis failed

to support the apparent differences. Of some Interest was the effect of

partila reinforcement on the three groups' performaance during thle first

two 2-min. blocks. Only during this interval could the effect easily be










observed. By the 8th min. of extinction, the groups were responding al-

most identically although there was some~ divergence during the later

mi nutes. During the last 2-min. block of acquisition the differences

among these groups had almost disappeared, yet, w~ith the removal of the

reinforcer, large differences reappeared during the first few minutes of

extinction. The behavior of the groups seemed to indicate that maximal~

performance differences occur as a function of the schedule of rein-
forcement and of change in conditions. In another study on operant con-

ditioning of the GSR, Kimmel and Hill (1960) used several difforont

response-contingent reinforcers and obtained maximusm effects ofrter rein-
force~ment was terminated. However, this largo change in resp~onding at

the beginning of extinction was not found in the other two operant GSR

conditioning studies (Fowlier and Kimnmel, 1962; Klnmmel and Kimmsel, 1963).

No explanation of this discrepancy between studies is apparent.

The present results support the hypothesis that partial rein-

forcement of the unsolicited GSR results in higher rates of responding

during acquisition and greater resistance to extinction than does coni-

tinuous reinforcement although the effect was apparent only during the

initial periods of each condition.


Hypothesis ]
The behavior of the Noncontingent groups, in addition to their

control use, was of interest eeL se. Fowler and Kimsel (1962) found

that their N~oncontingent group emitted fewer and fewer responses during

the acquisition period and responded significantly less than the Con-

tingent group during the Ist mln. of extinction. Kimrmel and Kimmal










(1963) obtained approximately the sam~e result during acquisition for

their Noncontingent group. In this latter study, the extinction period

was prolonged fromt L to 10 min. and it wasJ found that the Noncontingent

group showed a gradual increase in the number of responses emitted dur-

Ing extinction.

This trend of increased responding continued until, at the end

of the 10-min. period, the frequency of responses was almost as great as

the Initial base level. The interpretation given by Klrmmel (1962) for

this effect was that the SsJ In the Noncontingent group must have learned

not to respond during th~e acquisition period (the noncontingent light

was delivered only when the S~s were not responding). During extinction

the S~s stopped not 191Ponding and, th~us, their response frequency curve

gJradually rose. In the present study three different groups of 2,s were

run under the moKdified noncontingent procedure as previously noted (e.g.,

p. 47). Each of the three groups received a different number of noncon-

tingent reinforceIMents which accompanied the reinforcements delivered to

the Contingent Sis (cf. p. 48).

The thi rd hypothesis was formulated under the two assumptions

that the delivery of the noncontingent reinforement would result in do-

creased responding and that the amount of decrease would bear a direct

relationship to the number of reinforcements delivered,as it was deter-

mined that the Nloncontingent 5s were actually being trained on a m~odi-

fied variable-interval schedule of reinforcement w~ith the 100NC group

receiving the shortest Interval. Speciflwally. the hypothesis stated

that the 100DNC group would give the least responses, followed by the

G6NC group, and then the 33NC group. Examirnation of the performance










curves during the last 8 in.I1 of the acquisition period confirms the

hypothesis. Comparison of: the~se performance curvles wilth~ the results of:

the K;imma and K~immel study (1963) indicates that the terminal response

frequencies were very similar for the 100NC groups. The major diffor-

ences between the two, studies with respect to these two) groups was5 that

the rate of decrease in responding in the Kimrmel and Kimmel study (1963)

was more rapid than in the present study. This result wJould be expected

due to the meihthodological differences of noncontingent reinforcome~nt.

During extinction it was expected that the relative response frequencies

of the three Noncontingent groups would gradually increase. The r3sults

confirmed this expectation for only the! 66N~C and 33NC1 groups. The re-

sponse frequency curves for these two groups may be compared to the ex-

tinction curve of the floncontingent group of the Kimrsel and Kimmrel (1963)

study. Extinction occurred more rapidly in the prcsont study, however.

Th~e results indicated that extinction occurred during the first 8 min.

of the extinction period while, in the earlier study, a 10-min. period

was required. The major difference with respect to these groups in the

two studies Is that, in the present study, although giving similar shaped

extinction curves, the NSoncontingent Ep responded at a higher level

throughout the acquisition and the extinction period.

The behavior of the 100PtC group of the present study was cer-

tainly different from wrhat was expected. it should be noted, howre~ver,

that the over-all relative response frequency of this group (for the

entire 16 min.) wars below all other groups, as was predicted. There re-

mains to be explained, however, the behavior of this group during the

first part of extinction. It wrouild seem that an *expectancy"' or










"disEcriminationl" hypothesis could be used to account for thle immediateat"

extinction apparently given. Howerver, this hypothesis would fail to

handle the results of the earlier studies. Examination of the data for

the individual Ss indicates that 11 of the 15 S~s in this group increased

in response frequency during the first 2 min. of extinction when comn-

pared with the last 2 min. of acquisition. This fact indicates that the

apparent increase was not an artifact resulting from only one to two S~s

who might have shownr large response shifts. Two possibhili ti es remai ning

are that the shift was a chance occurrence or that the increased respond-

ing was related to the high frequency of reinforcement coupled with thes

particular method of delivery of noncontingent reinforcement used in the

present study.


Hypothesj s &
Fowicr and KimsRel (1962) and Klrmmel and KmelrR1 (1963) had demon-1

strated that the removal of the reinforcing stimulus resulted in a de-

crease in the magnitude of thle differences between the Contingent and

IMoncontingent groups over time. These decreasing differences were in-

terpreted as extinction effects and were clearly shown in the Kimsel and

KImselI (1963) study. In that study a significant Contingency x Minutes

interaction was obtained in the extinction data. Their extinction period

was markted by a gradual decrease in responding in the 1000 group and a

gradual increase in responding in the 100NC group. The fourth hypothesis

of the present study was, Tn effect, a prediction based upon these pre-

vious findings. It follows from the hypothesis that the differences










obtained among the six experimental groups would diminish or disappear

during the extinction period.

The extinction period was extended to 16, min. in the present

study to allowr for the possibility of greater resistance to extinction

in the partial reinforcement groups. The results during extinction were

presented in Figs. 2, 3, and 4. Inspection of these figures definitely

indicates a disappearance of group differences, and thereby, confirma-

tion of the hypothesis. Closer inspection of the data during the ex-

tinction period reveals several interesting trends. First, the statis-

tical analysis did not show a significant Contingency effect or a sig-

nificant Contingency x Minutes interaction, This fact alone could mean

thiat extinction occurred immediately upon termination of the reinforce-

ment period. The performance curves presented In Flgs. 3 and 4 support

the contention of: immeediate extinction only for group 100NC. On the

other hand, groups 66C and 66~NC definitely showed a more gradual extinc-

tlon. For the first i, mln. group 66C wJas responding above group 66N~C.

During thH next 4 mln. group 66C dropped sharply and thecn rose again to

a.point above the 100 percent base line. This fluctuation continued for

the remaining 8 min. of extinction. The behavior of group 66NIC during

the first 8 min. of the extinction period nicely illustrated the expected

gradual extinction curves. Beginning w~ith the first 2-mln. block, this

group increased in response rate in an almost linear fashion until the

83th min. Comp~arison of groups 330 and ~33NC during the first 8 min. of

extinction also showed decreasing responding for the 330 group and in-

creasing responding for the 33NC group; both groups reaching almost the

same response level at the end of 8 min. These apparent differences









during the First 8 min. of ex:tinction er6e accompanied by a significant

Contingency x Ratio x Minutes interaction (F a 2.37, df = 6/252, p <1

.05), a significant Contingency x Ratio interaction in the linear compo-

nont of the trends across extinction (F 3.41, df a 2/84, p < .05),

and a significant Contingency x Minutes interaction (F 3.05, df Z

3/168, p
the data and the statistical analyses, it may be concluded that signif!-

cant e::tinction effects were obtained during the first 8 min. of ex.-

tinction in all groups except the 1000 and 100NC. These results agree

to some extent with the previous fin~dings.

The explanation for the equivrocal results obtained In groups

1000 and IDDNC and also for the fluccuations occurring in some of the

other groups during the last 8 min. of extinction is not clear. Betwoon

Ss variances indicated large intersubject differences. Examination of

the data of individual is Indic~ated that many of threm ;howJed wide rang=s

of responding.


Incompzlete adaptation gand r~esponding hgigg
the 100 percent base IMSe

The Fowler and Kimsecl study~ (1962) had allowed only 4 nAn. of

adaptation and the records indicated continuously falling curves for

both Contingent and Noncontingent groups during the reinforcomnent period.

The Kinmel and KimseI study (1963) allowed 10 mln. of adaptation prior

to the beginning of the reinforca~ement period. This was done to allow

the Ss to adapt more completely to the exper::mosntal chamber and for

their response rates to stabilize. Since the mean performance curve of

the Contingent group in that study was above the 100 percent base level









during acquisition, they concluded that their extended adaptation period

had adequately handled the problem. However, upon examination of the

Kcirmmel (1962) data for relative response frequencies of each 2, thie

present investigator found that during the first minute of acquisition

only nine Ss of the 15 E~s actually were, responding abova their owrn 100

percent base rates, although the mean of the relative response frequen-

cies during the Ist min. was 126 percent! By the 2nd min. of acquisi-

tion only seven ss of the 15 ESs were responding above the 100 percent

base level. Examination of the remaining min. of acquisition for each

S Indicated that only about one-holf the Ss were responding above the

100 percent base. The extention of the adaptation period, then, only

partially handled the problem of adaptation whecn viewed with regard to

the individual S~s.

One of the differences between the present study and the elrlicr

studies is that a new laboratory facility was used. The experimental

room was smaller, more sourdproof, and these characteristics may have

stressed the Ss. Although a full 10 min. of adaptation wras given, and,

although the group response curves showed an adaptation effect, it wa~s

possible that adaptation did not fully occur.


Hypaothesis 2

The 0 percent control group of the present study was included to

determine the frequency output of the unelicited GSR for a time period

comparable to the experimental conditions. The mean percent transformed

response frequencies of this group were prresented in Table 8. The data

in Table 8 indicate partial confirmation of the hypothesis, In that










after an initial decrease in responding (adaptation) the mean response

frequency tended to stabilize. However, the mean response frequency for

the group showed a slight increase during min. 11-20. This increased

responding after about 10 min. may be an indication of stress in the

situation (nalmo, 1959). If the Ss wrere stressed by the lack of stimu-

lation, then the effect of the light stimulus may have functioned as

more than a reinforcer in thle Contingent and Noncontingent groups. The

light may also have given the S something to attend to in his otherwise

sensorily impoverished conditions and may have resulted in decreased re-

spondinrg. This could account for the higher rate of responding of the

0 percent control group in comparison to the Contingent groups during

acquisition. It is more likely that the major variable in the present

experiment wIhich was responsible for the over-all reduced responding of

the Contingent groups was the reinforcemeent of responses that we;re so

small that either (a) the S could not discriminate these small responses

from no responses at all, or (b) they could not be conditioned by in-

strumentall means (assumain3 that "small" and "large" unoticited GSRs are

due to different processes). Both of these hypotheses are developed

below.


Decreased resp~ondins IEEE 115 small1 response5

It was previously noted that the sensitivity of the present re-

cording equipment was greater than that of: the equipment used in the

prior studies. In addition, the E~s in the previous studies (Fowler and

Kimrmel, 1962-; Kimm~el and Hilli, 1960; and Kimmel and Kimmrel, 196j3) moni-

tored the GSR record visually and delivered the reinforceme~nt w~hen the










particular response criterion had beeni met. In the present study the

S~s' GSR was passed through a differential amplifier which automatically

delivered thle reinforcer when a "criterion" response occurred. This de-

vice was much more sensitive to the very small unelicited GSRs (i.e.,

10-25 ohms) than was possible in the earlier studios. Assuming, thus,

that the procedure in the present study resulted in thle delivery of rein-

forcements for smaller responses, the following additional assumption

becomes necessary to understand the effects this had. Klimsal (1952) as-

sumed that one of the conditions necessary for te~l successful operant

conditioning of an autonomic response is proprioceptive feedback from

that response. In other words, thle response murst produce stimuli of

sufficient magnitude to be above the physiological noise level of the S.

The mechanism for feedback from autonomic activity is provided by affor-

ent (sonsory) nerves in the structures controlled by the autonomic sys-

tem (Young, 1961). These afferent impulses go to the reticular forma-

tion and bypotbalmic centers and influence cortical activity. This is

the presumed feedback; system. With this system, and given the large num-

ber of small responses reinforced, we~ hacve a possible explanation for

the suppressed response rates. The very small responses emitted by an I,

were detected and reinforced by the apparatus in addition to the large

responses. The difference between the size of a very small response and

a very large response could be a factor of about 200 In some cases. If

we assume that these very small responses provided no or Ilttle feedback

then it follows that reinforcemaents were being delivered when no "effec-

tive" responses were being emitted. In this case we could expect the

response rate to decrease, since this condition is very similar to










noncontingenrt reinforcemeint. The data show that delivery of a rein-

forcement when no response is occurring definitely results in decreased

responding.

An interesting hypothesis regarding the nature of these small

responses can be developed. Lacey and Lacey (1958) report that sponta-

neous activity of living tissue results in discharges, no matter how

carefully controlled the stimulus conditions may be. In addition, they

indicate that this activity is stable over time. They have postulated

that the GSR shows this type of: spontaneous activity and that thle locus

of its Initiation Is in the ". . phiysiochemical pcrocesses at the level

of localized cell aggregates." (p. 163, italics not: In original) If

these responses are stable over time, do not show an adaptation effect,

and are a function rAQliey of local peripheral activity, it is highly

likely that they are not influenced by reinforcement.


Nature of the larcle_ response

Thle "large" unelicited GSR is probably cortically initiated and

innervated via the autonomic nervous system. Fulton (1943) reported

that autonomic and sometic nerves are interdependent in that there is

extensive overlapping between the two0 in the cortex. It may be that

only these cortically initiated GSRs adapt over time and are operantly

conditionable. The problem wJould be to find a means by which these two

types of responses can be differentiated.

Conditioning the AAREEl and.}gggg resPonscs

In the present study a dichotomry was made between large and

small responses by counting as a small response all pen deflections less










than 1 percent of an S's base resistance, and as a large response all

those responses greater than or equal to 1 percent of the base resist-

ance. Figs. 5 and 6 presented these data for groups 10J0 and 100NC

(chosen for illustrative purposes) during acquisition and extinction.

The performance curves support the reasoning developed above. Condl-

tioning appeared to occur in the large responses, but not in the small.

The results are contaminated in the present case, however, in that gll

responses (large gan small) werre reinforced.


Adaptation of 2111.1 91d Jagr~qe repnses

Further support for the stability of the small responses over

time and the adaptation of the large responses would obtain if it could

be shown that the frequency of the former response remained steady dur-

ing the adaptation period while the larger responses showed an adapta-

tion effect. To examine this possibility the adaptation data for the

six groups of the ma~in experiment were divided into small and large re-

sponses as explained abovt,. These relative response frequency curves

are presented in Fig. 7 in 2-min. blocks for the 10-min. adaptation pe-

ried. The smagi responses in the left half of the figure do not show sn

adaptation effect, as is indicated by their fluctuating around their 1003

percent base level. The right half of the figure shows the curves For

the igrrag responses. Immedioatly obvious is the decreasing relative re-

sponse frequencies over time. This decrease is interpreted to mrean that

adoption occurred for the large type of response. A test of the signif-

icance of the li near component of the trend of these data indicated a

large significant negative slope (F s 20.80, df a 1/84, p <2 .001) for

the large responses.





















































































X~uanbaij


)O lua~iad


I I I C


yC
aO

crO


., 9,

0 0-

O UT

O a,



-aO




OO
r-C

On acC


CCE

C aC




nLC






OO-
o ~c~cU


o c


ZZZ


ZZZ

ZZZ
000
ZZZ
com
De
ea.


C, a









The hypothesis regarding two types of unelicited GSRs finds sup-

port in these data. This finding can account for the relatively de-

pressed response rates observed in the present study. In addition, the

possibility now~ exists for obtolning a very strong conditioning effect,

in future studies, as well as achieving greater control over the unctic-

ited GSR. This task would require differential reinforcement of re-

sponses distinguished by a magnitude criterion. It wrould be necessary

to reinforce several different magnitudes to obtain a better understand-

ing of the phenomenon.


Conclusion

The results of the present study indicated that the unelicited

GSR can be differentially influenced by Contingency and Schedulos of re-

Inforcement. The effect of Contingency of reinforceme~nt shows that the

unelicited GSR canr bi; made instrumental, and, thereby operate upon the

environment. Formerly, this fact w~as thought to be true for somatically

medi ated responses only. in addition to the Contingency variable, row~-

over, the influence of other variables can be demonstrated with the more

conventional instrumental responses. For ex:ample, magnitude, delay, and

type of reides-eementn all influence the acquisition and extinction of

instrumentally conditioned somatic responses. Another important varia-

ble determining performance is the schedule of reinforcement. Showing

that schedules of reinforcemecnt have a similar effect upon instrumentally

conditioned autonomic responses allows for the possibility of an under-

lying mechanism and set of law~s which apply to both somatic and auto-

nomic response systems. The present study has demonstrated that










fixed-ratio schedules of reinforcemnent do differentially influence the

unclicited GSR in a fashion somew~hat similar to their influence on so-

matically mediated responses.

With these findings, in conjunction w~ith the results of Fowlerr

and Kimm~el (1962) and Kimmeol and Kimmeal (196i3), the hypothesis regarding

the nonconditionability of autonomic responses by instrumental means is

further weakened.

Operant conditioning of autonomic responses now requires addi-

tional research to determine the particular nature of the response which

can be conditioned, the influence of other types of schedules of rein-

forcement upon the response, and the study of additional parameters of

which instrumental sometically mediated responses have been shown to be

a function.













CHAPTER V


SUMMIARY


The purpose of the present study was to determine the influence

of partial reinforcement upon acquisition and extinction of the unelic-

ited GSR. Ninety college students were divided into two major groups

(Contingent [C) and N~oncontingent [HC]). The Contingent group was di-

vided further into three subgroups each of which received a different

schedule of partial reinforcement on a fixed-ratio (100 percent, 66 2/3

percent, and 33 1/2 percent). All S~s wrer run In a light-proof, sound-

proof experimental chamber. Each Noncontingent S was paired to a Con-

tingent S of the same sex and was run in another experimental chamber at

the same time. The delivery of the reinforcer (a dim white light), to

both S~s, occurred whenover the Contingent S had emitted the requi red

number of GSRs.

All S~s received a 10-min. period of adaptation, the last 4 min.

of whrtich provided a base measure of resting response. After the 10-min.

adaptation period was completed, a 16-min. period of reinforcemrent \ws

given, during which the Contingent groups received the light following

the emission of a GSR (the number of emitted GS~s required to receive a

light was determined by the schedule of reinforcement). Finally, there

followed a 16-m1in. period of extinction during which no stimull were

given.










The dependent variable wa~s the number of responses emitted per

2-min. block expressed as a percentage of the number of responses emitted

during the base resting period (all frequencies wr~6e first transfor-med



During reinforcement the over-all rate of responding was greater

in the Contingent group than in the Noncontingent group (p
Among the Contingent groups the effect of partial reinforcement was max-

imal during the initial minutes of reinforcement. A significant con-

vergence of these Contingent groups was obtained over the 16 nin. of re-

Inforcomecnt (p
vergence of the Contingent and the Noncontingent groups. It appeared

that extinction was completed by the first 8 min. (the Contingent groups

decreased responding whlle the Nloncontingent groups increased). How-

ever, the differences between groups 1000 and 100NC fluctuated widely.

During extinction the effect of partial reinforcement upon the Contin-

gent groups was most obvious at the beginning of the extinction period.

The highest responding occurred in group 33C followed by group 66C and

then group 100C. These differences had almost vanished at the end of

8 min. of extinction, except that group 33C continued to respond above

all other groups throughout the 16 min. of extinction.

A 0 percent control group (15 Ss) was also run. These S~s merely

sat in the dark soundproof ed experimental chamber with no stimulation

for 42 min. This group showed an initial adaptation effect, followed by

fairly steady (but unexpectedly high level) responding during the last

32 min.










The low~ response rates of the Contingent groups, compared to the

0 percent control and to a similar group in the Kirmmel and Kirmmel (1963)

study, were interpreted as due to reinforcement of very small GSRs which

were not reinforced in the previous study. An hypothesis was developed

regarding the nature of the small and large GSRs and their condition~-

bility by instrumental methods. Finally, further research on this prob-

lem was proposed.

It was concluded that (a) the effect of partial reinforcement

was greatest at the beginning of acquisition and at the beginning of ex-

tinction; (b) the present study confirmed and agreed with the previous

studies (Fowlor and Kimmel, 1962; and Kimsetl and Klmmel, 1963) which

showed that Contingent reinforcement resulted in higher response rates

than Noncontinge~nt reinforcfEment; and (c) the hypothesis stating that

autonomic responses are not conditionable by instrumental means (Kimrble,

1961) was weakened further.












REFERENCE


Barnes, G. W. and Doaron, A. Stimulus complexity and sensory reinforce-
ment J.* coQBR Physipl. Psychol., 1961, 54. 466-469.

Barnes, G. W. and Kish, 8. B. On some properties of visual reinforce-
mont. ABrE. Psychologitt 1958, 13, 417. (Abst ract)

D~oren, J. J. Resistance to extinction as a function of the fixed ratio.
de sa..Eackt.,196, G, 304-303.

Duncan, D. B. Multiple range and multiple F tests. Biometricy,_ 1955.


Edwards, A. L. Exsperimental_ desgin ilL rgycholggical research. New
York: Rinehart, 1960.

Ferster, C. B. and Skinner, B. F. Scheduled 91 reinforcement. New
York: Appleton-Century-Crofts, 1957.

Fowler, R. L. and Kimmecl, H. D. Operant conditioning of the GSR. J_.
gag. Psychol., 1962, 63, 563-567.

Fox, S. S. Maintained sensory input and sensory deprivation in monkeys:
A behavioral and neuroph~armacologi cal study. Doctoral thesis,
Univ. of Mich., 1959. In Solomon, P. cas~gJ, (Eds.) SefluEU
deprivation, Cambridge, Mass: Harvard Univ. Press, 1961.

Fulton, J. F. PhysioloAY of the nervous@ 9Y2190. NewJ York: Oxford
Press, 194~3. (2nd Ed.)

Gaito, J. and Turner, E. D. Error terms in trend analysis* ?22091.
QGul., 1963, 50, 664-r44.

Grant, D. A. Analysis of variance tests in the analysis and comparison
of curves. Ptiychpl. Bull., 1956, sy, 141-154.

Grings, U~. UJ. Laboratory Instrumentation in.Psychology. Palo Alto:
national Press, 195r4.

HarwEood, C. W. 95503EdL qtonomi c conditloning Research i l progress.
Dellinghamn, Wash.: Western Washington College of Education,
1959. 31-38.

Harwood, C. UJ. Operant heart rate conditioning. Pyhl g. 92









Holland, J. G. and Sk~inner, B. F* IThe nalyssi gi behavior. NwJ York:
McGrow-Hllt, 1961.

Kimble, G. A. Hi lqar~d_ nd Mtarqul' Conditioning gag~ Learning~c. Clow
York: Appleton-Contury-Crofts, 1961.

Kimserl, Ellen B. Operant conditioning of the galvanic skin response: a
replication. M. A. Thesis, 1962, Univ. of Florida.

Kimmel1, Ellen. Stability of individual differences in activation. Pa-
per read at meetings of the Southeastern Psychological Associa-
tion, Gatlinburg, Tenn., 1964.

Kimmlel, Elion and Kimse~l, H. D. A replication of operant conditioning
of the GSR. i. gag. Psyc-hol., 1963, (2, 212-213.

Klmmel, U. D. and Baxter, R. Avoidance conditioning of the GSR* i.* 22.*
Psycholl., 1954 (In press).

Kirmmel, H. D. and Hill, Fran~ces A. Operant conditioning of th~e GSR.
P~sy~chl_. RED.., 1960, L., 555-562.

Kish, G. B. Learning when th.? onset of illumnination is used as rein-
forc ing st imrulus. 4. J gga CCI oh IYsIo Psycho;., 1955, _4 >
261-264.

Lacey, J. 1. and Lacey, Beatrice C. The relationship of resting auto-
nomic activity to motor imnpulsivitY* ReA, Publ. AAA.* 9 323&* 913s.*
1958, 3), 144-209.

Malmio, R. R. Act ivat ion: A neuropsychologi cal dimension. Psychoi.
Rev.. 1959, 6 ., 367-386.

Handler, G., Preven, D. W., and Kubiman, Clementina, KS. Effects of op-
erant reinforcemlent on the GSR* J.* yl52 Anal. Bob., 1962, 5,
317-321.

Moore, J. W. and Gormnezono, 1. Yoked comparisons of instrumental and
classical eyelid conditionin9* i.* 22.* Psy~Chol., 1961, gg,
552-559.

Howrer, 0. H. Preparatory set (expectancy) a determinant in motiva-
tion and learning. Psychol. Reg., 1938, 95, 62-91.

Eowrer, 0. H. and Jones, Helen M. Habit strength as a function of the
pattern of reinforcempent. J. gag Psychol., 1945. 15,l 293-311.









Razran, G. The observable unconscious and the inferable conscious in
current soviet psychophysiology: Interoceptive conditioning,
sem~entic conditioning, and the orienting reflex* 1#1. 2*
1961, 68 81-147.

Shearn, D). W. Operant conditioning of heart rate. Unpublished doctoral
dissertation, Univ. of Indiana, 1960.

Skinner, B. F. Science Man 0021~ behavior. N~ewJ York: The Macmillan
Co., 1953.

Skinner, B. F. The behavi C o_( gymanls~s:g an experi ment@l :nalys i 9.
New~ York;: A\ppleoton-Century-Crofts 1938.

Skinner, B. F. The experimental anralysis of behavior. Amecr. Sci,
1957. 95, 31r3-371.

Snedecor, G. WI. Statistical agfthodgi. A~mes, lowa: iowa State College
Press, 1956. (5-~t-h EdS.)

Young, P. T. Motivation gand gotion: A survey of the determinanty of
hLMrAD.Agg animpi _activity. New York: Wly 91

































APPEN~DICES
































APPENDIX A












INSTRUCT IONS


Your task in this experiment is merely to relax<, remain as mo-

tlonless as possible, and pay attention to wrhat happens. Since Cwe wi ll

be recording your galvanic skin response w~ith very sensitive equipment,

it is absolutely imperative that~B you avoid abrupt movements, deep or un-

even breathing, etc.; because any activity of this sort will produce er-

rors in our data. We would like for you to get as comfortable as possi-

ble now,, before wec begin recording, so that you will be able to remain

still when we~ start. If it becomes necessary to move (for example to

scratch your nose), please do so with your free hand wJith as little

abruptness as possible. Don't cross or uncross your legs or move about

in the chair--just relax.

You will find that: the more relaxed you can get, the easier it

will be to remain still. Yet, it is necessary that you remain alert in

the sense of paying attention to w~hat may happen. That is, keep your

ears, eyes, etc. open. This is also an absolute necessity. Do not give

in to any tendency to become drowsy; do not close your eyes.

Do you have any question?































APPENDIX 8




























I


E
O
O



O


C

O

Z O"
t. r

Z



O


Sro
e


t


ZI

Zy I I
at w




Jn Z

. 1



Il J




ZW I I U

II
II Z


I
































APPENDIX C



















































~ihe recording
of adaptation.


TABLE 10

, NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 100
PERCEllF COlNTINGEfff SUBJECTS


__


S
Number

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15


2

5

6

12

2

I

17

11

*

15

15

9

5

20

26

17


apparatus was not started until the third minute





77




TABLE 11
NJUMBER OF RESPONJSES PER 2-MIN. DLOCK FOR REINFORCEMEfff FOR 100
PERCECFf COffflN~GENT SUBJECTS


Nuraber 2 14 6 8 10 12 114 16
I 2 3 3 4 4 4 4 to
2 8 9 10 7 9 7 10 9
3 5 8 6 5 1 2 2 3
4 5 2 4 2 3 3 2 4
5 3 7 811 11 8 4 9
6 9 11 8 10 7 9 8 10
7 II 2 7 8 5 8 O 4
8 9 11 13 14 14, th 15 11
9 10 9 87 11 11 8 10
to 8 15 12 11 10 13 813
11 11 4 5 9 II 10 th 14)
12 4 4 5 3 2 4 5 2
13 7 15 17 15 It 13 12 14~
14t 17 17 12 13 13 10 8 14
15 8 10 11 13 13 13 12 11












TABLE 12

NUMBER OF RESPONSES PER 2-MI~N. LOCK FOR EXTINCTION FOR( 100
PERCElffCONTINGENT SUBJECTS


Nurrrer 2 4 6 8 10 12 14 16

I 2 47 7 47 10 4
2 12 13 15 13 12 8 88

3 1 0 2 51 2 0 0

4 5 3 4 5 5 4 3 7

5 11 10 10 16 11 10 11 II

6 11 12 11 11 14 11 13 11

7 6 7 5 8 8 8 6 7
8 14 10 10 13 12 12 4 10

9 9 8 6 5 8 7 9 11

10 16 12 13 13 11 11 8 9

11 11 12 15 14 16 10 11 12

12 2 5 47 46 15 4

13 23 16 17 17 13 11 15 18

14 16 19 17 16 12 11 13 10

15 13 16 11 15 14 15 19 14
































APPENDIX D



















Number 2 4 6 8 10
1 21 10 8 7 18

2 1 0 1 3 2

3 9 9 6 6 7

4 13 17 20 18 15

5 8 9 8 9 11

6 7 7 2 6 5

7 5 43 3 5
8 9 12 15 13 18

9 25 19 18 18 19

10 13 5 7 5 6
11 2 9 I 8 II

12 17 14 14 11 13

13 18 11 16 15 12

14 17 12 12 it it

15 17 10 12 9 15


TABLE 13

NUMBER OF RESPONSES PER 2-MIN. BLOCK
PERCElnR COlnWEINGENT


FOR ADAFTATION FOR 66
SUBJECTS


















Nu~a~er 2 4 6 8 10 12 14 16

I 13 7 1, 5 3 11 16 16

2 3 43 46 8 55

3 8 7 6 9477 5
4 !E 16 12 12 19 18 13 17

5 10 13 66 7 6 5 1!

6 3 5 3 7 5 6 9 8

7 3 5 5 4 7 5 4 7

8 9 to 13 12 II II 12 11

9 12 13 15 19 17 th! 17 15

10 7 48 8 85 95
11' 6 53 46 56 2

12 3 8 9 10 12 12 13 15

13 13 14 13 IS 13 15 IS 11

14 7 10 10 11 10 8 II G

15 9 12 12 10 10 12 8 S


TABLE 14

RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMIErli FOR 66
PERCE~fffONTINGENT SUBJECTS


NUtMBERi OF












TALE 15
NUM13ER OF RESPONSES PER 2_-MIN. BLOCK FOR EXTINCTION FOR 66
PERCEfKi CO)RINJGEPHT SUBJECTS


~u~b eg 2 Ir 6 8 10 12 Ilb 16
i 16 to 7 11 12 15 10

2 3 7 3 1 6 5 8

3 6 4 8 7 7 3 6 3
4 7l* 23 24 29 18 17 27 19

5 2 1; 5 5 53 82
6 II 10 5 10 12 7 14 7

7 3 5 i 9 5 9 6 6
8 th 12 10 11 14 16 5 12

9 23 22 17 21 21 22. 20 14
10 4 7 4 6 )6 66

1 6 5 3 45 5 L5
12 15 11 1418 14 17 18 13
13 15 17 12 15 14, 18 24 10
14 812 9 7 11 7 12 5

15 12 11 lb 15 6 13 10

































APPENDIX E


















I .. ICeT1 CPe~O~b-a -~-~-- -. --~ -~---~-~_.~.-~--r--~--- -


TABLE IG

RESPONSES P'ER 2-WMI. BLOCK FOR ADAF-TATCION FOR 33
PERCEIBT CONTINIGEtff~ SUBJECTS


NUMBER OF


2

3








8

9

10

11

12

13

14

15


13

3



10

12

IG,



13





10



16


















S
Nu ~ar 2 14 6 8 10 12 14 16


TALE 17

RESFONSES P'ER 2-MNlt. BLOCK FOR RZEl:!FORCtEMENT FOR 33
PERCLITK C~tSINGjEr? SUBJECTS


1IJMB-ER OF












TAD3LE- 18

NUM~Er. OF T:ESPON~SES FEf: 2-itlrl. DLOCK FOR. EXTI::CTION FOR 33
ErCEN~T CD~Tff~lNGEIFf SUDJECTS


flu oir 2 4 6 10 12 lb; 1El
1 10 1L: 10 10 lb; 7 It 9

2 6 2 ii 2 12 7

3 17 10 11 12 14 12 13 11
4 10 6; 8 5 1 8 C

5 13 10 11 10 13 11 12 11
G 12 G 8 11 6 15 10 1

7 9 7 6 3 7 5 77
C 3 96 3 10 12 G9
3 12 Ilb 10 15 10 13 7 3
10 6 6 10 11 lb1 16 12 5

11 13 11 8, 2 2 4 5 I4
12 7 12 13 11 11+ 10 9 11
13 14; tl 9 12 13 12 13 14
14c 15 12 9 Il, 17 10 12 14,
15 55 557937 3


































AiPPENDIX F


















Number 2 4 6 8 10

1 14 8 9 5 6

2 10 13 10 11 11

3 4 t 2 1 0

4 21 14 15 13 14

5 9 6 to 9 7

6 3 5 7 11 9

7 11 6 7 4 5;

8 a 20 24 23 25

9 9 7 20 4 7

10 20 17 20 17 21

11 12 13 11 15 7

12 4 3 2 3 3

13 10 11 14 12 12

14 20 22 24 23 26

15 14 12 12 14 th


TABLE 19

RESPONSES PER 2-MIN. BLOCK FOR ADAffATION FOR 100
PERCENT NOXNCO TI1NGEt T SUB~JECTS


NUMBER OF


:"The recording
of adaptat Ion.


apparatus was not started untiI the third minute


















Humbers 2 4 6810 12 14 16

1 9 5 7 3 4 2 4

2 5 6 4718 7 ,6

3 63 00 22 3

4 15 13 12 9 12 11 10 3

5 2 6 3 9) 7 9 5 5

6 4 5 74 22 1

7 4 22 5 13 3
0 IG 13 it 10 15 10 12 Ilb

3 5 4 2 4 2 0 O
10 14 12 18 12 th14 13 12

11 12 2 4 12 3 r; 1

12 3 5 30 5 11

;3 4 9 8o 10 7 4

14 IS r6 14 1 3 21 25 2 19

15 7 11 10 lb 9 10 9


TABLE 20

RESPO)NSES PER; 2.-MLI. BLOCK FOR( REINFORCEMENT FOR 100
PERC;13ilu 0NCi;NTINGENT SUCJECTS


liUMBER OF




















10

5

8

I

6

II

8

15

15





7

3



21

II


CIII=~---
111 --


~I _I~


_


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


TASLE 21

RESPONlSES PER 2-Mi ll. BLOCK


FOR EXT I CT ION FOR 100


II'JMI3ER OF


PER\CENTF NONlCONTINIGEtT7 SUJECTS


S
flambeSrs



2

3



5









10

II

12

13



15


11




I~



13


14





17

13


12

































APPENDIX G

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 EKT8Z0XZ4_O42YLQ INGEST_TIME 2011-07-15T20:31:00Z PACKAGE UF00097930_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

THE INFLUENCE OF FIXED-RATIO SCHEDULES OF REINFORCEMENT ON ACQUISITION AND RESISTANCE TO EXTINCTION OF THE INSTRUMENTALLY CONDITIONED GSR By WILLIAM A. GREENE A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA August, 1964

PAGE 2

UNIVERSITY OF FLORIDA 3 1262 08552 2356

PAGE 3

ACKNOWLEDGMENTS The author wishes to express his appreciation to the members of his supervisory committee for their aid and critical perusal; to the VA Hospital at Coral Gables, Florida for providing time and facilities for data analysis and writing; and especially to my committee Chairman, Professor H. 0, Klmmel. Aclcnowl edgment is also given to the U. S. Public Health Service for their financial support (MH (^060-2). II

PAGE 4

TABLE OF COMTEMTS ACKNOWLEDGMENTS ii LIST OF TABLES . v LIST OF FIGURES vl 1 1 Chapter I INTRODUaiON I The evidence 2 Recent studies 2 The effect of partial reinforcement 5 Procedural differences of the present study .... 9 Hypotheses 10 II METHOD ^2 Subjects '2 Instructions '2 Apparatus '2 Design '3 Procedure ^^ Interview '5 III RESULTS ^7 Acquisition 21 Extinction 29 percent control group 33 Small and large emitted GSRs 37 IV DISCUSSION. M* Hypothesis 1 fS Hypothesis 2 ^ Hypothesis 3 50 Hypothesis k 53 ill

PAGE 5

TABLE OF COMTEMTS (Continued) Page IV incomplete adaptation and responding below the 100 percent base rate 55 Hypothesis 5 56 Oezrsaseri reiponJi.ig a.id the s.nall responses S7 Nature or th^ large response 59 Conditioning the small and large responses. ... 59 Adapiaiiion of smaW and larg^ reiponje^ 60 Conclusion 62 V SUMMARY 6
PAGE 6

LIST OF TABLES Table Page I MEANS ANO STANDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FRECIUENCIES OF EACH CONTINGENCY ANURATIOOF REINFORCEMENT GROUP DURING ADAPTATION (IN 2-MIN. BLOCKS) 20 2 MEANS AND STANDARD DEVIATIONS OF V X HI , WHERE X 15 THE AVERAGE OF THE LAST TWO 2-MIN. BLOCKS OF ADAPTATION, FOR EACH SUBJECT 22 3 MEANS ANO STANDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY AND RATIO OF REINFORCEMENT GROUP DURING ACO.UISITION (IN 2-MIN. BLOCKS) 26 k SUMMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCENT RESPONSE FREQUENCIES DURING THE !6 MIN. OF ACQUISITION (IN 2-MIN. BLOCKS) 27 5 MEANS ANO STANE>ARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY AND RATIO OF REINFORCEMENT GROUP DURING EXTINCTION (IN 2-MIN. BLOCKS) 30 6 SUMmRY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCENT RESPONSE FREQUENCIES DURING THE FIRST 8 MIN. OF EXTINCTION (IN 2-MIN. BLOCKS) 32 7 SUMMARY CF ANALYSIS OF VARIANCE OF LINEAR TREND OF TRANSFORMED PERCENT RESPONSE FREQUENCIES DURING THE FIRST 8 MIN. OF EXTINCTION (IN 2-MIN. BLOCKS) 3^ 8 MEANS AND STANDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FREQUENCIES FOR THE PERCENT CONTROL GROUP DURING THE ^2-MIN. SESSION (IN 2-MIN. BLOCKS) 36 9 SUWWRY OF ANALYSIS OF VARIANCE ON LARGE RESPONSES FOR THE LAST 8 MIN. OF ACQUISITION AND THE ENTIRE 16 MIN. OF EXTINCTION FOR THE CONTINGENT AND NONCONTINnEMT GROUPS RECEIVING 100 PERCENT REINFORCEMENT (IN 2-MIN. BLOCKS) *»2

PAGE 7

Table LIST OF TABLES ( Continued) Page 10 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 100 PERCENT CONTINGENT SUBJECTS 7b tl NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMENT FOR 100 PERCENT CONTINGENT SUBJECTS 77 12 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 100 PERCENT CONTINGENT SUBJECTS 78 I J NUMBER OF RESPONSES PER 2-MIM. BLOCK FOR ADAITATION FOR 6b PERCENT CONTINGEWT SUDJCCTS 80 |l» NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMENT FOR 66 PERCENT CONTINGENT SUBJECTS 81 15 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 66 PERCENT CONTINGENT SUBJECTS 82 16 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 33 PERCENT CONTINGENT SUBJECTS S^* 17 NUMBER OF RESPONSES PER 2-MlN. BLOCK FOR REINFORCEMEWT FOR 3? PERCENT CONTINGENT SUBJECTS 85 18 NUMBER OF RESPONSES PER 2-MIM. BLOCK FOR EXTINCTION FOR 32 f*VC^:^ CONTINGENT SUBJECTS 86 19 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 100 PERCEIfT NONCONTINGENT SUBJECTS 88 20 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMEWT FOR 100 PERCENT NONCONTINGENT SUBJECTS 89 21 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 100 PERCENT MONCOIfTINGENT SUBJECTS 90 22 NUMBER OF RESPONSES FOR 2-MIN. BLOCK FOR ADAPTATION FOR 66 PERCEfff NONCONTIMGEJIT SUBJECTS 92 23 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMENT FOR 66 PERCENT NONCONTINGENT SUBJECTS 93 24 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 66 PERCENT NONCONTINGENT SUBJECTS 9^ Vi

PAGE 8

LIST OF TABLES (Continued) Table Page 25 NUMBER OF RESPONSES PER 2-HIN. BLOCK FOR AOAPTATiON FOR 33 PERCEWr NOMCOtfTINGEMT SUBJECTS 96 26 NUHBER OF RESPONSES PER 2-MIN. BLOCK FOR REiNFORCEMErfT FOR 33 PERCENT NONCOi^lNGEMT SUBJECTS 97 27 NUMBER OF RESPONSES TER 2-HIN. BLOCK FOR EXTINCTION FOR 33 PERCENT NONCONTINGEMT SUBJECTS 9B 28 NUMBER OF RESPONSES PER Z-MIN. BLOCK FOR 1-10 MIN. FOR PERCENT CONTROL SUBJECTS 100 29 NUHBER OF RESPONSES PER 2-HIN, BLOCK FOR 11-26 HIN. FOR PERCENT CONTROL SUBJECTS 101 30 NUHBER OF RESPONSES PER 2-HIN. BLOCK FOR 27-^2 HIN. FOR PERCENT CONTROL SUBJECTS 102 31 HEANS AND STAN0A.10 DEVIATIONS OF THE NUHBER OF RESPONSES FOR EACH (iROUP DURING ADAPTATION (IN 2-MIN. BLOCKS) 104 32 MEANS ''m STANDARD DEVIATIONS OF THE NtiHBER OF RESPONSES FOR EACH GROUP DURING ACQUISITION (IN 2-HIN. BLOCKS) 105 33 HEANS AND STANDARD DEVIATIONS OF THE NUHBER OF RESPONSES FOR EACH GROUP DURING EXTINCTION (IN 2-HIN. BLOCKS) '06 vii

PAGE 9

LIST OF FIGURES Figure Pag« 1 Percentage of transformed response frequencies during the 10-min. adaptation period in 2-min. blocks for the Contingent and Noncont i ngent groups • 19 2 Percentage of transformed response frequencies during the 16 min. of reinforcement and the 16 min. of extinction for the confined Contingent and combined Noncont i ngent groups 2| 3 Percentage of transformed response frequencies during the 16 min. of reinforcement and the 16 min. of extinction for the Contingent groups Zh k Percentage of transformed response frequencies during the 16 min, ot reinforcement and 16 min. of extinction for the Noncont i ngent groups 85 5 Percentage of transformed response frequencies during the 16 min. of acquisition and the 16 rain, of extinction in 2-min. blocits for the large responses In the 100 percent groups 59 6 Percentage of transformed response frequencies during the 16 min. of reinforcement and the 16 min. of extinction in 2-min. bloclcs for the small responses of the 100 percent groups M> 7 Percentage of transformed response frequencies during the lO-min. adaptation period In 2-inin. b1ocl
PAGE 10

CHAPTER I INTRODUCTION The purpose of this study was to determine the Influence of partial reinforcement upon the rate of emission of the unellcited GSR, a response mediated by the autonomic nervous system. Only recently has any evidence been reported to support the contention that autonomically mediated responses may be influenced by the presentation of reinforcement following their emission. The present study, thus, was intended to provide further empirical evidence on operant conditioning of the GSR and, hopefully, to relate operant conditioning of autonomic responses more fully to conventional operant conditioning. Kimble (I96I) distinguishes between two types of conditioning — classical and instrumental. He supports the distinction by indicating the possible differences between the two types of conditioning. Among the differences marshalled Is the autonomic-classical and somaticInstrumental hypothesis. Kimble concluded, from the evidence he presents, that this distinction cannot be completely maintained on empirical grounds as some nonautonomic responses have been classically conditioned. His strongest statement on the matter, however. Indicates that even though some sl
PAGE 11

the conclusion that such responses can be modified by classical, but not instrumental training methods." (p. 100) Furthermore, the efficacy of this technique, with autonomic responses, was questioned by Skinner, for he wrote, Glands and smooth muscles do not naturally produce thte kinds of consequences involved in operant reinforcement, and when we arrange such consequences experimentally, operant conditioning does not take place. We may reinforce a man with food whenever he 'turns red,' but we cannot in this way condition him to blush. . . ." (1953, p. HU) Skinner's belief, however, was based upon meager factual evidence. The evidence Mowrer (1938) attempted to make the GSR instrumental in a shock avoidance situation. His attempt ended in failure, in another experiment Skinner and Oelabarre (Skinner, 1938) tried to condition vasoconstriction instrumental 1y. In this experiment a positive reinforcement was made contingent upon the response, but the authors reported no conclusive results. The Pavlovians (Kimble, 1961) simply have indicated that glandular responses cannot be conditioned instrumental ly, without ci ting any data. Recent studies More recently. Handler et al . (1962) attempted to influence the rate of occurrence of unelicited GSRs by rewarding the response with money, the earning of which was signalled by a light. They reported increased emission of GSRs for some of their Ss. The results reported by Shearn (I960) are also somewhat equivocal. He attempted to condition

PAGE 12

heart rate acceleration by making the criterion response Instrumental in avoiding a shock. His results Indicated that while the experimental group differed significantly from the control group in number of accelerations, there was no significant reduction in the number of shocks received. Harv«3od (1959, 1962) tried to condition heart rate deceleration by rewarding the criterion response. Although the results reported are not conclusive his research is continuing with improved instrumentation. in contrast to the negative findings of Skinner and Oelabarre (Skinner, 1938) reported above, Razran (1961) has reported a Russian study which showed instrumental conditioning of blood volume changes. in this study five human S,s were stimulated with painful electrical shock. This type of stimulation usually produces vasoconstriction, however, if vasodilation occurred the stimulus was terminated. When the §,s were allowed to observe their own pHctoplethysmographs (in real time) they learned to escape the siiock. In another study using ar avoidance conditioning paradigm, KImmel and Baxter (1964) found that an avoidance group gave superior perforr>i
PAGE 13

(1962) attempted to determine whether the unali cited GSR could be conditioned Instrumentally. These investigators found that the experimental groups receiving 16 min, of response-contingent training differed from the control group during acquisition and the early minutes of extinction. In addition, the relative response frequency trends during extinction between the two groups were opposite In direction, although this effect was not statistically significant. These results were Interpreted as providing further support for the conclusion that the unelicited GSR can be conditioned operantly. The essential aspects of the Fowler and Kiramel (1962) study were replicated by KImmel and KImmel (1963). In this study S$ were divided Into two groups, Contingent and Noncontlngent. Contingent S,s received a dim white Ijght fol 'owing each emitted GSR while Ss In the Noncontlngent group received the light only when no GSRs were being emitted. The dim white II gilt which served as the relnforcer was the sole source of visual stimulation. Each S, in the Noncontlngent group was "yoked" to an S, In the Contingent group to receive the same nun^er of visual stimuli per minute during the 16-mln. training period. Following training, all S,s received 10 mIn. of extinction, during which time no lights were presented. Before the actual training period was begun, S.s were allowed to adapt to the experimental room for 10 mIn. This allowed the S.s to achieve a more stable response rate than was obtained In the earlier studies using sliorter adaptation periods. Due to this longer adaptation period, considerable reduction In the amplitude of the emitted GSR was noted. Therefore, the criterion for counting responses and for presenting or not presenting the relnforcer was changed to Include all responses

PAGE 14

detectable by the naked eye. The grace period' following delivery of a reinforcer was changed from 5 sec. to 3 sec. to allow more responses to be counted. Kiinnel and KIntnel's results indicated that the Contingent groups responded more frequently during acquisition than during the last 5 min. of adaptation (base level) while the Noncont i ngent group responded less frequently during the reinforcement period than they had previously. The large obtained differences in response frequencies between the two groups were statistically significant. During extinction the trends of the two groups were converging toward the level which had obtained during adaptation, thus indicating opposite trends. Both the Contingency and the Contingency X Hinutes interaction effects were statistically significant during extinction* The effect of partial rei nforcement The previous studies (Fowler and Kiimel, 1962; Kimmel and Hill, I960; KImtnel and Kimmel, 1963; and Mandler et al .. 1962), using the operant conditioning technique with the GSR, have reinforced every response. Other criteria for the delivery of reinforcement may be followed. For example, every other response may be reinforced, every tenth response may be reinforced, or almost any number of responses may be required of the organism before reinforcement is delivered. In addition to the procedure of requiring a set number of responses to occur before reinforcement is delivered the experimental arrangement may be designed so that the organism obtains a reinforcement after a certain time 'a response e . li cited by the stimulus light was neither counted nor reinforced. Thus, any response occurring from 1 to ^ sec. following the light was in tiie grace period.

PAGE 15

Interval has passed, provided the criterion response occurs after the specified time period. The Influence of such schedules of reinforcement has been studied In detail (Ferster and Skinner, 1957). A f«w of the basic schedules of reinforcement are: (a) Fixedratio schedules, in which the reinforcement Is delivered after the specified number of responses have occurred, (b) Variable-ratio schedules. In which the number of responses required for reinforcement varies from reinforcement to reinforcement auout some mean value, (c) Fixed-Interval schedules, in which reinforcement Is delivered following the first response occurring at the end of some fixed-interval, (d) Variable-Interval schedules. In which the period of time during which no reinforcement is available varies between reinforcements around some mean value. Each of these schedules results In performance curves peculiar to the schedule. In the typical free operant Instrumental conditioning paradigm (e.g., bar pressing) high rates of responding can be maintained when the number of responses required to obtain a reinforcement is very large. Skinner (1957) has reported high rates of responding even when the number of responses required for one reinforcement was as high as 900. Such schedules of reinforcement (in this case a fixed-ratio schedule) have been found to result In high stable response rates during training with much greater resistance to extinction than can be obtained with continuous rel nf orcvnent . For example, Mowrer and Jones OSkS) found that rats trained to press a bar for food reward differed In resistance to extinction as a function of the schedule of reinforcement during acquisition. The greater the ratio of nonrelnforced responses to reinforced responses, the greater the resistance to extinction.

PAGE 16

Recdntly, Boren (1961) trained rats to bar press on several ftxed-ratio schedules and found increasing response rates with the increasing fixed-ratios used. More simpiy, the great»=-r the number of responses required for a reinforcement, the higher the rate of responding obtained. In addition, it was found that the number of responses emitted during extinction was approximately a linear increasing function of tha fixed-ratio used during acquisition. Both the Fowler and Kinmel (1962) and the Kimmel and Kinroel (1963) studies reported decreasing response rates during the reinforcement period for the rtoncontingent Ss. Kimmel (1962) hypothesized that the §,s who were receiving the light when they were not responding ({toncontingent §.s) may have ween learning not to respond. That is, those Ss were being reinforced for not making GSRs. Analysis of the stimulus conditions during the reinforcement period (by the present writer) of the Noncont i ngont group indicated that the S,s were being reinforced on a modified variable-interval schedule (modified because the mean interval changed from minute to minute depending upon the number of reinforcements delivered). Assuming that the hypothesized process of notresponding (not emitting GSRs) involves active inhibition, then reinforcing the inhibition would tend to Increase the probability that Inhibition would occur as reflected by decreased resoonding. In fhc prirvlous studies the nun^er of lights delivered to a Noncontingent S. for any given minute was determined by the number of lights received by the matched Contingent S. for the particular minute in question. The E then delivered the light to the Noncont i ngent S unsystematlcally (varied the interval), but only virtien the S was not responding.

PAGE 17

8 Assuming that this analysis is correct and that the ftoncont I ngent groups were actually being reinforced for Inhibiting the GSR on a modified variable-interval schedule, then predictions can be made regarding the behavior of the Noncont I ngent Ss In the present study. Each Noncont I ngent S. in the present study wi 1 1 be yoked lo a Contingent S. The number and pattern of lights received by the Noncontingant Ss will be irwnedlately determined by the Contingent Ss. Since there are three fixed-ratio schedules of reinforcement for the Contingent groups (100 percent, 66 2/3 percent, and 33 J/3 percent) the actual miHri9er of lights received by a Noncont I ngent S vi»l 1 1 vary according to the fixed-ratio schedule of the yoked-Contlngent S. Thus, group lOONC receives the most lights, followed by group 66NC, and, then group 33MC. The number of lights received will determine the modified variableinterval schedule a particular Noncont i ngent S receives; the greater the number of lights the shorter the variable-interval. Holland and Skinner (1961) report that a short variable-interval schedule resulted in higher response rates than a long variable-Interval schedule. For exanple, when the variable-interval during which no reinforcement was available Is long the organism tended to emit fewer responses during the Interval than when the Interval Is shorter. Assuming that group lOONC receives the shortest variable-Interval schedule then this group should be reinforced more for Inhibiting GSRs, and, thus respond the least (see Hypothesis 3).

PAGE 18

Procedural differencos of. the prescsnt study The present study employed procedures similar to those described by Kiinnel (1962) with a few major changes. These differences were as follows: 1. All of the previous recent GSR studies were done utilizing a partially soundproofed room and recording and amplifying equipment dif ferent from that used in the present study. Although the over-all difference was quantitatively unknown It should be noted that the present recording and amplifying equipment was more sensitive, and was able to monitor and record f rom Ss with higher base resistance levels. In addition a double-walled Industrial Acoustic Chamber (lAC) was used In the present study, in place of the partial soundproofing of the previous laboratory. 2. The responses to be reinforced were automatically determined by a differential amplifier. The amplifier was adjusted by the E during the adaptation period so that It would reinforce all noticeable deflections of the GSR recorder pen (all responses greater than or equal to a 10 ohm urop activated the circuit). The higher reliability of the reinforcement apparatus coupled with the greater sensitivity of the equipment resulted In the reinforcement of a larger nund>er of unellclted GSRs than In any previous study in the series. 3. A "true" yoked control procedure was utilized. In the previous studies each S was run separately, and the Moncont I ngent Ss received a reinforcement only when no GSRs were occurring. In the present study tvKt Ss were run simultaneously. Under these conditions the yokedcontrol S received the same number and pattern of reinforcements ^ iil£.

PAGE 19

10 same time as the Contingent Ss. This important modification resulted in the delivery of a reinforcement to the Noncontlngent S regardless of his ongoing GSR activity. Thus, Ss In the Noncontlngent groups might receive the relnforcer when they were roaklny a GSR or not, since delivery or nondelivery was completely determined by the behavior of th*^. Ss In the Contingent group. This yoking procedure is essentially the same as that used by Hoore and Gormezano (1961). k. The period of extinction was increased to 16 min. Since It was expected that the resistance to extinction Mould be Increased after partial reinforcement training, this increase was Introduced to enhance the likelihood of detecting the effect. 5. The previous studies all utilized a 100 percent schedule of reinforcement. The present study used several different ratios of partial and continuous reinforcement to determine their effects. 6. Since the effects of prolonged absence of stimulation on the unellclted GSR were unknovm for the present laboratory a special percent group was run. The £s In this group merely sat In the experimental chamber but received no stimulation. tiypQ^hese? The major hypotheses concerned the effects of Contingency of reinforcement (Contingent vs. ^k)ncontlngent) and Schedules of reinforcement (100 percent, 66 2/3 percent, 33 1/3 percent, and percent) upon the rate of emission of GSRs during the experimental sessions. The specific hypotheses were as follows:

PAGE 20

11 1, The rate of occurrence of emitted GSRs will vary as a function of Contingency of reinforcement. The Contingent groups will emit a greater number of responses during acquisition and extinction. 2, Within the Contingent groups, the rate of unellcited GSRs win vary ^'ivarsely with the percentage of reinforcement, both during acquisition and extinction. Thus, the highest rate of responding will be obtained in the 33 1/3 percent group, followed by the 66 2/3 percent group and 100 percent group, respectively. In addition, all of the Contingent groups will respond above the 100 percent base level during acquisition and extinction. 3, Within the Noncont I ngent groups, the rate of emission of unellcited GSRs will vary Inversely with the schedule of reinforcement received by the yoked-Contlngent Ss. Thus, the ordering of groups will be the same as In Hypothesis 2, but by Hypothesis 1, the over-all rates of the NoncontI ngent groups will be lower than the Contingent groups. k» During the extinction period the rate of GSR emission will tend to return to the 100 per cent level (base level). Thus, the Contingent groups will decrease responding, and the Noncontingent groups will increase responding during extinction. 5. The rate of unellcited GSRs In the percent group will tend to stabilize and remain constant for the experimental session.

PAGE 21

CHAPTER II METHOD Subjects One hundred and five undergraduate students at the University of Florida, 63 men and k2 women, volunteered as Ss. Each S. who served in the experiment was paid one dollar for participation. Sjs of the same sex were run In pairs and the proportion of males to females In each treatment was equal. Instructions' ^ The Ss were Instructed to relax, to remain as motionless as possible, but to remain awake and alert. An Interview was conducted at the end of the experimental session to determine whether these instructions had been followed. If it was determined that an S. had not followed instructions he/she and the yoked partner were rejected. Ss would be rejected due to experimenter error, equipment malfunction, or failure to follow instructions (sleeping, moving, closing the eyes, etc.). Apparatus A small, dim patch of white light (0.7 ft. candles), located in the direct line of sight approximately 30 in. in front of S., served as ^his support was provided by grant fW 06060-2. 3See Appendix A for complete instructions. 12

PAGE 22

13 the reinforcing stimulus. The experimental rooms were double-walled soundproofed lACs equipped with a padded chair, a small mini box housing the reinforcing light, and a spealcer which served as an intercom and for communication with the S,. The GSR was piclced up as a OC resistance change from the palm and bacl< of the S,'s left hand by 3/^ in. zinc electrodes. The electrodjf were covered with small gauze patches saturated with zinc sulfate solution and were housed in lucite cups filled with saline electrode paste (Grings, 195^). The constant current through the S, was 20 microamps. The control panel in an adjoining room contained two Biophysical GSR an^liflers. and Texas Instruments Co. Recti -Riters. The recorders were equipped with signal -magnet pens for recording the occurrence of the reinforcing stimulus. The circuit controlling the reinforcing stimulus contained a differential amplifier and a switching system for automatic delivery and programming of the stimulus, in addition a hold circuit prevented the delivery of the reinforcing stimulus more than one time in any k sec, period (the grace period). A blocic dies agram of the reinforcing system can be found in Appendix B. Design The basic experimental plan consisted of a 2 x 3 factorial design with 15 §,s assigned at random to each cell, with the restriction that each pair of Ss be of the same sex and the male/female ratio be constant between cells. Ss were run in pairs, one member being assigned to the Cont ' nqent group and one member to the Noncont i ngent group. The Independent variables were percentage of reinforcement (100 percent, 66 2/3 percent, and 33 1/2 percent) and Contingency of reinforcement

PAGE 23

Ill (Contingent and Noncontingent). The reinforcing circuit was automatically activated by any response (^10 ohms) of the Contingent S, and delivered to both members of a pair at the same time. In addition, another group of 15 S,s was run under conditions of no stimulation (O percent group), to determine the course of GSR emission without stimulation. Procedure The two Ss of a pair entered the laboratory via different doors and were seated in separate experimental chambers. The palm and bacl< of each S,'s left hand were then cleaned by acetone and the GSR electrodes were attached and held In position by a rubber band. The overhead light was extinguished and the experimental chamber doors were closed placing each S in total darioiess. The experimenter (E) then read the instructions over the Intercom. The S,'s questions were answered by rereading the appropriate part of the Instructions. During the entire experimental period the £s were monitored auditorily to detect any noises due to movements (shuffling, changing position in the chair, etc.), sneezes, coughs, etc. which might produce a GSR. All GSRs so Identified were not scored as responses. After a brief Interval, during which the instruments were checl(ed and adjusted a lO-mln. adaptation period was begun. At the conclusion of the lO-min. adaptation period, the 16-min. reinforcement period was begun. Responses made by Contingent S,s were passed through the differential amplifier which closed the circuit delivering reinforcements; the reinforcements were delivered to both S,s of a pair simultaneously. The

PAGE 24

15 programming apparatus delivered reinforcements on a flxed-ratfo schedule (i.e., 1:1, 2:3, and 1:3)* The circuit was designed to deliver a reinforcement at the point at which the particular GSR had reached Its maximum. The duration of the reinforcement was l/IO sec. Following the reinforcement period the S,s were given a I6<^ln. period of extinction during which no stimulation was given. At the conclusion of the extinction period, the electrodes were removed and the Interview was conducted separately for each S.. Interview Each S was asked, (a) "Old you have the In^ression at any tima during the experiment that the occurrence of the light depended upon something you might have done?"^ If the answer to question (a) was negative, the S, was told to make the hypothetical assumption that his behavior controlled the light, and then he was asked, (b) "Can you now think of any possible connection between your behavior and the presence or absence of the light?" Only the data of Ss whose answers indicated that they had followed the Instructions (particularly regarding movement) completely were used. Five Ss were rejected for failure to follow the Instructions regarding movement. ^ The S.s were then thanked, given a ^hree Ss, one from each Contingent group, Indicated that they thought the light came on when they expected it. txami nation of their records Indicated that these S.s were near the mean of their respective groups. 5Three ''ontingent Ss In the 100 percent group were rejected for moving. T\«3 Contingent S.s^in the 33 1/3 percent group were rejected, one for moving, the other .or deep and irregular breathing. The rejection of these Contingent Ss necessitated the rejection of their yoked controls; therefore, a total of 10 Ss were rejected.

PAGE 25

16 chtt exchangeable for one dollar, and Instructed not to discuss the experiment with anyone. The S,s of a pair left the laboratory by the doors through which they had entered.

PAGE 26

CHAPTER ill RESULTS The primary performance measure was the number of responses emitted per minute during adaptation, acquisition, and extinction* A response was defined as any deflection of the GSR recorder pen which in« dicated at least a 10 ohm drop In resistance. The reliability of this measure was determined separately for the Contingent and ^k}ncont i ngent groups. This was accomplished by having two judges (one of whom was the E) read an unsystematical ly selected minute from the record of each of the S,s in the main experiment. The second judge read the records "blindly," In that she did not icnow to virtiich group a record belonged. The correlation (Pearson) between the readings of the two judges was calculated for each group and both indicated high reliability (Contingent, r » .96; Noncont I ngent , r » .97). Both of these correlation coefficients were significantly different from zero (p
PAGE 27

18 transformed to "y/x •+• I to overcome the skewness of their frequency distributions (Snedecor, 1956). The transformed measures were then expressed as percentages of the transform of the average of the last two 2-min. bloclcs of the 10-min. adaptation period. These transformations v^re done separately for each 2-min. block for each S,. The group performance curves during adaptation were expected to indicate declining relative response frequencies as adaptation progressed, as well as no differences in average frequency at the end of adaptation. The relative response frequency curves, shown In Fig. 1, Indicate that the general trend of the curves Is downward, as expected. Tab/e 1 presents the means and standard deviations for the adaptation period for the transformed percent response frequencies In 2-mln. blocks. An analysis of variance of these data Indicated that only the Minutes effect was significant (F 7.02. df s V33^, P <: .001). 7 This significant effect, In conjunction with the shape of the curves In Fig. 1, indicated a significant adaptation trend. The response frequencies during the last 4 min. of adaptation were examined for each of the seven groups. It was necessary that these pretralning frequencies be comparable between groups so that any differences found after training could be attributed to the experimental treatment and not to sampling errors. The average response frequencies (and SDs) during the last two 2-mln. blocks of adaptation transformed to ^For one pair of S,s the recording apparatus was not started until the beginning of the third minute of adaptation. Thus, 2 missing data were replaced and 2 df were subtracted from the error term.

PAGE 28

19 z z z 1— tz z

PAGE 29

20 TABLE 1 MEANS AND STAMOARO DEVIATIONS OF TRANSFORMED PERCEfflT RESPONSE FREQUENCIES OF EACH CONTINGENCY AND RATIO OF REINFORCEMENT GROUP DURING ADAPTATION (IN 2'MIN. QLo:/.:^) Group Adaptation Cent i ngent X SO X SO X 100 66 33 SD 105

PAGE 30

21 ^ i 4I for alt groups are shown In Table 2, Analysis of variance Indicated no significant differences anong the groups (F < I, df s 6/98). To compare the effect of the relnforcer on the different Contingent and Noncont I ngent groups during acquisition and extinction, the performance curves for the different groups were plotted in 2-mlo. blocks for the 16 min. of acquisition and extinction. Figs. 2, 3* and k present these curves. Acquisitio n The first section of Fig. 2 shows the combined performance curve of the three Contingent groups and of the combined yoked Noncont I ngent controls during the 16 min. of acquisition. Inspection of these curves reveals that the over-all mean of the Contingent (95.83 percent) was higher than that of the hk>ncont I ngent groups (88.33 percent). Further examination of the acquisition curves indicates that the over-all difference between the Contingent and Noncont i ngent groups became larger during reinforcement. To evaluate the statistical significance of these differences the data obtained during acquisition were subjected to an analysis of variance. Table 3 presents the means and standard deviations for the 90 S.S of these groups for the 16 min. of reinforcement, and Table k summarizes the analysis of variance. The analysis indicated that the difference between Contingent and Noncont i ngent groups were statistically significant (F 6.23, df » ]/iik, p< .025). No other differences were found to be significant.

PAGE 31

22 TABLE 2 MEANS AND STANDARD DEVIATIONS OF VThFT, WHERE X IS THE AVERAGE OF THE LAST TWO 2-MIN. BLOCKS OF ADAPTATION, FOR EACH SUBJECT Group Mean Standard Deviation Cent i ngent 100 3.33 0.82 66 3.29 0.72 33 3.06 0.76 Noncont i ngent 100 3.3» 1.04 66 3.30 0.68 33 3.36 0.68 0% control 3.15 0.7b

PAGE 32

23 (rMsaa_%rMsai^OOl) XDU9nb9jj asuodsay paiujo^suDJi ^o juaDjaj

PAGE 33

2k t~ t— >z z z

PAGE 34

25 z z z o o o z z z "<:• ,-> -»f

PAGE 35

26 TABLE 3 MEANS AND STANDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FREQUENCIES OF EACH CONTINGENCY AND RATIO OF REINFORCEMENT GROUP DURING ACQUISITION (IN 2-MIN. BLOCKS) ~ '

PAGE 36

27 TABLE k SUMMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCENT RESPONSE FREQUENCIES DURING THE 16 MIN. OF ACQUISITION (IN 2 -MIN. SLOCKS) Source

PAGE 37

28 Fig, 3 Indicates that the differences among the Contingent groups were greatest during the first 2-min. block of ral nforcen^nt . These differences oscillated and became smaller as the number of reinforcements increased, until, at the end of reinforcement , there was almost no difference among these three groups. To evaluate this trend, the data for the Contingent groups were subjected to an additional separate analysis. Although this analysis Indicated no significant over-al 1 F ratios, t tests on the first block of 2 min. showed the response rate of the 33C group to be significantly greater than that of the 6oC and lOOC groups (jt » 1.93, p < .05. and L = 2.68, p < .025, respectively). These con^arlsons tend to support the observation that the largest differences among the Contingent groups occurred during the first 2mln. block of acquisition. Several facts about the data suggested that the linear component of the acquisition trend In the Contingent groups be examined. The 33C group's mean performance during the first 2-mln. block was about 100 percent, while both the 66C and lOOC were below 100 percent. Secondly, the over-all curves of the two groups that were below 100 percent at the beginning of acquisition appeared to Increase slightly as training progressed while the 33C group appeared to decrease. It was clear that the slopes of the lOOC and 66C groups were positive while that of the 33C group was negative. The differences among the linear components of these trends were significant (F « 8.91, df ?.A2, p < .001),^ thus ®The error term used In this analysis Is the partitioned error term suggested by Grant (1956) and more recently by Gal to and Turner (1963).

PAGE 38

29 establishing the statistical reliability of the apparent convergence of the Contingent groups during training. Contrasted to the behavior of the Contingent groups, were the performance curves of the Noncont i ngent groups. Fig. k indicates that the tendency of these three groups was toward fewer responses as training progressed. No statistically significant differences In linear trend were present in these data nor was their over-all trend significantly negative. Extinction The second half of Fig. 3 and the second half of Fig. k show the transformed response frequency curves for each Contingent and Noncontin* gent group, respectively, for the extinction period, in 2-min. blocks. The means and standar<^^ 'deviations of these data are presented in Table 3* The first thing to be observed in the extinction data is the general convergence of the groups. By the end of the first 8 min* of extinction, the differences that were present at the end of training had almost entirely disappeared, with the exception of group lOONC, which fluctuated greatly. Furthermore, the only group that remained consistently above the base level of 100 percent was group 33C. Of special Interest was the behavior of the Contingent groups as compared to the Noncontingent groups during the first 2 rain, of extinction. The Contingent groups separated considerably from each other, v\^lle the differences present during the last 2 mln. of acquisition for the Noncont i ngent groups almost entirely disappeared during the first 2-mln. bloctt of extinction. These changes were not statistically significant.

PAGE 39

30 TABLE 5 MEANS ANO STANDARD DEVIATIONS OF TRANSFORMED PERCEWT RESPONSE FREQUENCIES OF EACH CONTINGENCY ANO RATIO OF REINFORCEMENT GROUP DURING EXTINCTION (IN 2-MIN. BLOCKS)

PAGE 40

31 An analysis of variance was carried out on the data from the first 8 min. of extinction and is presented in Table 6. Tiie analysis sliowed several things* First (sea Table 5)» extinction" occurred to a sufficient extent that the over-all difference due to Contingency of reinforcement was not significant. Secondly, the Minutes X Contingency interaction was not significant, indicating again the extent to which extinction mitigated against significant group differences (see the significant modified Minutes X Contingency interaction below). To evaluate possible differences vriiich occurred early in the extinction period but soon disappeared, analysis of variance was carried out on the data of the first 2-mln. blocic of extinction. In this analysis the Contingency effect resulted In an F ratio of 3.81 (df » 1/8^, p < .!0). Further breal
PAGE 41

32 TABLE 6 SUHMARY OF ANALYSIS OF VARIANCE OF TRANSFORMED PERCEtir RESPONSE FREQUENCIES DURING THE FIRST 8 MIN. OF EXTINCTION (IN 2-MIN. BLOCKS) Source df MS (Between S^) Contingency Ratio C X R Error (b) (89) I 2 2 84 1,551.75 4,319.47 1,304.72 694.03 1.545.10 2.80 (Within Ss) Ml nutes M X C M X R M X C X R Error (w) (270) 3 3 6 6 252 199.4ij 153.74 261.46 181.77 459.07 193.48 1.35 2.37* *P < .05.

PAGE 42

data from these four groups during the first 8-mln. of extinction resulted In a significant Contingency x Minutes Interaction (F 3.05, df s 3/168, p < .05). This significant Interaction Indicated the convergence of these groups over minutes. Both Contingent groups decreased responding, and both Noncont i ngant groups increased responding. Table 6 shows that the Minutes X Contingency X Ratio effect was significant (F 2.37, df r 6/252, p < .05). This interaction may be clarified by examination of Figs. 3 and k. The performance curves Indicate that the differences between groups 33C, 33NC, and the differences between groups 66C, 66NC appear to have become smaller while the differences between groups lOOC and lOONC fluctuated. Further analysis of this significant Interaction was accanpllshed by analyzing the trend of the data for the first iJ min. of extinction. Table 7 presents the results of this analysis. A significant Contingency X Ratio linear component was found (F 3.^1, df = 2/3A, p < .05). This effect resulted largely from the fact that the difference between the linear extinction trends of the 33C and 33NC groups was significant (F 4.32. df « 1/84, p < .05). The triple Interaction of Table 6 was probably significant because of the convergence of groups 33C and 33NC, the partial convergOTce of groups 66C and 66NC, and the considerable fluctuation between groups lOOC and lOONC. percent control group It was hypothesized that the ^s coci^rlslng the no stimulation group, who merely sat In the experimental chamber, would adapt to the situation and then stabilize (Indicated by an initial decrease In

PAGE 43

y* TABLE 7 SUMMARY OF ANALYSIS OF VARIANCE OF LINEAR TREND OF TRANSFORMED PERCENT RESPONSE FREQUENCIES DURING THE FIRST 8 MIN. OF EATINaiON (IN 2-HIN. BLOCKS) Source

PAGE 44

35 relative response frequency followed by fairly constant responding). Table 8 presents the means and standard deviations of the transformed response percentages In blocks of 2 min, for the 13 Ss In thi» group* The data of the percent group are presented In the same time divisions as the earlier data (tO-min. adaptation, 16-min. acquisition, IG-mln extinction). Table 3 indicates an immediate decrease in responding dur* ing the first 8 min. followed by a slight increase during the next 2mln. bloclc. During the next 16-min. period (comparable to the acquisition period) the mean percentage responding remained approximately constant at a little above the 100 percent base line and then decreased below this base line toward the end of the period. The last 16 min. (comparable to the extinction period) were marked by a general decrease after a slight initial increase In responding. Statistical analysis of the first lO-min. period (adaptation) indicated that the variation of the means of the five 2-min. blocks was statistically significant (F » 2.o4, df 4/56, p < .05). This finding, in conjunction with the means shown In Table 8, indicated that adaptation did occur during the first 10 rain. Analysis of the variation of the 2-min. means from minutes 11-26 and minutes 27-^2 showed no significant differences. Indicating that the subjects had more or less stabi 1 1 zed. Another indication of the effect of the reinforcing stimulus was obtained by comparing the percent control group to all other groups In the main experiment during acquisition and extinction. Examination of Figs. 2, 3, 4, and Table 8 shows that, after the first 2-min. block of acquisition, the percent control group responded at a higher level of

PAGE 45

36 TABLE 8 MEANS AND liTANDARD DEVIATIONS OF TRANSFORMED PERCENT RESPONSE FREQUENCIES FOR THE PERCENT CONTROL GROUP DURING THE 42-MtN. SESSION (IN 2-MiN. BLOCKS) Mi nutes

PAGE 46

37 relative response frequency than all the other groups until the I4th minute. However, the percent Control group gradually gave fewer and fewer responses v\fh{le the Contingent groups were either gradually giving more responses or remaining about the same; all of the Noncont i ngent groups were gradually decreasing. Analys<;;s of variance carried out on these data indicated that the percent Control group gave significantly more responses during acquisition than: (a) lOOC group (F • 5.58, df a 1/28, p < .05), (b) lOONC group (F s 7. 61, df s 1/28, p < .025), (c) 66NC group (F 10.20, df 1/28, p < .005), and (d) 33NC group (F » 5.36, df 1/28, p < .025) . Paring extinction the percent control continued to decrease responding gradually. For the first 6 min. of extinction the percent control group's response level was equal to or above all the groups except group 33c. (Group 33C responded consistently above all other groups throughout extinction.) For the remainder of the extinction period the percent control responded consistently below or equal to all groups except lOONC. Small and large emitted GSfts It was expected that the relative response frequencies for the Contingent groups would be above the 100 percent (pret raining) level during the reinforcement period. Fig. 3 shows that the opposite occurred. In general, all of the Contingent groups responded below their own preliminary levels.

PAGE 47

38 There was the possibility that the sensitivity of the equipment used In the present study resulted In recording and reinforcing of very small responses wlilch may not have been conditlonable. Lacey and Lacey (1958) have pointed out that some GSR activity Is fairly stable over time and is not correlated with environmental stimulation. If these responses are actually not conditlonable, then their "reinforcement and inclusion in the frequency count for determination of the present dependent variable would contaminate the true" conditioning effects. To examine this possibility the total response frequency vuas dichotomized Into those responses greater than or equal to 1 percent of the base resistance of the S at the time a response occurred (large responses), and those responses less than 1 percent of the base resistance (small responses).'^ Base resistance in this case was defined as the resistance (in Ki lohns) of the S. Each S's base resistance was obtained for each minute of the experimental session and sro^l I and large responses were measured and tabulated accordingly. These were then transformed and expressed as percentages of the number of responses during the last k min, of the adaptation period. These transformations were carried out on the data for all S.s separately and the results for the lOOC and lOONC groups are presented In Fig. 5 (large) and Fig. b (small) in bloclcs of 2 rain. The first half of Fig, 5 (large responses) indlcot'-'j that the lOOC group responded consistently above the 100 percent base level for the ''The reason these small responses may not be subject to conditioning by instrumental methods wi 1 1 be dealt with In the discussion section. ^2Klnwel (SEPA, 1964).

PAGE 48

39 • • 100 NONCONTINGENT A

PAGE 49

40

PAGE 50

entire 16 min. of acquisition although no apparent trend is evident* During this same period the lOONC group responded consistently below the 100 percent base level and below the lOOC group (except for one data point). An analysis of variance over the last 3 min. of acquisition was carried out on the transformed relative response frequencies grouped Into Z-min. blocks and is summarized in Table 9. The analysis indicated that the obvious difference (for the last 8 min. of acquisition) between the two groups was significant (F 4.40, df = 1/28, p < .05). In addition, the apparent increasing separation between the two groups over time from the 10th through the 16th rain, was reflected In a significant Contingency X Minutes interaction (F 2.91, df = 3/84, p < *05). The second half of Fig. 5 shows the large response performance during the 16 min. of extinction. The figure shows that the lOOC group initial iy increased responding, reaching a peak at 8 min., and then responded below this peak for the remaining 8 min. The lOOC group responded above the 100 percent base level throughout the extinction period. The second half of Table 9 presents the analysis of variance for these data over the entire 16 rain, of extinction. The analysis indicated that the difference between the Contingent and Noncont I ngent groups was significant (F 4.98, df a 1/28, p < .025). The Contingency X Minutes interaction was not significant, Indicating that the differences between the two groups did not change greatly during the extinction period. Fig. 6 presents the curves of the mean transformed response frequencies of the lOOC and lOONC groups for responses less than 1 percent of the base resistance. The figure shows the performance curves In

PAGE 51

k2 en UJ CD < o (/) 3^ Cy UJ o • o z z sS GO O z ec & o o u. o yt Ul Ul z ^ H'.> a: o. o y> ti. Ul of z o < z Ul s Ul u a£ o Ui ^^ Hi o Ul a. • •as. 3 — 2f c o zfe O Ul UJ u. o o 1= u. — °Ui Si -J UJ iuj < X u.»o Ul o Ul \n a. C9 3 O K in o% <*> un — tn r>» CT\ \0 <*^ o «» * t. ^ o ^ CM CM u 3 o r^ o ^ i>» cs o • • • lR «^ r*. (** m u> * « * — vO — (!) c >-' 4) = e 4^ IC UJ o i*> ^ vD O in m •— trt • • • « — o «i m :?

PAGE 52

kZ blocks of 2 mtn. for the 16 mtn. of acquisition and 16 min. of extinction. Since this class of response was tentatively identified as nonconditionable, the course of change for these responses was expected to be uncorrelated with reinforcement contingency. The first half of Fig. 6 indicates unsystematic changes. Analysis of variance of these data, hov^ever, resulted in a significant Contingency X Minutes interaction (F a 2.97, df s 7/196, p
PAGE 53

CHAPTER IV DISCUSSION The purpose of the present study was to determine whether the course of conditioning and extinction of the uneliclted GSR could be differentially influenced by different schedules of reinforcement. The results indicate that a differential effect was obtained. One group of S^s received a response-contingent, dim, white light as a positive reinforcement. This Contingent group was subdivided into three equal subgroups. Each subgroup received the reinforcer on a different fixedratio reinforcement schedule, viz., 100 percent, 66 2/3 percent, 33 1/3 percent. Yoked to each S, in the Contingent group was an S who received the same number and pattern of reinforcements, but the delivery of the reinforcement was completely determined by the Contingent S. Thus, S^s in the Noncont i ngent groups received the reinforcer independent of their behavior. Several facts led to the assumption that the dim white light used in the present study would act as a positive reinforcer. The Ss sat in an experimental room which was soundand light-proof. Before actually beginning the delivery of the light, a 10-min. adaptation period was given during which time the S merely sat in the dark room. Thus, it would seem to follow that any z)ild changes in this sensorily deprived environment might serve as a reward. Several studies on animal S,s confirm this hypothesis. Barnes and Baron (1961), Barnes and Kish (1953),

PAGE 54

^5 and Kish (1955) found that when mice were confined to a dark chamber significant increases in number of bar contacts were obtained when light onset was made contingent upon bar contacts. Fox (Solomon, 1961) was able to train monkeys to press a lever to turn on a light when they were kept In a dark chamber. When the monkeys were next deprived of light for a period of hours and then given access to the bar, a large and significant increase In responding occurred. These studies help define the brief light used In the present study as a positive reinforcer. Perhaps even more important for the definition of the light as a positive reinforcer are the Investigations by Fowler and KImmel (1962) and Klmmel and KImmel (1963). Both of these studies were designed to condition the GSR Instrumantally with a brief flash of dim white light as the positive reinforcer. Both studies reported positive results. HypQthe${,g 1 The first hypothesis of concern In the present study stated that the Contingent groups would emit a greater number of responses during acquisition and extinction than the Noncontlngent groups. Analysis of the data indicated the confirmation of this hypothesis for the acquisition period. The over-all average relative response frequency for the Contingent groups was 95.8 percent while for the htoncontingent groups It was 88.3 percent. This superiority of the Contingent groups was statistically significant (F 6.23, df = 1/8U, p < .025). The previous studies (Fowler and Kltnnel, 1962; and KImmel and KImmel, 1963) on conditioning of the unellclted GSR also reported such differences during the acquisition period. Thus, the present result confirms the previous

PAGE 55

k6 findings and Is in agreement with them. However, an important procedural change introduced in the present study should be emphasized at this point. The earlier studies using this technique delivered noncontlngent reinforcement only at times when the S. was not making a response. That restriction precluded the delivery of the reinforcer fortuitously when a response was being emitted. However, in the present study an S, in the Noncontlngent group could receive a reinforcement at any time, whenever his Contingent mate received one. It had been anticipated that this would result in stnaller absolute differences between the Contingent and Noncontlngent conditions. This, in fact, did occur. T;j "etermlne the extent to which a Noncontlngent S was reinforced for emitting a GSR, the number of reinforcing lights (a response was considered reinforced if the light occurred contiguously with the response or within I sec. after) was divided by the total number of lights received by an S.. This score was then expressed as a percentage for each S.. The means and standard deviations of the percentages of the reinforcing lights for the three Noncontlngent groups were: lOONC, Mean 18.93 percent, SO » 10.25 percent; 66NC, Mean 20.07 percent, SO a 9.26 percent; 33NC, Mean a 22.33 percent, Sv s 12.05 percent. When the present study is compared to thd Klmmel and Kimmel (1963) study, the absolute effects were smaller. However, despite this procedural change, reliable differences were obtained in the present study. During the extinction the over-all difference between Contingent and Noncontlngent groups only partially supports the hypothesis. The mean of the Contingent group was 101 percent while that of the ftoncontingent group was 95 percent. However, this difference was not

PAGE 56

47 statistically significant. Analysis of tiie data of the ^ioncontingent groups during the first 8 min. of extinction showed that the 66NC and 33NC groups gradually increased in response frequency until, at the end of 3 mIn., their relative percent response was approximately equal to their base rates as determined during the last h min. of adaptation. These two groups showed extinction curves comparable to the extinction curve obtained in the Kimmel and Kimroel (1363) study for a lOONC group, although the extinction was faster in the present two groups. This more rapid extinction was probably due to the difference in the procedure associated with the delivery of the noncont i ngent rel nforceiner.t (t'ascrlbed above), and to the fact that the S.s in the 66NC and 33NC groups of the present study received considerably fewer lights^ than the S^s of the tOONC group of tiie other study. The main reason the hypothesis under consideration was only partially supported was due to the relative performances of the lOOC and lOONC groups. Of considerable interest was the fact that the lOONC group demonstrated what can best be termed "Immediate extinction. This group shifted from having the lowest relative response rate in the Noncont i ngent groups during the last 2 rain, of acquisition to having the highest relative response rate during the first 2 rain, of extinction. Furthermore, the lOONC group was actually responding higher than tha ?COC group during the first 2 rain, of extinction. This finding was not In agreement with either the Fowler and KImrael (1962) or KIramel and Kinwel (1963) study. In the Fowler and 13The mean number of lights received by the S,s In each Moncontingent group was: group lOONC, 67.73; group 66NC, 47.53; group 33NC, 22.^.

PAGE 57

ii8 Kimmel study the difference between the two groups (lOOC and lOONC) v^as found to be statistically significant in extinction during the first minute. In the Kimmel and Kimrael study large and significant differences occurred during the ent i re lO-min. period of extinction for the two groups (lOOC and lOONC) comprising that study. The contrast between these and the present results cannot be easily dismissed and will be dealt with below. It must be concluded, nonetheless, that the noncontingent reinforcement procedure used in the present study was not as effective in maintaining the difference between Contingent and Noncontlngent groups during extinction as was the procedure of delivering the reinforcement to Noncontingent S,s only when an £ was not making a resjsonse. Hypothesis 2 The second hypothesis tested in the present study was concerned with the influence of the three schedules of partial reinforcement upon the performance of the Contingent groups. The prediction specifically stated that the higiiest rate of responding would occur in the 33C group both during acquisition and extinction. The next highest level of respondln was aj
PAGE 58

The over-all differences among the Contingent groups were not statistically significant, but analysis of the first 2-niin. block of acquisition showed the 33C group to be significantly tnore responsive than the 66C group (i. 1.93, P < .05) and also greater than the lOOC group (t « 2.68, p
PAGE 59

50 observed. By the 8th min. of extinction, the groups were responding almost Identically although there was some divergence during the later minutes. During the last 2-mln. block of acquisition the differences among these groups had almost disappeared, yet, with the removal of the reinforcer, large differences reappeared during the first few minutes of extinction. The behavior of the groups seemed to indicate that maximal performance differences occur as a function of the schedule of reinforcement ^j^ of change In conditions. In another study on operant conditioning of the GSR, Kinmel and Hill (I960) used several different response-contingsjnt relnforcers arid obtained maximum effects after reinforcement was terminated. However, this large change In responding at the beginning of extinction was not found In the other two operant GSR conditioning studies (Fowler and Klnnel, 1962; Kimmel and KIramel, 1963)* No explanation of this discrepancy between studies is apparent. The present results support the hypothesis that partial reinforcement of the uneli cited GSR results In higher rates of responding during acquisition and greater resistance to extinction than does continuous reinforcement although the effect was apparent only during the Initial periods of each condition. Hypothesis i The behavior of the Noncont i ngent groups, in addition to their control use, was of Interest 2S£ SS.* Fowler and Kimmel (1962) found that their Moncont i ngent group emitted fewer and fewer responses during the acquisition period and responded significantly less than the Contingent group during the 1st mln. of extinction. Kimmel and Kimmel

PAGE 60

51 (1963) obtainod approximately the same result during acquisition for their Noncontingent group. In this lattor study, the ejitlnction period was prolonged from ^ to 10 min. and It was found that the Noncontingent group showed a gradual Increase In the number of responses emitted during extinction. This trend of increased responding continued until, at the end of the lO-mln. period, the frequency of responses was almost as great as the initial base level. The Interpretation given by Kimmel (1962) for this effect was that the S,s In the Noncontingent group nuist have learned not to respond during the acquisition period (the noncontingent light was delivered only when the S,s were not responding). During extinction the S,s stopped not respond I n;^ and, tlius, their response frequency curve gradually rose. In the present study throe different groups of S.s were run under the modified noncontingent procedure as previously noted (e.g., p. 47). Each of the three groups received a different number of noncontingent reinforcements which accompanied the reinforcements delivered to the Contingent Ss (cf, p. US), The third hypothesis was formulated under the two assumptions that the delivery of the noncontingent reinforcement would result In decreased responding and that the amount of decrease viould bear a direct relationship to the number of reinforcements delivered, as It was determined that the Noncontingent Ss were actually being trained on a modified variable-Interval schedule of reinforcement with the lOONC group receiving the shortest Interval. Specifically, the hypothesis stated that the lOONC group would give the least responses, followed by the 66NC group, and then the 33NC group. Examination of the performance

PAGE 61

52 curves during the last 8 rain, of the acquisition period confirms the hypothesis. Comparison of these performance curves with the results of the KJwnel and Kimnel study (1963) Indicates that the terminal response frequencies were very similar for the lOONC groups. The major differences between the two studies with respect to these two groups was that the rate of decrease In responding In the Klnmel and Kitnrjel study (1963) was more rapid than In the present study. This result v«uld be expected due to the met liodo logical differences of noncont I ngent reinforcement. During extinction It was expected that the relative response frequencies of the three Noncont I ngent groups would gradually Increase. The rir.ults confirmed this expectation for only the 6bNC and 33NC groups. The response frequency curves for these two groups may be compared to the extinction curve of the Nonconti ngent group of the KImmel and KImrael (1963) study. Extinction occurred more rapidly In the present study, however. The results Indicated that extinction occurred during the first 8 min. of the extinction period while, In the earlier study, a 10-min. period was required. The major difference with respect to these groups In the two studies Is that, in the present study, although giving similar shaped extinction curves, the Noncont i ngent ^ responded at a higher level throughout the acquisition and the extinction period. The behavior of the lOONC group of the present study was certainly different from what was expected. It should be noted, however, that the over-all relative response frequency of this group (for the entire 16 rain.) was below all other groups, as was predicted. There remains to be explained, however, the behavior of this group during the first part of extinction. It would seam that an expectancy" or

PAGE 62

53 "discrimination" hypothesis could be used to account for the "Immediate" extinction apparently given. However, this hypothesis would fall to handle the results of the earlier studies. Examination of the data for the individual S.s indicates that 11 of the 15 SjS In this group Increased In response frequency during the first 2 rain, of extinction when compared with the last 2 min. of acquisition. This fact indicates that the apparent Increase was not an artifact resulting from only one to two ^s who might have shown large response shifts. Two possibilities remaining are that the shift was a chance occurrence or that the increased responding was related to the high frequency of reinforcement coupled with the particular method of delivery of noncont I ngent reinforcement used In the present study. Hypothesis ii Fowler and Klnroel (1962) and KImmel and Kimpftel (1963) had demonstrated that the removal of the reinforcing stimulus resulted In a decrease in the magnitude of the differences between the Contingent and Noncontlngent groups over time. These decreasing differences were Interpreted as extinction effects and were clearly shown In the Klmmel and Klnriel (1963) study. In that study a significant Contingency x Minutes interaction was obtained In the extinction data. Their extinction period was marlted by a gradual decrease In responding In the lOOC group and a gradual Increase In responding In the lOONC group. The fourth hypothesis of the present study was, -n effect, a prediction based upon these previous findings. It follows from the hypothesis that the differences

PAGE 63

5* obtained among the six experimental groups v^ould diminish or disappear during the extinction period. The extinction period was extended to 16 m!n* in the present study to allow for the possibility of greater resistance to extinction in the partial reinforcement groups. The results during extinction were presented in Figs. 2, 3, and k. Inspection of these figures definitely indicates a disappearance of group differences, and thereby, confirmation of the hypothesis. Closer inspection of the data during the extinction period reveals several interesting trends. First, the statistical analysis did not sboM a significant Contingency effect or a significant Contingency x Minutes interaction* This fact alone could mean that extinction occurred immediately upon termination of the reinforcement period. The performance curves presented In Figs. 3 and k support the contention of immediate extinction only for group lOONC. On the other hand, groups 66C and &6NC definitely showed a more gradual extlnc tlon. For the first U min* group 66C was responding above group 66NC. During tha next A min. group 66C dropped sharply and then rose again to a point above the 100 percent base line. This fluctuation continued for the remaining 8 min. of extinction* The behavior of group 66NC during the first 8 min. of the extinction period nicely Illustrated the expected gradual extinction curves. Beginning with the first 2-min. block, this group Increased in response rate in an almost linear fashion until the 8th mtn. Comparison of groups 33C and 33NC during the first 8 min. of extinction also showed decreasing responding for the 33C group and increasing responding for the 33NC group; both groups reaching almost the same response level at the end of 8 min. These apparent differences

PAGE 64

55 during the first 3 min. of extinction were accompanied by a significant Contingency x Ratio x Minutes interaction (F 2.37t df s 6/252, p< .05), a significant Contingency x Ratio interaction in the linear component of the trends across extinction (F 3.^1, df » 2/34, p < .05), and a significant Contingency x Minutes interaction (F 3.05, df « 3/168, p < .05) for groups 33C, 33NC, 66C, and 66NC. On the basis of the data and the statistical analyses, it may be concluded that significant extinction effects were obtained during the first 8 min. of extinction in all groups except the I DOC and lOOMC, These results agree to scMTie extent with the previous findings. The e>q>lanation for the equivocal results obtained In groups 100C and lOONC and also for the fluccuatlons occurring In some of the other groups during the last 3 min. of extinction Is not clear. Between Ss variances indicated large Intersubject differences. Examination of the data of individual Ss indicated that many of them j>howed wide rang«$ of responding. incomplete ^daptat i on apd r' c^ . spondlng below the 100 percent base rate The Fowler and Kinnel stud^' (1362) had allowed only k min. of adaptation and the records indicated continuously falling curves for both Contingent and Noncont I ngent groups during the reinforcement period. The Klmmel and Kimmel study (1963) allowed 10 min. of adaptation prior to the beginning of the reinforcement period. This was done to allow the S.S to adapt more completely to the exp-r: *: mental chamber and for their response rates to stabilize. Since the mean performance curve of the Contingent group In that study was above the 100 percent base level

PAGE 65

56 during acquisition, tiiey concluded that their extended adaptation period had adequately handled the problem. However, upon examination of the Kimmel (1962) data for relative response frequencies of each S^ the present investigator found that during the first minute of acquisition only nine S,s of the 15 S,s actually were responding above their own 100 percent base rates, although the mean of the relative response frequencies during the 1st mln. was 126 percent! By the 2nd min. of acquisition only seven S_s of the 15 S,s were responding above the 100 percent base level. Examination of the remaining min. of acquisition for each S, indicated that only about one-half the S,s were responding above the 100 percent base. The extentlon of the adaptation period, then, only partially handled the problem of adaptation vri^.^n viewed with regard to the individual S,s. One of the differences between the present study and the eerlier studies is that a new 1aborator> facility was used. The experimental room was smaller, more sourdproof, and these characteristics may have stressed the Ss. Although a full 10 min. of adaptation was given, and, although the group response curves showed an adaptation effect, it was possible that adaptation did not fully occur. Hypothesis 5. The percent control group of the present study was included to determine the frequency output of the unellcited GSR for a time period comparable to the experimental conditions. The mean percent transformed response frequencies of this group wore presented in Table 8. The data in Table 8 Indicate partial confirmation of the hypothesis. In that

PAGE 66

57 aftar an initial decrease in responding (adaptation) the mean response frequency tended to stabilize. However, the mean response frequency for the group showed a slight increase during min. 11-20. This increased responding after about 10 mIn. may be an indication of stress in the situation (Maino, 1959). If the SjS were stressed by the lacl< of stimulation, then the effect of the light stimulus may have functioned as more than a reinforcer In the Contingent and Noncont i ngent groups. The light may also have given the S something to attend to In his otherwise sensori ly Impoverished conditio is and may have resulted In decreased responding. This could account for the higher rate of responding of the percent control group In comparison to the Contingent groups during acquisition. It is more likely that the major variable In the present experiment v*>tch was responsible for the over-all reduced responding of the Contingent groups was the reinforcement of responses that were so small that either (a) the S, could not discriminate these small responses from no responses at all, or (b) they could not be conditioned by Instrumental means (assumino that "staall" and "large" unellcited GSRs are due to different processes). Both of these hypotheses are developed below. Decreased responding and the ^mal 1 responses It was previously noted that the sensitivity of the present recording equipment was greater than that of the equipment used In the prior studies. In addition, the Es In the previous studies (Fowler and Klinnel, 1962; Klmmel and Hill, I960; and Kimmel and KImmel, 1963) monitored the GSR record visually and delivered the reinforcement vnhen the

PAGE 67

58 particular response criterion had been met. In the present study the S,s' GSR was passed through a differential amplifier which automatically delivered ttie reinforcer when a "criterion" response occurred. This device was much more sensitive to the very small unelicited GSRs (I.e., lO-ZS ohms) than was possible in the earlier studies. Assuming, thus, that the procedure In the present study resulted in the delivery of reinforcements for smaller responses, the following additional assumption becomes necessary to understand the effects this had. Kimmel (1962) assumed that one of the conditions necessary for the successful operant conditioning of an autonomic response is proprioceptive feedback from that response, in other words, the response must produce stimuli of sufficient magnitude to be aix>ve the physiological noise level of the S,. The mechanism for feedbaci< from autonomic activity Is provided by afferent (sensory) nerves in the structures controlled by the autonomic system (Young, 1961). These afferent impulses go to the reticular formation and hypothalmic centers and influence cortical activity. This is the presumed feedback system. With this systetn, and given the large number of small responses reinforced, we have a possible explanation for the suppressed response rates. The very small responses emitted by cui §. were detected and reinforced by the apparatus in addition to the large responses. The difference between the size of a very small response and a very large response could be a factor of about 200 In some cases. If we assume that these very small responses provided no or little feedback then it follows tliat reinforcements were being delivered when no "effective" responses were being emitted. In this case we could expect the response rate to decrease, since this condition is very similar to

PAGE 68

59 noncont S ngent reinforcement. The data show that delivery of a reinforcement when no response Is occurring definitely results in decreased responding. An interesting hypothesis regarding the nature of these sntall responses can be developed. Lacey and Lacey (1958) report that spontaneous activity of living tissue resuUs in discharges, no matter how carefully controlled the stimulus conditions may be. In addition, they indicate that this activity is stable over time. They have postulated that the GSR shows this type of spontaneous activity and that the locus of its Initiation is in the". . . physlochemical processes at the level of localized cell aggregates." (p. 163, italics not In original) If these responses are stable over time, do not show an adaptation effect, and are a function solely of local peripheral activity, it is highly likely that they are not Influenced by reinforcement. Nature of the large response The "large" unellclted GSR Is probably cortlcally Initiated and Innervated via the autonomic nervous system. Fulton (19^3) re|X3rted that autonomic and somatic nerves are Interdependent In that there Is extensive overlapping between the two In the cortex. It may be that only these cortlcally Initiated GSRs adapt over time and are operantly conditlonable. The problem would be to find a means by v^rfilch these two types of responses can be differentiated. Conditioning the small and large responses In the present study a dichotomy was made between large and small responses by counting as a small response all pen deflections less

PAGE 69

60 than I percent of an S,'s base resistance, and as a targe response all tix>se responses greater than or equal to I percent of the base resistance. Figs. 3 and 6 presented these data for groups lOJC and lOONC (chosen for illustrative purposes) during acquisition and extinction. The performance curves support the reasoning developed above. Conditioning appeared to occur in the large responses, but not in the small. The results are contaminated in the present case, ttowever, in that al I responses (large t^nd , small) were reinforced. Adaptation of smal I and large respons.e , ^ , Further support for the stability of the small responses over time and the adaptation of the large responses would obtain if it could be shown that the frequency of the former response remained steady during the adaptation period while the larger responses showed an adaptation effect. To examine this possibility the adaptation data for the six groups of the ntaiu experiment were divided into small and large responses as explained abcv^. These relative response frequency curves are presented in Fig. 7 In 2-min. blocks for the lO-min. adaptation period. The small responses in the left half of the figure do not show ^ adaptation effect, as is indicated by their fluctuating around their 100 percent base level. The right half of the figure shows the curves for the large responses. Immediately obvious is the decreasing relative response frequencies over time. This decrease is interpreted to m&aax that adaption occurred for the large type of response. A test of the significance of the linear component of the trend of these data indicated a large significant negative slope (F 20.80, df 1/84, p < .001) for the large responses.

PAGE 70

61 z z z 1I1o o o z z z z z z 4-' o o o z z z J'\ z z z o o < C o Q. O -o < 0) ^ 3 C Z o -1>' --'1 AV c o a o -o < 3 C ( rMS9y_%ijMS8i^ooi) XDuanbajj ssuodsay paaiJO^suDJj[ p ^uaDjaj 0) a) (/) ^ c M o CD l/l c a) .— L. L. O C7 1. i/l (0 (U — o -o c c 0) fO ZJ cr — 0) — sro ME 0) in (D c ^ O +J Q. tf> i (D C 3 o c CD c !-> E I <4CM O !-> C C 0) — o en o OJ -D C MOO c — z 0) T3 1Q. C 0) fD Q. C 1 O C 0) r^ (D en +j c . Q.._ ai 13 +j .--DC Li.
PAGE 71

62 The hypothesis regarding two types of unellcited GSRs finds support In these data. This finding can account for the relatively depressed response rates observed In the present study. In addition, the possibility now exists for obtaining a very strong conditioning effect* in future studies, as well as achieving greater control over the unellcited GSR. This tasl< would require differential reinforcement of responses distinguished by a magnitude criterion. It would be necessary to reinforce several different magnitudes to obtain a better understanding of the phenomenon. Conclusion The results of the present study Indicated that the unellcited GSR can be differentially Influenced by Contingency and Schedules of reinforcement. The effect of Contingency of reinforcement shows that the unellcited GSR can b^ made Instrumental, and, thereby operate upon the environment. Formerly, this fact was thought to be true for somatically mediated responses only, in addition to the Contingency variable, however, the Influence of other variables can be demonstrated with the more conventional instrumental responses. For example, magnitude, delay, and type of reif*fct cement all influence the acquisition and extinction of instrumental ly conditioned somatic responses. Another Important variable determining performance Is the schedule of reinforcement. Showing that schedules of reinforcement have a similar effect upon Instrumental ly conditioned autonomic responses allows for the possibility of an underlying mechanism and set of laws wliich apply to both somatic and autonomic response systems. The present study has demonstrated that

PAGE 72

63 fixed-ratio schedules of reinforcement do differentially influence the uneli cited GSR in a fashion somewhat similar to their influence on somatically mediated responses. With these findings, in conjunction with the results of Fowler and Kimmel (1962) and Kimmel and Kininel (1963), the hypothesis regarding the nonconditionabi lity of autonomic responsss by instrumental means is further weal
PAGE 73

CHAPTER V SUHHARY The purpose of the present study was to determine the Influence of partial reinforcement upon acquisition and extinction of the unellcited GSR. Ninety college students were divided into two major groups (Contingent [C] and Noncont i ngent [NC]). The Contingent group was divided further into three subgroups each of w5»ich received a different schedule of partial reinforcement on a fixed-ratio (100 percent, 66 2/3 percent, and 33 1/2 percent). All ^s were run in a light-proof, soundproof experimental chamber. Each ^ioncontingent S, was paired to a Contingent S. of the same sex and was run In another experimental chanber at the same time. The delivery of the rainforcer (a dim white light), to i>oth Ss, occurred whenever the Contingent S, had emitted the required number of GSRs. All §,s received a lO-mln. period of adaptation, the last k min. of which provided a base measure of resting response. After the 10-mIn. adaptation period was completed, a 16-min. period of reinforcement was given, during v^lch the Contingent groups received the light following the emission of a GSR (the number of emitted GSRs required to receive a light was determined by the schedule of reinforcement). Finally, there followed a 16-mln. period of extinction during which no stimuli were given. 6k

PAGE 74

65 The dependent variable was the nundser of responses emitted per 2inin. block expressed as a percentage of the nun^er of responses emitted during the base resting period (all frequencies were first transformed to Vx-i-i). During reinforcement the over-all rate of responding was greater In the Contingent group than in the Noncont i ngent group (p < .023)* Among the Contingent groups the effect of partial reinforcement was maximal during the initial minutes of reinforcement. A significant convergence of these Contingent groups was obtained over the 16 tviin. of reinforcement (p ncont i ngent groups increased). However, the differences between groups lOOC and 100NC fluctuated widely. During extinction the effect of partial reinforcement upon the Contingent groups was most obvious at the beginning of the extinction period. The highest responding occurred in group 33C followed by group 66C and then group lOOC. These differences had almost vanished at the end of 8 min. of extinction, except that group 33C continued to respond above all other groups throughout tiie 16 min. of extinction. A percent control group (15 Sjs) was also run. These S,s merely sat In the dark soundproofed experimental chandser with no stimulation for U2 min. This group showed an initial adaptation effect, followed by fairly steady (but unexpectedly high level) responding during the last 32 min.

PAGE 75

66 The low response rates of the Contingent groups* compared to the percent control and to a similar group in the Kinme! and Kfmmel (1363) study, were interpreted as due to reinforcement of very small GSRs which were not reinforced in the previous study. An hypothesis was developed regarding the nature of the small and large GSRs and their conditionabillty by instrumental methods. Finally, further research on this problem was proposed. It was concluded that (a) tiie effect of partial reinforcement was greatest at the beginning of acquisition and at the beginning of extinction; (b) the present study confirmed and agreed with the previous studies (Fowler and KImmel, 1962; and Kiinnel and Kimmel, 1963) which showed that Contingent reinforcement resulted in higher response rates than Noncontlngent reinforcement; and (c) the hypothesis stating that autonomic responses are not conditionable by Instrumental means (Kimble, 1961) was wealcened further.

PAGE 76

REFERENCES Barnes* G, W. and Baron, A, Stimulus complexity and sensory reinforcement. J. comp . phvsiol . PsvchoK . 1961, ^ ^66-469. Barnes, 6. W, and KIsh, B. B, On some properties of visual reinforcement. Amer . Psvclioloqlst. 1958, ii, 417. (Abstract) Boren, J.J. Resistance to extinction as a function of the fixed ratio, i. exa. Psvchol .. 1961, C±, 304-308. Duncan, 0. B. Hultiple range and multiple F tests. Biometrics. I955t Edwards, A, L. Experimental , design i .n, psvcIto log leal f;esearch . Now York: Rinehart, I960. Ferster, C. B. and Sl
PAGE 77

68 Holland, J. 6. and Skinner, B. F. The analysis of behavior . New York: McGraw-Hill, 1961. Kimble, G. A. Hilgard and Marauls ' Conditioning and Learning . New York; Appleton-Century-Crofts, 1961. Kimrael, Ellen B. Operant conditioning of the galvanic skin response: a replication. M. A. Thesis, 1962, Univ. of Florida. Kimmel, Ellen. Stability of Individual differences In activation. Paper read at meetings of the Southeastern Psychological Association, Gatllnburg, Tenn., 196^. Klnriel, Ellen and Kimmel, H. 0. A replication of operant conditioning of the GSR. I, S2£. Psvchol .. 1963, ^ 212-213. Kimmel, H. 0. and Baxter, R. Avoidance conditioning of the GSR. ^. jgsft. Psychol .. I96U (In press). Kitnnel, H. 0. and Hill, Frances A. Operant conditioning of the GSR. Psvchol . Rsa.. I960. Z, 555-562. Kish, G. B. Learning when th.i ons<£t of liluraination is used as reinforcing stitriilus. jJ.. como . phvsiol . Psvcho> .. 1955, M, 261-264. Lacey, J. I . and Lacey, Beatrice C. The relationship of resting autonomic activity to motor impulslvtty. Reg.. Publ . Agjg.. Nerv . 01^., 1958, i^ |J^-209. Malmo, R. R. Activation: A neuropsychological dimension. Psvchol. Rev .. 1959, ^ 367-386. Handler, G,, Preven, 0. W., and Kuhlman, Clementina, K. Effects of operant reinforcement on the GSR. ±. exp . Anal . Be^ .. 1962, ^ 317-321. Hoore, J. W. and Gormezano, I. Yoked comparisons of Instrumental and classical eyelid conditioning. ±, e j XP ,. Psvchol .. 1961, ^ 552-559. Mowrer, 0. H. Preparatory set (expectancy) a determinant In motivation and learning. Psvchol . Rev .. 1938, iti, 62-91. Nowrer, 0. H. and Jones, Helen M. Habit strength as a function of the pattern of reinforcement. ±, e^ . Psvchol .. 19^5, 21* 293-3H.

PAGE 78

69 Razran, G. The observable unconscious and the Inferable conscious in current soviet psychophyslology: Interoceptive conditioning* semantic conditioning, and the orienting reflex. P;sychol « Rev «. 1961, ^ 81-147. Shearn, 0. W. Operant conditioning of heart rate. Unpublished doctoral dissertation, Univ. of Indiana, I960. Skinner, B. F. Science and human behavior . New YorIc: TheMacmillan Co., 1953. Skinner, B. F. The l?ehavlor of organisms: ar^ experimental analvsU. New York: App I eton -Century -Crofts, 1938. Skinner, B. F, The experimental ^alysls of behavior. Amer . Scl .. 1957. iii. 3^3-371. Snedecor, G. W. Statistical methods . Ames, Iowa: Iowa State College Press, 1956. (5th Ed.) Young, P. T. ^totlvat^on and emotion: A survey Qf the determinants of human and animal activity . New York: Wi ley, 1961 .

PAGE 79

APPENDICES

PAGE 80

APPENDIX A

PAGE 81

INSTRUCTIONS Your task In this experiment is merely to relax, remain as motionless as possible, and pay attention to what happens. Since we will be recording your galvanic skin response with very sensitive equipment, it is absolutely imperative that you avoid abrupt movements, deep or uneven breathing, etc.; because any activity of this sort will produce errors in our data. We would like for you to get as comfortable as possible now, before we begin recording, so that you will be able to remain still when we start. If it becomes necessary to move (for exem^le to scratch your nose), please do so with your free hand with as little abruptness as possible. Don't cross or uncross your legs or move about in the chair~just relax. You will find that the more relaxed you can get, the easier It will be to remain still. Yet, it is necessary that you remain alert in the sense of paying attention to what may happen. That is, keep your ears, eyes, etc. open. This Is also an absolute necessity. Oo not give In to any tendency to become drowsy; do not close your eyes. Oo you have any question? 72

PAGE 82

APPENDIX 6

PAGE 83

J

PAGE 84

APPENDIX C

PAGE 85

TABLE 10 NUMBER OF RESPONSES PER 2-411 N. BLOCK FOR ADAPTATION FOR 100 PERCENT CONTINGENT SUBJECTS m ' '1 t II 1 « re s Number

PAGE 86

77 TABLE n NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMENT FOR 100 PERCEm* CONTINGENT SUBJECTS Number

PAGE 87

78 TABLE 12 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 100 PERCENT CONTINGENT SUBJECTS s Number

PAGE 88

APPENDIX

PAGE 89

TABLE 13 NUMBER OF RESPONSES PER 2-^IN. BLOCK FOR AOAfTATiON FOR 66 PERCENT CONTINGENT SUBJECTS s Number

PAGE 90

81 TABLE 14 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMEtfl' FOR 66 PERCEMT CONriNGEtfr SUBJECTS s %Hber

PAGE 91

82 TABLE 15 NUMBER OF RESPONSES PER 2-HIN. SLOCK FOR EXTiNCTiON FOR 66 PERCENT COKTINGENT SUBJECTS s Numbers

PAGE 92

APPENDIX E

PAGE 93

TABLE 16 NUHBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 33 PERCENT CONTINGENT SUBJECTS — _6I4— . ' IL.i ii'iilJM Number

PAGE 94

8S TABLE 17 NUHBEA OF RESPONSES PER 2-MIN. BLOCK FOR RSIJifORCEMEtff FOR 33 PERcemr comtingew subjects s Number

PAGE 95

86 TABLE 18 NUMBER OF RESPONSES PER 2-MIN, BLOCK FOR EXTIfJCTION FOR 33 PERCEtn" COMTINUEMT SUBJECTS s Nunfeer 2 4 6 8 10 12 14 16 1

PAGE 96

APPENDIX F

PAGE 97

TABLE 19 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 100 PERCENT N0NCO^^r!NGE^^ SUBJECTS Number

PAGE 98

89 TABLE 20 mmilK OF RESPONSES PER 2-Hni. BLOCK FOR REI NroaCEHE<mJ4GENT SUBJECTS s Hunji>ers

PAGE 99

90 TABLE 2> NUMBER OF RESPONSES PER 2-MJN. BLOCK FOR EXTINCTION FOR 100 PERCENT NONCOKTINGEOT SUBJECTS TTT r : ". T"r -*— Numbars

PAGE 100

APPENDIX G

PAGE 101

TABLE 22 NUNBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 66 PERCENT NGNCONTINGENT SUBJEaS s Muinber

PAGE 102

93 TABLE 23 NUHBER OF RESPONSES PER 2-HiN. BLOCK FOR REINFORCEMEHT FOR 66 PERCEMT NONCONTiNGENT SUSJECTS s Number

PAGE 103

TABLE 24 NUHBEfl OF RESPONSES PEifi 2-MlN* BLOCK FOR CATii4CTiwN FOR 66 PERCENT NOflCOin'iNGENr SUBJECTS s Ntrnuef

PAGE 104

APPENDIX H

PAGE 105

TABLE 25 OF RESfONSES PER 2-illM. ELO:fl FOR AOAFTATION FOR 33 PERCENT NO^ico^^'l^JGE^n^ subjects s NiMifer

PAGE 106

97 TABLE 26 NUMBER OF RESiPOMSES PER 2<^tlL BLXiCIC FOR REINFORCEMENT FOR 33 PERCENT NONCONTINQENT SUBJECTS s Numsijer

PAGE 107

98 TABLE 27 NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR EXTINCTION FOR 33 PERCENT NONCOIfTiNGENT SUBJECTS s Number

PAGE 108

APPENDIX I

PAGE 109

TABLE 28 NUreER OF RESPONSES PER 2-HIN. BLOCK FOR l-IO HI N. FOR PERCEKT CONTROL SUBJECTS S Number

PAGE 110

101 TABLE 29 NUMBER OF RESPONSES PEP. 2-MIN. BLOCK FOR 11-26 MIN. FOR PERCEKT COMTROL SUBJECTS s Numb«r

PAGE 111

102 TABLE 30 NUHBER OF RESPONSES PEk 2-HIN. BLOCK FOR 27'*42 MiN. FOR P£RCE»fr COKTROL SUBJECTS s. Number

PAGE 112

APPENDIX J

PAGE 113

TABLE 31 MEANS AND STANDARD DEVIATIONS CF THE NUHBER OF RESPONSES FOR EACH GROUP DURING ADAPTATION (IN l-tUU, niOCKS) Group

PAGE 114

lOS TABLE 32 MEANS AND STANDARD DEVIATIONS OF THE NUMBER OF RESPONSES FOR EACH GROUP DtmiNQ ACQUISITION (IN 2-MIN. BLOCKS) Group

PAGE 115

106 TABLE 33 MEANS AND STANDARD DEVIATIONS OF THE NUMBER OF RESPONSES FOR EACH GROUP DURtNG EXTINCTION (IN 2*MIH, BLOCKS) 100 66 33 Group 12 3^5678 Cont I ngent X 10.1 9.8 9.8 10.7 9.7 8.7 9.7 9.0 100 SD 5.9 5.1 ^.8 4,7 4.2 3.5 4.9 4.2 X 11.1 10.7 9.3 10.4 9.9 10.7 12.5 8.3 66 80 6.5 5.8 6.0 6.0 5.5 5.9 8.0 4.5 X 10.1 9.1 8.6 8.8 9.8 9.3 8.8 8.9 33 SD 4.0 3.7 2.8 4.2 4.7 4.4 2.9 3.1 Noncont i ngent X 9.8 8.3 10.1 8.7 8,6 8.6 9.0 8.9 SO 7.3 7.2 6,4 6.6 5.6 5.4 6.8 6.3 X 8.9 9.5 10.0 10.9 9.7 9.4 10,0 9,9 $0 5.5 5.3 5.6 5.74.6 4.8 5.4 5.3 X 9.2 9.3 9.3 10.7 9.8 10.5 10.4 11.4 SD 4.8 5.9 5.6 5.4 5.6 4.8 5.1 5.6 X 9.7 9.5 9.2 8.1 7.9 8.6 8.9 8.0 SO 5.9 4.1 2.9 3.3 3.7 4.1 4,6 4,7

PAGE 116

BIOGRAPHICAL SKETCH William Albert Groene was born September 13> 193^> in San Diego, California. In June, 1952, he was graduated from Point Loma Higii Sciiool. In August, 1956, he received the degree of Bachelor of Arts from San Olego State College, From 1957 until I960 he served in the Army Security Agency of the United States Army and was stationed in Germany. Following his discharge from the Array, he enrolled in the Graduate School of San Oiego State College. He worked as a research and teaching assistant until August, 1961. He received the Master of Science degree from San Oiego State College in June, 1963. From September, 1961, until the present time he has pursued his woric at the University of Florida toward the degree of Doctor of Phi losophy. William Albert Greene is married to the former Jean Lynell Hunt and is the father of one child. He is a member of Psi Chi and the Psychometric Society.

PAGE 117

This dissertation was prepared under the direction of the chairman of the candidate's supervisory committee and has been approved l^y «11 mombers of that coninittee. It was submitted to the Dean of the College of Arts and Sciences and to the Graduate Council, and was approved as partial fulfillment of the requirements for the degree of Doctor of Phi losophy. August 8, \S6k Dean, ColleQBf of/Aris and Sciences Dean, Graduate School Supervisory Committee: Chairman t, fAiK^

PAGE 118

^^ 2 1 76 7