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
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
LIST OF TABLES. .............. ...........
LIST OF FIG RS .... .... ....
I ~ INTRODUCTIGN ...... ......
The evidence. .. .. .. .. . ..
Recent studies. ...........
The effect of partial reinforcemelnt .
Procedural difference of the present
. . .
Hypot heses .
. . .
11 RETHOD. .. .... ..
Subjects. . ..
Apparatus .. .
Design. . .. .
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)
Incomplete adaptation and responding
below the 100 percomt base rate ... .. .
Hypothesis 5. . . . . ......
Dercrased responding anrd t:he s.nall
:Jature of thz large response. ..........
Conditioning the small and large responses. ..
Adaptation of small
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
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)
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
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 ........
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 ... ...
LIST OF TA\DLES Continue
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
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). .. .. .. . . ***
LIST OF FIGURES
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
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.
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.
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
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
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.
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.
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.
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.
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.
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
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.
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.
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
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.
Z Z ZJ cr
wu. u cc
u J w u J 15 ,~
ZZZO OO u-a
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.
I .I v O M
i : : c-
I I I Iu -
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:
Group 1 2 3 4 5
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.
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
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
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
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
z z z ..C9 -- -5E
v v. v -
,'~ t O--0
.d 0I -n N )
o 0 --
i O -- C
Z Z Z .E O
ooo c moJ
o Z o .; .- U
~~s I- .a 0
~~oo -- '
, vi *
.- O -
MEAllS AND STANDARD DEVIATIONS OF TRANSFORM1ED PERCENT RESPONSE
FREQUENCIES OF EACH CONTINGENCY AND RATIO OF
REHI~FORCEM3EIIT GROUP DURING ACQUISITION
(IN~ 2-MIN~. CLOCKS)
Group 1 2 34 5 G 7 8
_ __ __
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
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
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-
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.
MEANS AND STFANDARD DEVIATIONS OF TRANSFORMED PERCEFIT RESPONlSE
FREQUENCIES OF EACH CONTI7NGENCY AND RArTIO OF
REINFORCEMENT GROUP DURING EXTINCTION
(IN 2-MI~N. BLOCKS)
Group 1 2 3 5 6 7 8
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).
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
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
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
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 --
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
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
12Kimmael (SEPA, 1964).
0 - u .-
n 0 O 3 c
o .*s~ LIa1\0 -
X~a\ai -soda LOu~u i ola~a
o o o
o ,z cc,
X~uanbai~ asuodsatl paiuiolsuD11 10 iua~iad
y 0 *
crl *- O
-- *U a.
,::::::----- ~,' ;
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-
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
v m l
a a ni
- UO -
a a o
mt an -
LD m m C
- LA O
O N L
If" c O
*.ol n 3
4.8 .- L.
*- I 1
CCD 0 -
LI1 ce" uJ
..I O. O G.
(I J) b. 2
.. > s
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.
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
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.
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.
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.
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 =
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.
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-
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
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
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
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.
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
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
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.
I I I C
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.
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
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
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
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-
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
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.
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,
Harwood, C. UJ. Operant heart rate conditioning. Pyhl g. 92
Holland, J. G. and Sk~inner, B. F* IThe nalyssi gi behavior. NwJ York:
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 >
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,
Moore, J. W. and Gormnezono, 1. Yoked comparisons of instrumental and
classical eyelid conditionin9* i.* 22.* Psy~Chol., 1961, gg,
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
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
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?
Zy I I
ZW I I U
, NUMBER OF RESPONSES PER 2-MIN. BLOCK FOR ADAPTATION FOR 100
PERCEllF COlNTINGEfff SUBJECTS
apparatus was not started until the third minute
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
NUMBER OF RESPONSES PER 2-MI~N. LOCK FOR EXTINCTION FOR( 100
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
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
NUMBER OF RESPONSES PER 2-MIN. BLOCK
FOR ADAFTATION FOR 66
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
RESPONSES PER 2-MIN. BLOCK FOR REINFORCEMIErli FOR 66
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
I .. ICeT1 CPe~O~b-a -~-~-- -. --~ -~---~-~_.~.-~--r--~--- -
RESPONSES P'ER 2-WMI. BLOCK FOR ADAF-TATCION FOR 33
PERCEIBT CONTINIGEtff~ SUBJECTS
Nu ~ar 2 14 6 8 10 12 14 16
RESFONSES P'ER 2-MNlt. BLOCK FOR RZEl:!FORCtEMENT FOR 33
PERCLITK C~tSINGjEr? SUBJECTS
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
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
RESPONSES PER 2-MIN. BLOCK FOR ADAffATION FOR 100
PERCENT NOXNCO TI1NGEt T SUB~JECTS
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
RESPO)NSES PER; 2.-MLI. BLOCK FOR( REINFORCEMENT FOR 100
PERC;13ilu 0NCi;NTINGENT SUCJECTS
RESPONlSES PER 2-Mi ll. BLOCK
FOR EXT I CT ION FOR 100
PER\CENTF NONlCONTINIGEtT7 SUJECTS