See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281862637

Effects of amount and percentage of reinforcement and

number of acquisition trials on conditioning and
extinction.

Article · January 1961

CITATIONS READS

70 168

1 author:

Allan R Wagner
Yale University
111 PUBLICATIONS   10,952 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Human Learning View project

Taste Aversion Learning View project

All content following this page was uploaded by Allan R Wagner on 18 September 2015.

The user has requested enhancement of the downloaded file.

Journal of Experimental Psychology
1961, Vol. 62, No. 3, 234-242

EFFECTS OF AMOUNT AND PERCENTAGE OF REINFORCE-

MENT AND NUMBER OF ACQUISITION TRIALS
ON CONDITIONING AND EXTINCTION'
ALLAN R. WAGNER 2
State University of Iowa

The relative performance levels importance of anticipatory reward

obtained under partial and continu- (ra ~ sa) and anticipatory frustration
ous instrumental reward training have (r/ — Sf), both of which are assumed
been characterized by Goodrich (1959), to vary with reward magnitude and
Haggard (1956), and Spence (1960, number of acquisition trials.
Ch. 6) as including: (a) during early
acquisition trials, a decrement under METHOD
partial reinforcement which is more Subjects
pronounced or more persistent as
performance is measured closer to The 5s were 64 male and 64 female experi-
mentally naive hooded rats from the colony
the goal region; (b) during asymp- maintained by the Psychology Department
totic acquisition trials, superiority of the State University of Iowa. Their ages
of partial reinforcement on response ranged from 125 to 130 days at the beginning
measures relatively early in the re- of the experiment.
sponse chain; and (c) superior extinc-
tion performance following partial Apparatus
reinforcement. The apparatus was a straight, enclosed
Of primary interest in the present pine runway consisting of a 15-in. entry box,
a 12-in. startbox, a 33-in. alley, and a IS-in.
investigation was a determination of goalbox. All sections of the runway were
the way in which magnitude of reward 3 in. wide with the exception of the startbox
influences the two mentioned acquisi- which was narrowed to 2 in. Opaque doors
tion effects, and the way in which separated the adjacent sections but, with the
magnitude of reward and number of exception of the start door, were closed only
to prevent retracing. A clear glass ceiling
acquisition trials jointly influence the covered all sections of the runway at a height
extinction effect. Such determina- of 4 in. The walls and floor of the entry and
tions were thought to be particularly startboxes were painted a flat grey, those
relevant to those theoretical interpre- of the alley and goalbox a flat black.
The apparatus was located in a black,
tations (e.g., Amsel, 1958; Beier, sound-deadened room. Dim, diffuse illumi-
1958; Logan, 1960; Spence, 1960, nation was provided by the light reflected
Ch. 6) of partial reinforcement phe- from the black walls and ceiling and originat-
nomena which have stressed the ing from four shielded 150-w. fluorescent
tubes placed 4 ft. above the runway. A large
1
This research was conducted during the fan located directly beneath the runway
the author's tenure as a National Science provided a general masking noise.
Foundation Predoctoral Fellow at the State The apparatus provided three time meas-
University of Iowa, and comprises part of a ures of locomotor performance in the runway.
dissertation submitted to the Graduate Col- Starting time was measured from the opening
lege of that University, in partial fulfillment of the start door to the interruption of an
of the requirements of the PhD degree. The infrared beam located 6 in. distant. Running
author is indebted to Kenneth W. Spence time was measured from the interruption of
for his advice and assistance throughout the this first beam to the interruption of a second
course of the investigation. beam located 12 in. beyond the start door.
2
Now at Yale University. Goalbox time was measured between the
234
 CONDITIONING AND EXTINCTION 235

interruptions of infrared beams located 7 in. schedule 5s were tested under approximately
and 1 in., respectively, from the end of the 22J-23 hr. food deprivation.
runway. The last beam passed just above the Acquisition.—-Half of the 5s received
front lip of a stationary, J in. deep, metal 16 and the other half 60 acquisition trials,
food cup. 1 trial per day, during which the per-
centage reinforcement and magnitude of
Procedure reward were manipulated. The sequence of
reinforcement and nonreinforcement for the
Experimental design.—The experimental partial reinforcement groups was arranged
design included two levels of three variables: as follows: H h- - + + - - + + -
percentage reinforcement (100% vs. 50%), h+H K It should be noted that
magnitude of reward (.08 gm. vs. 1.0 gm.), this sequence has the following characteristics:
and number of acquisition trials (16 vs. 60). (a) two reinforced and two nonreinforced
The design may be schematized as a 2 X 2 X 2 trials appear in each successive block of four
factorial, 16 5s, 8 males and 8 females, being trials; (b) the first and last trials are rein-
randomly assigned to each of the eight forced; (c) never more than three reinforce-
experimental groups. ments or nonreinforcements occur in a row;
Habituation.—Thirty days prior to the (d) when this 20 trial sequence is repeated
first training day 5s were placed on a 24-hr, the first order sequential probabilities are
food deprivation schedule, with 30-min. counterbalanced, i.e., reinforcement equally
access each day to an ample supply of wet as often follows a reinforcement as it does
mash prepared from ground Wayne Lab Blox. a nonreinforcement.
This maintenance schedule was employed The reward pellets were pressed from wet
throughout the course of the experiment, mash prepared immediately preceding the
while water was continuously available except running of each squad of 5s. Two pellet-
during the experimental sessions. makers were used which were designed to
On the 2 days prior to experimental form pellets of .08 gm. and 1.0 gm. dry weight,
training 5s were allowed to explore the run- when the mash was prepared by mixing .5 ml.
way in groups of three for S-min. periods. distilled water with each gram of ground
After the exploration on each of these days Wayne Lab Blox.
5s were placed in individual grey goalboxes Extinction.—Extinction Trial 0 was ad-
and allowed to eat a number of reward pellets ministered to each group the day following
from a food cup identical to that used in the its last acquisition trial. Extinction Trials
runway. 1 to 32 were then run on consecutive days.
Experimental training.—A training trial On each extinction trial 5 remained in the
was initiated with the introduction of 5 nonreinforced goalbox for 20 sec.
into the entry box. When 5 had entered the
narrow startbox the retrace door was lowered RESULTS
and 3 sec. later the start door was raised.
(The 2-in. width of the startbox discouraged Acquisition
5 from turning so that all 5s remained
oriented toward the start door during the Mean starting, running, and goal-
3-sec. interval.) When 5 had entered the box speeds over blocks of four acquisi-
goalbox the retrace door was closed and 5 tion trials are plotted in Fig. 1 for the
was removed 20 sec. later on nonreinforced treatment groups receiving the four
trials or immediately after eating on reinforced
trials. combinations of percentage reinforce-
The 5s were tested in squads of 8, one 5 ment and magnitude of reward.
from each of the experimental groups. Eight Measures from all ,Ss, i.e., 16- and 60-
such squads of males were tested first on each trial groups combined, are included
experimental day followed by eight squads in the curves over the first four blocks
of females. The 5s were tested in the same
sequence in every experimental session, but of acquisition trials, whereas the
the order of testing was systematically varied subsequent blocks of trials represent
among the squads so that equal numbers of means from only those 5s in the 60-
5s from each group were assigned to each trial groups. Over the first 16 trials,
of the squad positions. Following the testing
of all 8 5s in a squad, they were returned to
in all three response measures, large
their home cages, and 30 min. later 5s were reward produced faster speeds than
given their daily feeding. According to this did small reward, while continuous
236 ALLAN R. WAGNER

STARTING
• • 100% - l.Og
. . so % - i.og
o o 100% -.08g
o o 50o/0
..08g

RUNNING

1.4 r

1.2

1.0

.8

.6
GOAL BOX
,4

4 5 6 7 8 9 10 II I2 13 14 IS
BLOCKS OF FOUR ACQUISITION TRIALS
FIG. 1. Mean acquisition speeds for the treatment groups differing in magnitude of reward
and percentage reinforcement. (The vertical line marks the point at which training is ter-
minated for half the 5s in each group.)

reinforcement produced faster speeds design was computed for the means
than did partial reinforcement. over Trials 1 to 16 for each of the
In order to evaluate the reliability three response measures. Since half
of these observations, as well as the of the 5s in each of the groups were
equivalence of the 16- and 60-trial male and half female, sex was also
acquisition groups, a separate analysis included in these and all subsequent
of variance appropriate for a factorial analyses as a "control" variable.
 CONDITIONING AND EXTINCTION 237

The results of these analyses showed that: (a) the effect of magnitude of
that: (a) magnitude of reward was reward attained significance in both
significant beyond the .005 level in the starting and running measure
the starting, running, and goalbox (F = 6.05, P < .025, and F = 9.53,
measures (F = 26.67, F = 23.74, and P < .005, respectively; df = 1/56)
F = 8.20, respectively; df = 1/112); but is associated with a probability
(V) percentage reinforcement was between .10 and .20 in the goalbox
highly significant in the goalbox measure (F = 2.67, df = 1/56); (&)
measure (F = 48.35, df = 1/112, 50% reinforcement led to significantly
P < .001), reached significance in the higher performance levels than did
running measure (7*'=4.05, df= 1/112, 100% reinforcement on both the start-
P < .05), but failed to approach ing and running measure (F = 4.03,
significance in the starting measure P < .05, and F = 5.03, P < .025,
(F < 1); (c) although the starting respectively; rf/=l/56), whereas 50%
and running speed curves suggest reinforcement led to a significantly
an interaction between magnitude of lower performance level than did
reward and percentage reinforcement, 100% reinforcement on the goal-
this tendency was not statistically box measure (F = 8.01, df = 1/56,
reliable in anyfof the response meas- P < .01); and (c) in none of the
ures; and (d) there were no reliable three response measures was there a
differences between the experimental statistically reliable interaction of
groups receiving only these 16 acquisi- percentage reinforcement and magni-
tion trials and those which subse- tude of reward.
quently received 44 additional trials.
With reference again to Fig. 1, it is Extinction
apparent on all three response meas- Since the several experimental
ures that each of the large reward groups began extinction at markedly
groups maintained its advantage over different levels, a comparison of the
the corresponding small reward group obtained extinction speeds appeared
throughout the 60 acquisition trials. to offer only minimal information
Likewise, in the case of the goalbox about the rate of extinction. A
measure the 100% reinforcement preferable index appeared to be ex-
groups maintained their advantage tinction speed expressed as a propor-
over the 50% groups. However, tion of the corresponding terminal
on both the starting and running acquisition speed. Consequently each
measures the performance of each of response speed for each S over Extinc-
the 50% groups may be seen to sur- tion Trials 1-32 was expressed as a
pass that of the corresponding 100% proportion of the mean speed 5 had
group by the end of training. attained on that response measure
In order to evaluate statistically over the last four acquisition trials
the asymptotic performance levels and Extinction Trial 0.
attained under the various experi- Figure 2 presents the mean starting,
mental treatments received by the running, and goalbox proportions over
60-trial 5s, a separate analysis of blocks of four extinction trials for
variance appropriate for a factorial the eight experimental groups. Of
design was computed for the means major concern is the comparison of
over the last 8 acquisition days the resistance to extinction of the 50%
on each of the response measures. and 100% groups following the vari-
The results of these analyses showed ous acquisition conditions. In all
238 ALLAN R. WAGNER

t.OG REWARD
.10

.00

.90

0 .80
.70

K
IO ._. 100% "60 V.
.90 . o- -o 50% ' 16 N a _-
100%-16
.40

I.IO
1.00
.90
<9
Z .80
|.TO
(t .60
.SO
.40 Jr

1.00 •--•

.90
.80
g.70
01
.60

*.»
O
19 .40
.30
.20 i
R e f 1 2 3 4 9 6 7 8 R«f 1 2 3 4 9 6 7 8
BLOCKS OF FOUR EXTINCTION TRIALS

FIG. 2. Mean extinction proportions as a function of the percentage reinforced (50% or

100%) and the number (16 or 60) of prior acquisition trials plotted separately for the groups
which had received 1.0-gm. or .08-gm. rewards during training. (The reference point includes
the final four acquisition trials and Extinction Trial 0.)

three response measures it may be still present in every comparison, was

observed that with acquisition condi- appreciably smaller. It should also
tions of 1.0-gm. reward, 50% rein- be noted that in none of the com-
forcement led to a markedly greater parisons did the difference in extinc-
resistance to extinction than did 100% tion performance between the 50%
reinforcement. However, when .08- and 100% groups appear to vary
gm. reward was employed, the superi- with the number of prior acquisition
ority of the 50% groups, although trials.
 CONDITIONING AND EXTINCTION 239

A separate analysis of variance Also of interest are the separate

appropriate for a factorial design was effects of magnitude of reward and
computed for the mean proportions number of acquisition trials on re-
over Extinction Trials 1-32 for each sistance to extinction. As noted
of the response measures. The re- previously, magnitude of reward in-
sults of these analyses with respect teracts with percentage reinforcement
to the percentage reinforcement vari- in determining resistance to extinc-
able showed that: (a) in none of the tion. Consequent to this finding,
measures did the triple interaction separate F tests were also computed
of percentage reinforcement, magni- comparing the 1.0-gm. and ,08-gm.
tude of reward, and number of acqui- treatment groups at the two percent-
sition trials reach significance; (&) ages of reinforcement on each of the
in none of the measures did the inter- three response measures. When com-
action of percentage reinforcement bined with a 50% reinforcement
with number of acquisition trials schedule 1.0-gm. reward, as compared
reach significance (F < 1 in each to .08-gm. reward, led to a greater
case); but (c) in all three response resistance to extinction which attained
measures the interaction of percentage statistical significance in the case of
reinforcement with magnitude of re- the starting measure (F = 17.85,
ward was statistically significant df = 1/112, P < .001) and the run-
(F = 15.70, P < .001; F= 14.97, ning measure (F = 10.69, df = 1/112,
P < .001; and F = 4.71, P < .05 for P < .005), but not the goalbox meas-
the starting, running, and goalbox ure (F < 1). When 100% reinforce-
measures, respectively; df = 1/112). ment was employed, however, 1.0-gm.
Since the effects of percentage rein- reward led to kss resistance to extinc-
forcement were reliably different when tion than did .08-gm. reward. The
combined with 1.0-gm. reward than latter finding was significant in the
with ,08-gm. reward, separate F tests case of the goalbox measure (F=5.71,
were computed of the differences df = 1/112, P < .025) and the run-
between the 50% and 100% treatment ning measure (F = 4.83, df = 1/112,
groups at the two levels of reward P < .025) but not in the starting
magnitude. The results of these measure (F = 1.87, df = 1/112,
comparisons for the starting and .10 < P < .20).
running measures indicated that the Statistical analyses revealed that
50% groups were significantly more an increase in the number of acquisi-
resistant to extinction than the 100% tion trials led to a reliable increase
groups when combined with 1.0-gm. in the extinction proportions only in
reward (F = 43.54, and F = 47.33, the case of those treatment groups
respectively; df = 1/112; P < .001 which received .08-gm. reward, and
in each case) but not when com- then only on the starting and running
bined with ,08-gm. reward (F < 1 and measures (F = 22.41, P < .001, and
F = 1.97, respectively; df = 1/112). F = 5.72, P < .025, respectively;
In the goalbox measure, however, df — 1/112) there being no reliable
percentage reinforcement was sta- effect of number of acquisition trials
tistically reliable when combined with upon the goalbox measure. Although
either 1.0-gm. reward (F — 59.65, number of acquisition trials seldom
df = 1/112, P < .001) or with .08- affected the extinction proportions it
gm. reward (F = 21.88, df = 1/112, would still be expected to affect extinc-
P < .001). tion speeds, since Ss began extinction
240 ALLAN R. WAGNER

at a higher level following 60 than on all response measures. If one makes

following 16 acquisition trials. When the special assumption with Beier (1958)
analyses of variance were performed that rg — sa under inconsistent reward,
on the untransf ormed extinction speeds approaches the mean of the ra — sts
significantly superior extinction per- appropriate to the several rewards re-
ceived, then rg — sa under random 50%
formance of those groups receiving reinforcement approaches 50% of the
60 acquisition trials, as compared level attained under 100% reinforcement
to those receiving only 16 acquisition and hence a greater partial reinforce-
trials, was obtained in all comparisons. ment decrement should be obtained with
large than with small reward. The
DISCUSSION present results are in this direction but
do not attain significance. The fact
Beier (1958) and Logan (1960) have that a similar nonsignificant tendency
assumed that under conditions of incon- was observed by Hulse (1957) in the
sistent reinforcement, rg — sg is raised only previous attempt at such a com-
or lowered on each trial toward the value parison, however, suggests that it may be
appropriate for the reward received on a reproducible effect.
that trial. Several other investigators The latter, postfrustration perform-
(e.g., Amsel, 1958; Spence, 1960, Ch. 6) ance is less determinant, depending
have asserted, in addition to their indi- upon the relative degree of r, — se
vidual assumptions concerning the course superiority of the continuous groups as
of ra — sg, that after some minimal num- compared to the r/ — Sf induced drive
ber of reinforcements, nonreinforcement level superiority of the partial groups,
will elicit a primary motivational re- as well as upon the degree of conditioning
sponse termed frustration, which is of Sf to the approach response under
directly related in intensity to the strength partial reinforcement. Since random
of ra — v An anticipatory or condi- partial reinforcement leads commonly
tioned form (>•/) of this emotional to greater asymptotic starting and
response is further assumed to generalize running speeds than does continuous
to stimuli which antedate the goal, and reinforcement, it must be argued not
is expected not only to lead to an in- only that the approach response has
crease in drive level, but also, through become effectively conditioned to the
its characteristic response produced stim- early alley cues plus s/, but also that the
uli (sf), to tend to elicit unlearned or drive level advantage of the partial
previously learned responses which are groups is greater than the rg — s,
antagonistic to the approach response. advantage of the continuous groups.3
It is assumed however, that during 3
random partial reinforcement training That the partial superiority is not found
that these cues may become conditioned on those response measures close to the goal
may be attributed to less effective condition-
to, and tend to elicit the locomotor ing of the approach response to the goalbox
response itself. cues plus Sf, and hence the persistance to
While these positions when taken some degree of competing responses elicited
together lack in many instances the by Sf. This argument appears quite reason-
degree of development necessary for able in the context of the prior assumptions
unequivocal deductions, they do allow if it is noted that r/ — s/ may be expected
for a consistent interpretation of a size- to become conditioned to the approach re-
able number of partial reinforcement sponse when the Sf cues are introduced
initially at a weak value, and hence with a
phenomena. For example, during the negligible tendency to elicit competing re-
early, prefrustration acquisition trials, sponses, and are then increased gradually
partial reinforcement should lead to a at the same time that the approach response
lower rg — sg than should continuous is being strengthened. This condition ob-
reinforcement and hence, as presently tains much more clearly in the early portions
obtained, a lower level of responding of the alley than in the goal region where
 CONDITIONING AND EXTINCTION 241

Since r/ — Sf is presumed to be a On both theoretical (Amsel, 1958) and

function of rg — sa, increasing the magni- empirical (Amsel, 1958; Koehler, 1956;
tude of reward should increase the partial Weinstock, 1957) grounds it was ex-
reinforcement drive level superiority. pected that a greater partial reinforce-
The present finding of no significant ment extinction superiority would be
asymptotic interaction between rein- observed following 60 than following 16
forcement schedule and reward magni- acquisition trials. In order for the
tude might be interpreted as casting present failure to observe such a tend-
doubt on the frustration interpretation, ency to be considered consistent with
or as lending additional plausability to frustration theory it must be proposed
the notion that the continuous rein- that Sf had already become effectively
forcement ra — sa advantage is also conditioned to running in the partially
increased with increased reward. How- reinforced groups by the end of 16 trials.
ever, that there was observed a sizeable The present results would more clearly
and consistent tendency for the starting appear to place restrictions on the
and running partial reinforcement asymp- assumptions that may be made concern-
totic superiority to be greater with large ing the nature of the contributions of
than with small rewards appears to number of acquisition trials and magni-
justify further study. tude of reward to the partial reinforce-
It follows most clearly from the theo- ment extinction effect. While it may be
retical formulation, that partially rein- assumed that an increase in the number
forced 5s that have been trained to of acquisition trials under certain con-
approach in the presence of frustration ditions may lead to an increase in the
stimuli will be more resistant to extinc- magnitude of the partial effect, it must
tion than will continuously reinforced also be assumed that the function ex-
5s that have competing responses pressing this relationship has a limiting
elicited to a greater degree by the value determined by the magnitude of
anticipatory frustration which develops reward employed.
during extinction. Furthermore, the While it is not unlikely that both
present finding, as well as that of Hulse frustration and incentive motivation are
(1958), of a greater partial reinforcement important determiners of partial rein-
extinction superiority following large forcement phenomena, the present for-
than following small rewards is not mulations are necessarily speculative.
unexpected. Nor is the observation in Further clarification awaits less equivocal
each of these studies, that this increased evidence of the relative values of these
effect is due, in part, to faster extinction factors in the various portions of training
following continuous reinforcement with and at different segments of the response
large as compared to small rewards. chain.
Since the magnitude of ra — sg would be
greater at the start of extinction with SUMMARY
large than with small rewards, it is The acquisition and extinction of loco-
consistent that larger rewards should motor behavior in various segments of a
lead to a slower rate of extinction for par- straight alley were investigated as a function
tial 5s due to increased motivation, but of percentage reinforcement (50% or 100%),
should lead to a faster rate of extinction magnitude of reward (.08 gm. or 1.0 gm.),
for continuous 5s owing to the greater and number of acquisition trials (16 or 60).
vigor of the competing responses elicited Sixteen naive hooded rats were assigned to
by the more intense Sf. each of the eight groups in a 2 X 2 X 2
experimental design. A 24-hr, intertrial
due to the proximity of the primary frustra- interval was observed during acquisition and
tion event, generalized r/ — s/ may be ex- during the 33 extinction trials which followed.
pected to occur earlier and to follow a course The relative acquisition performance levels
of greater intensity, and thus produce re- of partially and continuously reinforced 5s
sponses which compete more effectively with were found to depend upon the portion of
the approach response. training and the segment of the response
242 ALLAN R. WAGNER

chain that was observed. When the response BEIER, E. M. Effects of trial-to-trial varia-
measure was taken over the early segments tion in magnitude of reward upon an
(first and second 6 in.) of the 5-ft. alley, there instrumental running response. Unpub-
was observed an initial superiority of the lished doctoral dissertation, Yale Univer-
100% 5s but an asymptotic superiority of the sity, 1958.
50% 5s. When the response measure was GOODRICH, K. P. Performance in different
taken over the final 6 in. preceding the goal segments of an instrumental response chain
cup the superiority of the 100% 5s persisted as a function of reinforcement schedule.
over the entire course of acquisition. Large /. exp. Psychol, 1959, 57, 57-63,
as compared to small rewards, in addition HAGGARD, D. F. Acquisition and extinction
to producing higher overall response speeds, of a simple running response as a function
also tended to produce a greater early trials of partial and continuous schedules of rein-
50% decrement and a greater asymptotic forcement. Unpublished doctoral disserta-
50% superiority, although no statistical tion, State University of Iowa, 1956.
support for the latter two tendencies was HULSE, S. H. Schedule of reinforcement,
obtained. amount of reinforcement, and duration of
Of primary interest during extinction were goal confinement as variables in condi-
the effects of size of reward and number of tioning and extinction. Unpublished doc-
acquisition trials on the magnitude of the toral dissertation, Brown University, 1957.
commonly obtained superiority of partially HULSE, S. H. Amount and percentage of
reinforced as compared to continuously rein- reinforcement and duration of goal con-
forced 5s. The extinction results showed finement in conditioning and extinction.
that whereas the "partial reinforcement /. exp. Psychol., 1958, 56, 48-57.
effect" did not vary with number of prior KOEHLER, J., JR. Acquisition and extinction
acquisition trials, it was markedly greater of a running response as a function of the
following 1.0-gm. than following .08-gm. percentage of reinforcement and the num-
rewards. The increased partial reinforcement ber of acquisition trials. Unpublished
effect with larger rewards reflected not only doctoral dissertation, Tulane University,
greater resistance to extinction of partially 1956.
reinforced 5s but also less resistance to ex- LOGAN, F. A. Incentive. New Haven: Yale
tinction of continuously reinforced 5s with Univer. Press, 1960.
large as compared to small rewards.. SPENCE, K. W. Behavior theory and learning:
The acquisition and extinction results Selected papers. Englewood Cliffs, N. J.:
were considered in relation to frustration Prentice-Hall, 1960.
and incentive motivation theory. WEINSTOCK, M. B. A factorial study of some
variables affecting resistance to extinction
REFERENCES under partial reinforcement with spaced
trials. Unpublished doctoral dissertation,
AMSEL, A. The role of frustrative nonreward Indiana University, 1957.
in noncontinuous reward situations. Psy-
chol. Bull, 1958,55, 102-119. (Received June 11, 1960)

View publication stats