Journal of Experimental Psychology: Copyright 1982 by the American Psychological Association, Inc.

Animal Behavior Processes 0097-7403/82/0802-0131S00.75

1982, Vol. 8, No. 2, 131-141

Effect of a Stimulus Intervening Between CS

and US in Autoshaping
Robert A. Rescorla
Yale University

Five experiments explored the effect of placing a stimulus between the termi-
nation of a key light CS and the onset of a food US in an autoshaping preparation
with pigeon subjects. Whether that stimulus was itself a key light evoking pecking
or an auditory event failing to evoke pecking, it facilitated performance to the
key light CS. The use of a within-subjects design made differential conditioning
of background cues and stimulus generalization unlikely accounts. Experiment
3 found that extinction of the intervening stimulus did not remove the facilitative
effect. Experiment 4 found that pairing the CS with food and the intervening
stimulus did not produce the effect unless both occurred on the same trial.
Together, these experiments rule out an account in terms of second-order con-
ditioning of the CS by the intervening stimulus. Experiment 5 used an intervening
stimulus in a second-order conditioning design to demonstrate that the stimulus
acted to improve the association between the CS and the reinforcer. That ex-
periment found the intervening stimulus to make responding to the CS more
vulnerable to subsequent extinction of its reinforcer. Together, these experiments
document a kind of catalytic effect of the intervening stimulus on the CS-re-
inforcer association.

It is well known that Pavlovian condition-
ing is best when the conditioned stimulus
(CS) is temporally proximal to the uncon-
ditioned stimulus (US; e.g., Gormezano &
Moore, 1969; Mackintosh, 1974; Newlin
& LoLordo, 1976). When a temporal gap
intervenes between the CS and the US, con-
ditioning is depressed. However, there have
been a number of recent reports that this
deficit can be attenuated if a second stimulus
occupies the time period between the CS and
the US (e.g., Bolles, Collier, Bouton, &
Marlin, 1978; Kehoe, Gibbs, Garcia, & Gor-
mezano, 1979; Wasserman, Carr, & Deich,
1978; Balsam & Gibbon, Note 1; Kaplan,
Bottjer, & Hearst, Note 2). The effect of
this intervening stimulus (IS) agrees both
with an older literature on delay of instru-
mental reinforcement (e.g., Grice, 1942) and
with the intuition that a filled time period
This research was supported by National Science
Foundation Grant BNS 78-02752. Thanks are due
Paula Durlach, Janis Marcus, and Allan Wagner for
many helpful comments.
Requests for reprints should be sent to Robert A.
Rescorla, who is now at the Department of Psychology,
University of Pennsylvania, 3815 Walnut Street, Phil-
adelphia, Pennsylvania 19104.
131
may bring the CS and US together percep-
tually (e.g., Wertheimer, 1955). However,
the manner in which this stimulus has its
effect is not well understood.
The purpose of the present series of ex-
periments was to provide both additional
documentation and a partial analysis of this
effect in an autoshaping preparation. One
may consider two broad classes of explana-
tion. One possibility is that the intervening
stimulus acts to improve the first-order as-
sociation between the CS and the US. Fa-
cilitation of that association might be ac-
complished, for instance, if the IS reduced
the amount of background conditioning that
an otherwise poorly signaled US accom-
plished. This might reduce the degree to
which a well-conditioned background stim-
ulus was present concurrently with the CS
and thus permit more conditioning of that
CS (Balsam & Gibbon, Note 1). Another
possibility is that the CS becomes directly
associated with the IS which in turn bears
a favorable relation to the US (Rashotte,
1981). This second-order conditioning ac-
count is analogous to the most popular ac-
count for the action of intervening stimuli
in instrumental learning (e.g., Spence, 1947).
132 ROBERT A. RESCORLA
Experiments 1 and 2 provided an initial
documentation of the effect of various types
of intervening stimulus in an autoshaping
preparation. Experiments 3 and 4 evaluated
the contribution of second-order condition-
ing to the superiority produced by the in-
tervening stimulus. Experiment 5 provided
evidence that the effect results, in part, from
a potentiated CS-reinforcer association.
Experiment 1
This experiment used a within-subjects
design to evaluate the effect of placing a
stimulus at various points between a key
light CS and a food US in autoshaping. Each
bird received four different colored key
lights, each terminating 10 sec before the
delivery of food. The key lights differed in
that either the entire period, the first half,
the second half, or none of the period con-
tained another stimulus.
In addition to providing increased sensi-
tivity, a within-subjects design allows an
evaluation of two initial interpretations that
can be applied to previous between-subjects
experiments. The first is that the condition-
ing of IS per se facilitates responding to the
CS, whatever its relation to the CS. This
might come about because of stimulus gen-
eralization or because responding to the IS
generally elevates responding in the situa-
tion. The second possibility is that the effect
of the intervening stimulus is mediated by
differential conditioning of background cues.
In a within-subjects design, however, the IS
signals the same proportion of USs in all
animals, but the CSs bear different temporal
relations to the IS. Hence differences in re-
sponding to those CSs cannot be attributed
to differences in the value of the background
against which they are trained or tested or
to differences in the potential for stimulus
generalization from the IS.
Method
Subjects and apparatus. The subjects were 16 fe-
male Carneaux pigeons about 1 yr old, maintained at
75% of their free-feeding weights. They had previously
participated in a second-order autoshaping experiment
that used a response key and stimuli that differed from
those employed here. Treatments in the present exper-
iment were arranged to be orthogonal to those experi-
mental histories.
The apparatus consisted of four identical operant
chambers, each measuring 27 X 27 X 35 cm. The metal,
front panel of each chamber had a 5 X 5 cm food mag-
azine in its center, located 5 cm above the wire mesh
floor. Three' response keys, 2.5 cm in diameter, were '
located one directly above the hopper and one on either
side of the center of the front wall, 20 cm above the
floor. Located behind the right-hand key was an IEE
inline projector that permitted the transillumination of
the key with color and orientation stimuli. The left-hand
half of that key could be illuminated with blue (B),
green (G), red (R), or yellow (Y) light, at which time
the other half of the key remained black. In addition,
the entire key could be illuminated by a uniform white
(W) stimulus. These stimuli were generated by Ekta-
chrome slides of drawings composed of "Color-aid" ar-
tists paper. The remaining walls and ceiling of the cham-
bers were composed of clear Plexiglas.
These chambers were placed in sound- and light-at-
tenuating shells, with ventilation fans providing back-
ground noise of 62 dB (re 20 MN/m 2 ), On the rear wall
of those shells was mounted a 6-W bulb that was con-
tinuously illuminated during the session except during
the operation of the food hopper. The hopper was il-
luminated when it made available Purina Pigeon Grain.
Procedure. Because these birds had previously re-
ceived autoshaping in another experiment, no magazine
training was necessary. Each conditioning session con-
tained twelve 5-sec presentations of each color, each
presentation followed after a time interval of 10 sec by
5-sec availability of grain. The W stimulus filled the 10
sec between one color and food, occupied the initial 5
sec of that time for another color, occupied the terminal
5 sec for another color, and was not presented following
the fourth color. The birds were run in squads of four,
with the color assignments counterbalanced according
to a Latin square. The mean intertrial interval was 1
min. Conditioning continued for 28 sessions.
"/
Results
Figure 1 displays the mean response rates
during the various stimuli over the course of
conditioning. Although the levels of respond-
ing were low to all CSs, it is clear that re-
sponding developed most rapidly and to the
highest level in the CS for which the entire
time gap was filled. From the second 4-day
block onward, that CS evoked reliably
greater responding than did the CS without
any intervening stimulus (Wilcoxon Ts < 2,
p$ < .02). The two CSs with only part of the
interval filled gave intermediate results. Fill-
ing the first portion of the interval had the
greater effect early in Conditioning; only af-
ter the fourth block of days was that CS
inferior to the CS with the full interval filled
(Ts<4,ps< .05). The CS for which W oc-
cupied the last portion of the interval only
gradually developed more responding than
 INTERVENING STIMULUS
the unfilled CS. On the final block of trials,
the two CSs with partially filled intervals
were not different but were both superior to
the CS with its interval unfilled and inferior
to the CS with its full interval filled (Ts <
4, ps < .05).
Throughout the experiment, the most re-
sponding was observed to W, the filling stim-
ulus. This responding was not differential
depending upon its placement in the interval
and was at an overall mean of 68 responses
per minute.
Discussion
This experiment documents the facilita-
tive effect of an intervening stimulus on re-
sponding to a trape CS. It suggests that per-
formance is greater the more of the trace
interval is occupied by the intervening stim-
ulus. It further suggests that the initial por-
tion of the interval is the most important to
fill, especially early in the course of condi-
tioning.
These results appear to rule out differ-
ential conditioning of background stimuli
and simple stimulus generalization as inter-
pretations of the facilitative effect of the in-
tervening stimulus. However, they leave
open a number of alternative accounts. It
remains possible, for instance, that the in-
tervening stimulus has its effect primarily
by producing second-order conditioning of
the target CS. This interpretation is en-
couraged by the observation that more fa-
cilitation was produced by presenting the
intervening stimulus immediately after the
CS than by an equivalent duration presen-
tation later in the interval. Experiments 3
and 4 provide evidence on various aspects of
this second-order conditioning account.
The particular nature of the intervening
stimulus used here suggests, however, an-
other account. During an empty trace inter-
val, the animals presumably engage in var-
ious behaviors other than key pecking; for
instance, it seems likely that the animals
show approach to the food magazine during
this time. Many of these behaviors may be
incompatible with key pecking. Moreover,
since they immediately follow presentation
of the trace CS, these behaviors may become
conditioned to that CS. The role of the in*-
133
Ul 20
z>
I
15
,0
I0
2
</)
UJ
I 2 3 4 5 6
BLOCKS OF 4 DAYS
Figure 1. Mean responses per minute during four stimuli
all receiving a 10-sec trace relation to the food US. (All
[•], none [O], the first 5 sec [0], or the second 5 sec
[O] of the trace was filled with another stimulus.)
tervening key light might then be to prevent
the occurrence of such incompatible re-
sponses and/to protect the CS from becoming
associated with them. In order to evaluate
this possibility, Experiment 2 replicated the
basic result of Experiment 1 with the use of
an intervening stimulus that does not evoke
key pecking.
Experiment 2
When a pigeon receives an auditory CS
signaling food, it learns that relation, but it
does not exhibit the learning in the form of
key pecking (e.g., Nairne & Rescorla, 1981).
Informal observations in this laboratory sug-
gest that auditory signals may engender a
variety of behaviors, such as magazine ap-
proach, head bobbing, and pecking at the
floor. Consequently, in this experiment a
tone was used as the stimulus intervening
between a key light and food, so as to fill the
interval with a detectable stimulus that does
not evoke the same response as the key light.
In this and all subsequent experiments, the
trace interval was shortened to 5 sec to fa-
cilitate acquisition.
Method v
Subjects and apparatus. The subjects were 16 fe-
male Carneaux pigeons which had previously partici-
pated in another autoshaping experiment with different
stimuli. Their assignment to conditions in the present
experiment was orthogonal to their previous experimen-
134 ROBERT A. RESCORLA
UJ 40 •-
30
20
I
O
Q.
tn
UJ
cr 10
u auditory stimulus and an intervening visual
1 2 3 4 5
DAYS
Figure 2. Mean responses per minute during a key light
with a 5-sec trace relation to the food US. (The 5-sec
period was filled with a tone [•] or not [O].)
tal histories. They were maintained at 75% of their ad
lib body weights throughout the experiment. The ap-
paratus was a set of eight Skinner boxes identical in all
essentials to those of Experiment 1. The apparatus per-
mitted the presentation of the R and G stimuli used in
Experiment 1 as well as an 1800-Hz tone of 76 dB.
Procedure, In order to ensure low levels of initial
responding, all birds were first given four sessions in
which the R and G key lights and the tone (T) were
each presented 12 times for 5 sec without consequence.
On each of the next 5 days, all animals received 12
presentations each of R and G, followed after a S-sec
gap by 5 sec of food. For half of the animals, T inter-
vened between R and food, and for half it intervened
between G and food. The intertrial interval was 1 min.
Results and Discussion
The mean rates of responding to the key
light CSs on the last pretest day were less
than one per minute. Figure 2 shows the
mean rates of responding over the course of
conditioning. Although the rates of respond-
ing were similar on the first conditioning
day, the key light followed by the tone rap-
idly came to be pecked at a higher rate. By
the fifth day of conditioning, that difference
was highly reliable, T(\ 1) = 4,p < .02. Key
pecking during the tone occurred at a rate
of less than one response per minute.
These results indicate that the stimulus
intervening between a key light and food
need not evoke pecking in order to facilitate
responding to that key light. This suggests
that the role of the intervening stimulus is
not to prevent the development of anticipa-
tory responses that are incompatible with
key pecking. In fact, informal observations
suggested that the tone regularly evoked
such behaviors during the trace interval.
The rapid acquisition observed in this ex-
periment compared with the acquisition in
Experiment 1 is likely attributable to the
simpler procedure and the shorter trace in-
terval. This experiment was not designed, of
course, to evaluate the relative magnitude
of the effects produced by an intervening
stimulus, but it does indicate that a sub-
stantial effect can be obtained even when an
auditory stimulus is used. These results are
in agreement with those reported by Balsam
and Gibbon (Note 1) and by Kaplan, Bottjer,
and Hearst (Note 2).
Experiment 3
A highly plausible interpretation of the
effects observed in Experiments 1 and 2 is
that the intervening stimulus second-order
conditions the target CS. Previous experi-
ments (e.g., Nairne & Rescorla, 1981; Res-
corla, 1979) provide ample evidence that
both auditory and visual stimuli that have
been paired with a food US can produce sub-
stantial second-order conditioning of key
lights that they themselves follow. There are
several different forms that such second-or-
der conditioning could take. One possibility
is that a stimulus-stimulus (S-S) association
is formed between the CS and the IS. An-
other is that a stimulus-response (S-R) as-
sociation is formed between, the CS and some
emotional response evoked by the IS. The
available evidence suggests that in auto-
shaping in which both stimuli are key lights,
second-order conditioning is likely to be of
the former character. One relevant piece of
evidence is that after second-order auto-
shaping, extinction of the first-order stimu-
lus dramatically attenuates responding to
the second-order stimulus with which it had
been paired (e.g., Rescorla, 1979). Conse-
quently, when the stimulus intervening be-
tween a key light and food is another key
light, it is reasonable to anticipate that any
of its facilitative effect that is attributable
to second-order conditioning could be re-
 INTERVENING STIMULUS
duced by extinction of the intervening stim-
ulus.
Experiment 3 used that logic to evaluate
the contribution of second-order condition-
ing. Two target CSs were each presented in
a trace relation to food. A different inter-
vening stimulus was used for each CS. Then
one of those intervening stimuli was extin-
guished, and the target CSs were retested.
To the degree that the facilitative effect is
due to second-order conditioning, one would
expect it to be reduced in the CS for which
the intervening stimulus had been extin-
guished. For comparison purposes, each an-
imal also received a third CS without any
event intervening before reinforcement.
Method
Subjects and apparatus. The subjects were 16 fe-
male Carneaux pigeons with histories similar to those
in the previous experiments, maintained in the same
fashion. The apparatus was that of Experiment 2 except
that the right-hand response key was used. A projector
permitted the presentation of two colors, blue (B) and
yellow (Y), each of which was projected on the bottom
half of the key, and of two grid patterns each of which
could be projected on the top half of the key. One grid
consisted of 1-mm black lines spaced 2 mm apart on a
white background-and oriented 45° from the vertical;
the other grid consisted of 1-mm black lines spaced 1
mm apart on a white background and oriented -45°
from the vertical. In addition, the top half of the key
could be illuminated by the white background alone.
When either half of the key was illuminated, the other
half remained black. These stimuli are all readily dis-
criminable from one another and are regularly used in
this laboratory as CSs and reinforcers in second-order
conditioning experiments.
Procedure. In order to ensure low levels of initial
responding, all animals first received four sessions dur-
ing each of which they received twelve 5-sec nonrein-
forced presentations each of the five stimuli to be used
in the study, blue, yellow, the two grids, and white. On
each of the next 4 days, all birds received 12 reinforced
presentations each of the two grids and white. The stim-
uli were 5 sec long and were followed by food 5 sec after
their termination. Blue intervened between one of the
stimuli and food, and yellow intervened between a sec-
ond and food; the third stimulus had no intervening
event. Animals were assigned to1 groups so as to balance
across the white and the two grid stimuli whether or not
the stimulus would be followed by a color and if so which
color.
Over the next 4 days, all animals received a treatment
designed to extinguish responding to one of the colors
but not to the other. Half of the animals received on
each day 6 reinforced blue and 24 nonreinforced yellow
presentations. The other half of the animals received the
same treatment with blue and yellow interchanged.
On the next day, all animals first received continued
135
l 80
60
£ 20
PRE 1 2 3 4 + - C
TEST CONDITIONING TEST
(DAYS)
Figure 3. Mean responses per minute during two key
lights with a 5-sec trace relation to the food US. (For
one stimulus [•], that trace was filled with another key
light [A]; for another [O], it was not. To the right is
responding to the CS with a filled interval after the filler
had been extinguished [-] or not [+] as well as re-
sponding to the CS conditioned without the filler [O].
color discrimination for 3 reinforced and 12 nonrein-
forced trials; they then received 12 nonreinforced test
presentations each of the white background and the
grids. The issue was whether extinction of a color would
reduce responding to the stimulus for which it had
served as the intervenor before food.
Results and Discussion
Figure 3 shows the mean responding dur-
ing the various CSs and to the IS over the
course of conditioning and during the final
test. The data were collapsed over stimuli to
show responding during the patterns that
had been followed or not by a color as well
as during the colors. To the far left are the
rates of responding on the final preexposure
day. These were, of course, quite low. The
middle panel shows the course of condition-
ing, which was similar to that of the previous
experiment. Responding was greater to those
stimuli (two per animal) that had colors fol-
lowing them than to the stimulus that did
not, 7(16) = 0, p < .01. As in Experiment
1, responding was greatest of all to the filling
color, indicated in the figure by triangles.
The subsequent discrimination between
colors was rapidly learned. By the final day
of discrimination, the mean response rate
during the reinforced and nonreinforced col-
ors were 99 and 5 responses per minute, re-
spectively. The data of most interest, from
the final test of the patterns, are shown in
the final panel of Figure 3. The data are
136 ROBERT A. RE&CORLA
Table 1
Design of Experiment 4
A X US
A to take the form of an association between
B US
B X
Note. All animals received all four trial types. A, B, and
X are key lights; the US is food.
separated according to whether the color
that followed the pattern had been rein-
forced (+) or extinguished (-) or whether
no color had followed the pattern during
training (C). Both pattern stimuli that had
previously been followed by colors continued
to show greater responding than did the un-
filled pattern, whether or not the color had
been extinguished (Ts < 17, p < .02). But
there was no reliable difference in respond-
ing between the two patterns that previously
had been followed by colors.
These data suggest that there was little
learning of an association between the ori-
entations and colors which was reflected in
responding to those orientations. Although
that sort of learning has been documented
to occur in second-order conditioning exper-
iments with these stimuli (e.g., Rescorla,
1979; Rescorla & Durlach, in press), it ap-
parently is not substantial in sequential pro-
cedures that terminate in a US. This finding
is in agreement with results recently re-
ported by Holland (1980) for an activity
conditioning preparation in rat subjects. In
any case, associations between the orienta-
tions and colors were clearly not substantial
enough to account for the facilitative effect
of the intervening stimulus.
It may be noted that this and the previous
experiments used experimentally experi-
enced birds. Previous experience may change
the general level of performance, but ar-
ranging for the current treatments to be or-
thogonal to the earlier experiences makes
unlikely any simple interaction with the ef-
fects observed here. It is of particular rele-
vance to note that in this laboratory we
regularly observe substantial second-order
conditioning of an S-S variety in birds, with
various experimental histories. / filled condition, then the two CSs should
However, the absence of S-S learning in
this experiment need not mean that there
was no contribution of any form of second-
order conditioning to the present effect. Sec-
ond-order conditioning is known sometimes
one stimulus and the response evoked by a
second. We do not have a full understanding
of when second-order conditioning will take
an S-S or an S-R form, but we do know
that even in autoshaping it is sometimes S-
R in nature (Nairne & Rescorla, 1981).
Under those circumstances, one would not
expect reduction in the value of the color to
adversely affect responding to the pattern
despite the fact that the color acted to pro-
duce second-order conditioning of the pat-
tern. Consequently, Experiment 4 provides
an alternative evaluation of the second-order
possibility.
v
Experiment 4
This experiment attempted to generate
the effect of an intervening stimulus simply
by combining the effects on a target CS of
second-order conditioning from an IS with
those of first-order conditioning from the
ultimate reinforcer. A schematic of the de-
sign is shown in Table 1. The logic was to
use two target CSs (A and B), each followed
after a time period by the US on half of their
occurrences. In addition, each CS was fol-
lowed by the same intervening stimulus (X)
on half of its occurrences. The CSs differed
in that for Stimulus A, X and the US oc-
curred on the same trial (on the other trials
the A was presented alone), whereas for
Stimulus B, X and the US occurred on dif-
ferent trials. The result was that both CSs
were followed the same number of times by
the same IS and by the same US but that
they differed in the degree to which the IS
filled the gap between the CS and the US.
Consequently, the IS had the same oppor-
tunity to condition both CSs but different
opportunities to fill the time gap. Similarly,
the US had the same opportunity to condi-
tion both CSs but had the preceding gap
filled only for one. Consequently, if only the
addition of first- or second-order condition-
ing was generating greater responding in the
show similar performance. If they differ, this
would suggest that there is something im-
 INTERVENING STIMULUS 137

portant about the IS and US occurring on

the same trial which goes beyond the action
of independent processes.
Method
Subjects and apparatus. The subjects were 16 fe-
male Carneaux pigeons with histories similar to those
of the previous experiments. The apparatus was that of
Experiment 1, modified so that the right-hand key could I
ui
be illuminated by a green (G), yellow (Y),< or white (W) 2 3
light. DAYS
Procedure. On the first day of the experiment, all
animals were preexposed to the stimuli, receiving 12 Figure 4. Mean responses per minute to a key light
nonreinforced presentations each of G and Y and 24 followed on the same trial by a filler and a food US
nonreinforced presentations of W. All stimuli were 5 sec (•) or followed by those events on separate trials (O).
long and were delivered at a mean intertrial interval of
1 min.
directly as a reinforcer for it but rather by
For the next 4 days, the animals were divided into
somehow facilitating its association with the
two groups both of which received autoshaping. One
group of eight birds received 12 trials on which G was
ultimate US. Such a facilitation would pre-
followed immediately by W and then food, 12 trials on
sumably best take place when both the IS
which G was presented without consequence, 12 trials
and the US occur on the same trial.
on which Y was followed by W but no food was given,
These results also speak against an ac-
and 12 trials on which Y was followed by a S-sec gap
count in which the gap filler slows the de-
that terminated in food. The result was that G and Y
were both followed by food on half of the trials and by
velopment of latent inhibition to the CS. It
W on half of the trials; but for G, the W and food were
is possible that presenting the CS in a trace
positively correlated, and for Y they were negatively
relation to the US allows a reduction in CS
correlated. The other group of eight birds received the
salience and hence in its conditionability.
same treatments with the roles of G and Y interchanged,
Following the CS by another might, how-
Results and Discussion ever, retard that process (Lubow, Schnur,
& Rifkin, 1976) and result in the present
Prior to conditioning, the mean rates of effect. However, in this experiment both CSs
responding to G and Y were one and two were followed with equal frequency by the
responses per minute, respectively. Figure 4 same stimulus and hence should not differ
shows the results of principal interest, the
mean rates of responding to the colors during in Itthisshould
regard.
be noted that certain current
conditioning. The results were collapsed notions of conditioning would have antici-
across the particular stimulus identities and pated a quite different outcome for this ex-
separated according to whether W filled the periment. To the degree that the adminis-
gap before food or occurred on other trials. tration of the US interferes with the ability
It is clear from the figure that only the stim- of the IS to produce second-order condition-
ulus for which W actually filled the gap ing, the CS followed by W and food on the
showed a high level of responding. By the
final day of conditioning, the difference be- same trial should have been inferior (e.g.,
Cheatle & Rudy, 1978; Holland, 1980).
tween the two stimuli was highly reliable, Apparently, the facilitatiye impact of the IS
7X16) = 2, p< .02.
These results do not support a second-or- on the CS-US association is substantial
enough to overcome the loss of second-order
der conditioning interpretation. Whether the
form of that second-order conditioning is S- conditioning elsewhere observed when the
US occurs orrthe same trial as an effective
S or S-R, we would have expected substan- first-order CS.
tially more responding than was actually
observed to the stimulus for which W and Experiment 5
food occurred on separate trials. Thus these
results suggest the possibility that the IS fa- Experiments 3 and 4 present negative ev-
cilitates responding to the CS not by serving idence on a second-order conditioning inter-
138 ROBERT A. RESCORLA
Table 2
Design of Experiment 5
n
TT
A.
V reinforced presentations each of B and X and 24 non-
nJ3
G X
n
¥J
D|J
Note. All animals received all four trial types. R, G, B,
and Y are key colors; H is a horizontal grid.
pretation and by inference support the pos-
sibility that the IS serves as a kind of catalyst
to promote the action of the reinforcer. Ex-
periment 5 attempted to provide more direct
evidence on this possibility. It used the find-
ing, referred to above, that in second-order
autoshaping one can identify an association
between an S2 and the SI with which it has
been paired by measuring the change in re-
sponding induced to S2 when SI is extin-
guished. In order to exploit that fact, an in-
tervening stimulus was used to fill the time
interval between an S2 and one SI but not
between that S2 and another SI. Then one
of the Sis was extinguished, but the other
was not. The expectation was that if S2 is
better associated with the SI for which its
gap had been filled, then extinction of that
SI should especially depress responding
toS2.
Method
Subjects and apparatus. The subjects were 16 -naive
female Carneaux pigeons about 6 mo old at the start
of the experiment. They were maintained at 75% of their
free-feeding weight throughout the experiment. The
apparatus was that of Experiment 2- In this experiment,
the left-hand key was activated. A projector located
behind that key permitted the presentation of red (R)
or green (G) on the left-hand side, a grid of 1-mm black
horizontal lines spaced 1 mm apart on a white back-
ground (H) on the right-hand side, or the full illumi-
nation of the key with either blue (B) light or an X
composed of 1-mm black lines on a white background.
Procedure. The animals were initially trained to eat
food from the magazine. On the first day, they were
placed in the chambers, and the food hoppers were ac-
tivated until the birds had eaten for about 15 sec. Then
the hopper presentations were progressively shortened
and given less frequently until the birds consistently ate
during a 5-sec presentation. On each of the next 3 days,
the animals were given 44 hopper deliveries spaced at
a mean intertrial interval of 1 min.
Table 2 illustrates the remainder of the experiment
for half of the color counterbalancing. The birds were
given first-order conditioning designed to produce au-
toshaped responding to X and B while yielding low rates
of responding to G, R, and H. On the first 3 days of
this stage, all birds received 30 reinforced presentations
each of X and B. On the next 2 days, they received 12
reinforced presentations of H. Finally, all birds received
2 days on which they received 16 presentations each of
X and B, 12 presentations of H, and 6 presentations
each of R and G. On those days, H, G, and R were
never reinforced, but 4 of the 16 presentations of X and
B were followed by food. This partial reinforcement of
X and B was designed to match the frequencies of re-
inforcement they would be given during their subsequent
use as reinforcers in the second-order phase of the ex-
periment.
On each of the next 4 days, B and X were used as
reinforcers to second-order condition R and G; H was
employed as the intervening stimulus. All animals re-
ceived three second-order trials with each of four pair-
ings: R-X, R-B, G-X, and G-B. In each case, stimulus
presentations were for 5 sec, and 5 sec intervened be-
tween stimuli. For half of the animals, H filled the time
interval between R and X and the time interval between
G and B; for the other half of the animals, H filled the
gap between R and B and that between G and X. In
addition, all animals received two separate food-rein-
forced presentations each of B and X, to maintain their
level of first-order conditioning. During this phase of the
experiment, the intertrial interval was 2 min.
The consequence of these treatments was that if H
facilitates the formation of an association between S2
and SI, then for half of the animals R should be es-
pecially well associated with X, and G with B. Notice,
however, that the frequencies of presentation of the in-
dividual events and the pairings of R and G with B, X,
and H were all matched, so as to make possible the
inference that any observed effects are due to the cor-
relation of H with the occurrence of a particular pairing.
On each of the next 7 days, the animals were given
discriminative first-order conditioning with B and X. In
each of the two groups, half of the animals received 6
reinforced presentations of B and 24 nonreinforced pre-
sentations of X on each day; the other half had the roles
of B and X interchanged. Finally, two test sessions were
given; The first half of each contained continued first-
order discrimination for 3 reinforced and 12 nonrein-
forced trials; the second half contained 6 nonreinforced
presentations each of R and of G. The expectation was
that if, say, R is especially well associated with X and
G is well associated with B, then R should be more
sensitive than is G to an extinction treatment of X.
Results and Discussion
First-order conditioning of B and X pro-
ceeded rapidly, as did the discrimination of
G, R, and H. By the last day preceding sec-
ond-order conditioning, the mean responses
per minute for X and B were 83 and 80,
whereas those for G, R, and H were 6, 7,
and 1, respectively. The course of second-
order conditioning of R and G is shown on
the left-hand side of Figure 5, which presents
 INTERVENING STIMULUS
the mean number of trials with a response,
a measure that I find more stable in describ-
ing the results of second-order conditioning
experiments. By the end of second-order con-
ditioning, H had attained a percentage of
trials with a response of 80. Throughout sec-
ond-order conditioning, responding to B and
X remained high and similar on the various
trial types.
The subsequent discrimination between B
and X was reasonably rapid. By the final day
of this phase, the mean responses per minute
for the reinforced and nonreinforced first-
order stimuli were 100 and 7, respectively.
Responses occurred on 85% and 3% of the
trials, respectively. The results of most in-
terest, from the final test sessions with G and
R, are shown on the right-hand side of Fig-
ure 5. These results are separated according
to whether the first-order stimulus with
which the color had previously had a filled
relation was subsequently reinforced (+) or
extinguished (-). Since both R and G had
been paired with each first-order stimulus
but only one with a filled interval, each an-
imal contributed data to each type of stim-
ulqs. These results show a greater level of
responding to the S2 whose filled SI had
been reinforced and whose nonfilled SI had
been extinguished than to the S2 with the
converse treatment, T(10) = 3,/j < .02. This
suggests that filling the interval actually im-
proved the strength of the association be-
tween an S2 and an SI.
It may especially be noted that since H
had been equally paired with R and (fJ as
well as with X and B, it would not be possible
for changes in the value of H to mediate
differential responding to R and G.
General Discussion
This series of experiments makes it clear
that the deficit produced by a time interval
between a CS and a US can be attenuated
if another stimulus fills that interval. More-
over, that attenuation does not appear to be
attributable to a variety of plausible inter-
pretations. It can be demonstrated under
circumstances in which neither differential
background conditioning, differential oppor-
tunity for competing responses, nor differ-
ential stimulus generalization from the filler
139
MOO
BO
F+
F-
20
2
DAYS TEST
Figure 5, Mean percentage of trials with a response to
a second-order CS paired in a trace relation with two
different first-order stimuli. (For one of those pairings,
another stimulus filled the trace interval. To the right
is responding to the second-order CS after the first-order
stimulus with which it had a filled interval received ei-
ther extinction [F-] or additional reinforcement [F+].)
to the target CS seems to be a potential ac-
count. Moreover, it is not fully due to the
filler's serving as a reinforcer for the CS in
the manner of second-order conditioning.
Apparently, the filler serves a kind of cata-
lytic function actually promoting the asso-
ciation between the CS and the reinforcer.
This conclusion need not mean that sec-
ond-order conditioning never contributes to
the effect of an intervening stimulus in trace
conditioning. The results of Experiments 3
and 4 suggest that second-order conditioning
is not a complete account. But the finding
of Experiment 1, that filling the early portion
of the interval produces the best effect, sug-
gests that there may be some contribution
of second-order conditioning, especially early
in conditioning.
It is of interest to compare the effects of
filling a trace interval not only with the re-
sults of simple trace conditioning but also
with those of a delayed conditioning proce-
dure. Indeed, one way to think of delayed
conditioning is as a variation of the present
procedures in which the intervening stimulus
is maximally similar to the target stimulus.
Although the present experiments have not
explored this case, other evidence (e.g.,
Newlin & LoLordo, 1976) suggests that a
delayed procedure generates responding dur-
140 ROBERT A. RESCORLA

ing the interval which is superior to that ob- ditional stimuli occur when the stimuli are
served with a filled trace procedure. A nat- present concurrently with the CS whereas
ural explanation can be given in terms of the present effects were demonstrated with
stimulus generalization, although it is pos- sequential presentation might encourage that
sible that an intervening stimulus that is speculation. Most theories of memory seem
highly similar to the target CS would serve to anticipate that the filler would harm
as a superior catalyst. rather than help the short-term memory for
The present findings extend the list of the CS (e.g., Wagner, in press). However,
ways in which one stimulus can influence it is possible that the filler somehow protects
conditioning of another in compound con- the memory of the CS from the interfering
ditioning situations. There is a good deal of effects of other stimuli that the organism
evidence that under some conditions a sec- might receive during the trace interval. Al-
ond stimulus can reduce conditioning of a ternatively, the filler might enhance the
CS paired with a reinforcer. This may occur power of the US. A US that is well signaled
either when the second stimulus is originally by the filler might be more effective than a
neutral (overshadowing) or when it has a US that is unsignaled. This interpretation
history of pretraining with the same US seems less plausible in the light of the small
(blocking). One interpretation of such find- facilitative effect in Experiment 1 of filling
ings is that two stimuli must compete for the the last portion of the interval. Moreover,
association with the US (see Dickinsqn & current empirical and theoretical work on
Mackintosh, 1978; Rescorla & Holland, the effectiveness of USs anticipates the op-
1982). There is also evidence that two stim- posite conclusion (e.g., Rescorla & Holland,
uli in a compound become associated with 1982).
each other. Such associations may be either Consequently, at present the evidence al-
S-R (e.g., Rizley & Rescorla, 1972) or S- lows little more than the acknowledgment
S (e.g., Rescorla, 1979) in character.' At that these data agree with a kind of percep-
least one case of such within-compound as- tual intuition, that events which are bridged
sociations is responsible for a potentiation in time by a third event appear to go to-
of conditioning like that observed here. Dur- gether. It may be that this is analogous to
lach and Rescorla (1980) found that asso- another perceptual influence, that of simi-
ciations between odors and tastes were partly larity, which has received some additional
responsible for the enhanced aversion that analysis (Rescorla & Gillan,;1980).
a poison was capable of producing to an odor
when they were paired in the presence of a Reference Notes
taste. Finally, there is evidence that associ- 1. Balsam, P. D., & Gibbon, J. Associative factors un-
ations between the elements of a compound derlying trace decrements in autoshaping. Paper pre-
interact with those between the elements and sented at the meeting of the Eastern Psychological
the reinforcer. Under some circumstances Association, Washington, April 1978.
that interaction is such that it attenuates 2. Kaplan, P. S,, Bottjer, S. W., & Hearst, E. Filling
a gap in our understanding of trace conditioning:
both associations (e.g.,-Cheatle & Rudy, Discrimination of the intertrial interval from the
1978; Holland, 1980; Rescorla, 1981). "empty" period before VS. Paper presented at the
But the present data suggest that there meeting of the Eastern Psychological Association,
are also cases in which the presence of a Hartford, Connecticut, April 1980.
second stimulus can truly facilitate the as-
sociation between a stimulus and the US. References
There are several possible ways in which Bolles, R. C., Collier, A. C., Bouton, M. E., & Marlin,
such a facilitation could be accomplished. N. A. Some tricks for ameliorating the trace-condi-
The presence of the filling stimulus may tioning deficit. Bulletin of the Psychonomic Society,
somehow enhance the salience of the target 1978, ;/, 403-406.
CS. It might, for instance, better preserve Cheatle, M. D., & Rudy, J. W. Analysis of second-order
odor-aversion conditioning in neonatal rats: Impli-
it in a short-term memory so as to make it cations for Kamin's blocking effect. Journal of Ex-
more available for association with the US. perimental Psychology: Animal Behavior Processes,
The fact that most interfering effects of ad- 1978, 4, 237-249.
 INTERVENING STIMULUS
Dickinson, A., & Mackintosh, N. J. Classical condi-
tioning in animals. Annual Review of Psychology,
1978, 29, 587-612.
Durlach, P. J,, & Rescorla, R. A. Potentiation rather
than overshadowing in flavor-aversion learning: Ah
analysis in terms of within-compound associations.
Journal of Experimental Psychology: Animal Be-
havior Processes, 1980, 6, 175-187.
Gormezano, I., & Moore, J.-W. Classical conditioning:
Beyond the basic parameters. In M, H. Marx (Ed.),
Learning: Processes. New York; Macmillan, 1969,
Grice, G. R, An experimental study of the gradient of
reinforcement in maze learning. Journal of Experi-
mental Psychology, 1942, 30, 475-489.
Holland, P. C. Second-order conditioning with and with-
out unconditioned stimulus presentation. Journal of
Experimental Psychology: Animal Behavior Pro-
cesses, 1980, 6, 238-250.
Kehoe, E. J., Gibbs, C. M., Garcia, E., & Gormezano,
I. Associative transfer and stimulus selection in clas-
sical conditioning of the rabbit's nictitating mem-
brane response to serial compound CSs. Journal of
Experimental Psychology: Animal Behavior Pro-
cesses, 1979, 5, 1-18.
Lubow, R. E., Schnur, P., & Rifkin, B. Latent inhibition
and conditioned attention theory. Journal of Exper-
imental Psychology: Animal Behavior Processes,
1976, 2, 163-174.
Mackintosh, N, J. The psychology of animal learning,
New York: Academic Press, 1974.
Nairne, J. S., & Rescorla, R. A. Second-order condi-
tioning with diffuse auditory reinforcers in the pigeon.
Learning and Motivation, 1981, / 2, 65-91.
Newlin, R. J., & LoLordo, V. M. A comparison of peck-
ing generated by serial, delay, and trace autoshaping
procedures. Journal of the Experimental Analysis of
Behavior, 1976, 25, 227-241.
Rashotte, M. E. Second-order autoshaping: Contribu-
tions to the research and theory of Pavlovian rein-
forcement by conditioned stimuli. In C. M. Locurto,
141
H. S. Terrace, & J. Gibbon (Eds.), Autoshaping and
conditioning theory. New York: Academic Press,
1981.
Rescorla, R. A. Aspects of the reinforcer learned in
second-order Pavlovian conditioning. Journal of Ex-
perimental Psychology: Animal Behavior Processes,
1979,5,79-95.
Rescorla, R. A. Within-signal learning in autoshaping.
Animal Learning & Behavior, 1981, 9, 245-252.
Rescorla, R. A., & Durlach, P. J. Within-event learning
in Pavlovian conditioning. In N. S, Spear & R. Miller
(Eds.), Information processing in animals: Memory
mechanisms, Hillsdale, N.J.: Erlbaum, in press.
Rescorla, R. A., & Gillan, D. J. An analysis of the
facilitative effect of similarity on second-order con-
ditioning. Journal of Experimental Psychology: An-
imal Behavior Processes, 1980, 6, 339-351.
Rescorla, R. A., & Holland, P. C. Behavioral studies
of animal learning and memory. Annual Review of
Psychology, 1982, 33, 265-308.
Rizley, R. C., & Rescorla, R. A. Associations in second-
order conditioning and sensory preconditioning. Jour-
nal of Comparative and Physiological Psychology,
1972,*;, 1-11.
Spence, K. W. The role of secondary reinforcement in
delayed reward learning. Psychological Review, 1947,
54, 1-8.
Wagner, A. R, SOP: A model/ of automatic memory
processing in animal behavior. In N. S. Spear & R.
Miller (Eds.), Information processing in animalsf
Memory mechanisms. Hillsdale, NJ.: Erlbaum, in
press.
Wasserman, E. A., Carr, D. L., & Deich, J. D. Asso-
ciation of conditioned stimuli during serial condition-
ing by pigeons. Animal Learning & Behavior, 1978,
6, 52-56.
Wertheimer, M. Laws of organization in perceptual
forms. In W. D. Ellis (Ed.), A source book ofgestalt
psychology. New York: Humanities Press, 1955.
Received May 12, 1981 •