Joulmalof Experimental Psychology: Copyright 1998 by the American PsychologicalAssociation,Inc.

Animal Behavior Processes 0097-7403/98/$3.00

1998, VoL 24, No. 1, 47-59

Transfer Across Delayed Discriminations:

II. Differences in the Substitutability of Initial Versus Test Stimuli
Peter J. Urcuioli and T h o m a s B. DeMarse T h o m a s R. Zentall
Purdue University University of Kentucky

In 2 experiments, pigeons were trained on, and then transferred to, delayed simple
discriminations in which the initial stimuli signalled reinforcement versus extinction
following a retention interval. Experiment 1 showed that discriminative responding on the
retention test transferred to novel test stimuli that had appeared in another delayed simple
discrimination but not to stimuli having the same reinforcement history off-baseline. By
contrast, Experiment 2 showed that performances transferred to novel initial stimuli whether
they had been trained on-baseline or off-baseline. These results suggest that the test stimuli in
delayed simple discriminations acquire control over responding only in the memory task itself.
On the other hand, control by the initial stimuli, if coded as outcome expectancies, does not
require such task-specific training.

Comparing performances across different delayed discrimi-
nations has been a popular way to identify what animals
remember during the retention interval of a working memory
task. This comparison has been frequently driven by the
belief that animals not only code the physical attributes of
stimuli appearing prior to the retention interval (retrospec-
tive coding) but also anticipate events occurring with
predictable regularity after the retention interval (prospec-
tive coding; cf. Honig & Thompson, 1982; Wasserman,
1986). Indeed, there are considerable data supporting this
position (e.g., Capaldi, Birmingham, & Alptekin, 1995;
Cook, Brown, & Riley, 1985; Peterson, 1984; Roitblat,
1980; Urcuioli & Zentall, 1990, 1992). The present experi-
ments were designed in part to make this prospective-
retrospective distinction even clearer.
A good example of the distinction can be found in studies
of delayed simple discriminations (e.g., Honig & Wasser-
man, 1981; Urcuioli & Zentall, 1990). In this task, the
stimuli appearing prior to the retention interval (the initial
stimuli) signal, by themselves, the reinforcement contingen-
cies on the forthcoming retention test. In a go/no-go delayed
simple discrimination with pigeons, for instance, a red initial
stimulus might signal that pecks will be reinforced to
whichever line orientation (or test stimulus) appears after the
retention interval, whereas a green initial stimulus might
signal that the trial will end in nonreinforcement (i.e., that
pecking the line-orientation test stimulus will be extin-
guished). Note that red and green carry all of the information
Peter J. Urcuioli and Thomas B. DeMarse, Department of
Psychological Sciences, Purdue University; Thomas R. Zentall,
Department of Psychology, University of Kentucky.
This research was supported in part by National Institute of
Mental Health Grant 1 RO1 MH45979 (subcontract).
Correspondence concerning this article should be addressed to
Peter J. Urcuioli, Department of Psychological Sciences, 1364
Psychology Building--Room 3180, Purdue University, West Lafay-
ette, Indiana 47907-1364. Electronic mail may be sent via Internet
to uche @psych.purdue.edu.
necessary to respond appropriately on the retention test (i.e.,
to peck the test stimulus on reinforced trials and not to peck
it on nonreinforced trials). Compare this with a delayed
conditional discrimination involving the same stimuli but in
which the reinforcement contingencies are determined by
particular combinations of initial and test stimuli. For
example, with a red initial stimulus, reinforcement might be
delivered for responding to the vertical but not to the
horizontal test stimulus, and vice versa following a green
initial stimulus. In this task, red and green do not, by
themselves, signal whether or not reinforcement is available
after the retention interval. That can be determined only
when the test stimulus appears.
Typically, delayed simple discrimination performances
are much more accurate over longer retention intervals than
delayed conditional discrimination performances (Cohen,
Galgan, & Fuerst, 1986; Honig & Wasserman, 1981; Ur-
cuioli & Zentall, 1990; Weisman, Bruce, & Beninger, 1987).
Given the same stimuli and stimulus sequences across tasks,
this has been taken as evidence for different working
memory codes (Honig & Dodd, 1986; Wasserman, 1986). In
particular, it has been argued that in delayed simple discrimi-
nations, animals prospectively code the initial stimuli in
terms of the end-of-trial outcomes that they signal (viz., food
versus none). By contrast, in delayed conditional discrimina-
tions, the memory codes must be something other than
differential outcome expectancies because outcomes are
uncorrelated with the initial stimuli.
This prospective coding or outcome expectancy interpre-
tation of delayed simple discrimination performances is
supported by other data in the literature. For instance, the
usual advantage of simple over conditional discrimination
performances disappears if all trials end in reinforcement,
albeit contingent on either responding or not responding to
the test stimuli (Urcuioli & Zentall, 1990). Under these
conditions, the same expectancy (i.e., of food) would always
be generated by the initial stimuli; thus, there would be no
differential expectancies to cue responding on the retention
test. More convincingly, perhaps, animals trained on delayed

47
48 URCUIOLI, DEMARSE, AND ZENTALL
simple discriminations in which trials end in either reinforce-
ment or extinction (i.e., in different outcomes) readily
transfer their discriminative performances to novel initial
stimuli with the same outcome associations as the initial
stimuli they replace (Urcuioli & Zentall, 1992, Experiment
1). This transfer-of-control effect resembles findings used to
establish that outcome expectancies play a role in other
delayed discrimination paradigms (Edwards, Jagielo, Zen-
tall, & Hogan, 1982; Honig, Matheson, & Dodd, 1984;
Peterson, 1984; Urcuioli, 1990), and its origins were the
main focus of the present study.
Table 1 schematically illustrates the procedure for demon-
strating such transfer. In each of two training phases,
animals learn a delayed simple discrimination in which the
initial stimuli presented at the beginning of a trial (desig-
nated S1-S4) signal whether responding to individual test
stimuli (designated T1-T4) will ( + ) or will not ( - ) be
reinforced. Note that the two training tasks involve different
initial stimuli and different test stimuli. In the transfer-test
phase, the initial stimuli from the first task replace those
used in the second, thus creating novel combinations of
initial and test stimuli. In previous research (Urcuioli &
Zentall, 1992), it has been found that at long retention
intervals, go versus no-go responding to the test stimuli
transfers virtually without decrement across initial stimuli.
This finding is consistent with the idea that the same
working memory codes mediate performances in the two
training tasks and, more specifically, that each set of initial
stimuli gives rise to differential outcome expectancies (viz.,
food vs. no food). It is these differential outcome expectan-
cies that ostensibly determine whether or not animals
respond to the retention-test stimuli. Stated in another
fashion, for any set of test stimuli, it does not appear to
matter what initial stimuli generate the outcome expectan-
cies.
An alternative interpretation of the positive transfer
observed in this design, however, is that animals learn in
training to retrospectively code the initial stimuli and, on
that basis, to then respond (or not respond) to whatever test
stimulus appears after the retention interval. For example,
the animal might learn in Phase 1 that when remembering $1
(e.g., red), it should peck at the next stimulus appearing after
the retention interval, whereas when remembering $2 (e.g.,
Table 1
A Delayed Simple Discrimination Design Producing
Positive Transfer of Performance
Training
Phase 1 Phase 2 Testing
S1--*Ti+ S3--*T3+ SI--~T3+
SI~T2 + $3-'*T4+ S1---*T4+
$2"-"T1- S4--~T3 - S2--*T 3 -
52---~T2- S4--*T4- $2--*T4-
Note. SI~S4 = initial stimuli; T r T 4 = test stimuli; + and -
indicate end-of-trial food presentation and no food, respectively,
for responding to the test stimuli. Arrows indicate the retention
interval.
green), it should not peck following the retention interval.
This alternative seems quite plausible given that the test
stimuli, unlike the initial stimuli, are uncorrelated with the
reinforcement contingencies. If it is correct, then the finding
that discriminative performances continue to be maintained
at or near baseline levels following a swap in initial stimuli
is hardly surprising. In other words, if performance is
independent of the test stimulus presented on each trial, then
substituting the initial stimuli from one delayed simple
discrimination to another creates, from the animal's perspec-
tive, a task no different from one on which they have already
been trained.
Nonetheless, other aspects of the previously reported
transfer results do not easily fit this retrospective coding
explanation. For instance, it has been found that perfor-
mances during testing were maintained at baseline levels
only following long retention intervals and deteriorated
considerably when the retention interval was very short (see
Figure 2 in Urcuioli & Zentall, 1992). There is no obvious
reason to expect this difference if an animal simply responds
(or does not respond) to whatever stimulus appears after the
delay. Furthermore, transfer is not obtained if the initial
stimuli from the two training tasks do not share the same
differential outcome associations. For example, if one set of
initial stimuli signals reinforcement versus extinction whereas
the second set signals reinforcement contingent on respond-
ing versus not responding to the test stimuli, then the two
sets of initial stimuli are not interchangeable. This too seems
contrary to the altemative hypothesis because if animals
merely learn to retrospectively code the initial stimuli of a
delayed simple discrimination and to effectively treat the test
stimuli as irrelevant, then their go versus no-go perfor-
mances should transfer across initial stimuli in this situation
too.
Nevertheless, a formal evaluation of the impact of the test
stimuli in transfer of delayed discriminations is important
because the discrepant data mentioned above could probably
be accomodated by ad hoc arguments. Experiment 1 pro-
vided such an evaluation.
Experiment 1
If transfer across delayed simple discriminations occurs
because responding on the retention test is independent of
what test stimulus appears there, then the substitution of any
familiar stimuli for the test stimuli used in training should
not disrupt performances. Specifically, this view predicts
immediate positive transfer of performance to trials consist-
ing of initial stimuli trained within a delayed simple
discrimination and of test stimuli with which the animal has
had experience but which have never appeared in such a
task. By contrast, if the properties by which the test stimuli
support discriminative performances are acquired only in the
task itself, then replacing them with stimuli that have not
been trained in a delayed discrimination should cause
performances to collapse.
 TRANSFER ACROSS DELAYED DISCRIMINATIONS 49

Me~od
Anima~
Sixteen White Carneaux pigeons (Columba livia; retired breed-
ers) obtained from the Palmetto Pigeon Plant (Sumter, SC) were
used in the experiment. All had previous experience in two-choice
matching-to-sample and were randomly assigned to groups in a
manner that balanced those histories. The pigeons were housed
individually in stainless-steel cages in a common colony room on a
14:10-hr light-dark cycle. Grit and water were always available in
the home cages, but food availability was restricted to maintain
each pigeon's body weight at approximately 80% of its free-
feeding value. Typically, the pigeons obtained enough food in each
experimental session to maintain those weights. When they did not,
supplemental feedings were provided in the home cage. In addi-
tion, each pigeon was fed in the home cage on the 1 day per week
that the experiment was not conducted.
Apparatus
The two experimental chambers used for this experiment were
identical to those used by Urcuioli and Zentall (1992). Each
chamber (Model SEC-002; BRS/LVE Inc., Laurel, MD) contained
a panel with three equally spaced pecking keys 2.5 cm in diameter
(Model PIP-016), although only the center response key was used.
Behind the center key was a BRS/LVE in-line stimulus projector
(Model IC-901-IDD), which could display the following stimuli:
red, green, yellow, and white homogeneous fields; a single white
vertical or horizontal line on a black background; an open white
triangle or white circular annulus on black backgrounds; and an
open blue square or blue X on black backgrounds. The latter two
stimuli were constructed by inserting a blue Wratten filter in line
with the square and the X elements on the photographic film
(Pattern No. 696) mounted inside the projector. The reason for
presenting the latter two forms as blue (instead of the usual white)
was to make them as distinct as possible from the other form
stimuli.
A 5.0-cm × 5.8-cm opening in each panel provided access to a
rear-mounted grain magazine, which when operated was illumi-
nated by an ESB-28 lightbulb located in a metal cover surrounding
the magazine. General chamber illumination was provided by a
General Electric No. 1829 lightbulb located inside a small metal
housing 7.6 cm above the center key. The opening in this housing
was positioned to direct light toward the ceiling. A continuously
running exhaust fan attached to the outside of the chambers
provided ventilation and masking noise. All experimental events
were controlled and recorded by individual Cromemco Z-2D
microcomputers. One half of the pigeons in each group described
below were assigned to each chamber.
Procedure
The experiment proper consisted of two delayed simple discrimi-
nation training phases and a transfer test phase (see Table 2). Each
training phase was preceded by preliminary training sessions not
shown in the table but described below.
Preliminary training for Phase 1. Because all pigeons had
previous experimental experience, magazine training and shaping
of the keypeck response was not required. The first part of
preliminary training, then, consisted of reinforcing single pecks to
the stimuli that would later appear in the delayed simple discrimina-
tions. Every pigeon initially received one session during which
pecking red and green hues on the center key was reinforced by 3-s
access to grain. The center-key stimuli for the next four sessions
were yellow and white hues, vertical and horizontal lines, triangle

Table 2
Design of Experiment I
Training
Group Phase 1 Phase 2 Testing Phase 3
P R ---, VFI 5s T ---4 YFI 5s R ---* YFI 5s
R---. HFI 5s T ---, WFI 5s and b S - F I 5s/EXT R --* WFI 5s
G ---* V EXT C ---* Y EXT b X - F I 5s/EXT G ---* Y EXT
G ---" H EXT C ~ W EXT G --* W EXT

PC R---~ VFI 5s T ---* bSFI 5s R ---, YFI 5s

R---* HFI 5s T ---. bXFI 5s and Y-FI 5s/EXT R ~ WFI 5s
G---* V EXT C ~ bS EXT W - F I 5s/EXT G ---- Y EXT
G - - . H EXT C ---, bX EXT G ---* W EXT

N R -...¢ VFI 5s T ----. YFI 5s R --~ Y EXT

R - - . HFI 5s T ---. WFI 5s and b S - F I 5s/EXT R ---. W EXT
G---" V EXT C ~ Y EXT b X - F I 5s/EXT G --~ YFI 5s
G---* H EXT C ~ W EXT G ---* WFI 5s

NC R ---* VFI 5s T --, bSFI 5s R ---* Y EXT

R ---* HFI 5s T ---4 bXFI 5s and Y - F I 5s/EXT R ~ W EXT
G ~ V EXT C --~ bS EXT W - F I 5s/EXT G ---. YFI 5s
G --. H EXT C --* bX EXT G --- WFI 5s
Note. P = positive; PC = positive control; N = negative; NC = negative control; R = red; G =
green; V = vertical lines; H = horizontal lines; T = triangle; C = circle; Y = yellow; W = white;
bS = blue square; bX = blue X. FI = fixed-interval schedule; EXT = extinction. Arrows denote
retention interval separating initial stimuli (on the left) and test stimuli (on the right). Counterbalanc-
ing of initial stimuli and reinforcement schedules has been omitted. See text for further details.
50 URCUIOLI, DEMARSE, AND ZENTALL
and circle forms, and blue square and blue X color-form com-
pounds, in that order. Each stimulus in a session appeared 30 times
in random order, and successive stimulus presentations were
separated by a 10-s intertrial interval (ITI).
During the next five sessions, pecking red and green on the
center key was reinforced on modified fixed-interval (FI) sched-
ules. A single peck to the hue appearing on each trial started the FI,
and the first peck after the interval expired produced food
reinforcement. The FI value was 2 s for the first, 3 s for the second,
and 5 s for the third session, and 10 s for the fourth and fifth
sessions. In all sessions, each hue appeared 30 times in random
order with the restriction that neither appeared more than three
times in a row. Successive trials were again separated by a 10-s ITI,
the first 9 s of which was spent with the houselight off. The
houselight was turned on for the last 1 s of the ITI and remained on
until the end of the reinforcement cycle. Reinforcement duration
was constant within a session but was varied from 2-5 s across both
subjects and sessions to maintain each pigeon's body weight at or
near its 80% value.
Pecking vertical and horizontal center-key lines was reinforced
during the final two preliminary training sessions by using the
modified FI 2-s and FI 5-s schedules, respectively. These sessions
were identical to those described above except that only one half of
the trials with each stimulus ended in food. On the remaining trials,
the line stimulus (and the houselight) went off automatically after
the fixed interval had expired. Then, following a dark interval equal
in duration to the reinforcement cycle, the ITI began. Partial
reinforcement was used with the lines to mimic the overall
schedule of food presentation that the pigeons would later experi-
ence following these same stimuli in the Phase 1 delayed simple
discrimination.
Phase 1 training. Following preliminary training, all pigeons
learned a delayed simple discrimination with red and green initial
stimuli and vertical and horizontal test stimuli. Each delayed
simple discrimination trial began with a 10-s presentation of either
red or green on the center key, which was timed from the first peck
to the initial stimulus. The first key peck after the 10-s interval had
elapsed turned off the initial stimulus and produced either the
vertical or horizontal line on the same key following an intervening
blank interval of 500 ms (hereafter referred to as the 0-s retention
interval). For half of the pigeons, the trial ended in reinforcement
(i.e., the first peck after 5 s to whatever test stimulus appeared
produced food) if the initial stimulus had been red, whereas the trial
ended in nonreinforcement (i.e., the test stimulus went off automati-
cally after 5 s) if the initial stimulus had been green. The remaining
pigeons experienced the opposite contingencies. Training for each
pigeon in this Phase 1 delayed simple discrimination continued
until the pigeon reached a criterion of five of six consecutive
sessions in which 90% or more of all test-stimulus responses
occurred on reinforced trials.
The four possible combinations of the red and green initial
stimuli and the vertical and horizontal test stimuli occurred 15
times in random order in each session. Each session also contained
10 additional trials, 5 with red and 5 with green, which ended
immediately with food reinforcement following offset of the hue
(i.e., there was no test-stimulus presentation). These extra trials
were scattered throughout each session and were included to
maintain comparable rates of key pecking to both initial stimuli (cf.
Honig & Wasserman, 1981). All trials were separated from one
another by a 10-s ITI of which the first 9 s was spent in darkness.
The houselight came on for the last 1 s of the ITI and remained on
until the end of food presentation on reinforced trials or until
stimulus offset on nonreinforced trials.
After reaching criterion with the 0-s retention interval, each
pigeon then received 15 additional sessions of mixed-delay train-
ing. These sessions were conducted in exactly the same fashion as
described above except that the blank-key delay separating the
initial stimuli from the test stimuli was either 0, 5, or l0 s. Each
retention interval occurred equally often with each combination of
initial and test stimuli.
Preliminary training for Phase 2. Next, each pigeon received
preliminary training to peck the stimuli that would later appear in
the second delayed simple discrimination task. For the first five of
these sessions, pecking triangle and circle forms on the center key
was reinforced according to the modified FI schedules described in
Preliminary training for Phase 1. All trials with these stimuli ended
with food, and the sequence of FI values across sessions was
identical to that used for preliminary training with the red and green
hues. Each pigeon then received two sessions of partial-
reinforcement training with yellow and white center-key hues
followed by two similar sessions with the blue square and blue X
compounds. These latter sessions were run in the same fashion as
the corresponding ones that preceded the Phase 1 delayed simple
discrimination.
Phase 2 training. In Phase 2, all pigeons learned another
delayed simple discrimination involving initial and test stimuli
different from those used in Phase 1. In addition, each training
session alternated over days with sessions during which pecking
two other stimuli was partially reinforced. As shown in the middle
column of Table 2, all groups were trained with the same initial
stimuli in their Phase 2 delayed simple discrimination (viz., the
triangle and circle forms) but were exposed to different test stimuli
and to different stimuli during the alternating off-baseline sessions.
For the positive group (Group P) and the negative group (Group
N), yellow and white hues served as test stimuli in the delayed
simple discrimination. For half of the pigeons in these two groups,
the first test-stimulus peck after 5 s produced food when the
preceding initial stimulus was the triangle, whereas test-stimulus
pecking was extinguished (i.e., the test stimulus went off automati-
cally after 5 s with no food reward) when the initial stimulus was
the circle. These contingencies were reversed for the remaining
subjects. During the alternating off-baseline sessions, pecking the
blue square and blue X compounds was reinforced after 5 s on one
half of all trials. The remaining trials ended automatically after 5 s
without food (i.e., in extinction). Each color-form compound
appeared 30 times in random order during these sessions.
For two control (C) groups, Group PC and Group NC, the blue
square and blue X compounds served as the test stimuli in the
delayed simple discrimination with the reinforced initial stimulus-
test stimulus combinations balanced across subjects. The alternat-
ing off-baseline sessions for Groups PC and NC consisted of 30
presentations each of yellow and white hues on the center key with
food occurring after 5 s on one half of each trial type. Extinction
was in effect on the remaining trials.
For all four groups, training on the Phase 2 discrimination was
initially conducted with a 0-s retention interval and was continued
until each pigeon reached a criterion of five of six consecutive
sessions in which 90% or more of its test-stimulus responses
occurred on reinforced trials. At that point, 15 additional sessions
of mixed-delay training with 0-, 5-, and 10-s retention intervals
were provided. The mixed-delay sessions continued to alternate
with each pigeon's respective off-baseline task.
Because of difficulties in the acquisition of the Phase 2 task, the
alternating session procedure was discontinued for 1 pigeon in each
group after 30 sessions. Two of these pigeons eventually reached
the 90% criterion. Two others, however, did not reach this criterion
after 10 to 20 additional consecutive training sessions; however,
they were advanced to the mixed-delay procedure because their
discriminative performances were consistently and highly accurate
(i.e., more than 80% of their test-stimulus responses occurred on
reinforced trials). For all of these pigeons, the alternating session
routine was reinstituted once mixed-delay training began.
 TRANSFER ACROSS DELAYED DISCRIMINATIONS 51

Refresher training and Phase 3 testing. Immediately prior to substantial performance difference between Groups P and N but
testing, subjects received refresher training on their Phase 1 and none between Groups PC and NC.
Phase 2 tasks to ensure that both baseline delayed simple discrimi- Alternatively, if pigeons learn in training to respond or not
nation performances were intact. The sequence of refresher ses- respond independently of what stimuli appear on the retention test,
sions was as follows: (a) training on the Phase 1 delayed simple then Group PC should, like Group P, show positive transfer in
discrimination with 0-s delays until the 90% performance criterion testing, and Group NC should, like Group N, show negative
was met for a single session, (b) mixed-delay training on the Phase transfer. The rationale here is that for both Groups P and PC, red
2 delayed simple discrimination until delay performances recov- and green continue to signal the same end-of-trial reinforcement
ered to levels comparable to those observed at the end of Phase 2, contingencies for responding in transfer as they did in Phase 1
(c) a single partial-reinforcement session with the off-baseline training, and those signalling properties alone should determine for
stimuli, and (d) mixed-delay training on the Phase 1 delayed simple the pigeons whether to peck after the retention interval. On the
discrimination until delay performances on this task likewise other hand, the reinforcement contingencies vis-~i-vis the red and
recovered to levels comparable to those at the end of Phase 1. If the green initial stimuli for Groups N and NC in testing are the opposite
mixed-delay performances on the latter task did not fully recover of what they were in Phase 1 training, so these pigeons should
within three sessions, then the last three components of the entire respond more to the retention-test stimuli on nonreinforced than on
refresher sequence were repeated in that order until such recovery reinforced trials (at least initially).
was obtained.
Finally, each pigeon was tested on a new delayed simple Results
discrimination involving red and green initial stimuli (from Phase
1) and yellow and white test stimuli (from Phase 2). The Baseline (Training) Performances
mixed-delay procedure was used throughout the five test sessions,
with trial sequencing, stimulus durations, and so on identical to Acquisition of the common Phase 1 task proceeded at
those used in training. Note that for Groups P and N, yellow and comparable rates in all groups. The average number of
white hues had previously appeared as test stimuli in another training sessions needed to reach the .90 DR criterion ranged
delayed simple discrimination, whereas for Groups PC and NC, between 10.0-12.5 sessions and did not differ significantly
they had not. Note also that for Groups P and PC, the relations between groups, F(3, 12) = 0.22. Likewise, the four groups
between the red versus the green initial stimulus and the end-of-
performed similarly during the mixed-delay sessions follow-
trial outcomes (i.e., reinforcement versus extinction) were identical
to those in effect during the baseline delayed simple discrimination ing acquisition. The DRs averaged over the last five
in which these stimuli had originally appeared (i.e., the Phase 1 mixed-delay sessions, shown in the left half of Table 3, were
task). By contrast, this relation was reversed for Groups N and NC. very near to or above .90 at all three delays in all groups. A n
analysis of variance (ANOVA) showed no significant group
effect, F(3, 12) = 0.09, nor a Group × Delay interaction,
Data Analysis and Predictions
F(6, 24) = 0.22. Only the overall effect of delay was
The primary dependent measure was a discriminationratio (DR) significant, F(2, 24) --- 12.34.
calculated by dividing the total number of key pecks to the test For the various groups, the Phase 2 delayed simple
stimuli on reinforced trials by the total number of key pecks to the discrimination involved the same initial stimuli (triangle and
test stimuli over all trials. Little or no discrimination is indicated by circle) but different test stimuli (either yellow and white
DRs in the vicinity of .50 (i.e., pecking occurs as often on hues or blue square and blue X compounds). Despite this
nonreinforced as on reinforced trials). The DR approaches a difference, acquisition of the Phase 2 task was comparable
maximum of 1.00 as performances become more and more
across groups: The average number of sessions needed to
discriminative (i.e., as test-stimulus responding becomes increas-
reach a .90 DR ranged from 20.0-25.2 and did not differ
ingly confined to reinforced trials). A DR less than .50 indicates
discriminative performance opposite to that reinforced by the significantly among groups, F(3, 12) = 0.13. Phase 2
contingencies (i.e., pecking the test stimuli more often on nonrein- acquisition was slower than in Phase 1, but part of the reason
forced than on reinforced trials). For all statistical analyses, Type I for this may have been that the delayed simple discrimina-
error rate was set as .05 on a per-decision basis by using the tabled tion sessions in Phase 2 were not run consecutively as they
values reported by Rodger (1975a). were in Phase 1. Instead, each session alternated with
Different predictions regarding how the various groups should off-baseline partial-reinforcement training with two other
perform in testing were derived from the different views concern-
ing the role of the test stimuli in delayed simple discriminations. If
the test stimuli play an integral part in supporting delayed simple Table 3
discrimination performances and if their function is acquired Discrimination Ratios by Delay for Each Group in
within that context, then Group P should show positive transfer of Experiment 1 Averaged Over the Last Five Sessions
performance to the new task (i.e., its DRs should be greater than of Their Phase 1 and Phase 2 Delayed
.50) given that their red and green initial stimuli continued to signal Simple Discriminations (DSDs)
the same reinforcement contingencies as they had in Phase 1. On
the other hand, Group N should initially show negative transfer Phase 1 DSD: Delay Phase 2 DSD: Delay
(i.e., its DRs should be below .50) because their red and green Group 0s 5s 10 s 0s 5s 10 s
initial stimuli signalled the opposite contingencies in the transfer
test. By contrast, although the corresponding initial stimulus- P .97 .94 .90 .93 .94 .92
outcome relations held for Groups PC and NC, neither positive nor PC .97 .94 .92 .83 .85 .86
negative transfer, respectively, should be apparent in their test N .96 .92 .88 .93 .93 .91
performances (i.e., their DRs should be near .50) because the NC .97 .92 .90 .85 .85 .83
yellow and white test stimuli had not previously appeared in a Note. P = positive; PC = positive control; N = negative; NC =
delayed simple discrimination. In short, this view predicts a negative control.
52 URCUIOLI, DEMARSE, AND ZENTALL
stimuli. Also, the triangle and circle initial stimuli were
probably more difficult for pigeons to discriminate than the
red and green hues in Phase 1 (cf. Carter & Eckerman, 1975;
Urcuioli & Zentall, 1986).
Mixed-delay performances for the Phase 2 task are
summarized in the right half of Table 3. Once again,
discriminative performances were highly accurate across
delays in all four groups, and an ANOVA showed no
significant effects, all Fs < 1.31.
Transfer-Test Performances
The right panel of Figure 1 shows DRs on the first Phase 3
test session for each group. For comparison purposes,
performances on the last Phase 1 mixed-delay refresher
session (baseline) are shown in the left panel.
Baseline performances were comparable across groups as
confirmed by an ANOVA, which revealed no significant
effects of group or delay. There was a significant Group X
Delay interaction, F(6, 24) = 1.70, which apparently arose
from the slightly lower DR in Group PC at the 5-s delay.
Nevertheless, this represents a minor variation in discrimina-
tive performances that were otherwise uniformly high in all
four groups.
In contrast, very substantial between-group differences
were apparent on the first Phase 3 test session. Specifically,
Group P showed strong positive transfer of its delayed
simple discrimination performance to the new task consist-
ing of red and green hues (from Phase 1) as initial stimuli
and yellow and white hues (from Phase 2) as the retention-
test stimuli. The DRs for this group were well above .50 at
1.0
BASELINE
.9
©
~ .8
z
~ .7
OP
r~ ON
• PC
A NC
.4
"f i i
0 5
Figure 1. Discrimination ratios by delay for the four groups in Experiment 1 on the mixed-delay
Phase 1 (baseline) refresher session preceding transfer (left panel) and on the first session of Phase 3
transfer (fight panel). P = positive group; N = negative group; PC = positive control group; NC =
negative control group.
all three delays, with performances at the longer delays as
accurate in testing as they were on the last baseline refresher.
By contrast, none of the other groups showed a similar
effect. Indeed, for them, the average DRs clustered around
.50 with somewhat lower DRs in Group N than in Groups
PC and NC. An ANOVA on the combined Day 1 test data
revealed that all three factors--group, delay, and their
interactioniwere significant, F(3, 12) = 142.26, F(2, 24) =
11.48, and F(6, 24) = 12.21, respectively.
Separate ANOVAs compared first-session test perfor-
mances in Group P with those in Group PC and in Group N
with those in Group NC. For the former pair, the relation
between the red and green initial stimuli and the end-of-trial
outcomes was the same in testing as it was in the Phase 1
(baseline) task (cf. Table 1). For the latter pair, this relation
was reversed. An ANOVA on the data from the two positive
groups confirmed that discriminative performances were
significantly more accurate overall in Group P than in Group
PC, F(1, 6) = 142.40. The Group × Delay interaction was
also significant, F(2, 12) = 14.62. An ANOVA on the data
from the two negative groups also revealed a significant
group effect. In this case, the average DRs in Group N were
significantly lower than they were in Group NC, F ( l , 6) =
7.08. The Group × Delay interaction was also significant,
F(2, 12) = 3.79, indicative of the fact that the lower DRs in
Group N were apparent at the 5- and 10-s delays but not at
the 0-s delay (compare the open circles with the open
triangles in Figure 1).
To evaluate whether the positive transfer seen in Group P
was apparent from the outset of testing and to evaluate
PHASE 3 TEST - DAY 1
1.0
.9
.8
.7
.6
.4
t "f I I i
10 0 5 10
DELAY (sec)
 TRANSFER ACROSS DELAYED DISCRIMINATIONS
whether Group N, Group NC, or both may have shown
initial negative transfer, we computed discriminative perfor-
mances for the first 12 delayed simple discrimination test
trials for each group. These DRs were .84, .52, .34, and .44
for Groups P, PC, N, and NC, respectively. Clearly, Group P
showed immediate positive transfer of performance. Addi-
tionally, the very low DR for Group N (.34) indicates
negative transfer: These pigeons tended to peck the yellow
and white test stimuli more often following the initial
stimulus now associated with extinction (but formerly
associated with reinforcement) than following the initial
stimulus now associated with reinforcement (but formerly a
signal for extinction). By contrast, the data for Groups PC
and NC indicate that these pigeons responded about equally
often to the test stimuli on both reinforced and nonreinforced
trials. Post hoc contrasts (Rodger, 1975a) on these data
showed that the DRs were not significantly different in
Groups PC and NC, F(3, 11) = 0.54, but they were
significantly greater in Group P than in Group N, F(3, 11) =
17.3. Moreover, the average DR for Groups PC and NC
combined did not differ significantly from the average for
Groups P and N, F(3, 11) = 1.78. Together, these statistical
results imply that the DRs for the first 12 test trials were
ordered from highest to lowest as follows: Group P > Group
PC = Group NC > Group N (see Rodger, 1975b, Equation
23).
Discussion
The results of this experiment show very clearly that
transfer across delayed simple discriminations does not
occur simply because pigeons learn during training to
respond on the retention test independently of the test
stimulus that appears. Two findings in particular are inconsis-
tent with this inattention account.
The first is the failure of the two control groups, PC and
NC, to show the same transfer effects as those exhibited by
Groups P and N, respectively. Indeed, there was no evidence
of transfer in Groups PC and NC. This result occurred
despite the fact that these two groups had received just as
much delayed simple discrimination training during Phases
1 and 2 as Groups P and N and, more important, that their
history of partial reinforcement for responding to the test
stimuli used in transfer was identical to that for Groups P and
N. In short, the difference in performance between the
control and experimental groups cannot be explained either
by differences in the amount of delayed simple discrimina-
tion training they received or by differences in familiarity
with the test stimuli used in transfer. With these two
variables controlled, the data indicate instead that transfer
across delayed discriminations is precluded unless the test
stimuli in transfer have been previously trained in another
delayed simple discrimination. This is clearly contrary to the
inattention account: If discriminative performances on the
retention test are independent of the test stimuli, the
performances in Groups PC and NC should have mirrored
those observed in Groups P and N.
The second inconsistent result was the finding that
discriminative performances by Group P (and less so by
53
Group N) during transfer varied as a function of the retention
interval. For Group P, DRs were much higher at the two
longest retention intervals than at 0 s, which replicates a
previously reported result from an experiment that used a
design similar to that used here (see Urcuioli & Zentall,
1992, Experiment 1; cf. Group EE). Again, there was no
reason to anticipate such a difference if pigeons simply
responded or did not respond on the retention test solely on
the basis of the prior initial stimulus.
Clearly, then, the test stimuli play an integral role in
delayed simple discrimination performances despite the fact
that they are uncorrelated with reinforcement. Moreover, the
degree of control they exert varies as a function of the
retention interval. The latter finding--that discriminative
control gets stronger (rather than weaker) as the retention
interval is lengthened---seems peculiar at first glance. How-
ever, it is understandable given the training procedure and
the nature of the working memory code that is thought to
govern performance.
It has been previously argued (Urcuioli & Zentall, 1990,
1992) that when the initial stimuli of a delayed simple
discrimination are correlated with the presence versus
absence of end-of-trial reinforcement, those initial stimuli
give rise to differential outcome expectancies, which in turn
provide a discriminative cue for whether to respond to the
test stimuli. In other words, the working memory code is an
expectancy of food versus no food. This assertion is
consistent with the fact that delayed simple discrimination
performances are more accurate when the end-of-trial out-
come is differential with respect to the initial stimuli than
when all trials end in food, albeit contingent on different
patterns of test-stimulus responding (Urcuioli & Zentall,
1990). It is also consistent with the ability of other stimuli
with similar differential outcome associations to readily
substitute for the initial stimuli used in training (Urcuioli &
Zentall, 1992).
The poorer and seemingly anomalous transfer at the 0-s
delay probably arises because differential outcome expectan-
cies cannot fully develop until shortly after the initial stimuli
go off, The reason for this is that our training procedure
included a small proportion of trials in which food immedi-
ately followed the presentation of either initial stimulus.
Thus, the initial stimuli by themselves were not consistent
signals for food versus no food. It was only after their offset,
and after the absence of immediate food, that they reliably
signalled upcoming reinforcement versus extinction for
responding to the test stimuli. In short, the primary working
memory code on 0-s delay trials was likely to be different
from the code on longer delay trials; it may have been based
on the physical attributes of the initial stimuli at 0 s (a
retrospective code), but based on an outcome expectancy at
longer delays (a prospective code).
Of course, why there should be an interaction between the
nature of the working memory code and the nature of the test
stimuli in transfer remains to be answered. First, however,
we were interested in determining whether the on- versus
off-baseline training effects seen here were unique to test
stimuli or whether they would also occur with the initial
stimuli.
54 URCUIOLI, DEMARSE, AND ZENTALL

Experiment 2 of this experiment balanced as closely as possible those prior

E x p e r i m e n t 1 s h o w e d that transfer across delayed discrimi-
nations occurs o n l y if the test stimuli h a v e p r e v i o u s l y
appeared in a d e l a y e d simple discrimination. S i m p l y arrang-
ing off-baseline r e i n f o r c e m e n t o f those stimuli, e v e n in a
fashion identical to that o c c u r r i n g within a d e l a y e d simple
discrimination, does not suffice to m a k e t h e m substitutable
for other test stimuli. In E x p e r i m e n t 2, w e asked w h e t h e r the
s a m e rule applies to the initial stimuli. In other words, does
transfer across d e l a y e d discriminations also require that the
substituted initial stimuli be trained within a delayed simple
discrimination, or w o u l d off-baseline training suffice to
e m p o w e r t h e m w i t h the properties r e q u i r e d for transfer?
C o n s i d e r i n g that d e l a y e d simple discriminations with
r e i n f o r c e m e n t versus extinction are differential o u t c o m e
tasks (Urcuioli & Zentall, 1990), w e anticipated that transfer
to n e w initial stimuli w o u l d o c c u r e v e n if those stimuli w e r e
trained outside the d e l a y e d simple discrimination context.
This prediction was based on findings f r o m other differential
o u t c o m e studies that showed, for e x a m p l e , that in two-
c h o i c e tasks such as m a t c h i n g - t o - s a m p l e , pigeons i m m e d i -
ately transfer their m a t c h i n g p e r f o r m a n c e s to stimuli trained
off-baseline with the s a m e o u t c o m e associations as the
samples they replace (e.g., Peterson, 1984; Urcuioli, 1990;
U r c u i o l i & D e M a r s e , 1996). T h e c o r r e s p o n d i n g manipula-
tion o f substituting off-baseline stimuli for the initial stimuli
in a d e l a y e d s i m p l e discrimination, h o w e v e r , has not been
done. E x p e r i m e n t 2 thus p r o v i d e s important c o m p a r a t i v e
data to c o m p l e m e n t those obtained in E x p e r i m e n t 1.
Method
Animals and Apparatus
Eight White Carneaux retired breeder pigeons (Columba livia)
from the Palmetto Pigeon Plant (Sumter, SC) served in this
experiment. They were housed and maintained in the same fashion
as the pigeons in Experiment 1. Each pigeon had previously
participated in a matching-to-sample study (see DeMarse &
Urcuioli, 1993, Experiment 2). Their assignment to the two groups
histories. The apparatuses and stimuli were the same as those used
in Experiment 1.
Procedure
The design of the experiment, summarized in Table 4, consisted
of two training phases followed by two transfer tests. Training and
testing are described below along with preliminary training ses-
sions not depicted in the table.
Preliminary training. Each pigeon was first trained to peck the
center-key stimuli that would later appear as initial or test stimuli in
the delayed simple discriminations. The stimuli for these sessions,
the order in which the sessions were conducted, and all other details
were exactly the same as those described for the corresponding
sessions in Experiment 1.
All pigeons were then trained for five sessions to peck triangle
and circle center-key stimuli on the modified FI interval schedules
previously described in Preliminary training for Phase 1. As in
Experiment 1, the FI value was increased from 2 to 10 s across
sessions with food presentation occurring after the completion of
every interval. For the last two preliminary training sessions,
pecking yellow and white hues on the center key was partially
reinforced on the FI 2-s and FI 5-s schedules, respectively.
Approximately one half of the trials with each stimulus ended with
food.
Phase ] training. Next, all pigeons received delayed simple
discrimination training with triangle and circle initial stimuli and
yellow and white test stimuli. Acquisition of this task was
conducted with a 0-s retention interval separating the initial stimuli
from the test stimuli, until each pigeon reached a criterion of five of
six consecutive sessions in which 90% or more of its test-stimulus
responses occurred on reinforced trials. At that point, each pigeon
received 15 additional sessions of mixed-delay training with
retention intervals of 0, 5, and 10 s occurring equally often in each
session. Two pigeons, both in the initial-stimulus training group
(Group IST; see below), did not reach criterion after 40 and 50
sessions, respectively, but were moved to the mixed-delay proce-
dure nonetheless because their DRs were consistently high (e.g.,
they averaged .83 and .92, respectively, over the last five acquisi-
tion sessions). The details of both the acquisition and mixed-delay
sessions were identical in every respect to those for the correspond-
ing sessions in Experiment 1.

Table 4
Design of Experiment 2
Training Testing
Group Phase 1 Phase 2 Phase 3 Phase 4
IST T ---* YFI 5s R ~ bSFI 5s R --* YFI 5s V ~ YFI 5s
T--.WFI5s V---* ( + ) and R---~bXFI5s R--,WFI5s V---~WFI5s
C --~ Y EXT H ~ (-) G --4 bS EXT G ---* Y EXT H ---* Y EXT
C --~ W EXT G ~ bX EXT G ---* W EXT H ~ W EXT

OB T ---* YFI 5s V~ bSFI 5s R --~ YFI 5s V --~ YFI 5s

T --, WFI 5s R --~ (+) and V ~ bXFI 5s R - 4 WFI 5s V ---. WFI 5s
C ---* Y EXT G ~ (-) H~ bS EXT G -~ Y EXT H ---* Y EXT
C ~ W EXT H~ bX EXT G ~ W EXT H ~ W EXT
Note. IST = initial-stimulus training; OB = off-baseline training; T = triangle; C = circle; Y =
yellow; W --- white; R = red; G = green; V = vertical lines; H = horizontal lines; bS = blue square;
bX = blue X; FI = fixed-interval schedule; EXT = extinction. Arrows denote retention interval
separating initial stimuli (on the left) and test stimuli or end-of-trial food presentation ( + ) versus no
food ( - ) (on the right). Counterbalancing of initial stimuli and reinforcement schedules has been
omitted. See text for further details.
 TRANSFER ACROSS DELAYED DISCRIMINATIONS
Phase 2 training. During Phase 2, pigeons learned a new
delayed simple discrimination, which alternated over days with
sessions involving two other center-key stimuli not appearing in the
delayed simple discrimination. For Group IST, the Phase 2 delayed
simple discrimination consisted of trials with red and green hues as
initial stimuli and the blue square and blue X color-form com-
pounds as test stimuli. The stimuli appearing during the alternating
sessions for this group were vertical and horizontal lines. For the
off-baseline group (Group OB), the Phase 2 delayed simple
discrimination consisted of trials with vertical and horizontal lines
as initial stimuli and the two color-form compounds as test stimuli.
The alternating sessions for this group involved red and green
center-key stimuli. The group labels reflect the fact that in the
Phase 3 transfer test that followed, the initial stimuli from the Phase
2 delayed simple discrimination were substituted for those in Phase
1 for Group IST, whereas the corresponding substitution for Group
OB involved the two stimuli trained off-baseline.
For both groups, delayed simple discrimination acquisition in
Phase 2 was conducted with a 0-s retention interval and with the
same criterion as in Phase 1. Once each pigeon reached criterion, it
received 15 additional sessions of mixed-delay training. As in
acquisition, the mixed-delay sessions continued to alternate with
off-baseline training involving either the vertical and horizontal
lines (Group IST) or the red and green hues (Group OB).
The off-baseline sessions in this experiment were run differently
from those of Experiment 1 to compensate for the fact that transfer
involved the initial stimuli rather than the test stimuli. During these
70-trial sessions (35 with each off-baseline stimulus), the first peck
to the stimulus appearing on a trial started a 10-s interval that ended
with the first peck after the interval had expired. On 10 trials (5 with
each stimulus), food was presented immediately following stimu-
lus offset. On the remaining 60 trials, offset of the center-key
stimulus was followed by a delay of either 0, 5, or 10 s during
which the houselight remained on. At the end of the delay, food was
presented response independently if the trial had begun with one
stimulus, or the houselight simply went off without food (i.e., the
ITI began) if the trial had begun with other stimulus. These
contingencies are represented in second column of Table 4 by the
plus and minus signs, respectively, following vertical and horizon-
tal lines or red and green hues. The stimulus followed by food on
delay trials was counterbalanced across the 4 pigeons in each
group. Note that for both groups, the delivery of food (or lack
thereof) during the off-baseline sessions was structured in the same
way as in the mixed-delay delayed simple discrimination task: (a)
on most trials, food was presented some time after the initial
stimulus had gone off but only on trials beginning with one initial
stimulus and not the other; and (b) on the remaining trials, food was
contingent on the first peck to either stimulus after 10 s.
Refresher training and Phase 3 testing. In preparation for the
first substitution test, each pigeon received the following sequence
of refresher sessions: (a) training on the Phase 1 delayed simple
discrimination with 0-s delays until the DR was at or above .90 for
a single session, (b) refreshers on the Phase 2 delayed simple
discrimination with mixed delays until performances recovered for
a single session to the level observed at the end of Phase 2, (c) a
single session with the off-baseline stimuli and contingencies, and
(d) refreshers on the Phase 1 delayed simple discrimination with
mixed delays until performances recovered for a single session to
the level observed at the end of Phase 1. If more than three sessions
were required to recover performances on this last task, then the
last three components of the entire refresher sequence were
repeated prior to testing.
During the five Phase 3 transfer sessions, pigeons in both groups
were tested on a delayed simple discrimination with red and green
initial stimuli and yellow and white test stimuli. For all subjects, the
end-of-trial reinforcement contingencies associated with red and
55
green hues matched those associated with either (a) the red and
green initial stimuli in the Phase 2 delayed simple discrimination
(Group IST), or (b) contingencies in effect during the off-baseline
sessions with these same stimuli (Group OB). In other words, if red
had preceded the delivery of food during Phase 2 and green had not
(or vice versa), the same thing was true during Phase 3 testing. All
other details of the delayed simple discrimination transfer task
were identical to those for the delayed simple discriminations used
in the prior training phases. Note too that the Phase 3 transfer task
in this experiment was identical to the corresponding task in
Experiment 1.
Refresher training and Phase 4 testing. The second transfer
test involved substituting vertical and horizontal lines for the
triangle and circle as initial stimuli in the Phase 1 delayed simple
discrimination. The five Phase 4 test sessions were otherwise
identical to all other mixed-delay sessions and were preceded by
the following refresher sequence: (a) training on the Phase 1
mixed-delay task until performances recovered for a single session
to the level observed at the end of Phase 1, (b) training on the Phase
2 mixed-delay task until a similar recovery was achieved, (c) a
single session with the off-baseline stimuli, and (d) a second
recovery of Phase 1 delayed simple discrimination performances
with mixed delays.
The Phase 4 test differed in one notable respect from the Phase 3
test: It reversed the type of initial stimulus substitution that each
group received. In other words, for Group IST, the initial stimuli in
the Phase 1 delayed simple discrimination were replaced by the
off-baseline stimuli from Phase 2, whereas for Group OB, those
same initial stimuli were replaced by the delayed simple discrimi-
nation task stimuli from Phase 2.
Results
Baseline (Training) Performances
Initial acquisition and m i x e d - d e l a y p e r f o r m a n c e s on the
c o m m o n Phase 1 task w e r e v e r y similar in the t w o groups.
For e x a m p l e , the average n u m b e r o f sessions n e e d e d to
reach a D R criterion o f .85 (to c o m p e n s a t e for the 2 pigeons
that failed to reach the .90 criterion) was 13.2 and 15.2 for
Groups I S T and OB, respectively, F ( 1 , 6) = 0.14. T h e left
h a l f o f Table 5 s h o w s that the D R s for the last five
m i x e d - d e l a y sessions w e r e at or a b o v e .90 at e v e r y delay in
both groups. A n A N O V A s h o w e d no significant effects, all
F s < 1.13.
The Phase 2 delayed simple discrimination was also
acquired at similar rates e v e n though the initial stimuli w e r e
not the same for both groups. T h e n u m b e r o f sessions n e e d e d
to reach the .90 D R criterion in Phase 2 was 17.0 and 15.8 in
G r o u p s I S T and OB, respectively, a nonsignificant differ-
Table 5
Discrimination Ratios by Delay for Each Group in
Experiment 1 Averaged Over the Last Five Sessions
of Their Phase 1 and Phase 2 Delayed
Simple Discriminations (DSDs )
Phase 1 DSD: Delay Phase 2 DSD: Delay
Group 0s 5s 10 s 0s 5s 10 s
IST .90 .90 .90 .93 .92 .88
OB .94 .94 .91 .94 .97 .95
Note. IST = initial-stimulus training; OB = off-baseline training.
56 URCUIOLI, DEMARSE,AND ZENTALL

ence, F(1, 6) = 0.04. The average DRs for the last five response rates to the subsequently reinforced line (2.54
mixed-delay sessions, shown in the right half of Table 5, pecks/s) than to the nonreinforced line (3.90 pecks/s),
again depict uniformly high discriminative performances. although this difference was not significant, F(1, 3) = 3.36.
Here, an ANOVA revealed a significant effect of delay and a Finally, pecking during the 5- and 10-s delay intervals (i.e.,
Group x Delay interaction, Fs(2, 12) = 5.48 and 4.36, at a dark center key) was generally infrequent in both
respectively, the latter no doubt reflecting the higher DRs at groups. For the 3 pigeons that did peck noticeably during the
the 5- and 10-s delays in Group OB than in Group IST. Once delays, all pecked more often when food was forthcoming
again, however, this variation seems rather minor. than when it was not.
Table 6 shows how each pigeon responded to the off-
baseline stimuli in Phase 2, both during presentation of the
Transfer Performances
stimuli themselves and during the subsequent delay inter-
vals. The data represent average response rates over the last Figure 2 shows the Day 1 test data for both the Phase 3
five off-baseline sessions during Phase 2 training (i.e., and Phase 4 transfer tests (right panel) and the delayed
during the last five sessions that alternated with the mixed- simple discrimination performances for the last baseline
delay task). The stimulus that signalled that food would be refresher session preceding each test (left panel). The filled
presented after the delay is indicated by an asterisk. and open circles plot the Phase 3 results, and the filled and
The off-baseline contingencies for Group OB produced open triangles plot the Phase 4 results. Note that the
moderate rates of pecking to both the red and green stimuli, conditions for the Phase 3 test reflect the original group
although rates to the stimulus signalling food after a delay labels (cf. Table 4): for Group IST, the substitution that
were slightly, but consistently and significantly, higher (3.34 created the transfer task involved initial stimuli trained
pecks/s) than to the stimulus signalling no food (2.67 within a delayed simple discrimination, whereas for Group
pecks/s), F(1, 3) = 17.09. The overall rates to the vertical OB, the substitution involved the off-baseline stimuli. On
and horizontal lines in Group IST showed more variability the other hand, the group legend in Figure 2 for the Phase 4
across subjects, but interestingly, the pattern was for lower test reflects the fact that the nature of the substitution was
reversed relative to Phase 3 (i.e., OB-IST indicates that
Group OB was tested in Phase 4 with initial stimuli trained
Table 6 within a delayed simple discrimination, whereas IST-OB
Response Rates (in Peck~s) to the Off-Baseline Stimuli indicates that Group IST was tested with its off-baseline
During Presentation and During the Delay Intervals stimuli).
Following Their Presentation in Experiment 2 Baseline performances on the refresher sessions immedi-
Response rate ately preceding each test were similar in the two groups. On
neither refresher was there a significant overall difference
Group and During After presentation between groups, Fs(1, 6) = 2.59 and 0.00 for the Phase 3
pigeon Stimulus presentation 5-s delay 10-s delay and Phase 4 baselines, respectively, nor a significant Group x
Group OB Delay interaction, Fs(2, 12) = 0.18 and 1.76 for Phase 3 and
1 R* 3.20 0.03 0.01 Phase 4 baselines, respectively.
G 2.66 0.04 0.03 By contrast, there was a substantial and significant
between-group difference on the first Phase 3 transfer test.
2 R 2.19 0.01 0.00 Specifically, although both groups showed positive transfer
G* 2.66 0.02 0.02
of performance to the novel initial stimuli (i.e., their DRs
3 R* 3.74 0.01 0.00 were greater than .50), the effect was greater in Group OB
G 2.59 0.06 0.04 than in Group IST, F(1, 6) = 8.09. There was also a
significant overall effect of delay, F(2, 12) = 39.42, but no
4 R 3.25 0.20 0.12 Group X Delay interaction, F(2, 12) = 1.19.
G* 3.76 1.80 1.93 The Phase 4 transfer test showed the same trend. The
Group IST group for which the novel initial stimuli in the delayed
1 V* 2.39 1.95 2.00 simple discrimination had been trained off-baseline (Group
H 2.92 0.22 0.18
IST-OB) showed stronger positive transfer of performance
2 V* 1.32 0.00 0.00 than the group for which those same stimuli had been
H 1.39 0.00 0.01 previously trained as initial stimuli in a delayed simple
discrimination (Group OB-IST). However, this difference,
3 V 5.24 0.30 0.15 although numerically large, was not significant in an ANOVA,
H* 3.82 0.09 0.08 F(1, 6) = 1.26. The overall effect of delay and the Group x
Delay interaction also were not significant, Fs(2, 12) = 0.72
4 V 6.03 0.18 0.10 and 0.69, respectively.
H* 2.61 1.29 1.42
Note. Data are averaged over the last five off-baseline sessions of
Phase 2 training. OB = off-baseline training; IST = initial-stim- Discussion
ulus training; R = red; G = green; V = vertical lines; H =
horizontal lines; * = stimulus followed by food after the delay The results of the first (Phase 3) test session showed that
interval. delayed simple discrimination performances transferred to
 TRANSFER ACROSS DELAYED DISCRIMINATIONS 57

1.0 -~ BASELINE
1.0
TESTING - DAY 1

.9 .9
©

2.8
z

• IST - (Phase 3)
r...) .6 (3 O B - (Phase 3)
r~ • O B - I S T - (Phase 4)
I S T - O B - (Phase 4)

I I I I I I
0 5 10 0 5 10
DELAY (sec)

Figure 2. Discrimination ratios by delay for the two groups in Experiment 2. The left panel plots
performances on the last mixed-delay Phase 1 (baseline) refresher sessions preceding testing. The
right panel plots performances on the first Phase 3 transfer test (circles) and the first Phase 4 transfer
test (triangles). The conditions of testing for each group reversed from the Phase 3 to the Phase 4 test.
IST = initial-stimulus training group; OB = off-baseline training group; OB-IST = Group OB was
tested with initial stimuli trained within a delayed simple discrimination; IST-OB = Group IST was
tested with its off-baseline stimuli.

new initial stimuli in both groups. Replacing the initial
stimuli from one delayed simple discrimination either with
the initial stimuli from another (Group IST) or with stimuli
trained off-baseline but with similar outcome associations
(Group OB) disrupted the go versus no-go performances
relative to baseline but nonetheless maintained those discrimi-
nations well above the level expected by chance alone (i.e.,
.50). Moreover, and in contrast to Experiment 1, the
substitution of initial stimuli trained off-baseline did not
produce weaker transfer than the substitution of the same
stimuli trained within another delayed simple discrimina-
tion. Indeed, the Phase 3 transfer test showed a significant
effect in the opposite direction: Stimuli trained off-baseline
were better able to maintain delayed simple discrimination
performances in testing than those trained on-baseline. The
same general pattern of results was reproduced in the second
(Phase 4) test, although in this case the on- versus off-
baseline training effect was not significant.
The finding that initial stimuli trained off-baseline are
effective substitutes for those that have acquired their
properties in a delayed simple discrimination itself is not
surprising given the transfer effects reported elsewhere in
the differential outcome literature. In matching-to-sample
paradigms, for instance, pigeons immediately transfer their
choice performances at high levels of accuracy to novel
samples, which, through off-baseline Pavlovian pairings,
share the same food versus no-food associations as the
samples they replace (Peterson, 1984; Urcuioli, 1990; Ur-
cuioli & DeMarse, 1996). Apparently, in go/no-go delayed
simple discrimination tasks such as those studied here, the
cues governing performance on the retention test are much
the same as those in two-alternative forced-choice tasks:
namely, the outcome expectancies to which the initial
(sample) stimuli give rise. The proposed parallel is all the
more compelling given that Honig et al. (1984, Experiment
2) showed that pigeons showed immediate positive transfer
of their go/no-go delayed conditional discrimination perfor-
mances to novel initial stimuli trained off-baseline if those
stimuli were associated with the same outcomes as the ones
they replaced. Honig et al. also found immediate negative
transfer if the off-baseline stimuli were substituted for those
having the opposite outcome associations.
What is not immediately clear is why the transfer
observed here was somewhat stronger to initial stimuli
trained off-baseline than to those same stimuli trained within
a task like that used in testing. One possibility is that
differential outcome expectancies develop more fully or
more strongly when discrete events, such as the test stimuli
in a delayed simple discrimination, do not intervene between
the presentation of the signalling stimuli and the end-of-trial
outcomes. Another possibility is that the between-group
effect is more apparent than real given that the level of
transfer by Group IST was lower than that which has been
observed in similar groups in previous experiments (cf.
Group P in Experiment 1 of the present study; also cf. Group
EE in Urcuioli & Zentall, 1992, Experiment 1). Neverthe-
less, this ambiguity should not obscure the main finding:
namely, that on- versus off-baseline training does not have
the same effect on transfer to novel initial stimuli as it does
58 URCUIOLI, DEMARSE, AND ZENTALL
on transfer to novel test stimuli in delayed simple discrimina-
tions.
General Discussion
The two experiments of the present study reveal substan-
tial differences in the substitutability of new (but familiar)
stimuli for the initial and test stimuli of a delayed simple
discrimination. Experiment 1 showed that accurate go/no-go
delayed simple discrimination performances could be main-
tained when the test stimuli were replaced by others that had
been trained in another delayed simple discrimination but
not by stimuli with an equivalent reinforcement history
acquired off-baseline. Experiment 2, on the other hand,
showed that delayed simple discrimination performances
transferred immediately to novel initial stimuli with the
same differential outcome associations as the initial stimuli
they replaced and, more important, that this transfer did not
depend on whether those associations were acquired within
another delayed simple discrimination or off-baseline.
The results of Experiment 1 mean that transfer of delayed
simple discriminations cannot be explained by a combina-
tion of retrospective coding of the initial stimuli plus
inattention to the test stimuli. In other words, performances
in these tasks depend on the particular stimuli appearing on
the retention test even though the retention-test stimuli are
completely irrelevant to the reinforcement contingencies. By
contrast, the results of Experiment 2 indicate that following
training on a delayed simple discrimination with differential
outcomes, it does not matter what initial stimuli signal those
outcomes nor does it matter how the to-be-substituted
stimuli have been trained (viz., on- or off-baseline). To-
gether, the combined results provide further support for a
prospective coding (outcome expectancy) interpretation of
delayed simple discrimination performances. What remains
to be determined is the precise mechanism by which the test
stimuli allow those outcome expectancies to exert appropri-
ate discriminative control on the retention test.
The present findings also have parallels in other areas of
the conditioning and memory literatures. Regarding initial
stimulus substitutability, for example, we have already
mentioned a number of other differential outcome studies
using two-choice and go/no-go paradigms in which transfer
of performance was observed to novel initial or sample
stimuli that acquired differential outcome associations out-
side of the general task in which they were tested (e.g.,
Honig et al., 1984; Peterson, 1984; Urcuioli, 1990). Further-
more, transfer has also been observed when the initial or
sample stimuli used in testing were trained within the
context of the target task (e.g., Edwards et al., 1982).
Together, this entire collection of results indicates that when
discriminative performances come under control of the
animal's differential outcome expectancies, it does not
matter in what context those expectancies develop. As long
as the initial or sample stimuli reliably signal different
end-of-trial outcomes, the animal's anticipation of those
outcomes can continue to guide performances that have
previously developed in their presence.
The contrasting dependence of continued discriminative
responding on the nature of the test stimuli--in particular,
the prior context in which the test stimuli were trained--
resembles similar findings in the Pavlovian literature on
occasion setting. For example, in both serial feature-positive
and serial feature-negative discriminations, the enhance-
ment and suppression, respectively, of conditioned respond-
ing to the target by the feature does not transfer to new
targets that have been partially reinforced off-baseline
(Holland, 1986, Experiment 3; Lamarre & Holland, 1987,
Experiment 2; Wilson & Pearce, 1990). By contrast, these
same studies have shown that transfer to new targets will
occur if the target stimuli have been trained within other
feature-positive or feature-negative discriminations. One
exception to this pattern of results was reported by Rescorla
(1989) in a serial feature-negative autoshaping experiment
with pigeons. Rescorla found that the feature did transfer its
suppressive properties to a new target that was reinforced
(continuously) off-baseline. However, this discrepant result
could have arisen from generalization between the character-
istics shared by the training and transfer targets (i.e., both
appeared as lights on the top as opposed to the bottom half of
the pecking key).
Indeed, it is worthwhile to consider the similarities and
differences between the Pavlovian occasion-setting para-
digm and the delayed simple discrimination task. For
example, in a serial feature-positive task, the target stimulus
is reinforced if preceded by another stimulus (the feature)
but not otherwise. In a delayed simple discrimination,
responding to the test stimulus is reinforced if it is preceded
by one initial stimulus but not if it is preceded by another.
Thus, the delayed simple discrimination task becomes a
serial feature-positive discrimination if the initial stimulus
on the nonreinforced trials is omitted. Conversely, omitting
the initial stimulus on the reinforced trials would change a
delayed simple discrimination into a serial feature-negative
discrimination (i.e., reinforcement would follow the test
stimulus unless it was preceded by the initial stimulus or
feature). Could it be, then, that these procedural differences
are inconsequential and that the processes governing perfor-
mances in the two paradigms are similar if not identical?
The parallel between the on- versus off-baseline effects on
transfer to new test or target stimuli clearly encourages such
a view. But there also seem to be important differences in
regard to the substitutability of the initial stimuli or features.
For example, in the occasion-setting literature, serial feature-
positive performances do not transfer to novel features that
have been trained as simple Pavlovian CS + s (e.g., Rescorla,
1985, 1987). In other words, even though the feature in
training uniquely signals upcoming reinforcement, other
stimuli with a similar association acquired off-baseline do
not augment responding to the target stimulus. This indicates
that responding to the target in the training task is not
controlled by any reinforcement expectancy that might be
conditioned to the feature. If it were, then any stimulus
producing a similar expectancy ought to readily substitute
for that feature. Although features are interchangeable with
one another if all are trained in feature-positive tasks
(Rescorla, 1985, 1987; see also Davidson, Aparicio, &
Rescorla, 1988), this effect is apparently confined to such
on-baseline conditions. The same is clearly not true for the
initial stimuli in the delayed simple discriminations studied
 TRANSFER ACROSS DELAYED DISCRIMINATIONS
here and in a variety of other differential outcome tasks.
These initial stimuli readily substitute for the ones used in
training independently of whether their differential associa-
tion to reward is the result o f on- or off-baseline training.
This difference does not preclude overlapping associative
mechanisms, of course. It could very well be, for instance,
that outcome expectancies that are known to be effective
discriminative cues in many working memory tasks might
also act as occasion setters in Pavlovian paradigms if the
conditions of training promoted such control. To take a
simple example, if two feature-positive discriminations were
trained concurrently such that one feature indicated that its
target would be followed by one reinforcer (e.g., food),
whereas the other feature indicated that its target would be
followed by a different reinforcer (e.g., water), responding to
each target might be specifically augmented by (i.e., transfer
to) new features that have been trained as Pavlovian CS + s
for those same reinforcers (cf. Brodigan & Peterson, 1976;
Jenkins & Moore, 1973). Such a demonstration, if success-
ful, would suggest that the associative as well as the physical
attributes of stimuli can serve as occasion setters in Pavlov-
ian situations.
References
Brodigan, D. L., & Peterson, G. B. (1976). Two-choice conditional
discrimination performance of pigeons as a function of reward
expectancy, prechoice delay, and domesticity. Animal Learning
& Behavior, 4, 121-124.
Capaldi, E. J., Birmingham, K, M., & Alptekin, S. (1995).
Memories of reward events and expectancies of reward events
may work in tandem. Animal Learning & Behavior, 23, 40-48.
Carter, D. E., & Eckerman, D. A. (1975). Symbolic matching by
pigeons: Rate of learning complex discriminations predicted
from simple discriminations. Science, 187, 662-664.
Cohen, J. S., Galgan, R., & Fuerst, D. (1986). Retrospective and
prospective short-term memory in delayed response tasks in rats.
Animal Learning & Behavior, 14, 38-50.
Cook, R. G., Brown, M. F., & Riley, D. A. (1985). Flexible memory
processing by rats: Use of prospective and retrospective informa-
tion in the radial maze. Journal of Experimental Psychology:
Animal Behavior Processes, 11, 453-469.
Davidson, T. L., Aparicio, J., & Rescorla, R. A. (1988). Transfer
between Pavlovian facilitators and instrumental discriminative
stimuli. Animal Learning & Behavior, 16, 285-291.
DeMarse, T. B., & Urcuioli, P. J. (1993). Enhancement of matching
acquisition by differential comparison-outcome associations.
Journal of Experimental Psychology: Animal Behavior Pro-
cesses, 19, 317-326.
Edwards, C. A., Jagielo, J. A., Zentall, T. R., & Hogan, D. E.
(1982). Acquired equivalence and distinctiveness in matching to
sample by pigeons: Mediation by reinforcer-specific expectan-
cies. Journal of Experimental Psychology: Animal Behavior
Processes, 8, 244-259.
Holland, P. C. (1986). Transfer after serial feature positive discrimi-
nation training. Learning and Motivation, 17, 243-268.
Honig, W. K., & Dodd, P. W. D. (1986). Anticipation and intention
in working memory. In D. E Kendrick, M. E. Rilling, & M. R.
Denny (Eds.), Theories of animal memory (pp. 77-100). Hillsdale,
NJ: Erlbaum.
Honig, W. K., Matheson, W. R., & Dodd, P. W. D. (1984). Outcome
expectancies as mediators for discriminative responding. Cana-
dian Journal of Psychology, 38, 196-217.
59
Honig, W. K., & Thompson, R. K. R. (1982). Retrospective and
prospective processing in animal working memory. In G. H.
Bower (Ed.), The psychology of learning and motivation:
Advances in research and theory (Vol. 16, pp. 239-283). New
York: Academic Press.
Honig, W. K., & Wasserman, E. A. (1981). Performance of pigeons
on delayed simple and conditional discriminations under equiva-
lent training procedures. Learning and Motivation, 12, 149-170.
Jenkins, H. M., & Moore, B. R. (1973). The form of the auto-
shaped response with food or water reinforcers. Journal of the
Experimental Analysis of Behavior, 20, 163-181.
Lamarre, J., & Holland, P. C. (1987). Transfer of inhibition after
serial feature negative discrimination training. Learning and
Motivation, 18, 319-342.
Peterson, G. B. (1984). How expectancies guide behavior. In H. L.
Roitblat, T. G. Bever, & H. S. Terrace (Eds.), Animal cognition
(pp. 135-147). Hillsdale, NJ: Erlbaum.
Rescorla, R. A. (1985). Conditioned inhibition and facilitation. In
R. R. Miller & N. E. Spear (Eds.), Information processing in
animals: Conditioned inhibition (pp. 299-326). Hillsdale, NJ:
Erlbaum.
Rescorla, R. A. (1987). Facilitation and inhibition. Journal of
Experimental Psychology: Animal Behavior Processes, 13,
250-259.
Rescorla, R. A. (1989). Simultaneous and sequential conditioned
inhibition in autoshaping. Quarterly Journal of Experimental
Psychology, 41B, 275-286.
Rodger, R. S. (1975a). The number of non-zero, post hoc contrasts
from ANOVA and error rate. I. British Journal of Mathematical
and Statistical Psychology, 28, 71-78.
Rodger, R. S. (1975b). Setting rejection rate for contrasts selected
post hoc when some nulls are false. British Journal of Mathemati-
cal and Statistical Psychology, 28, 214-232.
Roitblat, H. L. (1980). Codes and coding processes in pigeon
short-term memory. Animal Learning & Behavior, 8, 341-351.
Urcuioli, P. J. (1990). Some relationships between outcome expect-
ancies and sample stimuli in pigeons' delayed matching. Animal
Learning & Behavior, 18, 302-314.
Urcuioli, P. J., & DeMarse, T. B. (1996). Associative processes in
differential outcome discriminations. Journal of Experimental
Psychology: Animal Behavior Processes, 22, 192-204.
Urcuioli, P. J., & Zentall, T. R. (1986). Retrospective coding in
pigeons' delayed matching-to-sample. Journal of Experimental
Psychology: Animal Behavior Processes, 12, 69-77.
Urcuioli, P. J., & Zentall, T. R. (1990). On the role of trial outcomes
in delayed discriminations. Animal Learning & Behavior, 18,
141-150.
Urcuioli, E J., & Zentall, T. R. (1992). Transfer across delayed
discriminations: Evidence regarding the nature of prospective
working memory. Journal of Experimental Psychology: Animal
Behavior Processes, 18, 154-173.
Wasserman, E. A. (1986). Prospection and retrospection as pro-
cesses of animal short-term memory. In D. E Kendrick, M. E.
Rilling, & M. R. Denny (Eds.), Theories of animal memory (pp.
53-75). Hillsdale, NJ: Erlbaum.
Weisman, R. G., Bruce, R., & Beninger, R. J. (1987). Simple and
conditional discrimination in rats: The effects of delay and
scopolamine. Learning and Motivation, 18, 274-287.
Wilson, E N., & Pearce, J. M. (1990). Selective transfer of
responding in conditional discriminations. Quarterly Journal of
Experimental Psychology, 42B, 41-58.
Received September 18, 1996
Revision received March 28, 1997
Accepted March 31, 1997 •