Behavioural Processes xxx (2006) xxx–xxx

Training delays reduce the choose-short effect with

keylight duration samples in pigeons
Douglas S. Grant ∗
Department of Psychology, University of Alberta, Edmonton, Alta., Canada T6G 2E9
Received 11 August 2005; accepted 30 November 2005

Abstract
Pigeons were trained to matching 2- and 8-s keylight samples. The delay on training trials was either 0 s (group 0sF), 2 s (group 2sF), or varied
between 1 and 3 s (M = 2 s, group 2sV). Testing at delays of 10 and 20 s revealed a choose-short tendency in all three groups. The magnitude of
this tendency was largest in group 0sF and was highly similar in magnitude in groups 2sF and 2sV. In Experiment 2, the training delay remained
at 0 s in group 0sF and was increased to 5 s in group 2sF (now group 5sF). Group 2sV (now group 5sV) received variable training delays ranging
from 2 to 8 s (M = 5 s). Testing at a 20-s delay and at a delay that exceeded the training delay by 15 s for each group revealed a robust choose-short
effect only in group 0sF. Groups 5sF and 5sV both demonstrated a weak and statistically nonsignificant choose-short tendency. When different
durations of keylight are employed as samples, training with a nonzero delay, whether fixed or variable, reduces the magnitude of the choose-short
effect, with longer training delays producing a greater reduction in the choose-short effect.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Timing; Memory; Matching-to-duration; Training delays; Choose-short effect; Pigeons

Trials in the delayed matching-to-sample procedure begin
with presentation of one of two or more stimuli as a sam-
ple stimulus. The sample is followed, typically after a delay
interval that varies in duration across trials, by two or more
comparison stimuli. Which comparison is correct (and hence
pecking it is reinforced) on any particular trial depends upon
which sample was presented at the beginning of the trial.
Across trials, both the stimulus presented as the sample and
the spatial position of the correct comparison stimulus are
varied.
Although the delayed matching task was introduced by
Blough in 1959, the first study to use the delayed matching
task as an analytical tool to investigate short-term memory in
pigeons was published by William A. Roberts shortly after my
arrival in his laboratory to pursue graduate studies in the sum-
mer of 1971 (Roberts, 1972). I was fortunate to participate in
many of those pioneering studies of memory conducted in Bill’s
laboratory; including studies of sample exposure effects (Grant,
1976; Roberts and Grant, 1974), proactive (Grant, 1975; Grant
and Roberts, 1973) and retroactive (Grant and Roberts, 1976;
∗ Tel.: +1 780 492 5299; fax: +1 780 492 1744.
E-mail address: douglas.grant@ualberta.ca.
Roberts and Grant, 1978a) interference, and compound sample
effects (Roberts and Grant, 1978b).
Today, almost a quarter century after the publication of
Roberts’ seminal paper in 1972, the delayed matching procedure
has been employed in certainly many hundreds, and likely thou-
sands, of studies analyzing short-term retention in pigeons. The
symbolic or arbitrary delayed matching procedure has proven
to be a particularly useful tool in the analysis of short-term
retention in pigeons. Because the relation between sample and
correct comparison is arbitrary in this procedure, the symbolic
matching procedure permits the analysis of memory for events
that could not be readily presented as comparison stimuli. Such
events include different tones (e.g., Kraemer and Roberts, 1984),
number of events (e.g., Roberts, 1995, 2005; Roberts et al., 1995;
Roberts and Mitchell, 1994), duration of events (e.g., Kraemer et
al., 1985; Roberts et al., 1989; Spetch and Wilkie, 1982, 1983),
and the presence or absence of particular stimuli (e.g., Colwill,
1984; Grant, 1991; Maki, 1979).
Of particular relevance to the present research is the use of the
symbolic matching procedure to study memory for the duration
of an event. In the first such study, Spetch and Wilkie (1982)
examined pigeons’ memory for 2- and 10-s durations of food
access and houselight illumination. Following training with a
0-s delay, pigeons were tested with delays ranging from 0 to

0376-6357/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.beproc.2006.01.005

BEPROC-1565; No. of Pages 10

2 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
20 s. For both food and light samples, accuracy was greater for
short samples than long samples after longer delays. That is,
subjects showed a strong tendency to choose the comparison
stimulus associated with the short sample after longer delays.
This observed tendency was called the choose-short effect. The
choose-short effect is a robust phenomenon and has been demon-
strated in a large number of studies (e.g., Gaitan and Wixted,
2000; Grant and Kelly, 1996, 1998; Grant and Spetch, 1991,
1993, 1994; Kraemer et al., 1985; Santi et al., 1993, 2003; Spetch
and Rusak, 1989; Spetch and Wilkie, 1982, 1983).
Several studies have assessed memory for duration follow-
ing training with a nonzero delay. Spetch (1987) used 2- and 8-s
durations of food access as samples and training delays of 0, 10
and 20 s across three successive stages of training. Each train-
ing phase was followed by a retention test and each revealed a
choose-short effect at a delay that was 10 s longer than the train-
ing delay, a result that was replicated using training delays of 0,
5 and 10 s by Spetch and Rusak (1992, Experiment 2a). Spetch
and Rusak (1992, Experiment 2b) trained one group with a con-
stant 5-s delay and a second with a variable delay ranging from
2 to 8 s in 2-s increments. The two groups did not differ either
during acquisition or retention testing, and both demonstrated
a strong choose-short effect at a 20-s delay. Kelly and Spetch
(2000) also observed a choose-short effect at delays of 10 and
20 s after training with 2- and 6-s durations of food access and a
constant delay of 5 s, and Spetch and Rusak (1989) observed a
choose-short effect at delays longer than the 10-s training delay
after training with either food or houselight durations. Finally,
Grant and Kelly (1998) employed 2- and 8-s keylight samples
and variable delays that ranged between 1 and 3 s during train-
ing and found a choose-short effect during retention testing at
delays of 15 and 30 s.
Although the studies reviewed in the preceding paragraph
suggest that the choose-short effect is affected little, if at all,
by the use of a nonzero delay during training, other studies
question that conclusion. For example, Dorrance et al. (2000)
trained pigeons on an event duration task with 2- and 10-s sam-
ples of keylight and delays of 0, 1, 2 and 4 s. When tested at
extended delays of 8 and 16 s, the typical choose-short effect
was not found, although there was a trend in that direction. Con-
sistent with the findings of Dorrance et al., Grant and Talarico
(2004, Experiment 1) and Talarico and Grant (2006, Experiment
1) failed to find a statistically reliable choose-short effect after
training with a variable delay ranging from 1 to 3 s. More-
over, when Talarico and Grant (2006, Experiment 2) retrained
using a 0-s training delay pigeons who had previously failed
to demonstrate a choose-short effect following training with
variable delays, a statistically reliable choose-short effect was
obtained.
Given the apparently conflicting findings reviewed above,
it is difficult to draw a firm conclusion in regard to what effect
training with nonzero delays has on the magnitude of the choose-
short effect. It may be noted that to date no research has directly
compared the magnitude of the choose-short effect after training
with a 0-s delay with that obtained after training with a nonzero
delay. In an effort to clarify the role of training with a nonzero
delay on the magnitude of the choose-short effect, the present
experiments provided such a comparison. In both experiments
reported in this article, one group was trained with a fixed 0-s
delay and two groups were trained with a nonzero delay, one
with a fixed delay on each trial and the other with a variable
delay the mean of which matched that of the fixed delay. Thus,
the present experiments allowed the assessment of the effects of
nonzero training delays and variability in those training delays
on the magnitude of the choose-short effect.
A second purpose of the present experiments was to assess a
suggestion by Dorrance et al. (2000) that pigeons might use dif-
ferential sample responding as a cue for choice in the matching-
to-duration task. They reported that there was significantly more
pecking in the presence of long samples than in the presence of
short samples and, furthermore, that rate of sample pecking was
greater on long-sample trials than on short-sample trials. It is
possible, therefore, that a high rate of pecking and/or a large
number of pecks controlled choice of the long-associated com-
parison, and a low rate of pecking and/or a small number of
pecks controlled choice of the short-associated comparison. On
this view, the choose-short effect arises because during delay
intervals pigeons typically are not pecking and, hence, when the
comparisons are presented either the low rate of pecking or the
fact that a large number of pecks has not recently been emitted
leads to choice of the short-associated comparison. To assess this
possibility, sample responding was recorded during all retention
testing sessions.
1. Experiment 1
Three groups of pigeons were trained with 2- and 8-s key-
light (i.e., black dot on a white background) samples. Half of the
pigeons within each group received line orientation comparisons
(vertical and horizontal bar) and the remaining birds received
color comparisons (red and green). The groups differed in terms
of the delay interval that intervened between termination of the
duration sample and presentation of the comparison stimuli dur-
ing training. In group 0sF, a fixed 0-s delay occurred on all trials
and hence the comparisons were presented immediately upon
termination of the sample. In a second group, 2sF, a fixed 2-s
delay occurred on all trials. In the final group, 2sV, a delay in
the range of 1–3 s (M = 2 s) occurred on all trials.
Following acquisition, all three groups received two retention
tests. In each test, 25% of the trials involved an extended delay
that was equally often 10 and 20 s. The remaining trials involved
a baseline delay of 0 s in the first test and 2 s in the second test.
1.1. Method
1.1.1. Subjects
Twenty-four naive, adult (6 months to 1 year) Silver King
pigeons served as subjects. They were reduced to and maintained
at 80% of their free-feeding weight throughout the duration of
the experiment. The pigeons were housed individually in wire
mesh cages between sessions and were provided with unlimited
amounts of water and grit. The colony room in which the birds
were housed was maintained on an alternating 12-h light–dark
cycle, in which light onset was at 6 a.m. Eight pigeons were
 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
assigned at random to each of the three groups: group 0sF, group
2sF, and group 2sV.
1.1.2. Apparatus
Training and testing was conducted in eight identical oper-
ant chambers, each measuring 29.0 cm × 29.0 cm × 24.0 cm
(height × length × width). In each chamber, a horizontal align-
ment of three circular pecking keys was centered along one end
wall. The key alignment was raised 22.5 cm from the barred-
floor base of the chamber. A force greater than 0.15 N applied
to any key was recorded as a keypeck. Affixed behind each key
was an Industrial Electronics Inc. (Van Nuys, CA) in-line pro-
jector which was used to illuminate the keys with a black dot on
a white background, horizontal and vertical lines, and red and
green colors. A 5.5-cm high × 5.0-cm wide rectangular open-
ing, which provided access to a retractable food magazine, was
located 9.0 cm directly beneath the edge of the middle key. A
28-volt lamp, within the magazine opening, was activated when
the food magazine was raised. Each chamber was enclosed in
a sound- and light-attenuating booth. Within each booth, an
exhaust fan provided ventilation and an external white noise gen-
erator provided masking auditory stimulation. All experimental
booths were isolated in the same darkened running room. The
only illumination inside the chamber was provided by the acti-
vation of keylights and the magazine light. Experimental events
were controlled from, and responses were recorded by, a micro-
computer located in an adjoining room. Experimental sessions
were conducted 6 days per week, and began at approximately
the same time each day for each of the three groups.
1.1.3. Procedure
1.1.3.1. Preliminary training. All 24 birds were trained to eat
from the magazine and then autoshaped to peck at red, green, a
horizontal line, and a vertical line presented on the center key
within their individual operant chambers. Once all birds were
reliably pecking each of the stimuli presented on each of the 3
pecking keys and eating from the feeders, training on duration
matching began.
1.1.3.2. Training. Trials began with an interval sample of 2 or
8 s. During the interval sample, the center key was illuminated
by a black dot on a white background. The 3 groups differed in
terms of the delay that intervened between sample termination
and onset of the comparison stimuli. For group 0sF, a 0-s delay
occurred on all trials and hence the comparisons were presented
immediately upon sample termination. For group 2sF, a 2-s delay
occurred on all trials. For group 2sV, a variable delay (M = 2 s,
range of 1–3 s) was presented on each trial. On each trial, the
delay length was computed by generating a random double-
precision value between 0 and 1, and equally often adding or
subtracting this value from “2”, thus producing a rectangular
distribution of delays within the 1–3 s range.
In each of the 3 groups, the 8 subjects were further divided
into 2 subgroups that received different comparison stimuli that
were presented on the outer two pecking keys. In the hori-
zontal/vertical subgroup, 2 birds were reinforced, with 2.5-s of
magazine-illuminated access to mixed grain from the food hop-
3
per, for pecking the horizontal line on short-sample (2 s) trials
and the vertical line on long-sample (8 s) trials. The remain-
ing 2 subjects in the subgroup were reinforced for the opposite
designations. For birds in the red/green subgroup, 2 birds were
reinforced for pecking the red key on short- and the green key
on long-sample trials. The remaining 2 subjects were reinforced
for the opposite designations. On all trials a peck to either of
the comparison stimuli resulted in the termination of both, fol-
lowed by either reinforcement (2.5-s of illuminated access to
grain) or non-reinforcement (2.5-s of darkness). None of the
results reported in this article were affected by the dimension of
the comparison stimuli and this variable will not be mentioned
further.
Sessions in each of the three groups consisted of an equal
number of short- and long-sample trials. The correct compari-
son stimulus was randomly displayed on either the right or the
left pecking key across trials but it was ensured that the correct
comparison appeared equally often on each side for each sample
duration within a session.
All sessions contained 64 trials, 32 with each sample duration,
each concluding with the onset of an intertrial interval (ITI)
consisting of a random duration that varied between 10 and 30 s
in 5-s increments (M = 20 s). The houselight remained off during
all sessions, hence both ITIs and delays were spent in darkness.
All subjects received 92 sessions of training at which point each
subject had consistent accuracy levels of 85% or higher.
1.1.3.3. Extended-delay test 1. All aspects of the extended-
delay testing phase were identical to the training phase for each
experimental group except that the delay between sample termi-
nation and comparison onset varied across 3 values. The 3 dif-
ferent delay values were 0 (baseline), 10, and 20 s. The sessions
were comprised of 75% of trials, which were randomly chosen,
involving the baseline delay (i.e., 0 s). The remaining 25% of tri-
als were equally divided into trials involving a 10- or 20-s delay.
Each of the 3 delay values occurred equally often with short and
long samples, and the position of the correct comparison stimu-
lus was balanced within sample-type and delay-interval factors.
Pecks to each of the samples were recorded and was used to
compute responses per minute to each of the 2 samples. A total
of 12 testing sessions were conducted. Each testing session after
the first was preceded by 2 sessions identical to those of training
in order to maintain performance.
1.1.3.4. Extended-delay test 2. Eight training sessions inter-
vened between the 2 extended-delay tests. All aspects of the
second delay test were identical to the first delay test except that
the baseline delay was 2 s rather than 0 s.
In all analyses reported in this article, p < 0.05 was adopted
as the criterion for statistical significance.
1.2. Results
1.2.1. Training
Fig. 1 shows data from the 92 sessions of training. Acquisition
was somewhat more rapid in group 0sF than in either of the two
groups receiving nonzero delays. Accuracy in the three groups
4 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
Fig. 1. Percentage of correct responses as a function of blocks of four sessions
for each of the three groups in Experiment 1.
converged by Block 17 and remained above 90% throughout
the remainder of training. A Sample Duration × Block × Group
analysis of variance (ANOVA) revealed a significant main effect
of block, F(22, 462) = 88.09, and a significant Block × Group
interaction, F(44, 462) = 2.39.
1.2.2. Extended-delay test 1
Accuracy during the first retention test, in which the base-
line delay was 0 s, are shown in Fig. 2. All three groups
showed a choose-short tendency at both extended delays,
although this tendency was more marked in group 0sF than in
either of the groups trained with a nonzero delay. A Sample
Duration × Delay × Group ANOVA revealed significant main
effects of sample duration, F(1, 21) = 12.50, and delay, F(2,
42) = 456.38. A significant Delay × Group interaction, F(4,
42) = 7.74, reflected the finding that accuracy at the 0-s delay was
highest in group 0sF whereas this group had the lowest accuracy
Fig. 2. Percentage of correct responses as a function of sample duration and
delay during the first extended-delay test in each group in Experiment 1. Error
bars show S.E.M.
Fig. 3. Percentage of correct responses as a function of sample duration and
delay during second extended-delay test in each group in Experiment 1. Error
bars show S.E.M.
at the extended delays. A significant Sample Duration × Delay
interaction, F(2, 42) = 12.52, revealed that the overall choose-
short tendency was significant. Although the Sample Dura-
tion × Delay × Group interaction was not significant, planned
separate Sample Duration × Delay ANOVAs revealed a signifi-
cant choose-short effect in groups 0sF and 2sF, Fs(2, 14) = 5.62
and 4.15, respectively, but not in group 2sV, F(2, 14) = 3.37.
1.2.3. Extended-delay test 2
Accuracy during the second retention test, in which the base-
line delay was 2 s, are shown in Fig. 3. All three groups showed
a choose-short tendency at both extended delays. This tendency
was largest in group 0sF and smallest in group 2sF. A Sample
Duration × Delay × Group ANOVA revealed significant main
effects of sample duration, F(1, 21) = 14.08, and delay, F(2,
42) = 318.24. The Delay × Group interaction was again sig-
nificant, F(4, 42) = 4.60, as was the Sample Duration × Delay
interaction, F(2, 42) = 18.53. In contrast to test 1, the Sample
Duration × Delay × Group interaction, F(4, 42) = 3.89, was sig-
nificant. Planned separate Sample Duration × Delay ANOVAs
revealed that the choose-short effect was statistically reliable in
group 0sF, F(2, 14) = 24.98, but not in either group 2sF or 2sV,
Fs(2, 14) = 1.23 and 2.41, respectively.
1.2.4. Sample response rates
Responses per minute to each of the two samples was com-
puted in each of the extended-delay tests. In test 1 rates on
short- and long-sample trials, respectively, were 35.4 and 59.4
in group 0sF, 50.4 and 58.3 in group 2sF, and 63.2 and 45.0
in group 2sV. Two tailed t-tests revealed that none of these
differences in response rate were statistically reliable (largest
t(7) = 1.45). Moreover, although response rate was numerically
higher on long-sample than on short-sample trials in group 0sF,
two birds who demonstrated the opposite difference in response
rate nonetheless demonstrated a large choose-short effect. In
test 2 rates on short- and long-sample trials, respectively, were
32.6 and 33.7 in group 0sF, 62.1 and 43.4 in group 2sF, and
 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
50.1 and 25.3 in group 2sV. In test 2, the difference in rate of
responding to the two samples had largely disappeared in group
0sF and was again statistically nonsignificant (t(7) < 1). More-
over, the four birds that demonstrated a higher rate of responding
on short- than on long-sample trials all demonstrated a choose-
short effect. Rate of responding was statistically higher on short-
than on long-sample trials in groups 2sF, t(7) = 3.28, and 2sV,
t(7) = 2.50.
1.3. Discussion
Both extended-delay tests provided some evidence that train-
ing with a nonzero delay reduces the magnitude of the choose-
short effect. In both tests, the choose-short effect was larger in
magnitude in group 0sF than in either of the groups trained with a
nonzero delay. Moreover, the choose-short effect demonstrated
in group 0sF was statistically reliable in both tests. In contrast,
the choose-short effect was statistically reliable only in test 1 in
group 2sF and it failed to reach significance in either test in group
2sV. There was little evidence that whether the delay was fixed
or variable had any major effect on the magnitude of the choose-
short effect. The magnitude of the choose-short effect was very
similar in groups 2sF and 2sV in test 1, and was only moderately
higher in group 2sV than in group 2sF in test 2. In the interest
of expository economy, further discussion of Experiment 1 will
postponed until Section 3.
2. Experiment 2
Experiment 1 provided some evidence that training with a
nonzero delay reduces the magnitude of the choose-short effect.
It was anticipated that use of longer nonzero delays might reveal
that effect even more dramatically. It was also anticipated that
use of a wider range of variable delays would allow any effect
of variability in delay to be more readily detected. The same
subjects and task employed in Experiment 1 was also employed
in the present experiment. Training sessions were identical to
those in Experiment 1 for group 0sF. The length of the fixed
delay was increased from 2 to 5 s on all training trials in group
2sF, now referred to as group 5sF. Finally, both the length and
range of the variable delay was increased in group 2sV, now
referred to as group 5sV. In this group, delays ranged from 2 to
8 s with a mean of 5 s.
After accuracy had stabilized in all three groups, two suc-
cessive extended-delay tests were conducted. The first involved
a 0-s baseline and delays of 5 and 20 s. Because 5-s was the
training delay in group 5sF and was the mean of the training
delays in group 5sV, little asymmetry in retention on short- and
long-sample trials was anticipated at this delay in either group.
Interest in these groups focused on the 20-s delay and, in partic-
ular, whether the choose-short effect at the 20-s delay would be
larger in group 0sF than in either group 5sF or 5sV.
It could argued that the larger magnitude choose-short effect
obtained in group 0sF relative to that obtained in either group 2sF
or 2sV in Experiment 1, and a potential similar effect obtained
in test 1 of the present experiment, might reflect the fact that the
extended delays exceeded the baseline training delay to a greater
5
extent in group 0sF than in the groups trained with a nonzero
delay. In the case of test 1 in the present experiment, a 20-s
extended delay exceeded the training baseline by 20 s in group
0sF, but by only 15 s in groups 5sF and 5sV. The second retention
test was designed to test retention at the baseline training delay
and at an extended delay that exceeded the training baseline by
15 s in each group. For all subjects the baseline delay during
the second retention test corresponded to the baseline delay (or
mean thereof) employed during training. Hence, the baseline
delay during test 2 sessions was 0 s in group 0sF and was 5 s
in groups 5sF and 5sV. The extended delay for all subjects was
15 s greater than the baseline delay. Hence, the extended delay
was 15 s in group 0sF and was 20 s in groups 5sF and 5sV.
2.1. Method
2.1.1. Subjects and apparatus
The subjects and apparatus were the same as in
Experiment 1.
2.1.2. Procedure
2.1.2.1. Training. For subjects in group 0sF, sessions were
identical to training sessions in Experiment 1. For subjects in
group 2sF, now group 5sF, sessions were identical to training
sessions in Experiment 1 except that the fixed nonzero delay
presented on each trial was 5 s rather than 2 s. For subjects in
group 2sV, now group 5sV, sessions were identical to training
sessions in Experiment 1 except that the range of delays was
2–8 s (rather than 1–3 s) and the mean of those delays was 5 s
rather than 2 s. On each trial, the delay length in group 5sV
was computed by generating a random double-precision value
between 0 and 1 and equally often multiplying that number by
1, 2 or 3. The resulting value was then equally often added or
subtracted from 5, thus producing a rectangular distribution of
delays within the 2–8 s range. Subjects in all 3 groups received
32 sessions of training.
2.1.2.2. Extended-delay test 1. Sessions were identical to those
of extended-delay test 1 in Experiment 1 except that the 10-s
delay was replaced by a 5-s delay. Hence, the 3 delays employed
were 0 s (baseline, 75% of trials), 5 s (12.5% of trials) and 20 s
(12.5% of trials). Eight testing sessions were conducted and each
test session after the first was preceded by two sessions identical
to those of training.
2.1.2.3. Extended-delay test 2. Four training sessions inter-
vened between the end of the first and beginning of the second
extended-delay test. This test was identical to the first extended-
delay test with 2 exceptions. First, the baseline delay in testing
corresponded to the baseline delay (or mean thereof) used in
training. Hence, the baseline delay was 0 s in group 0sF and
was 5 s in groups 5sF and 5sV. Second, only 1 extended delay
was employed and this delay was 15 s longer than the base-
line delay. Hence, the extended delay was 15 s in group 0sF
and was 20 s in groups 5sF and 5sV. In all 3 groups, the base-
line delay was presented on 75% of trials and the extended
delay was presented on the remaining 25% of trials. Eight
6 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
testing sessions were conducted and each test session after
the first was preceding by two sessions identical to those of
training.
2.2. Results
2.2.1. Training
As expected, accuracy remained high and stable through-
out the 32 sessions of training in group 0sF. Both groups that
now experienced longer training delays showed an increase in
accuracy across blocks of four sessions and reached asymptote
in Block 5 or 6. A Sample Duration × Block × Group ANOVA
performed on the data from groups 5sF and 5sV revealed a sig-
nificant effect of block, F(7, 98) = 21.50, and a significant effect
of group, F(1, 14) = 6.87, reflecting higher overall accuracy in
group 5sV (86.8) than in group 5sF (79.0).
2.2.2. Extended-delay test 1
Accuracy during the first retention test, in which the base-
line delay was 0 s, are shown in Fig. 4. Group 0sF showed a
marked choose-short effect at both delays that exceeded the
training baseline delay (i.e., 5 and 20 s). As expected, neither
group 5sF nor 5sV, for which the training delay was, or aver-
aged, 5 s showed a marked bias at the 5-s test delay. At the
20-s extended delay, both groups showed a modest choose-short
effect.
A Sample Duration × Delay × Group ANOVA revealed sig-
nificant main effects of sample duration, F(1, 21) = 21.75, delay,
F(2, 42) = 133.19, and group, F(2, 21) = 5.78, the latter reflect-
ing higher accuracy in groups 5sF (72.7) and 5sV (74.5) than in
group 0sF (67.4). A significant Delay × Group interaction, F(4,
42) = 8.60, reflected the finding that accuracy at the 0-s delay
was highest in group 0sF whereas this group had the lowest
accuracy at the 5- and 20-s delays. A significant Sample Dura-
tion × Group interaction, F(2, 21) = 4.41, reflected the finding
that the accuracy difference between short- and long-sample tri-
Fig. 4. Percentage of correct responses as a function of sample duration and
delay during the first extended-delay test in each group in Experiment 2. Error
bars show S.E.M.
als was considerably larger in group 0sF (20.7 percentage points)
than in either group 5sF (4.7 percentage points) or group 5sV
(7.5 percentage points). A significant Sample Duration × Delay
interaction, F(2, 42) = 6.12, revealed that the overall choose-
short tendency was significant. Although the Sample Dura-
tion × Delay × Group interaction was not significant, planned
separate Sample Duration × Delay ANOVAs revealed a signifi-
cant choose-short effect in group 0sF, F(2, 14) = 8.59, but not in
groups 5sF or 5sV, both Fs < 1.
2.2.3. Extended-delay test 2
Accuracy during the second retention test, in which the base-
line delay was equivalent to that in training (or the mean of the
training delays) in all three groups are shown in Fig. 5. Group
0sF showed a marked choose-short effect at the extended delay
of 15 s. In contrast, groups 5sF or 5sV demonstrated only a very
modest choose-short tendency at the extended delay of 20 s.
A Sample Duration × Delay × Group ANOVA revealed sig-
nificant main effects of sample duration, F(1, 21) = 14.65, delay,
F(1, 21) = 710.60, and group, F(2, 21) = 7.16, the latter likely
reflecting higher accuracy in groups 0sF (74.4) and 5sV (72.7)
than in group 5sF (67.1). A significant Delay × Group inter-
action, F(2, 21) = 12.51, reflected the finding that accuracy
at the baseline delay was highest in group 0sF whereas this
group had the lowest accuracy at the extended delay. A sig-
nificant Sample Duration × Group interaction, F(2, 21) = 3.52,
reflected the finding that the accuracy difference between short-
and long-sample trials was considerably larger in group 0sF
(20.5 percentage points) than in either group 5sF (3.9 percent-
age points) or group 5sV (6.9 percentage points). A significant
Sample Duration × Delay interaction, F(1, 21) = 11.98, revealed
that the overall choose-short tendency was significant, and a
significant Sample Duration × Delay × Group interaction, F(2,
21) = 3.71, revealed that the magnitude of the choose-short effect
differed between groups. Separate Sample Duration × Delay
ANOVAs revealed a significant choose-short effect in group 0sF,
Fig. 5. Percentage of correct responses as a function of sample duration and
delay during second extended-delay test in each group in Experiment 2. “B”
(baseline) is 0 s in group 0sF and 5 s in groups 5sF and 5sV. “B + 15 s” is 15 s in
group 0sF and 20 s in groups 5sF and 5sV. Error bars show S.E.M.
 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
F(1, 7) = 23.85, but not in groups 5sF or 5sV, Fs(1, 7) = 2.60 and
<1, respectively.
2.2.4. Sample response rates
Responses per minute to each of the two samples was com-
puted in each of the extended-delay tests. In test 1 rates on short-
and long-sample trials were highly similar in both groups 0sF
(43.0 and 44.3) and 5sF (51.3 and 53.3). In group 5sV, rate
of responding was somewhat higher on short- than on long-
sample trials (48.3 and 33.7). Two tailed t-tests revealed that
none of these differences in response rate were statistically reli-
able (largest t(7) = 1.13). In test 2 response rates were higher
on short- than long-sample trials in all three groups (0sF: 46.9
and 41.1, 5sF: 56.1 and 33.8, and 5sV: 46.2 and 24.1), and two-
tailed t-tests revealed that none of the differences in response
rates were significant, largest t(7) = 2.29.
2.3. Discussion
Both extended-delay tests provided evidence that training
with a nonzero delay reduces the magnitude of the choose-short
effect. In both tests, the choose-short effect was larger in mag-
nitude in group 0sF than in either of the groups trained with a
nonzero delay. Moreover, the choose-short effect demonstrated
in group 0sF was statistically reliable in both tests. In contrast,
the choose-short effect was not statistically reliable in either
retention test in group 5sF or group 5sV. Importantly, data from
the second retention test demonstrate that training with a nonzero
delay reduces the magnitude of the choose-short effect even
when groups are equated on the extent to which the extended
delay exceeds the baseline training delay. In the second retention
test, the extended delay exceed the baseline training by 15 s in
all three groups. Nonetheless, a marked and statistically reliable
choose-short effect was obtained only in group 0sF.
There was little evidence that whether the delay was fixed or
variable had any major effect on the magnitude of the choose-
short effect. The magnitude of the choose-short effect was only
slightly larger in group 5sV than in group 5sF in both retention
tests. Accuracy during training was reliably higher in group 5sV
than in 5sF. Moreover, accuracy at the 5-s delay in both extended-
delay tests was somewhat higher in group 5sV than in group 5sF.
It may be that the relatively short delays (e.g., between 2 and
4 s) experienced by subjects in group 5sV facilitated acquisition
and maintenance of accurate performance.
3. General discussion
As noted in the introduction to this article, the literature
concerning the effects of training with a nonzero delay on the
magnitude of the choose-short effect is certainly mixed. Some
studies suggest that such training has little, if any, effect on the
magnitude of the choose-short effect (Grant and Kelly, 1998,
Experiment 1; Kelly and Spetch, 2000; Spetch, 1987; Spetch
and Rusak, 1989, 1992). However, other studies have failed
to find a statistically reliable choose-short effect after train-
ing that included one or more nonzero delays (Dorrance et al.,
2000, Experiment 1; Grant and Talarico, 2004, Experiment 1;
7
Talarico and Grant, 2006, Experiment 1). Moreover, the find-
ing that retraining using a 0-s training delay pigeons who had
previously failed to demonstrate a choose-short effect following
training with variable delays resulted in a robust and statistically
reliable choose-short effect (Talarico and Grant, 2006, Experi-
ment 2), suggests that use of nonzero delays in training does
reduce the magnitude of the choose-short effect, at least under
some conditions.
The research reported in the present article is the first to
provide a within-experiment comparison of the magnitude of
the choose-short effect after training with a 0-s delay with that
obtained after training with a fixed or variable nonzero delay.
None of the four retention tests reported in this article provide
any evidence that whether a nonzero training delay is fixed or
variable has any marked effect on the magnitude of the choose-
short effect. On the other hand, whether a zero or nonzero
training delay was employed strongly influenced the magni-
tude of the choose-short effect. In all four retention tests, the
magnitude of the choose-short effect was greater in group 0sF
than in either group trained with a nonzero delay. Moreover, the
choose-short effect was statistically reliable in group 0sF in all
extended-delay tests whereas, with the exception of group 2sF
in test 1 in Experiment 1, a reliable choose-short effect was not
obtained in groups trained with a nonzero fixed or variable delay.
Finally, comparing the magnitude of the choose-short tendency
in the two groups trained with a nonzero delay in Experiments
1 and 2 suggests that increasing the delay, or mean delay, from
2 to 5 s reduced the choose-short tendency.
It may be noted that some studies have reported a choose-long
effect, a tendency to choose the long-associated comparison on
the majority of trials, when birds are tested at delays shorter than
the training delay (e.g., Spetch, 1987; Spetch and Rusak, 1989).
Given that choose-short and choose-long effects are viewed as
arising from the same set of processes (e.g., Spetch, 1987), it is
not surprising that groups failing to demonstrate a robust choose-
short effect at extended delays in the present experiments also
failed to demonstrate a choose-long tendency at delays shorter
than the training delay.
Although there are no doubt several procedural differences
between studies suggesting that training delays have little effect
on the choose-short effect and those suggesting that training
delays reduce the magnitude of the choose-short effect, perhaps
the most obvious of those differences involves the event that con-
veys the duration. In particular, with the exception of Grant and
Kelly (1998), studies that have obtained a robust choose-short
effect after training with a nonzero delay (Kelly and Spetch,
2000; Spetch, 1987; Spetch and Rusak, 1989, 1992) have used
different durations of houselight or access to food. In contrast,
studies which have obtained a reduced or eliminated choose-
short effect after training with delays (Dorrance et al., 2000;
Grant and Talarico, 2004; Talarico and Grant, 2006; the exper-
iments reported in this article) have used different durations of
exposure to keylight. The lone exception was the study reported
by Grant and Kelly (1998, Experiment 1) in which a reliable
choose-short effect was obtained at extended delays of 15 and
30 s after training with 2- and 8-s keylight samples and variable
delays in the range of 1–3 s (the same procedure as employed
8 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
in group 2sV in Experiment 1 of the present series). The Grant
and Kelly finding is perhaps not particularly surprising given the
present findings that use of training delays reduces the choose-
short tendency in relation to their length, but even a 5-s training
delay does not entirely eliminate the choose-short tendency.
Three primary accounts of the choose-short effect have been
offered, and each appears to encounter some difficulty explain-
ing the three primary findings that have emerged from research
on the effect of training delays on the choose-short effect. Those
findings are that: (1) training with delays reduces the choose-
short effect when keylight samples are employed; (2) the mag-
nitude of the choose-short effect appears to be affected little, if
at all, by whether the training delay is fixed or variable; and (3)
training with delays has little effect on the choose-short effect
when food or houselight samples are employed. As discussed
below, none of the three accounts of the choose-short effect
appears to offer a ready interpretation of why the effect of train-
ing delays on the choose-short effect should be dependent upon
the nature of the event that conveys the duration (i.e., keylight
exposure or food access). In addition, the present finding that
whether training delays are fixed or variable has little effect on
the choose-short effect appears to present a particular challenge
to the subjective shortening account of the choose-short effect.
The first account of the choose-short effect was developed by
Spetch and Wilkie (1983; for a review, see Grant et al., 1997) and
postulates that remembered duration subjectively shortens as the
event becomes more temporally remote. The subjective short-
ening model maintains that pigeons code durations analogically
in a retrospective manner (e.g., counts that accumulate during
an interval) rather than categorically (e.g., short and long). The
choose-short effect is produced by the shortening of this analog-
ical representation (e.g., loss of counts) during a delay interval
greater than that of training. According to this interpretation, the
working memory representation of the sample becomes system-
atically shorter during a delay interval that intervenes between
sample termination and presentation of the comparison stimuli.
Testing a pigeon immediately following termination of a long
sample (i.e., at a 0-s delay) results in a high proportion of correct
choices because the analogical representation of the sample in
working memory corresponds closely to the reference memory
representation of a long sample. Because the working memory
representation is held to subjectively shorten, at longer delays
(e.g., 10 and 20 s) the working memory representation of a long
sample corresponds less closely to the reference memory rep-
resentation of a long sample and corresponds more closely to
the reference memory representation of a short sample, thereby
leading to an increased tendency to choose the short-associated
comparison as delay increases.
The operation of a subjective shortening process would ren-
der an analogical coding process less effective in a variable delay
training procedure than in one with a fixed, zero or nonzero,
delay. In the present case, for instance, there would be consid-
erably fewer counts in working memory on short and long trials
with a 7- or 8-s delay than on trials with a 2- or 3-s delay. More-
over, the amount of subjective shortening that would occur on
long-sample trials with a 7- or 8-s delay might render an analogi-
cal representation difficult to discriminate from one generated on
a short-sample trial with a 2- or 3-s delay. Hence, a variable train-
ing delay would increase the difficulty of discriminating between
analogical representations generated on short- and long-sample
trials, and hence might encourage birds to abandon analogical
coding and to adopt categorical coding (see Grant and Kelly,
1998 for further discussion). Although this perspective clearly
anticipates that variable training delays would reduce the mag-
nitude of the choose-short effect, particularly as the range of
delays increases, it is less clear why a fixed delay would be
equally effective in doing so. It is also not immediately apparent
why use of training delays has little effect on the choose-short
effect when the durations involve timed access to food or house-
light illumination.
The results reported in this article can also be considered in
relation to two recently developed accounts that do not appeal
to processes of analogical coding and subjective shortening to
explain the choose-short effect. First, Zentall and his associates
(Dorrance et al., 2000; Sherburne et al., 1998) have proposed
an ambiguity account of the choose-short effect. According to
this view, the choose-short effect arises when pigeons are unable
to discriminate between a delay interval and the intertrial inter-
val and hence treat the delay as an intertrial interval. Because
the comparisons have never been presented at the end of an
intertrial interval, the animal responds as if no sample had been
presented. Because no sample is judged to be more similar to
a short sample than to a long sample, pigeons tend to choose
the short-associated comparison following a delay. To account
for the findings reported in this article, the ambiguity hypoth-
esis could maintain that use of a nonzero training delay helps
the animal to discriminate between the intertrial interval and the
delay interval. After nonzero-delay training, therefore, the ani-
mal is less likely to treat an extended-delay trial as a no-sample
trial, and hence the choose-short effect is weakened in magni-
tude or eliminated. This account also makes no strong prediction
in regard to the effect of variability in delays on the magnitude of
the choose-short effect. If the role of training delays is to disam-
biguate retention intervals from intertrial intervals, it is unclear
why training delays would appear to do so when the durations
are keylight presentations but not when the durations are food
presentations or houselight illuminations.
A second alternative account of the choose-short effect has
been offered by Wixted and his associates (Dougherty and
Wixted, 1996; Gaitan and Wixted, 2000). This detection account
maintains that whenever samples differ markedly in salience,
pigeons attempt to retrieve the memory for the more salient at
testing, responding to the alternative comparison in the absence
of memory for the more salient sample. On this view, the choose-
short effect arises because the long sample is more salient and,
therefore, presentation of the comparisons results in a search of
memory for the long sample. When the long sample memory
is not retrieved, either because the long sample was not pre-
sented (i.e., on a short-sample trial) or because the presentation
of a long sample had been forgotten (i.e., at longer delays), the
pigeon chooses the comparison associated with the short sam-
ple. To account for the present findings, the detection account
would need to argue that a nonzero delay during training causes
animals to abandon a strategy of trying to detect memory for the
 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
more salient sample at comparison onset and instead adopt an
alternative strategy. The problem with this tactic is that there is
no obvious reason why a nonzero delay during training would
cause pigeons to abandon a detection strategy. Moreover, it is
unclear why training delays would have different effects when
the durations are keylight presentations than when the durations
are presentations of food or houselight.
Responding to the two samples was recorded in the present
experiments to assess the suggestion by Dorrance et al. (2000)
that choice in the duration matching task might be controlled by
differential sample responding. Because only group 0sF demon-
strated a robust choose-short effect in all four extended-delay
tests reported in this article, only sample responding from this
group is relevant to Dorrance et al.’s suggestion. Recall that this
suggestion entails the notion that the choose-short effect might
be mediated by differential sample responding to the extent that
long samples generate higher rates of responding or more total
pecking than do short samples. If a low rate of responding or a
low number of recently emitted pecks came to control choice of
the short-associated comparison, pigeons would tend to choose
that comparison after a delay because there is little responding
during the delay. Rate of responding to the long sample was
numerically, but not statistically, higher only in test 1 in the
first experiment. In test 2 in that experiment and in test 1 of the
second experiment rate of responding to the two samples was
highly similar. In test 2 in the second experiment rate of respond-
ing was numerically, but not statistically, higher on short- than
long-sample trials. Hence, these data are not consistent with
the idea that differential rate of sample responding plays a role
in producing the choose-short effect. Moreover, although most
birds did emit more keypecks in the presence of long samples
than in the presence of short samples, this was not the case for
one bird in group 0sF that nonetheless demonstrated a robust
choose-short effect in all four retention tests. In addition, other
birds in this group typically demonstrated very little pecking dur-
ing sample presentation (i.e., less than 10 responses per minute),
hence rendering unlikely the possibility that differential numbers
of pecks on long- and short-sample trials could have served as a
cue for comparison choice. Thus, the data from the experiments
reported in this article suggest that differential sample respond-
ing, either in terms of rate or total number of keypecks, is not
necessary to produce the choose-short effect.
A reviewer of an earlier version of this article asked whether
the lack of a robust CSE at extended delays after training with
a nonzero delay might reflect individual differences in timing
strategy. In particular, if some of the birds trained with a nonzero
delay stopped timing when the keylight terminated whereas oth-
ers continue to time during the delay until presentation of the
comparisons, at extended delays the former birds would demon-
strate a choose-short effect whereas the latter would demonstrate
a choose-long effect. On this view, the group result would reflect
the number of birds adopting each timing strategy. Inspection of
individual data from the two groups trained with a nonzero delay
during each of the four retention tests revealed that no bird in
either of these groups consistently demonstrated a choose-long
effect at extended delays. Moreover, if some birds were choosing
the short-associated comparison and others the long-associated
9
comparison at extended delays, then the S.E.M.s at those delays
would be markedly higher in the groups trained with a nonzero
delay than in the group trained with a zero delay. Inspection of
Figs. 2–5 reveals that that was not the case. Hence, the failure to
obtain a robust choose-short effect after training with a nonzero
delay does not reflect individual differences in timing strategy.
At present, therefore, it is unclear why training with a nonzero
delay reduces the magnitude of the choose-short effect when dif-
ferent durations of keylight are used as samples but appears not
to do so when different durations of houselight or food access
are used as samples. A perhaps likely next step in the analysis is
a within experiment, and preferably within subject, comparison
of the effects of using a nonzero training delay on the choose-
short effect obtained with keylight durations versus houselight or
food access durations. If such an experiment revealed qualitative
differences in the effect of a nonzero training delay on the mag-
nitude of the choose-short effect, it would provide convincing
evidence that the mechanisms mediating discrimination or reten-
tion in duration matching tasks differ as a function of whether
keylight or houselight/food access samples are employed.
Acknowledgement
This research was supported by a grant from the Natural
Sciences and Engineering Research Council of Canada (OGP
0443).
I was William A. Roberts first graduate student and I would
like to thank Bill for helping me develop the knowledge and
skills necessary to a productive research career.
References
Blough, D.S., 1959. Delayed matching in the pigeon. J. Exp. Anal. Behav.
2, 151–160.
Colwill, R.M., 1984. Controlled processing in pigeons. Anim. Learn. Behav.
12, 285–291.
Dorrance, B.R., Kaiser, D.H., Zentall, T.R., 2000. Event-duration discrimi-
nation by pigeons: the choose-short effect may result from retention-test
novelty. Anim. Learn. Behav. 28, 344–353.
Dougherty, D.H., Wixted, J.T., 1996. Detecting a nonevent: delayed presence-
versus-absence discrimination in pigeons. J. Exp. Anal. Behav. 65, 81–92.
Gaitan, S.C., Wixted, J.T., 2000. The role of “nothing” in memory for event
duration in pigeons. Anim. Learn. Behav. 28, 147–161.
Grant, D.S., 1975. Proactive interference in pigeon short-term memory. J.
Exp. Psychol. Anim. Behav. Process. 1, 207–220.
Grant, D.S., 1976. Effect of sample presentation time on long-delay matching
in the pigeon. Learn. Motiv. 7, 580–590.
Grant, D.S., 1991. Symmetrical and asymmetrical coding of food and no-
food samples in delayed matching in pigeons. J. Exp. Psychol. Anim.
Behav. Process. 17, 186–193.
Grant, D.S., Kelly, R., 1996. The role of minimum wait time and sample
discriminability in the coding of event duration in pigeons. Learn. Motiv.
27, 243–259.
Grant, D.S., Kelly, R., 1998. The effect of variable-delay training on coding
of event duration in pigeons. Learn. Motiv. 29, 49–67.
Grant, D.S., Roberts, W.A., 1973. Trace interaction in pigeon short-term
memory. J. Exp. Psychol. 101, 21–29.
Grant, D.S., Roberts, W.A., 1976. Sources of retroactive inhibition in pigeon
short-term memory. J. Exp. Psychol. Anim. Behav. Process. 2, 1–16.
Grant, D.S., Spetch, M.L., 1991. Pigeons’ memory for event duration: differ-
ences between choice and successive matching tasks. Learn. Motiv. 22,
180–199.
10 D.S. Grant / Behavioural Processes xxx (2006) xxx–xxx
Grant, D.S., Spetch, M.L., 1993. Analogical and nonanalogical coding of
samples differing in duration in a choice-matching task in pigeons. J.
Exp. Psychol. Anim. Behav. Process. 19, 15–25.
Grant, D.S., Spetch, M.L., 1994. The role of asymmetrical coding of duration
samples in producing the choose-short effect in pigeons. Learn. Motiv.
25, 413–430.
Grant, D.S., Spetch, M.L., Kelly, R., 1997. Pigeons’ coding of event duration
in delayed matching-to-sample. In: Bradshaw, C.M., Szabadi, E. (Eds.),
Time and Behavior 4: Psychological and Neurobehavioural Analyses.
Elsevier Science B.V., Amsterdam, Netherlands, pp. 217–264.
Grant, D.S., Talarico, D.C., 2004. Processing of empty and filled time inter-
vals in pigeons. Learn. Behav. 32, 477–490.
Kelly, R., Spetch, M.L., 2000. Choice biases in delayed matching-to-sample
duration with pigeons: manipulations of ITI and delay illumination. Q. J.
Exp. Psychol. 53B, 309–323.
Kraemer, P.J., Mazmanian, D.S., Roberts, W.A., 1985. The choose-short effect
in pigeon memory for stimulus duration: subjective shortening versus
coding models. Anim. Learn. Behav. 13, 349–354.
Kraemer, P.J., Roberts, W.A., 1984. Short-term memory for visual and audi-
tory stimuli in pigeons. Anim. Learn. Behav. 12, 275–284.
Maki, W.S., 1979. Pigeons’ short-term memories for surprising vs. expected
reinforcement and nonreinforcement. Anim. Learn. Behav. 7, 31–37.
Roberts, W.A., 1972. Short-term memory in the pigeon: effects of repetition
and spacing. J. Exp. Psychol. 94, 74–83.
Roberts, W.A., 1995. Simultaneous numerical and temporal processing in the
pigeon. Curr. Dir. Psychol. Sci. 4, 47–51.
Roberts, W.A., 2005. How do pigeons represent number? Studies of number
scale bisection. Behav. Process. 69, 33–43.
Roberts, W.A., Cheng, K., Cohen, J.S., 1989. Timing light and tone signals
in pigeons. J. Exp. Psychol. Anim. Behav. Process. 15, 23–35.
Roberts, W.A., Grant, D.S., 1974. Short-term memory in the pigeon
with presentation time precisely controlled. Learn. Motiv. 5, 393–
409.
Roberts, W.A., Grant, D.S., 1978a. An analysis of light-induced retroactive
inhibition in pigeon short-term memory. J. Exp. Psychol. Anim. Behav.
Process. 4, 219–236.
Roberts, W.A., Grant, D.S., 1978b. Interaction of sample and comparison
stimuli in delayed matching-to-sample with the pigeon. J. Exp. Psychol.
Anim. Behav. Process. 4, 68–82.
Roberts, W.A., Macuda, T., Brodbeck, D.R., 1995. Memory for number of
light flashes in the pigeons. Anim. Learn. Behav. 23, 182–188.
Roberts, W.A., Mitchell, S., 1994. Can a pigeon simultaneously process tem-
poral and numerical information? J. Exp. Psychol. Anim. Behav. Process.
20, 66–78.
Santi, A., Bridson, S., Ducharme, M.J., 1993. Memory codes for temporal
and nontemporal samples in many-to-one matching by pigeons. Anim.
Learn. Behav. 21, 120–130.
Santi, A., Hornyak, S., Miki, A., 2003. Pigeons’ memory for empty and filled
time intervals signaled by light. Learn. Motiv. 34, 282–302.
Sherburne, L.M., Zentall, T.R., Kaiser, D.H., 1998. Timing in pigeons: the
choose-short effect may result from pigeons’ “confusion” between delay
and intertrial intervals. Psychon. Bull. Rev. 5, 516–522.
Spetch, M.L., 1987. Systematic errors in pigeons’ memory for event duration:
interaction between training and test delay. Anim. Learn. Behav. 15, 1–5.
Spetch, M.L., Rusak, B., 1989. Pigeons’ memory for event duration: intertrial
interval and delay effects. Anim. Learn. Behav. 17, 147–156.
Spetch, M.L., Rusak, B., 1992. Temporal context effects in pigeons’ memory
for event duration. Learn. Motiv. 23, 117–144.
Spetch, M.L., Wilkie, D.M., 1982. A systematic bias in pigeons’ memory for
food and light durations. Behav. Anal. Lett. 2, 267–274.
Spetch, M.L., Wilkie, D.M., 1983. Subjective shortening: a model of pigeons’
memory for event duration. J. Exp. Psychol. Anim. Behav. Process. 9,
14–30.
Talarico, D.C., Grant, D.S., 2006. Effect of training delays and start and
stop markers on the choose-short effect in pigeons. Behav. Process. 71,
98–106.