Pigeons Exhibit Contextual Cueing to Both Simple and Complex Backgrounds

Abstract

Repeated pairings of a particular visual context with a specific location of a target stimulus facilitate target search in humans. We explored an animal model of this contextual cueing effect using a novel Cueing-Miscueing design. Pigeons had to peck a target which could appear in one of four possible locations on four possible color backgrounds or four possible color photographs of real-world scenes. On 80% of the trials, each of the contexts was uniquely paired with one of the target locations; on the other 20% of the trials, each of the contexts was randomly paired with the remaining target locations. Pigeons came to exhibit robust contextual cueing when the context preceded the target by 2 s, with reaction times to the target being shorter on correctly-cued trials than on incorrectly-cued trials. Contextual cueing proved to be more robust with photographic backgrounds than with uniformly colored backgrounds. In addition, during the context-target delay, pigeons predominately pecked toward the location of the upcoming target, suggesting that attentional guidance contributes to contextual cueing. These findings confirm the effectiveness of animal models of contextual cueing and underscore the important part played by associative learning in producing the effect.

As do other discriminative stimuli, environmental contexts can be said to set the occasion for operant behaviors to be reinforced. However, an even more active role may be played by contexts, especially when they give clear direction to the behavior in question (Kelly & Spetch, 2012). Several studies by Spetch, Cheng, Blaisdell, and their colleagues have shown that visual landmarks can effectively guide animals toward the location of an invisible target (Cheng & Spetch, 1995; Leising & Blaisdell, 2009; Spetch, Cheng, & Mondloch, 1992). Thus, a pigeon might learn to peck 2 cm to the left of a small red square projected on a computer monitor in order to receive food reinforcement, with the location of those pecks being pinpointed by a touchscreen just in front of the monitor.

Computer control and touch technology can also be deployed to study an important and related form of contextual control—so-called contextual cueing. Within the realm of human vision science, Chun and Jiang (1998) discovered that the spatial configuration of objects in a visual display can serve as a powerful contextual cue, facilitating an individual's search for a visible target item. Chun and Jiang found that, when the precise location of a target was consistently paired with a particular configuration of distractor items in the visual display, reaction times (RTs) to process the target became reliably shorter than when the target appeared in newly generated configurations of those same distractor items on each trial. This RT benefit defines the contextual cueing effect. Chun and Jiang went on to hypothesize that contextual cueing results from an observer's attention being directed toward the location of the target on consistently-paired trials, owing to learned associations between the context and the location of the target.

This emphasis on associative learning in contextual conditioning suggests an interesting question: Do nonhuman animals also exhibit contextual cueing? Although animal models of contextual cueing have been under development for several years (for pigeons: Brooks, Rasmussen, Hollingworth, & Wasserman, 2008; for rhesus monkeys: Brooks, Dai, & Sheinberg, 2011), only the recent study by Goujon and Fagot (2013: for baboons) has so far been published.

Goujon and Fagot (2013) used the same basic methodology as Chun and Jiang. They trained 25 baboons to touch a single target stimulus (the letter T) embedded within configurations of seven distractor stimuli (the letter L) that either did or did not predict the target's location. Reliable contextual cueing—shorter RTs to predictive than to non-predictive configurations—was seen after only 24 training trials, and was retained after a 6-week delay.

Our own earlier work (Brooks et al., 2008) had found that pigeons too exhibit contextual cueing with the target-distractor paradigm of Chun and Jiang. So, in the present study, we report the results of a different variant of the contextual cueing task, in the hope of shedding further light on the situational generality and comparative mechanisms of contextual cueing.

Specifically, we devised a novel Cueing-Miscueing design to explore contextual cueing in pigeons. We compared pigeons’ latency to peck a target stimulus on two types of trials: correctly-cued trials and incorrectly-cued trials. We also compared two types of cueing contexts: solid color backgrounds which lacked any landmarks (Figure 1, top four panels) and photographic scene backgrounds which provided rich landmarks (Figure 1, bottom four panels).

Figure 1.

(Top) The uniformly colored contextual stimuli that were paired with the target stimulus superimposed in the four possible spatial locations: red (top-left), green (top-right), yellow (bottom-left), and purple (bottom-right). All four pigeons in the Color group received the same color context-target location assignments. (Bottom) The photographic scene contextual stimuli that were paired with the target stimulus superimposed in the four possible spatial locations for one of the four pigeons in the Scene group: house (top-left), mountain (top-right), lake (bottom-left), and cemetery (bottom-right). The other three pigeons in the Scene group had other counterbalanced scene context-target location pairings.

We expected that pairing specific target locations with particular contexts would encourage faster detection when the target occurred in the correctly-cued location than when it appeared in one of three possible incorrectly-cued locations; this correct-versus-incorrect cueing tactic has proven to be very effective in the study of human visual processing (e.g., Posner, Snyder, & Davidson, 1980). We also expected that the landmark-rich photographic scene backgrounds would support stronger contextual cueing effects than would the landmark-poor solid color backgrounds (Brockmole & Henderson, 2006). Landmarks close to invisible targets exert strong control over pigeons’ pecking behavior (Spetch, 1995), so we expected the same to hold true for visible targets as well. Indeed, without any internal landmarks at all, there was a chance that no contextual cueing would occur with the uniform color backgrounds.

In addition, we conducted contextual cue training with a 2-s gap between presentation of the context and presentation of the context plus the target (an instance so-called Stimulus Onset Asynchrony or SOA). We correspondingly monitored pecking during the delay to see if any anticipatory responding occurred during the SOA that might provide useful clues as to the nature of contextual cueing. In a final stage of the experiment, we systematically varied the SOA to see what effects might be observed on both anticipatory responding and target-directed responding.

Method

Animals

We kept eight feral pigeons at 85% of their free-feeding weights. The pigeons were divided into two groups: Color (colored backgrounds as the context) and Scene (scene photographs as the context). The birds had participated in several unrelated studies before beginning this project. All housing and training procedures had been approved by the Institutional Animal Care and Use Committee at the University of Iowa.

Apparatus

We used eight 36- × 36- × 41-cm conditioning chambers (detailed by Castro, Kennedy & Wasserman, 2010) located in dark rooms with continuous white noise. Each chamber was equipped with a 15-in LCD monitor behind a resistive touchscreen. The viewable portion of the screen was 28.5 cm × 17 cm. Pecks to the touchscreen were processed by a serial controller outside the chamber. A rotary dispenser delivered 45-mg food pellets through a vinyl tube into a plastic cup in the center of the rear wall opposite the touchscreen. Illumination during experimental sessions was provided by a houselight on the upper rear wall of the chamber. The pellet dispenser and houselight were controlled by a digital I/O interface. Each chamber was controlled by an iMac computer. Programs were developed in MatLab with Psychtoolbox-3 extensions (Brainard, 1997; Pelli,1997; http://psychtoolbox.org/).

Stimulus Materials

In Pretraining, we taught the eight pigeons to peck a target stimulus in four possible positions on the computer monitor. For the four pigeons later trained with color contexts (Color group), the Pretraining context was an otherwise black screen; for the four pigeons later trained with photograph contexts (Scene group), the Pretraining context was a single color photograph (snow and ice).

For the Color group, the contexts were four different colored backgrounds: red, green, yellow, and purple (Figure 1, top four panels). For the Scene group, the contexts were four different photographs: two (mountain and lake) depicted undeveloped nature scenes, whereas the other two (house and cemetery) also included humanmade items (Figure 1, bottom four panels). The 16.5- × 16.5-cm scenes were displayed in full color at a resolution of 555 × 555 pixels.

The same target stimulus was used in Pretraining and Training—an abstract shape consisting of nine white concentric circles superimposed on a dark gray square (3 × 3 cm). The target could appear in the top-left, top-right, bottom-left, and bottom-right corners of the Color and Scene contexts, 1.4 cm from the borders of the contextual image.

Pretraining

To familiarize the pigeons with the basic behavioral task before beginning Training, the birds were required to peck once at the target stimulus, which randomly appeared in one of four possible locations, in order to obtain food reinforcement. Pretraining comprised 240 trials (60 randomly scheduled presentations of the target in each corner of the contextual image) and continued for 2 or 3 days. Because of the pigeons’ prior experimental histories, they had no difficulty completing Pretraining without special measures having to be taken.

Training Design

Training involved two types of contextual cueing sessions: half of the sessions scheduled both correctly-cued and incorrectly-cued trials (Table 1), whereas the other half of the sessions scheduled only correctly-cued trials. In those sessions involving both correctly-cued and incorrectly-cued trials, the contextual cues correctly predicted the location of the target on 80% of the trials; on the remaining 20% of the trials, the contextual cues randomly and incorrectly predicted the target in one of the three remaining locations. Those incorrectly-cued trials were critical because they allowed us to see if the pigeons were faster to peck the target when it was presented in the most likely location than when it appeared in the other, much less likely locations. Of course, those incorrectly-cued trials necessarily weakened the predictive relation between the context and the target. That is precisely why we also trained the pigeons with interpolated sessions that provided unambiguous context-target information; we wanted to sustain the birds’ context-target associations as best we could even though those correctly-cued sessions afforded us little useful data (indeed, we do not include these sessions in any of our reported analyses). After giving the birds both of these types of sessions in strict alternation, we discontinued the 100%-predictive sessions to see if we could nevertheless maintain a robust contextual cueing effect. The results will show that we could.

Table 1.

Number of correctly-cued and incorrectly-cued trials in daily sessions

Correct Target Location	Number of Target Presentations in Four Possible Locations
	Upper Left	Upper Right	Lower Left	Lower Right
Upper Left	48	4	4	4
Upper Right	4	48	4	4
Lower Left	4	4	48	4
Lower Right	4	4	4	48

Here was our overall experimental regime. For the first 20 days, we alternated Correct/Incorrect-mixture sessions (odd days) and Correct-only sessions (even days). For the next 10 days, we presented just Correct/Incorrect-mixture sessions. Finally, we gave the pigeons 30 days of SOA testing with values of 0, 1, 2, 4, and 8 s to see what, if any, effect these different SOAs would have on the birds’ behavior.

Training Trials

Daily training sessions comprised 240 trials. Each trial began with a start stimulus—a black plus sign in the center of a small white square (3 × 3 cm)—in the center of the screen. After one peck to the start stimulus, the context appeared in the center of the black screen. The context was presented for 2 s, during which we recorded the number and location of any pecks that occurred—none were required. After 2 s, the target was superimposed on the context in one of four possible locations. The pigeons had to peck the target once, after which food was delivered (one 45-mg pellet) and a 10-s intertrial interval (ITI) ensued, during which the screen went black. During Training (and later SOA Testing), trials advanced only after the pigeons pecked the target stimulus. The time between target onset and the pigeons’ pecking it was recorded—RT was thus our prime behavioral measure. Training lasted 30 days.

SOA Testing

The procedure during SOA Testing remained the same, except that the delay between onset of the context and onset of the target varied among 0-, 1-, 2-, 4-, and 8-s values. On 0-s SOA days, the context and target were simultaneously presented; on 1-, 2-, 4-, and 8-s SOA days, the context was presented 1, 2, 4, and 8 s, respectively, before the target. Different SOA values were randomized across days within each block of 5 days for a total of 6 blocks. The order of SOA presentation was consistent for birds in each group; the four birds in the same group (Color or Scene) received the same SOA value on any given day, but the orders for the Color and Scene groups differed across blocks. SOA testing lasted 30 days.

Data analysis

We used analyses of variance (ANOVAs) in which an alpha level of .05 was adopted. Because of the factorial nature of all our ANOVAs, we report partial eta squared (${\eta }_p^2$) as the measure of standardized effect size and we include 95% confidence intervals (CIs) for ${\eta }_p^2$. In order to compute upper and lower boundaries of the CIs for ${\eta }_p^2$, we obtained the noncentrality parameter of the noncentral F distribution, as well as CIs for the noncentrality parameter using methods implemented in the Methods for the Behavioral, Educational, and Social Sciences (MBESS; Kelley, 2007a,b) R statistical package (R Development Core Team, 2007) and followed procedures described by Smithson (2003).

Results

RT Analysis

We excluded from all RT analyses any trials on which the RT to the target equaled or exceeded 20 s. Fewer than 0.27% of the trials were excluded across Training and Testing. Before statistical analysis, RTs were subjected to logarithmic transformation in order to bring the RT distributions into closer accord with normality.

Anticipatory Responding in SOA

The location of any anticipatory responding to the contexts alone was also analyzed; such responding was recorded in all sessions with 1-, 2-, 4-, and 8-s SOAs. Because we were concerned with where any recorded anticipatory pecks were directed, trials without any pecks (56.56% of all trials) were not considered. For this analysis, the entire area in which anticipatory pecks could be made was divided into equal-sized quadrants. If a peck was made in the quadrant where the target was next to appear, then it was deemed to be a correct anticipatory response; if the peck was made outside that quadrant, then it was deemed to be an incorrect anticipatory response. Because an incorrect anticipatory response was three times more likely to occur by chance than was a correct anticipatory response, incorrect anticipatory pecks were divided by 3 to equate their a priori probability of occurrence.

Reaction Time: Training

Figure 2 shows that RTs on Correctly-Cued and Incorrectly-Cued trials (on days with Correct/Incorrect-mixture training) very rapidly diverged for pigeons in both the Color (Figure 2, top) and Scene (Figure 2, bottom) groups. That marked RT disparity continued throughout interpolated training (in which 10 Correct/Incorrect-mixture sessions alternated with 10 Correct-only sessions) and beyond (in which 10 Correct/Incorrect-mixture sessions alone were given). Consistently faster RTs on Correctly-Cued than Incorrectly-Cued trials clearly document contextual cueing in pigeons trained with both color and scene backgrounds.

Figure 2.

(Top) Mean Log RT scores (ms) to Correctly-Cued targets and to Incorrectly-Cued targets for pigeons in the Color group as a function of days of training. (Bottom) Mean Log RT scores (ms) to Correctly-Cued targets and to Incorrectly-Cued targets for pigeons in the Scene group as a function of days of training. Error bars represent the standard error of the mean. The vertical line separates the initial 20 days (with alternating Correct/Incorrect-mixture sessions) from the subsequent 10 days (with Correct-only sessions). As a guide to relating Log RTs to raw RTs: Log RT of 6.0 = 403 ms, Log RT of 6.5 = 665 ms, Log RT of 7.0 = 1,097 ms, and Log RT of 7.5 = 1,808 ms.

RTs in the Color and Scene groups were separately submitted to repeated-measures ANOVAs with cueing (Correctly-Cued vs. Incorrectly-Cued) and day (1 to 20) as within-subjects factors. In the Color group, the main effect of day was significant, F(19, 57) = 3.29, p = .0003, ${\eta }_p^2=.52$, 95% CI [.15, .53], with RTs falling across days; so too was the main effect of cueing, F(1, 3) = 40.87, p = .0078, ${\eta }_p^2=.93$, 95% CI [.23, .96], with Correctly-Cued trials (M = 1,745 ms, SE = 77.28) supporting briefer RTs than Incorrectly-Cued trials (M = 2,851 ms, SE = 106.78). The Cueing × Day interaction was also significant, F(19, 57) = 2.83, p = .0013, ${\eta }_p^2=.48$, 95% CI [.10, .49], with Correctly-Cued vs. Incorrectly-Cued RTs diverging early in training. In the Scene group, the main effect of cueing was significant, F(1, 3) = 67.40, p = .0038, ${\eta }_p^2=.95$, 95% CI [.03, .97], with Correctly-Cued trials (M = 1,136 ms, SE = 50.29) supporting briefer RTs than Incorrectly-Cued trials (M = 2,353 ms, SE = 63.74); so too was the Cueing × Day interaction, F(19, 57) = 7.31, p < .0001, ${\eta }_p^2=.71$, 95% CI [.44, .72], with Correctly-Cued vs. Incorrectly-Cued RTs diverging early in training. The main effect of day was not significant, F < 1.

Because Goujon and Fagot had found evidence of contextual cueing in baboons after only 24 training trials (12 Predictive trials and 12 Random trials), we took a closer look at responding on Day 1 to see just how quickly our pigeons might have shown contextual cueing. The contextual cueing effect proved to be statistically significant on Day 1 in both the Color and Scene groups, but we were unable to detect reliable signs of learning even earlier in that session.

For this analysis, we first divided Day 1 into 4 blocks of 60 trials each. Second, we separately submitted RTs in the Color and Scene groups to repeated-measures ANOVAs with cueing (Correctly-Cued vs. Incorrectly-Cued) and block (1 to 4) as within-subjects factors. In the Color group, the main effect of cueing was significant, F(1, 3) = 169.92, p < .0001, ${\eta }_p^2=.98$, 95% CI [.24, .99], confirming that the color background supported contextual cueing on Day 1. The main effect of block was also significant, F(3, 9) = 4.77, p = .0287, ${\eta }_p^2=.61$, 95% CI [.05, .71]. Tukey post hoc analyses revealed that RTs in the second block were significantly lower than those in the first block, but not lower than those in the other two blocks. No interaction was observed, F < 1. In the Scene group, the main effect of cueing was also significant, F(1, 3) = 10.41, p = .0484, ${\eta }_p^2=.77$, 95% CI [.01, .86], confirming that the scene background supported contextual cueing on Day 1. No other main effect or interaction was significant. So, only 192 Correctly-Cued trials were necessary for pigeons to associate the four backgrounds with the four target locations (the other 48 were Incorrectly-Cued trials for comparison purposes), whether those backgrounds were colors or scenes.

Comparing the top and bottom panels of Figure 2 suggests that the contextual cueing effect may have been even stronger in the Scene group (Figure 2, bottom) than in the Color group (Figure 2, top). To properly evaluate this possibility, each pigeon's mean daily RT score on Correctly-Cued trials was subtracted from its mean daily RT score on Incorrectly-Cued trials. These RT difference scores are depicted in the top panel of Figure 3, with larger RT disparities signifying stronger contextual cueing. These RT disparities between Correctly-Cued and Incorrectly-Cued trials (on days with Correct/Incorrect-mixture training) very rapidly rose from zero for pigeons in both groups; notably, that rise was faster and higher in the Scene group than in the Color group.

Figure 3.

(Top) Mean Log RT (ms) difference scores between Correctly-Cued trials and Incorrectly-Cued for pigeons in the Color group and the Scene group as a function of days of training. (Bottom) Mean response rate difference scores (pecks per s) between Correct and Incorrect quadrants for pigeons in the Color group and the Scene group as a function of days of training. Error bars represent the standard error of the mean. The vertical line separates the initial 20 days (with alternating Correct/Incorrect-mixture sessions) from the subsequent 10 days (with Correct-only sessions).

These RT disparity scores were submitted to a repeated-measures ANOVA with group (Color vs. Scene) and day (1 to 20) as factors. The main effect of group was significant, F(1, 6) = 62.53, p < .0001, ${\eta }_p^2=.91$, 95% CI [.47, .95], with the Scene backgrounds (M = 1,213 ms, SE = 47.45) supporting greater RT disparity scores than the Color backgrounds (M = 1,106 ms, SE = 64.34). Also significant was the main effect of day, F(19, 114) = 3.16, p < .0001, ${\eta }_p^2=.34$, 95% CI [.10, .37]. The interaction was not significant, F < 1.

Anticipatory Responding: Training

During training, we also monitored the pigeons’ rates of SOA pecking to the quadrant in which the target was next to appear and to the average of the other three quadrants. Response rate to the Correct quadrant should exceed response rate to the Incorrect quadrants if the pigeons were effectively anticipating the location of the upcoming target.

Figure 4 shows that response rates to the Correct and Incorrect quadrants (on days with Correct/Incorrect-mixture training) rapidly diverged for pigeons in both the Color (Figure 4, top) and Scene (Figure 4, bottom) groups. That disparity persisted throughout interpolated training (in which 10 Correct/Incorrect-mixture sessions alternated with 10 Correct-only sessions) and beyond (in which 10 Correct/Incorrect-mixture sessions alone were given). Consistently faster response rates to the Correct quadrant than to the Incorrect quadrants clearly divulge discriminative anticipatory responding by pigeons trained with both color and scene backgrounds.

Figure 4.

(Top) Mean rate of anticipatory responding (pecks per s) in Correct and Incorrect quadrants for pigeons in the Color group as a function of days of training. (Bottom) Mean rate of anticipatory responding (pecks per s) in Correct and Incorrect quadrants for pigeons in the Scene group as a function of days of training. Error bars represent the standard error of the mean. The vertical line separates the initial 20 days (with alternating Correct/Incorrect-mixture sessions) from the subsequent 10 days (with Correct-only sessions).

The adjusted rate of anticipatory responding to the Color and Scene contexts was separately analyzed using a repeated-measures ANOVA with response location (correct vs. incorrect quadrant responding) and day (1 to 20) as within-subjects factors. In the Color group, the main effect of response location was significant, F(1, 3) = 11.37, p = .001, ${\eta }_p^2=.79$, 95% CI [.00, .87], with correct quadrant responding (M = 0.36 pecks per s, SE = 0.04) exceeding incorrect quadrant responding (M = 0.21 pecks per s, SE = 0.01). No other main effect or interaction was significant. In the Scene group, the main effect of response location was significant, F(1, 3) = 146.55, p < .0001, ${\eta }_p^2=.98$, 95% CI [.20, .99], with correct quadrant responding (M = 0.75 pecks per s, SE = 0.08) exceeding incorrect quadrant responding (M = 0.19 pecks per s, SE = 0.02). No other main effect or interaction was significant.

Contrasting the top and bottom panels of Figure 4 suggests that discriminative anticipatory responding may have been even stronger in the Scene group (Figure 4, bottom) than in the Color group (Figure 4, top). To properly assess this possibility, each pigeon's mean daily response rate score to the Incorrect quadrants was subtracted from its score to the Correct quadrant. These response rate difference scores are depicted in the bottom panel of Figure 3, with larger rate disparities signifying stronger discriminative anticipatory responding. These response rate disparities between Correct and Incorrect quadrants (on days with Correct/Incorrect-mixture training) rapidly rose from zero for pigeons in both groups; notably, that rise was much faster and higher in the Scene group than in the Color group.

These response rate difference scores were submitted to an ANOVA with group (Color vs. Scene) and day (1 to 20) as factors. The main effect of group was significant, F(1, 6) = 52.75, p < .0001, ${\eta }_p^2=.89$, 95% CI [.42, .94], with the Scene backgrounds (M = 1.12 pecks per s, SE = 0.19) supporting a greater response rate disparity than the Color backgrounds (M = 0.30 pecks per s, SE = 0.09). Neither the main effect of day nor the interaction was significant.

Reaction Time: SOA Testing

Pigeons’ RT scores at each of the five SOAs on Correctly-Cued and Incorrectly-Cued trials is pictured in Figure 5, with scores from the Color group (Figure 5, top) and the Scene group (Figure 5, middle) separately portrayed. In each case, RTs fell with increases in the SOA and RTs were faster on Correctly-Cued trials than on Incorrectly-Cued trials (cf. Jiang, Sigstad, & Swallow, 2013).

Figure 5.

(Top) Mean Log RT scores (ms) to Correctly-Cued targets and to Incorrectly-Cued targets for pigeons in the Color group as a function of SOA. (Middle) Mean Log RT scores (ms) to Correctly-Cued targets and to Incorrectly-Cued targets for pigeons in the Scene group as a function of SOA. (Bottom) Mean Log RT (ms) difference scores between Correctly-Cued trials and Incorrectly-Cued trials for pigeons in the Color group and the Scene group as a function of SOA. Error bars represent the standard error of the mean.

RTs in the Color and Scene groups were separately submitted to a repeated-measures ANOVA with cueing (Correctly-Cued vs. Incorrectly-Cued), SOA (0, 1, 2, 4, and 8 s), and block (six 5-day blocks) as within-subjects factors. In the Color group, the main effect of cueing was significant, F(1, 3) = 82.37, p < .0001, ${\eta }_p^2=.96$, 95% CI [.06, .98], with Correctly-Cued trials (M = 1,430 ms, SE = 40.44) supporting briefer RTs than Incorrectly-Cued trials (M = 1,922 ms, SE = 43.32). The main effect of SOA was also significant, F(4, 12) = 12.55, p < .0001, ${\eta }_p^2=.81$, 95% CI [.36, .86]. No other main effect or interaction was significant. In the Scene group, the main effect of cueing was significant, F(1, 3) = 183.58, p < .0001, ${\eta }_p^2=.98$, 95% CI [.25, .99], with Correctly-Cued trials (M = 1,245 ms, SE = 44.95) supporting briefer RTs than the Incorrectly-Cued trials (M = 2,299 ms, SE = 70.44). The main effect of SOA was also significant, F(4, 12) = 12.93, p < .0001, ${\eta }_p^2=.81$, 95% CI [.37, .86]. No other main effect or interaction was significant.

Comparing the top and middle panels of Figure 5 suggests that contextual cueing may have been stronger in the Scene group (Figure 5, middle) than in the Color group (Figure 5, top). To properly evaluate this possibility, each pigeon's mean daily RT score on Correctly-Cued trials was subtracted from its mean daily RT score on Incorrectly-Cued trials at each of the five SOAs. These RT difference scores are depicted in the bottom panel of Figure 5, with larger RT disparities signifying stronger contextual cueing. These RT disparities between Correctly-Cued and Incorrectly-Cued trials were greater than zero at all SOAs for pigeons in both the Color and Scene groups; the disparities were even greater in the Scene group than in the Color group. In addition, pigeons in both the Color and Scene groups exhibited greater RT disparities at the 1- and 2-s SOAs than at the 0-, 4-, and 8-s SOAs.

These RT disparity scores were analyzed using a repeated-measures ANOVA with group (Color vs. Scene), SOA (0, 1, 2, 4, and 8 s), and block (six 5-day blocks) as factors. The main effect of group was significant, F(1, 6) = 237.89, p < .0001, ${\eta }_p^2=.97$, 95% CI [.81, .98], with the Scene backgrounds (M = 1,058 ms, SE = 41.76) supporting greater RT disparity scores than the Color backgrounds (M = 492 ms, SE = 24.13). Also significant was the main effect of SOA, F(4, 24) = 16.82, p < .0001, ${\eta }_p^2=.73$, 95% CI [.45, .81]. No other main effect or interaction was significant.

Anticipatory Responding: SOA Testing

During testing, we also monitored the pigeons’ rates of SOA pecking to the quadrant in which the target was next to appear and to the average of the other three quadrants; these response rate scores came from SOAs of 1, 2, 4, and 8 s. Pigeons’ response rates at each of the four SOAs is pictured in Figure 6, with scores from the Color group (Figure 6, top) and the Scene group (Figure 6, middle) separately portrayed; in each case, response rates fell with increases in the SOA, with response rates higher in the Correct quadrant than in the Incorrect quadrants.

Figure 6.

(Top) Mean rate of anticipatory responding (pecks per s) in Correct and Incorrect quadrants for pigeons in the Color group as a function of SOA. (Middle) Mean rate of anticipatory responding (pecks per s) in Correct and Incorrect quadrants for pigeons in the Scene group as a function of SOA. (Bottom) Mean response rate difference scores (pecks per s) between Correct and Incorrect quadrants for pigeons in the Color group and the Scene group as a function of SOA. Error bars represent the standard error of the mean.

These response rate scores in the Color and Scene groups were separately submitted to a repeated-measures ANOVA with response location (correct vs. incorrect quadrant responding), SOA (1, 2, 4, and 8 s), and block (six 5-day blocks) as within-subjects factors. In the Color group, the main effect of response location was significant, F(1, 3) = 12.71, p < .0005, ${\eta }_p^2=.81$, 95% CI [.00, .88], with correct quadrant responding (M = 0.31 pecks per s, SE = 0.02) exceeding incorrect quadrant responding (M = 0.21 pecks per s, SE = 0.01). The main effect of SOA was also significant, F(3, 9) = 7.98, p < .0001, ${\eta }_p^2=.72$, 95% CI [.11, .82]. No other main effect or interaction was significant. In the Scene group, the main effect of response location was significant, F(1, 3) = 73.98, p < .0001, ${\eta }_p^2=.96$, 95% CI [.05, .98], with correct quadrant responding (M = 0.48 pecks per s, SE = 0.03) exceeding incorrect quadrant responding (M = 0.15 pecks per s, SE = 0.01). The main effect of SOA was also significant, F(3, 9) = 7.33, p = .0001, ${\eta }_p^2=.71$, 95% CI [.08, .81]. No other main effect or interaction was significant.

Contrasting the top and middle panels of Figure 6 suggests that discriminative anticipatory responding may have been stronger in the Scene group (Figure 6, middle) than in the Color group (Figure 6, top). To properly assess this possibility, each pigeon's mean daily response rate score to the Incorrect quadrants was subtracted from its score to the Correct quadrant at each of the four SOAs. These response rate difference scores are depicted in the bottom panel of Figure 6, with larger rate disparities signifying stronger discriminative anticipatory responding. These response rate disparities between Correct and Incorrect quadrants were greater than zero in both the Color and Scene groups; the disparities were even greater in the Scene group than in the Color group. In addition, pigeons in the Color and Scene groups tended to show greater response rate disparities at the 1- and 2-s SOAs than at the 4- and 8-s SOAs.

These response rate disparity scores were analyzed using a repeated-measures ANOVA with group (Color vs. Scene), SOA (1, 2, 4, and 8 s), and block (six 5-day blocks) as factors. The main effect of group was significant, F(1, 6) = 50.86, p < .0001, ${\eta }_p^2=.89$, 95% CI [.41, .94], with the Scene backgrounds (M = 0.33 pecks per s, SE = 0.05) supporting a greater response rate disparity than the Color backgrounds (M = 0.09 pecks per s, SE = 0.03). Also significant was the main effect of SOA, F(3, 18) = 3.32, p = .0217, ${\eta }_p^2=.35$, 95% CI [.01, .50]. No other main effect or interaction was significant.

Discussion

The results of the present study clearly indicate that the contextual cueing effect can readily be produced in pigeons by deploying our novel Cueing-Miscueing design. During the initial learning phase, pigeons quickly came to exhibit robust contextual cueing when the context preceded the target by 2 s, with reaction times to the target becoming far shorter on correctly-cued trials than on incorrectly-cued trials (Figure 2). These results add to the already expansive human literature and to the initial baboon report (Goujon & Fagot, 2013) underscoring the comparative generality of contextual cueing.

Either uniformly colored backgrounds (Figure 1, top; Figure 2, top) or photographic scene backgrounds (Figure 1, bottom; Figure 2, bottom) served as effective contextual cues; however, the contextual cueing effect proved to be much stronger with the photographic scene backgrounds than with the uniformly colored backgrounds (Figure 3, top). This contextual cueing superiority is likely to be due to the availability of salient spatial landmarks contained within the photographic scene backgrounds (Figure 1, bottom) and to the lack of spatial landmarks in the uniformly colored backgrounds (Figure 1, top). Indeed, with human participants, Rosenbaum and Jiang (2013) found that contextual control by photographic scenes reliably overshadowed contextual control by superimposed patterns of distractor items, further attesting to the salience of rich scene details.

Also during the initial learning phase, within the 2-s SOA delay, pigeons quickly came to peck predominately toward the location of the upcoming target on trials with uniformly colored backgrounds (Figure 4, top) as well as on trials with photographic scene backgrounds (Figure 4, bottom), suggesting that attentional guidance contributes to the contextual cueing effect (Chun & Jiang, 1998). That attentional guidance was much more pronounced with the photographic scene backgrounds than with the uniformly colored backgrounds (Figure 3, bottom).

Indeed, comparing the top and bottom panels of Figure 3 underscores the striking correspondence between the prominence of discriminative anticipatory pecking within the 2-s SOA delay and the strength of the contextual cueing RT effect. The more inclined the pigeons were to peck the correct quadrant in anticipation of the upcoming target stimulus, the greater the disparity in the birds’ RT to the target when it appeared in that (correctly-cued) quadrant than in another (incorrectly-cued) quadrant.

More detailed analysis of Day 1 performance attested to the striking speed of associative learning in our contextual cueing task (albeit a bit slower than the contextual learning of baboons; cf. Goujon & Fagot, 2013); visual context-target location learning was clearly manifested within only 1 day of training (192 Correctly-Cued trials plus 48 Incorrectly-Cued trials). Such rapid and robust learning strongly recommends contextual cueing for researchers seeking new and efficient paradigms to study associative learning processes. It also suggests that, even though all that the pigeons had to do to procure food reinforcement in this task was to peck the target when it appeared, the birds visually processed and remembered the preceding context, and associated it with the location of the subsequent target stimulus. Such S-S learning is highly reminiscent of anticipatory spatial learning in autoshaping (Wasserman, Carr, & Deich, 1978).

Following initial acquisition, we found that contextual cueing was demonstrable at SOAs ranging from 0 s to 8 s (Figure 5), with generally stronger contextual cueing being seen at the 1-s and 2-s SOAs (Figure 5, bottom). Again, either uniformly colored backgrounds (Figure 5, top) or photographic scene backgrounds (Figure 5, middle) served as effective contextual cues; however, the contextual cueing effect again proved to be much more robust with the photographic scene backgrounds than with the uniformly colored backgrounds (Figure 5, bottom).

Furthermore, we once more found that discriminative anticipatory pecking may effectively register attentional guidance on trials with SOAs greater than 0 s (Figure 6). Although overall rates of anticipatory SOA pecking declined as sessions involved progressively longer SOAs (perhaps because of the longer delays to food reinforcement and the corresponding lower overall rates of food delivery that were scheduled in these sessions), pigeons predominately pecked toward the location of the upcoming target both on trials with uniformly colored backgrounds (Figure 6, top) and on trials with photographic scene backgrounds (Figure 6, middle). Discriminative anticipatory pecking once more proved to be more pronounced with the photographic scene backgrounds than with the uniformly colored backgrounds as well as at the 1-s and 2-s SOAs than at the 4-s and 8-s SOAs (Figure 6, bottom); both of these findings agree with the RT results (Figure 5, bottom).

The present report of contextual cueing with pigeons thus joins that of Goujon and Fagot (2013) with baboons in suggesting that the processes of associative learning and attentional guidance in contextual stimulus control may be quite general. These positive results clearly warrant further comparative research into the contextual cueing effect; they also merit investigation into the biological mechanisms underlying contextual cueing (Chun, 2000; Lazareva, Shimizu, & Wasserman, 2012).

Highlights.

A novel Cueing-Miscueing design revealed robust contextual cueing by pigeons

Contextual cueing was more robust with photo backgrounds than with color backgrounds

Anticipatory pecking toward the locus of the target contributed to contextual cueing