Emotion Labeling in Pigeons.

Emotion Labeling in Pigeons - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Affect Sci . 2024 Dec 12;6(2):356–368. doi: 10.1007/s42761-024-00287-5 Emotion Labeling in Pigeons Jordan Nerz 1 , Cheyenne R Elliott 1 , Marisa Melo 1 , Tanner Raab 1 , Sophia Jones 1 , W David Stahlman 2 , Kenneth J Leising 1, ✉ Author information Article notes Copyright and License information 1 Department of Psychology, Texas Christian University, 2800 S. University Dr.Fort Worth, Box 298920, Attn, TX 76129 USA 2 Department of Psychological Science, University of Mary Washington, Fredericksburg, VA USA ✉ Corresponding author. Received 2023 Dec 23; Accepted 2024 Nov 21; Collection date 2025 Jun. © The Society for Affective Science 2024. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. PMC Copyright notice PMCID: PMC12209487  PMID: 40605948 Abstract At an early age, humans learn words to communicate their affective states and generalize their identification across differing external conditions. One might call this collection of effects emotion labeling. Previous research investigating analogous effects in nonverbal animal models has largely used drug-induced affective states or metacognitive states that are difficult to control. Some progress has been made by conceptualizing affect as a point within a two-dimensional space with pleasure on one axis and arousal on the other. Distinct affective states may thus serve as cues that control both human and non-human behavior. If conditions supporting responding produce distinct affect, labeling in one circumstance may transfer to novel conditions that elicit the same affective state. The present experiment examined the labeling of affective states induced in pigeons via outcomes of either the delivery of food or its omission. These outcomes followed responses (pecks) made in the presence of three visual stimuli ( A , B , and C ). On A trials, each outcome was followed by two labels; pecks to one of two labels, depending on whether food was delivered or omitted, were required to progress to the next trial. All pigeons labeled accurately. Consistent with human emotion labeling, pigeons learned to select a label associated with different core affect-inducing outcomes, transferred appropriately to novel conditions (i.e., trials with B and C in Test 1 and 2), and probably relied on more than just external cues (Test 3). Keywords: Core affect, Discrimination, Animal cognition, Emotion Children begin to use words to identify their affective state between 18 and 36 months, suggesting that the interoceptive stimuli (e.g., heart rate, hunger, pain) correlated with these states are discriminable and contribute to successful interactions with others (Bretherton & Beeghly, 1982 ). A functional interpretation follows naturally: Labeling an affective state, for example, may facilitate meeting a child’s needs, even if an explicit reinforcer does not immediately follow. Emotion labeling could be based on discrimination of the internal correlates of different emotions (e.g., arousal vs. lethargy), but it could also rely on discrimination of the cues that elicit them (e.g., food delivery vs. food omission) or their behavioral correlates (e.g., social approach vs. withdrawal). Successful labeling also requires that the labels (e.g., words or pictures) themselves be discriminable. Some children (e.g., many with Autism Spectrum Disorder [ASD]) have difficulty labeling emotions in themselves and in others. Interventions for such deficits typically start with connecting a child’s behavior to visual representations of emotion (Baron, 1992 ; Giddan, et al., 1995 ; Ogden & Minton, 2000 ; Saarni, 1999 ; Timler, 2003 ). Promoting Alternative Thinking Strategies (PATHS, Greenberg & Kusche, 1993 ) teaches labeling by asking clients to match their feelings to pictures of facial expressions on a strip on their desk on which is written, “I feel…” Emotions are learned about according to a hierarchy, beginning with primary emotions (Lewis, 1992 ). The early emergence and universality of labeling different primary emotions are due to the ease with which they can be discriminated against by children and caregivers. Much of the recent literature on emotion refers to subjective feelings as affective states that can be identified within a two-dimensional space with pleasure on one axis and arousal on the other (Russell, 2003 ). Core affect is the neurophysiological state underlying a particular point within the two-dimensional space. The use of these two dimensions and core affect enables operational definitions that can be utilized in research (Lewis, 1992 ; Russell, 2003 ). Many behaviors in non-human animals appear to be emotional (e.g., “fear” in response to a loud noise). With an absence of language abilities, however, one might question whether the affect in non-human animals differs qualitatively from that of humans. Recent literature has attempted to bridge the gap between humans and non-humans. Some researchers have argued for a distinction in levels of consciousness (i.e., affective consciousness vs. cognitive consciousness; see Panksepp, 2005 for a review) in affect within non-human animals. Mendl and colleagues have suggested emotion research utilizes operational definitions that align with dimensional concepts of core affect (Mendl & Paul, 2020 ; Mendl et al., 2010 , 2022 ). An organism’s internal state may serve as a discriminative stimulus which can be directly manipulated by drug administration. For example, an animal may learn to press a lever for food only when in a state of high arousal. Drug discrimination studies typically involve training animals to make one response (e.g., a left lever press) when under the influence of some drug (e.g., morphine) and a different response (e.g., a right lever press) when injected with saline (Colpaert, 1978 ; Griffiths et al., 1985 ). Colpaert trained rats to discriminate a 0.04 mg/kg dose of fentanyl from saline. Learned drug discriminations have been found to transfer to other narcotic drugs that produce similar internal states to the training drug, even if the route of administration (e.g., injection vs. oral administration) differed from that of training. This transfer suggests that the discrimination is based on the induced internal state rather than cues associated with drug administration. An affective state can also serve as a cue that modulates responding to an ambiguous discriminative stimulus, just as emotions in humans have been found to do (Matthews & MacLeod, 1994 ; Mendl et al., 2022 ; Mendl & Paul, 2020 ; Russel, 2003 ). In humans, for example, depressed individuals are more likely to interpret ambiguous scenarios as negative (Matthews & McLeod, 1994 ). In non-humans, a drug-induced affective state can modulate the consumption of a flavored liquid (e.g., Goddard & Holland, 1996 ; Holland & Forbes, 1982 ; Martin et al., 1990 ; Skinner & Martin, 1992 ; Skinner et al., 1998 , 2003 ). Martin et al. ( 1990 ) found that rats exhibited a conditional aversion to a flavor following morphine administration, but not saline. This aversion transferred to the consumption of plain water delivered after an injection of morphine, suggesting control by the induced affective state. Additional evidence of sensitivity to internal states has been found in metacognition research with monkeys (Brady & Hampton, 2021 ; Hampton, 2001 ; Kornell, 2009 ; Smith et al., 2006 ), dolphins (Smith et al., 2003 ), and pigeons (Iwasaki et al., 2018 ) by providing an uncertainty response that allows them to skip difficult trials or seek more information (Smith, 2009 ). It is difficult to experimentally isolate whether affective and/or non-affective states contribute to metacognitive decisions. For example, from a core affect perspective (Mendl & Paul, 2020 ), an animal might select to take a memory test when in a high arousal and high pleasure core affect state (i.e., anticipating a reinforcer) and opt out of the test when in a high arousal and low pleasure core affect state (i.e., anticipating the absence of a reinforcer). Overall, there is evidence that non-human animals discriminate between internal states, and discrimination transfers to states elicited by other stimuli. Previous work, however, has (1) been unable to pit behavioral and interoceptive cues against one another; (2) primarily focused on drug-induced affect (Mendl et al., 2010 ); (3) implicated metacognitive states (i.e., memory strength or confidence) whose dimensions are difficult to identify; and (4) often explicitly reinforced correct labeling behavior (Colpaert, 1978 ; Griffiths et al., 1985 ; Hampton, 2001 ; Lubinski & Thompson, 1987 ; Martin et al., 1990 ). Explicit reinforcement for correct labeling may shift the animal’s attention to the consequences of the response; in humans, labeling of affective states may happen in the absence of any obvious reinforcement contingency. The current experiment evaluated nonreinforced labeling of affective states associated with food delivery or omission in pigeons. Pigeons were trained to label their core affect and tested for transfer of labeling to affect elicited by other stimuli. Pecks to three discriminative stimuli (S D s A , B , and C ) were followed by a reinforcer on some trials, but subsequent labels were presented only on A trials. The outcome (reinforcer delivery or omission) following pecks to S D A determined which label must be pecked to continue the session. Test 1 evaluated transfer to S D B and C , which had not previously been followed by visual labels. Test 2 sought to determine whether the labels acquired different biological values. Test 3 aimed to determine which cues were used to label reinforcer delivery and omission by presenting subsets of stimuli (e.g., sound of food delivery) correlated with reinforcer delivery. Method Subjects Eight experimentally naïve White Carneau pigeons ( Columba livia ) approximately one year old acquired from Double T Farm (Glenwood, IA) served as subjects. Studies of comparative cognition with pigeons commonly use a sample size between 6 and 12 (e.g., Cheng et al., 1997 , Lechelt & Spetch, 1997 ), and many studies use five or fewer pigeons per study (e.g., Got & Watanabe, 2020 ; Gray et al., 2021 ; Navarro et al., 2020 ; Roberts, 2021 ; Watanabe, 2020 ; Wilkinson & Kirkpatrick 2020 ; Wright et al., 2021 ). Pigeons were maintained at 83–85% of their free-feeding weights. Experiments testing touchscreen-based responding for mixed grain with pigeons frequently utilize a food deprivation level of 80 to 85% of free-feeding weight (e.g., Cheng, et al., 1997 ; Lechelt & Spetch, 1997 ). Pigeons were individually housed in a colony with a 12-h light–dark cycle and had free access to water and grit. Experimental procedures occurred during the light portion of the cycle. All animal procedures were approved by the Institutional Animal Care and Use Committee at Texas Christian University. Apparatus Training and testing were conducted in a flat-black Plexiglas chamber (45 cm wide × 41 cm deep × 46 cm high). All stimuli were presented by a computer on a color LCD monitor (L1750, Hewlett Packard, Palo Alto, CA) visible through a 33 × 40 cm viewing window in the middle of the front panel of the chamber. The bottom edge of the viewing window was 12 cm above the chamber floor. Pecks to the monitor were detected by an infrared touch screen (EZ-170-WAVE; Ezscreen, Houston, TX) mounted on the front panel. A 28-V houselight located in the ceiling of the box was illuminated at all times. A food hopper (Coulbourn Instruments, Allentown, PA) was located in the center of the front panel, its access hole flush with the floor. All experimental events were controlled and recorded with a Hewlett Packard computer (Palo Alto, CA; Intel Core Duo E8400/1.96 GB RAM). A video card controlled the monitor in the SVGA graphics mode (800 × 600 pixels). Stimulus Displays Pigeons experienced eight different visual stimuli throughout training (see Fig. 1 ). The autoshaping stimulus was a white circle measuring 7.62 cm in diameter and was used to train pigeons to peck the touchscreen. A white plus sign stimulus (6.35 cm wide × 6.35 cm high) was present during all intertrial intervals (ITIs). The ready stimulus was a white circle measuring 6.35 cm in diameter with a black plus sign in the center. The three discriminative stimuli included a cyan circle (7.62 cm in diameter), a green triangle (7.62 cm wide × 7.62 cm high), and an orange star (7.62 cm high × 7.62 cm wide). Lastly, the label stimuli were two squares (7.62 cm wide × 7.62 cm high), depicting either a yellow “Happy” face or a blue “Sad” face. Fig. 1. Open in a new tab a Experimental stimuli and b positions of stimuli on touchscreen. a The ready stimulus used only in pretraining (1), the white plus presented during all intertrial intervals (2), the autoshaping stimulus (3), the discriminative stimuli A, B, and C (4–6), and the label stimuli (7–8). b All stimuli appeared on a black background. The white squares (1–3) represent the positions in which stimuli appeared on the screen. In pretraining, all stimuli (the three discriminative stimuli (S D s), the ready stimulus, and the autoshaping stimulus) appeared in all three locations randomly across trials. The ITI plus always appeared in position 1. In Phase 1 onward, the S D s always appeared in position 1 and the labels appeared in positions 2 and 3. During errorless learning, only one label was presented in either location 2 or 3. In all phases involving labeling, all labels appeared equally as often in the two possible locations Procedure Autoshaping The autoshaping phase was intended to establish reliable pecking to a visual stimulus (a digital image) and retrieval of mixed grain prior to introducing the discriminative stimuli used in emotion labeling training. A white circle was displayed in the center of the monitor and reinforced according to a mixed Pavlovian-instrumental schedule. Pigeons were initially trained with a white circle that would terminate with food delivery after either 8-s (i.e., a Pavlovian contingency) or a peck to the white circle (i.e., an operant contingency). A trial with a peck was followed by a 10-s ITI, whereas a trial without a peck was followed by a 40-s ITI. The response schedule was increased across sessions from a fixed-ratio (FR) 1 schedule (i.e., one response resulted in the delivery of grain) to a variable-ratio (VR) 3 schedule (i.e., an average of three responses were required to earn the reward) to encourage higher and steadier rates of responding. Sessions terminated after 64 trials or 60 min, whichever came first. Once subjects were reliably pecking the autoshaping stimulus on a VR-3 schedule, they advanced to pretraining. Pretraining The goal of pretraining was to establish reliable pecking to all of the S D s for mixed grain and to familiarize the pigeons with pecking all locations where stimuli (S D s and labels) would be located during subsequent phases. At the beginning of each trial, either the autoshaping circle, the ready stimulus, or one of the three S D s was randomly selected to appear in the center of the screen. Pecks were reinforced on an FR-1 schedule. A peck to the stimulus caused the removal of the stimulus and reinforcer delivery, consisting of 4-s access to grain, after which a 20-s ITI ensued. The start of an ITI was signaled by the presentation of the white plus stimulus located in the center of the screen, which remained on the screen for the duration of the ITI; this was true for all phases. A session was terminated after 68 trials or 60 min, whichever came first. Once pigeons were reliably responding to each of the three S D s, these stimuli began to be presented randomly equally often at each of three locations on the screen (center, bottom-left, and bottom-right, see Fig. 1 b); at this point, the autoshaping stimulus and ready stimulus were no longer presented. Subjects advanced to Phase 1 after a minimum of four sessions in which 68 trials were completed. Phase 1. Errorless Reinforcement Labeling Phase 1 presented the S D s and labels in the specific locations to be used throughout the rest of the experiment. In this phase, only the correct label was presented on each trial to facilitate the association of the labels with the outcome-induced affective state. The onset of a trial was marked by the presentation of one of the three S D s ( A , B , and C ) in the center of the screen, to which pigeons could peck once (FR-1) for the trial to progress. A response caused the screen to turn black and the delivery (or omission) of the reinforcer. The reinforcer was the activation of the hopper and illumination of the hopper light for 4-s, during which pigeons could retrieve grain. On nonreinforced trials, the screen remained black for 4-s post-trial. On both trial types, labels were presented immediately after the 4-s period following the termination of the S D A. A session terminated after completion of 60 trials (20 trials per S D ) or 60 min, whichever came first. An even number of trial types were programmed for trials 1–30 and 31–60. The correct label appeared equally often in each location for each outcome. The computer program randomized the order of trials for each session, with the exception that the correct label could not appear in the same location for more than five consecutive trials. The probability of the reinforcer following S D s A and B was gradually reduced until both stimuli resulted in grain on a random half of trials; trials with C were always followed by grain. S D s B and C were trained to serve as transfer stimuli and were not followed by labels in training. S D C was always followed by reinforcement to examine whether pigeons would be able to accurately label reinforcer omission, even though pigeons had never experienced an omitted outcome following S D C. Pigeons received several sessions in which the probability of the reinforcer following each of A and B was .80 followed by a minimum of four sessions where the probability was reduced to .60. On A trials, this was followed by the presentation of one of the two label stimuli at a randomly selected bottom corner of the display. One label (a “Happy” face) always followed reinforcer delivery, and the other label (a “Sad” face) always followed reinforcer omission. Birds were thus always presented with only the stimuli ultimately assigned as correct for that trial’s outcome. One peck to the label was required to terminate the stimulus and continue to a 20-s ITI, followed by the next trial. When the reinforcer probability for A and B was 80%, a programming error led to sessions in which A trials were followed by the correct label 50% of the time and the incorrect label on the other 50% of A trials and trials with S D C were omitted. This error was corrected when the probability of the reinforcer was reduced to .60. Pigeons completed a minimum of four sessions with a .60 probability of the reinforcer following A and B trials before they advanced to Phase 2. Phase 2. Intermixed Reinforcement Labeling The goal of Phase 2 was to introduce the choice between both labels and gradually increase the response requirement for reinforcer delivery, as long as labeling accuracy was high. From this point on, the probability of the reinforcer following A and B was .50 and one of each label was displayed on A trials following the reinforcer or its absence. The location of the labels varied across trials. To reduce switching behavior (i.e., pecking left and then right) on the label stimuli and to produce steady rates of responding, the response schedule for both the S D and the label stimuli was increased gradually from a FR-1 to a VR-2. All subjects completed 25 sessions in which a single peck was required for both S D and label stimuli, followed by 10 sessions with a VR-2 for S D and an FR-1 for label stimuli, and finally 15 sessions on a VR-2 for both S D and label stimuli. Because performance was not improving in the initial 25 sessions with an FR-1 schedule, the hopper duration and reinforcer omission period were increased from 4-s to 5-s when the response schedule was made leaner in an attempt to make reinforcer delivery and omission more salient. The label stimuli remained on the screen until the response requirement to the correct label stimulus was met, after which the stimuli disappeared, and a 20-s ITI ensued. This was true for all subsequent training phases. A session terminated after completion of 60 trials or 60 min, whichever came first. Phase 3. Blocked Reinforcement Labeling with S D A Pigeons showed little improvement from the first 5 sessions (1–5) of training to the final 5 sessions (46–50) of Phase 2 (see Fig. 2 a). As a result, the goal of Phase 3 was to facilitate labeling by giving sessions that included only A label trial types (i.e., no B or C transfer trial types) with reinforced-only and nonreinforced-only trial blocks (four trials/block) for one-half of the session (32 trials) and interspersed presentations of reinforced and nonreinforced trials for the remaining half of the session (32 trials; i.e., a computer randomized order of 16 reinforced and 16 nonreinforced trials). The order of the eight reinforced and nonreinforced blocks in each session was randomized by the computer program, and whether the first half of trials were blocked or interspersed was pseudorandomly determined. If blocked training were required for good performance on interspersed trials, then we would observe poor labeling during the half of interspersed trials that occurred at the start of a session. If performance during interspersed trials at the start of the session were equal to that of blocked trials at the end of the session, then the pigeons could be advanced to sessions with trials interspersed throughout. All other aspects of the session were the same as described previously. Fig. 2. Open in a new tab Proportion of correct first-label responses across bins. Proportion correct (first response) across phases. Bins 1–24 consisted of five sessions, whereas bin 25 consisted of three sessions and bin 26 consisted of six sessions. The dashed line represents chance level (.50) responding. Error bars represent the standard error of the mean. (a) In Phase 2, all trial types were interspersed within each session. (b) Phase 3 consisted of sessions of blocked and interspersed reinforced and nonreinforced A trials only. (c) In Phase 4, all trial types were again interspersed within each session Each session consisted of 64 trials or 60 min, whichever came first. The response schedule for S D and label stimuli was gradually increased from a VR-2 (S D and labels; 59 sessions) to a VR-3 (S D and labels; 3 sessions), to a VR-4 (S D and labels; 4 sessions) to a VR-5 (S D and labels; 3 sessions), and lastly to a VR-6 (labels, VR-5 S D ). In the first 59 sessions of this phase (i.e., VR-2 S D and labels), the hopper duration and reinforcer omission period were increased from 5-s to 6-s. This duration was decreased back to 5-s for all remaining training sessions. Following the completion of four sessions of the final response schedule (i.e., VR-5 S D and VR-6 labels), pigeons were advanced to the final training phase. Phase 4. Intermixed Reinforcement Labeling Retraining The goal of Phase 4 was to reintroduce sessions with all trials intermixed throughout. All three S D s were reintroduced and all three trial types and reinforcer outcomes (reinforced vs. nonreinforced) were interspersed, as in Phase 2. The session proceeded in the same manner as Phase 2, except the response schedules for S D , and label stimuli were VR-5 and VR-6, respectively. Once all pigeons had an average proportion correct (first response) of .80 on the most recent two sessions, pigeons were advanced to Test 1. This criterion was met after six sessions. However, one bird inadvertently received an additional training session prior to testing; the data from this training session were not included in analyses. Test 1 The goal of Test 1 was to evaluate transfer to novel conditions. The pigeons had no experience labeling affect following trials of B and C and additionally had no experience with a reinforcer absent outcome following C . Pigeons received two transfer test sessions consisting of 64 trials each. Each session began and ended with a ten-trial block of baseline training trials (i.e., trials presented with ( A ) or without ( B and C ) labels, as in training; A + , A − , B + , B–, C + ; two of each). The remaining 44 trials were presented in four 11-trial blocks that included six probe trials (A + , A–, B + , B–, C + , C–) per block, in which all S D s ( A , B , and C ) were followed by labels. This was true for all subsequent tests as well. On probe trials, the positions of the label stimuli were counterbalanced. The response schedule for S D s was the same as in Phase 4 (a VR-5) but the response schedule for label stimuli was changed from a VR-6 to an FR-2 + fixed-interval (FI) 4-s. Any two responses (correct or incorrect) would satisfy the FR2 component, and the first response (correct or incorrect) after the FI expired would terminate the probe trials. This schedule was used to reduce the likelihood of extinction by minimizing the number of responses required and better controlling the total time the labels were present on probe trials. Furthermore, the response requirements were met regardless of whether a response was correct to maintain response tendencies established during training. This was true for all subsequent tests. A 20-s ITI separated each trial. Pigeons received one session of retraining on Phase 4 contingencies in between each test session. Pigeons received a minimum of 20 sessions of retraining on Phase 4 contingencies before advancing to Test 2. Test 2 The goal of Test 2 was to evaluate whether proximity to reinforcer delivery (one label) or reinforcer omission (the other label) altered the biological value of the labels themselves. Pairings close in time (albeit backwards in time) to reinforcer delivery could result in some excitatory value (based on proximity) or inhibitory value (as a signal for the end of the reinforcer delivery). Large differences would alter the latency to peck each and could complicate an interpretation of the results as labeling. Pigeons were presented with transfer trials in which each S D was followed by either two “Happy” labels (H–H) or two “Sad” labels (S–S). In addition, pigeons were presented with “trials” in which the labels simply appeared after the end of the ITI, and in the absence of the S D and a reinforcing/nonreinforcing outcome (No S D ). Pigeons received four test sessions with one test trial per session for each of the 15 trial types ( A + H–H/S–S, A − H–H/S–S, B + H–H/S–S, B − H–H/S–S, C + H–H/S–S, C − H–H/S–S, and No S D H–S/S–S/H–H). Sessions consisted of three, ten-trial blocks that each contained five probe trials. Pigeons received one to two retraining sessions in Phase 4 in between each test session. Pigeons advanced to Test 3 after completing 14 retraining sessions on Phase 4, or when pigeons reached the criteria of a proportion correct (first response) of 0.80 on both reinforced and nonreinforced trials, whichever came first. Test 3 The goal of Test 3 was to determine whether accurate labeling was possible in the presence of the external stimuli associated with reinforcer delivery. Pigeons were tested with each S D with four different test trial types which varied in the way reinforcers were delivered. In training, the duration of reinforcer delivery or its absence was 5-s. Due to a programming error, this duration was set to 4-s for all baseline trials. On one trial type (Light-Noise-Food 4-s [LNF 4-s]), reinforcers were delivered in the same manner as it was in training, with the presentation of the hopper light (L), the sound of the hopper raising (noise, N), and 4-s of access to food (F). On LNF 1-s trials, all stimuli associated with reinforcer delivery were presented for 1-s (hopper sound, hopper light, and food available), which consequently prevented access to food. On NF 4-s trials, the hopper was raised (sound and access to food) without the hopper light. Lastly, the fourth trial type (L 4-s) consisted of the presentation of the hopper light only for 4-s. Because grain could not be delivered in the absence of the hopper noise, pigeons could not be presented with an L-F combination. Pigeons received four test sessions, each consisting of three 11-trial blocks that each contained six test trials and five baseline training trials. In all sessions, pigeons received one test trial per session of LNF 4-s and LNF 1-s for each of the six trial types. For two of the sessions, pigeons received one test trial per session of NF 4-s and L 4-s for each of the six trial types. Data Analysis Repeated measures analysis of variances (ANOVAs) and paired-samples t- tests were performed on proportion correct (first-label response), proportion of responses, and response latencies using SPSS (version 29. IBM, New York, NY). Planned comparisons were performed as follow-up tests based on a priori hypotheses using Statistica (TIBCO, CA, USA). Effect sizes for ANOVAs and t -tests were generated by SPSS, and effect sizes for planned comparisons were computed using SPSS code published by Smithson (2001) for all statistical results of p < .10. A Greenhouse–Geisser correction was used for data that violated the assumption of sphericity. The datasets generated during the study are available in the “Emotion Labeling in Pigeons” OSF repository, found here: https://osf.io/s29ac/?view_only=e5d4efda06c54f5389cdf2414aeee6bb . This research was not preregistered. Results Reinforcement Labeling Training Pigeons completed a mean of 155 total daily sessions (Min = 101, Max = 168) from Phase 1 to 4 (excluding pre-/post-test retraining sessions). One pigeon failed to respond reliably during pretraining, and another pigeon failed to perform above-chance levels on label trials by the end of Phase 4. The remaining six pigeons contributed data to all subsequent analyses. The need to train trial types in a blocked fashion during acquisition was not anticipated. After 50 sessions, we observed little improvement in performance when the proportion correct (first response) was averaged across five-session bins for Phase 2 (Fig. 2 a). A paired-samples t -test comparing the proportion correct on bin 1 to bin 10 revealed that performance was only marginally higher in bin 10, t (5) = 2.10, p = .09, d = 0.86, 95% CI (− 0.12, 1.78). After switching to blocked and interspersed training (Phase 3, Fig. 2 b), performance appeared to improve. A paired-samples t -test comparing bin 11 (i.e., the first bin of Phase 3) to bin 1 revealed that the proportion correct was significantly higher in bin 11 than in bin 1, t (5) = 4.28, p = .008, d = 1.75, 95% CI (0.40, 3.04). A comparison between the first (bin 11) bin and final bin (bin 25) of Phase 3 also revealed improved performance, t (5) = 5.05, p = .004, d = 2.06, 95% CI (0.56, 3.51). Lastly, pigeons’ accuracy was unaffected when all trial types ( A , B , and C ) were reintroduced and interspersed (Phase 4, Fig. 2 c). This observation was confirmed with a paired-samples t -test comparing bin 25 (i.e., the final bin of Phase 3) and bin 26 (i.e., Phase 4), t (5) = 1.33, p = .24, d = 0.54, 95% CI (− 0.34, 1.39). A chi-square analysis revealed that there were no differences in the proportion of correct responses as a function of the previous trial for a prior reinforcer outcome, χ 2 (5, N = 6) = 0.06, p = 1.00, or no reinforcer outcome, χ 2 (5, N = 6) = 0.36, p = 1.00, suggesting that only the most recent outcome influenced labeling accuracy. Test 1 During training, only trials of A were followed by labeling. Test 1 evaluated whether labeling would transfer to reinforcer present or absent outcomes with B and C . Furthermore, during C trials, the labels and reinforcer absent outcome were both novel. A comparison of performance during trained (e.g., trials with A ) and transfer (e.g., trials of B and C ) is often used to evaluate whether a behavior is supported by a specific set of conditions. Labeling accuracy was not expected to be similar between trials with A and those with B and C . Figure 3 displays the proportion correct (first response) across each reinforced and nonreinforced test trial type ( A , B , and C ). A repeated measures ANOVA was performed on proportion correct with trial type (baseline test [ A ] and transfer [ B and C ]) and trial outcome (reinforced vs. nonreinforced) as repeated measures revealed that pigeons’ labeling accuracy did not differ significantly across trial type, F (2, 12) = 1.32, p = .30, η p 2 = 0.18, 95% CI (0, 0.44), or trial outcome, F (1, 6) = 0.30, p = .60, η p 2 = 0.05, 95% CI (0, 0.42), which indicates successful transfer to reinforcer present and absent outcomes with B and C . And, as expected, no interaction between these factors was observed, F (2, 12) = 0.76, p = .49, η p 2 = 0.11, 95% CI (0, 0.37). Accuracy did not differ between labeling after trained and transfer S D s. A reinforcer present or absent outcome was always followed by two labels (i.e., chance is 50%); a series of t -tests comparing accuracy to the chance level found better than chance performance during every trial type, t s(5) > 9.80, p s < .001, d s > 4.00, 95% CIs (1.48 < d < 6.53). Fig. 3. Open in a new tab Proportion of correct first-label responses on Test 1 (transfer). A, B, and C represent the three S D s that preceded reinforcer delivery (plus sign) or omission (minus sign). The dashed line represents the chance level. Asterisks indicate statistical significance ( p ≤ .05) from chance level. Error bars represent the standard error of the mean Test 2 Test 2 was designed to evaluate whether one label acquired more excitatory or inhibitory value than the other. One label was consistently paired close in time with the reinforcer (i.e., potential excitatory value), but it also signaled the end of the opportunity to consume the reinforcer (i.e., potential inhibitory value). A higher excitatory value should translate into a shorter latency. To evaluate this possibility, two identical labels (both “Happy” or “Sad”) followed each reinforcer outcome. If one label acquired more excitatory or inhibitory value, then we should see faster responses on trials with both of those labels irrespective of the most recent reinforcer outcome. However, if the latency to peck a label varied as a function of the preceding outcome, then the value of the labels would be dependent on the induced affective state. Figure 4 a displays the mean latencies (s) for pigeons to make a first response when the labels were two “Happy” (H–H) vs. two “Sad” (S–S) labels on each reinforced and nonreinforced S D ( A , B , and C ). In contrast to one label always eliciting a faster response, responses were faster when the labels were congruent with the recent reinforcer outcome. This was confirmed with a repeated measures ANOVA performed on mean latency with the trial outcome (reinforced vs. nonreinforced) and label (H–H vs. S–S) as repeated measures. There was a main effect of trial outcome, F (1, 5) = 8.42, p = .03, η p 2 = 0.63, 95% CI (0, 0.81), and an interaction between trial outcome and label, F (1, 5) = 14.11, p = .01, η p 2 = 0.74, 95% CI (0.03, 0.86). The main effect of the label was nonsignificant, F < 1. A priori predictions led us to compare two of the six possible comparisons using paired-samples t -tests, which revealed pigeons were faster to peck one of the two “Sad,” rather than “Happy,” labels following a reinforcer absent outcome (nonreinforced S–S vs. nonreinforced H–H), t (5) = 2.66, p = .05, d = 1.09, 95% CI (0.02, 2.09), whereas response latencies were marginally faster for pecking one of the two “Happy” labels following a reinforcer outcome (reinforced H–H vs. reinforced S–S), t (5) = 2.12, p = .09, d = 0.87, 95% CI (− 0.11, 1.79). The combination of these results strongly supports that differences in the value of each label were not significantly different and thus could not support accurate labeling on their own or complicate an affective labeling interpretation. Fig. 4. Open in a new tab Test 2 latency to first peck. Data are shown as a function of trial outcome (reinforced and nonreinforced), trial type (A, B, C, and No S D ), and the pair of labels presented on a trial (happy-happy, sad-sad, and happy-sad). ( a ) A, B, and C represent the three discriminative stimuli (S D ) that preceded reinforcer delivery (plus sign) or omission (minus sign). ( b ) The No S D latencies are presented in a separate panel and were analyzed independently because all No S D trials were nonreinforced. Error bars represent standard error of the mean We also included some trials where the ITI expired and the labels were present in the absence of any S D or reinforcer outcome. These No S D trial types were a special case since (1) only reinforcer absent outcomes occurred and (2) we included some trials with one “Happy” and one “Sad” label during Test 2 to evaluate the pigeons’ perception. The presentation of the labels without an S D or reinforcer could be perceived by the pigeons as a reinforcer absent outcome, which would result in a faster latency to peck one of the two sad face labels than happy face labels and a greater proportion of pecks to the sad face when a sad and happy face label were present. However, the absence of the S D means that a reinforcer was not expected. Consequently, if unexpected reinforcer omission is supporting labeling, then latency to peck a label should appear similar across trials, and pecking the sad face should be equal to chance levels when both labels are present. A repeated measures ANOVA performed on mean latencies during the No S D label trial types (see Fig. 4 b) revealed that the mean latencies across H–S, H–H, and S–S did not differ statistically (using a Greenhouse–Geisser correction), F (1.07, 5.35) = 13.58, p = .48, η p 2 = 0.73, 95% CI (0.04, 0.86). Figure 5 displays the mean proportion of first responses to the “Sad” label on No S D H–S label trials across the four test sessions for each pigeon. Interestingly, despite finding no difference in latency, pigeons emitted a higher proportion of responses to the “Sad” face than would be expected by chance (0.50), t (5) = 4.00, p = .01, d = 1.63, 95% CI (0.34, 2.87). Fig. 5. Open in a new tab Test 2 proportion of responses to the “Sad” label on No S D test trials. The mean proportion of responses to the sad label across the four test sessions during the No S D test trials with a Happy and Sad label present. No S D preceded the presentations of the labels and no reinforcer was delivered. The dashed line represents the chance level Test 3 A reinforcer outcome included a series of relays activating a light (L) located near the food and a noise (N) from the solenoid used to raise the grain to make it accessible, as well as access to the mixed grain (F). To evaluate whether the stimuli that accompanied a reinforcer outcome was supporting labeling accuracy, we presented trials of subsets of reinforcer-associated stimuli. If the affective state induced by food or no food alone controls labeling, then the “Sad” face should be pecked in every condition in which eating does not occur. At the other extreme, if only one stimulus, or a subset of stimuli, controls labeling (e.g., the light), then we should observe pecking to the “Happy” face when that stimulus is present and to the “Sad” face when it is absent. Finally, if reinforcer-associated stimuli contribute, but do not control, accurate labeling, then the results should be mixed across trials with different subsets of stimuli. Figure 6 displays the proportion correct (first response) as a function of test trials (LNF 4-s, LNF 1-s, NF 4-s, and L 4-s). A repeated measures ANOVA performed on proportion correct with test trial (LNF 4-s, LNF 1-s, NF 4-s, and L 4-s) as the repeated measure revealed that proportion correct differed across trials, F (3, 15) = 6.93, p = .004, η p 2 = 0.58, 95% CI (0.11, 0.71). Planned comparisons evaluated LNF 4-s against LNF 1-s, NF 4-s, and L 4-s. The results revealed that LNF 4-s was significantly different from NF 4-s and L 4-s, t s(5) > 2.90, p s < .03, d s > 1.18, 95% CIs (0.08 < d < 2.23), but only differed marginally from LNF 1-s, t (5) = 2.40, p = .06, d = 0.98, 95% CI (− 0.40, 1.94). This suggests no single event-controlled labeling. Furthermore, proportion correct on LNF 4-s trials was significantly above chance, t (5) = 6.14, p < .001, d = 2.51, 95% CI (0.79, 4.19), and NF 4-s was significantly below chance, t (5) = 2.91, p = .02, d = 1.19, 95% CI (0.08, 2.23), but LNF 1-s and L 4-s did not differ from chance, t s(5) < 1.06, p s > .34, d s < 0.43, 95% CIs (− 0.43 < d < 1.26). These mixed results again suggest no single reinforcer-associated stimulus was supporting accurate labeling. Fig. 6. Open in a new tab Test 3 proportion correct of first-label response as a function of reinforcement type. Test 3 manipulated reinforcement-associated stimuli (hopper light [L], the hopper noise [N], and the availability of food [F], which was manipulated by the hopper duration [4-s or 1-s]). LNF 4-s was the typical reinforcement outcome pigeons experienced throughout all of the training. A correct choice was defined as a peck to the “Happy” label on all trials. The dashed line represents chance level and asterisks indicate statistical significance ( p ≤ .05) from chance level. Error bars represent the standard error of the mean There was some concern that the absence of the light on NF 4-s trials might reduce the likelihood that the pigeons would enter the magazine to search for food. To evaluate this concern, we compared the latencies for the head of the pigeon to break the photobeam located at the entrance to food delivery location during each test trial type. Figure 7 displays the mean latency to search for food across the different test types. Pigeons did not enter the magazine during LNF 1-s or NF 4-s trials. Additionally, pigeons took longer to enter the feeding niche on L 4-s trials compared to LNF 4-s trials (i.e., typical reinforcer delivery), t (5) = 2.95, p = .03, d = 1.20, 95% CI (0.09, 2.25). These results suggest that the light did control search behavior for food, but neither it, nor any other reinforcer-associated stimulus directly controlled labeling. Combining the results, however, it does appear that search behavior elicited by the light may to some degree counteract the absence of eating, which led to labeling the L 4-s only and LNF 1-s outcomes with both labels equally, compared to reliable labeling of the NF 4-s only outcome as reinforcer omission. Fig. 7. Open in a new tab Test 3 latency (s) to check for food as a function of reinforcement type. Test 3 manipulated reinforcement-associated stimuli (hopper light [L], the hopper noise [N], and the availability of food [F], which was manipulated by the hopper duration [4-s or 1-s]). LNF 4-s was the typical reinforcement outcome pigeons experienced throughout all of the training. Pigeons did not enter the magazine on LNF 1-s or NF 4-s trials, and, as a result, these trials are excluded. Error bars represent the standard error of the mean Discussion The present experiment trained pigeons to respond discriminatively to labels that corresponded to recent reinforcer delivery. After training with only the A trials, all pigeons acquired the labeling response and transferred labeling to novel conditions (Test 1). When presented with the same label in both locations during Test 2, pigeons were slower to respond to the “Sad” labels than “Happy” labels, but only when the reinforcer had been delivered. Lastly, Test 3 revealed a deficit during all variations of reinforcer delivery that excluded consumption, suggesting that correct labeling requires more than just the external cues associated with the delivery or absence of the reinforcer. When all trial types were interspersed, pigeons showed only marginal improvement between the mean of the first 5 sessions and the mean of the final 5 sessions of Phase 2, despite having 50 sessions of training. When pigeons were given sessions with only A trials, there was a significant improvement compared to a bin of the first five sessions, which continued in the last bin of Phase 3. As the blocking manipulation is confounded with additional session, it is unclear whether the blocked training was required for acquisition. Blocked and interspersed training is consistent with techniques used to train emotion labeling in humans. Through their experiences and interactions with caregivers, typically developing children begin with a few primary emotions before introducing secondary emotions (Way et al., 2007 ). Furthermore, emotion labeling interventions, such as the PATHS curriculum, start by teaching children to label single feelings before progressing to labeling multiple feelings (Greenberg & Kusche, 1993 ; Way et al., 2007 ). The results of Test 1 revealed that pigeons successfully transferred labeling to both transfer stimuli and trial outcomes. Pigeons were equally accurate on reinforced and nonreinforced C trials, indicating that the surprising experience of reinforcer omission after an extensive history of reinforcer delivery following S D C did not interfere with correct labeling. Accurate labeling and transfer could reflect that pigeons are using an affective state to label the reinforcing and nonreinforcing outcomes. These outcomes correspond directly to different points along the pleasure (and likely arousal) dimension of core affect and thus would be expected to be discriminable and serve as discriminative stimuli. In Test 2, pigeons were presented with transfer trials in which each S D was followed by either two “Happy” labels (H–H) or two “Sad” labels (S–S). If behavior were contingent on the affective state and not the S D directly, then response latencies should be shorter when the label was congruent with the most recent outcome and longer when incongruent. We observed the latter result, which indicates that reinforcer delivery or omission controlled labeling behavior, independent of any general excitatory or inhibitory value associated with each label. Test 2 also included trials of the labels presented in the absence of an S D or reinforcing outcome. Response latencies on No S D trials were statistically similar whether the two labels were “Happy,” “Sad,” or one of each (i.e., “Happy”- “Sad”). However, on trials with one “Happy” and one “Sad” label, the “Sad” label was selected more often than would be expected by chance. Though it is not surprising the pigeons perceived their affective state on these tests as more like reward loss than delivery, the absence of a reinforcer is not the same as the affective state induced by reward loss (as was the case for the other test trials). Test 3 revealed that in the absence of a distinct affective state, latency indicates no control by outcome-associated cues, but the percent label chosen indicated that the absence of the light was most likely to influence label selection. Perhaps, one could think of latency as confidence and selection as the best guess at what to do; a similar argument has been made for the value of an ambiguous landmark (Leising et al., 2015 ; Sawa et al., 2005 ). Recent tests have presented ambiguous stimuli to identify certain traits (e.g., optimism) in non-human animals (see Lagisz et al., 2020 for a review). Given the labels were ambiguous in the absence of any relevant events on No S D label trial types during Test 2, differences between subjects may suggest distinctions between “optimistic” and “pessimistic” pigeons. We did find one bird among the bunch that had no preference, perhaps the “realist” of the group. Test 3 investigated what cues might be supporting accurate labeling by presenting combinations of the external cues associated with reinforcer delivery: the sound of the hopper (Noise [N]), the hopper light (Light [L]), and eating (Food [F]). Most external stimuli did not influence labeling behavior. However, when the hopper light was off, labeling accuracy was below chance. The illuminated hopper light led to food checking, but did not produce above-chance level pecking to the label corresponding to reinforcer delivery. These results reveal that accurate labeling was driven mostly by contact with food and interoceptive cues (the interoceptive sensations associated with consumption, kinesthetic feedback from eating, and/or the corresponding affective state of pleasure). The logistics of isolating these cues prevented any additional testing. Further research is needed to fully disentangle affective and non-affective (e.g., contact with food) interpretations of these findings. Pigeons learned to peck labels following different core affect-inducing outcomes. With respect to the goals outlined in the introduction, we were able to (1) pit behavioral and interoceptive cues against one another when the LNF 1-s, L 4-s, and NF 4-s were presented without eating in Test 3, which led to chance performance on LNF 1-s and L 4-s and below-chance performance on NF 4-s; (2) establish accurate labeling in the absence of drug-induced states; (3) control reinforcer presence or absence with precision, leading to accurate labeling of both events within a single session; and (4) establish accurate labeling behavior in the absence of explicit reinforcement. The procedure presented here provides an option for those interested in studying affect in nonverbal species, as well as those interested in techniques to facilitate emotional competence in humans. Overall, the findings of the present experiment provide evidence that pigeons may be able to label naturally elicited states of core affect and transfer labeling behavior to novel situations. Additional Information Funding Not applicable. Competing Interests The authors declare no competing interests. Data Availability The datasets generated during the current study are available in the “Emotion Labeling in Pigeons” Open Science Framework (OSF) repository. Author's Contributions K.L. and W.S. conceived of the experiment. C.E. and J.N. prepared a document specifying all of the parameters to be used in the experiment and stages of training/testing. K.L. and C.E. programmed the experiment in Visual Basic and completed the initial data analysis. K.L., C.E., J.N., M.M., T.R., and S.J. were involved in data collection. K.L., C.E., J.N., M.M., T.R., S.J., and W.S. were involved in decisions regarding the experiment's progress. J.N., C.E., and K.L completed the final data analyses and first draft of the manuscript. K.L., J.N., S.J., and W.S. were involved in manuscript revisions. Ethical Approval The studies were performed in line with the principles of the Declaration of Helsinki and the Animal Welfare Act. Approval for the research was granted by the Institutional Animal Care and Use Committee at Texas Christian University. Informed Consent Not applicable. Footnotes Handling editor: Ralph Adolphs References Baron, J. (1992). The effect of normative beliefs on anticipated emotions. Journal of Personality and Social Psychology, 63 (2), 320–330. [ DOI ] [ PubMed ] [ Google Scholar ] Brady, R. J., & Hampton, R. R. (2021). Rhesus monkeys (Macaca mulatta) monitor evolving decisions to control adaptive information seeking. Animal Cognition, 24 (4), 777–785. 10.1007/s10071-021-01477-5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Bretherton, I., & Beeghly, M. (1982). Talking about internal states: The acquisition of an explicit theory of mind. Developmental Psychology, 18 (6), 906–921. 10.1037/0012-1649.18.6.906 [ Google Scholar ] Cheng, K., Spetch, M. L., & Johnston, M. (1997). Spatial peak shift and generalization in pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 23 (4), 469–481. 10.1037/0097-7403.23.4.469 Colpaert, F. C. (1978). Discriminative stimulus properties of narcotic analgesic drugs. Pharmacology Biochemistry & Behavior, 9 , 863–887. [ DOI ] [ PubMed ] [ Google Scholar ] Giddan, J. J., Bade, K. M., Rickenberg, D., & Ryley, A. T. (1995). Teaching the language of feelings to students with severe emotional and behavioral handicaps. Language, Speech & Hearing Services in Schools, 26 (1), 3–10. 10.1044/0161-1461.2601.03 [ Google Scholar ] Got, K. & Watanabe, S. (2020). The whole is equal to the sum of its parts: Pigeons (Columba livia) and crows (Corvus macrorhynchos) do not perceive emergent configurations. Learning and Behavior, 48 (1), 53–65. 10.3758/s13420-020-00413-w [ DOI ] [ PubMed ] Goddard, M., & Holland, P. (1996). Type of feature affects transfer in operant serial feature-positive discriminations. Animal Learning & Behavior, 24 (3), 266–276. 10.3758/BF03198975 [ Google Scholar ] Gray, S. L., Qadri, M. A. J., & Cook, R. G. (2021). Towards describing scenes by animals: Pigeons’ ordinal discrimination of objects varying in depth. Learning & Behavior, 49 (1), 85–98. 10.3758/s13420-020-00444-3 [ DOI ] [ PubMed ] [ Google Scholar ] Greenberg, M. T., & Kusche, C. A. (1993). Promoting social and emotional development in deaf children. the PATHS project . University of Washington Press. Griffiths, R. R., Lamb, R. J., Ator, N. A., Roache, J. D., & Brady, J. V. (1985). Relative abuse liability of triazolam: Experimental assessment in animals and humans. Neuroscience and Biobehavioral Reviews, 9 (1), 133–151. 10.1016/0149-7634(85)90039-9 [ DOI ] [ PubMed ] [ Google Scholar ] Hampton, R. R. (2001). Rhesus monkeys know when they remember. Proceedings of the National Academy of Sciences - PNAS, 98 (9), 5359–5362. 10.1073/pnas.071600998 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Holland, P., & Forbes, D. (1982). Control of conditional discrimination performance by CS-evoked event representations. Animal Learning & Behavior, 10 (2), 249–256. 10.3758/BF03212278 [ Google Scholar ] Iwasaki, S., Watanabe, S., & Fujita, K. (2018). Pigeons (Columba livia) know when they will need hints: Prospective metacognition for reference memory? Animal Cognition, 21 (2), 207–217. 10.1007/s10071-017-1153-z [ DOI ] [ PubMed ] [ Google Scholar ] Kornell, N. (2009). Metacognition in humans and animals. Current Directions in Psychological Science: A Journal of the American Psychological Society, 18 (1), 11–15. 10.1111/j.1467-8721.2009.01597.x [ Google Scholar ] Lagisz, M., Zidar, J., Nakagawa, S., Neville, V., Sorato, E., Paul, E. S., Bateson, M., Mendl, M., & Løvlie, H. (2020). Optimism, pessimism and judgement bias in animals: A systematic review and meta-analysis. Neuroscience and Biobehavioral Reviews, 118 , 3–17. 10.1016/j.neubiorev.2020.07.012 [ DOI ] [ PubMed ] [ Google Scholar ] Lechelt, D. P., & Spetch, M. L. (1997). Pigeons’ use of landmarks for spatial search in a laboratory arena and in digitized images of the arena. Learning and Motivation, 28 (3), 424–445. 10.1006/lmot.1997.0975 [ Google Scholar ] Leising, K. J., Hall, J. S., Wolf, J. E., & Ruprecht, C. M. (2015). Occasion setting during a spatial-search task with pigeons. Journal of Experimental Psychology: Animal Learning and Cognition, 41 (2), 163–178. 10.1037/xan0000048 [ DOI ] [ PubMed ] [ Google Scholar ] Lewis, M. (1992). Commentary: The self in self-conscious emotions. Monographs of the Society for Research in Child Development, 57 (1), 85. [ Google Scholar ] Lubinski, D., & Thompson, T. (1987). An animal model of the interpersonal communication of interoceptive (private) states. Journal of the Experimental Analysis of Behavior, 48 (1), 1–15. 10.1901/jeab.1987.48-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Martin, G. M., Gans, M., & van der Kooy, D. (1990). Discriminative properties of morphine that modulate associations between tastes and lithium chloride. Journal of Experimental Psychology Animal Behavior Processes, 16 (1), 56–68. 10.1037/0097-7403.16.1.56 [ PubMed ] [ Google Scholar ] Matthews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45 , 25–50. 10.1146/annurev.ps.45.020194.000325 [ DOI ] [ PubMed ] [ Google Scholar ] Mendl, M., & Paul, E. S. (2020). Animal affect and decision-making. Neuroscience and Biobehavioral Reviews, 112 , 144–163. 10.1016/j.neubiorev.2020.01.025 [ DOI ] [ PubMed ] [ Google Scholar ] Mendl, M., Burman, O. H. P., & Paul, E. S. (2010). An integrative and functional framework for the study of animal emotion and mood. Proceedings of the Royal Society B, Biological Sciences, 277 (1696), 2895–2904. 10.1098/rspb.2010.0303 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Mendl, M., Neville, V., & Paul, E. S. (2022). Bridging the gap: Human emotions and animal emotions. Affective Science, 3 , 703–712. 10.1007/s42761-022-00125-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Navarro, V. M., Wasserman, E. A., & Slomka, P. (2020). Taking pigeons to heart: Birds proficiently diagnose human cardiac disease. Learning & Behavior, 48 (1), 9–21. 10.3758/s13420-020-00410-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Ogden, P., & Minton, K. (2000). Sensorimotor psychotherapy: One method for processing traumatic memory. Traumatology, 6 (3), 149–173. 10.1177/153476560000600302 [ Google Scholar ] Panksepp, J. (2005). Affective consciousness: Core emotional feelings in animals and humans. Consciousness and Cognition, 14 , 30–80. 10.1016/j.concog.2004.10.004 [ DOI ] [ PubMed ] [ Google Scholar ] Roberts, W. A. (2021). An operant analog of food caching in the pigeon (Columba livia). Learning & Behavior, 50 (1), 82–88. 10.3758/s13420-021-00477-2 [ DOI ] [ PubMed ] [ Google Scholar ] Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110 (1), 145–172. 10.1037/0033-295X.110.1.145 [ DOI ] [ PubMed ] [ Google Scholar ] Saarni, C. (1999). The development of emotional competence . Guilford Press/Guilford Publications Inc. [ Google Scholar ] Sawa, K., Leising, K. J., & Blaisdell, A. P. (2005). Sensory preconditioning in spatial learning using a touch screen task in pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 31 (3), 368–375. 10.1037/0097-7403.31.3.368 [ DOI ] [ PubMed ] [ Google Scholar ] Skinner, D. M., & Martin, G. M. (1992). Conditioned taste aversions support drug discrimination learning at low dosages of morphine. Behavioral and Neural Biology, 58 (3), 236–241. 10.1016/0163-1047(92)90538-F [ DOI ] [ PubMed ] [ Google Scholar ] Skinner, D. M., Thornton, J. A., & Holland, P. C. (2003). The effects of feature type in operant serial feature-positive discriminations. Learning and Motivation, 34 (4), 389–409. 10.1016/S0023-9690(03)00036-5 [ Google Scholar ] Skinner, D. M., Goddard, M. J., & Holland, P. C. (1998). What can nontraditional features tell us about conditioning and occasion setting? In P. C. Holland, & N. A. Schmajuk (Eds.), Occasion setting: Associative learning and cognition in animals (pp. 113–144). American Psychological Association. 10.1037/10298-004 Smith, J. D. (2009). The study of animal metacognition. Trends in Cognitive Sciences, 13 (9), 389–396. 10.1016/j.tics.2009.06.009 [ DOI ] [ PubMed ] [ Google Scholar ] Smith, J. D., Shields, W. E., & Washburn, D. A. (2003). The comparative psychology of uncertainty monitoring and metacognition. The Behavioral and Brain Sciences, 26 (3), 317–339. 10.1017/S0140525X03000086 [ DOI ] [ PubMed ] [ Google Scholar ] Smith, J. D., Beran, M. J., Redford, J. S., & Washburn, D. A. (2006). Dissociating uncertainty responses and reinforcement signals in the comparative study of uncertainty monitoring. Journal of Experimental Psychology, 135 (2), 282–297. 10.1037/0096-3445.135.2.282 [ DOI ] [ PubMed ] [ Google Scholar ] Timler, G. R. (2003). Reading emotion cues: Social communication difficulties in pediatric populations. Seminars in Speech and Language, 24 (2), 121–130. 10.1055/s-2003-38903 [ DOI ] [ PubMed ] [ Google Scholar ] Watanabe, S. (2020). Discrimination of artificial starry sky by pigeons. Learning & Behavior, 48 (1), 22–26. 10.3758/s13420-020-00420-x [ DOI ] [ PubMed ] [ Google Scholar ] Way, I., Yelsma, P., Van Meter, A. M., & Black-Pond, C. (2007). Understanding alexithymia and language skills in children: Implications for assessment and intervention. Language, Speech & Hearing Services in Schools, 38 (2), 128–139. 10.1044/0161-1461(2007/013) [ DOI ] [ PubMed ] [ Google Scholar ] Wilkinson, A., & Kirkpatrick, K. (2020). Tracking of unpredictable moving stimuli by pigeons. Learning & Behavior, 48 (1), 27–40. 10.3758/s13420-020-00412-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Wright, A. A., Kelly, D. M., & Katz, J. S. (2021). Same/different concept learning by primates and birds. Learning & Behavior, 49 (1), 76–84. 10.3758/s13420-020-00456-z [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement The datasets generated during the current study are available in the “Emotion Labeling in Pigeons” Open Science Framework (OSF) repository. Articles from Affective Science are provided here courtesy of Springer ACTIONS View on publisher site PDF (1013.8 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Birds may represent a useful animal model for studying human mental disorders.#148040
• Investigating Emotions as Functional States Distinct From Feelings#114816
• "You are an expert annotator": Automatic Best-Worst-Scaling Annotations for Emotion Intensity Modeling#1011964
• Introspection of Emotions#719709
• THE EFFECTS OF LATE SIGN LANGUAGE ACQUISITION ON EMOTION EXPRESSION IN THE NARRATIVE RETELLINGS OF DEAF AND HARD-OF- HEARING CHILDREN AND ADULTS#439654