Transitive reasoning in the adult domestic hen in a six-term series task

Transitive reasoning in the adult domestic hen in a six-term series task | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Transitive reasoning in the adult domestic hen in a six-term series task Rachel Degrande, Océane Amichaud, Benoît Piégu, Fabien Cornilleau, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4431359/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 19 Nov, 2024 Read the published version in Animal Cognition → Version 1 posted 9 You are reading this latest preprint version Abstract Transitive inference (TI) is a disjunctive syllogism that allows an individual to indirectly infer a relationship between two components, by knowing their respective relationship to a third component (if A > B and B > C then A > C). The common procedure is the 5-terms series task, in which individuals are tested on indirect, unlearned relations. Few bird species have been tested for TI to date, which limits our knowledge of the phylogenetic spread of such reasoning ability. Here we tested TI in adult laying hens using a more solid methodology, the 6-terms series task, which has not been tested in poultry so far. Six hens were trained to learn direct relationships in a sequence of six arbitrary items (A > B > C > D > E > F) in a hybrid training procedure. Then, 12 testing sessions were run, comprising 3 non-rewarded inference trials each: BD, BE, and CE. All subjects showed TI within 12 inference trials and were capable of TI whatever the relative distance between the items in the series. We found that TI performance was not impacted by the reinforcement ratios of the items for most individuals; thus, making it harder to support a purely associative-based resolution of the task. We suggest that TI is based on the same cognitive processes in poultry ( Galloanserae ) than in modern flying birds ( Neoaves ), and that the cognitive strategy to solve the task might be driven mainly by individual parameters within species. These results contribute to a better understanding of transitive reasoning in birds. transitive inference relational memory cognition chicken Gallus gallus domesticus Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 INTRODUCTION To adapt efficiently to their living environment, animals generally use specific cognitive capacities, which may depend on species-specific physical or social constraints. In large groups, the ability to recognize conspecifics and to position themselves in the hierarchy enables individuals to avoid potential injuries due to predictable conflicts (Zuberbühler and Byrne 2006 ; Croney and Newberry 2007 ). In free-living conditions, hens form social groups organized according to a strict linear hierarchy, maintained by displays of dominance, also known as “pecking order” (Craig 1978 ). They are known to use a wide range of socio-cognitive skills to navigate such a complex social structure. For example, hens are able to recognize their peers based on several individual features (Abeyesinghe et al. 2009 ), and to adapt their behavior towards an unknown individual, depending on whether this individual won or lost the interaction with a known, dominant individual (Hogue et al. 1996 ). Among these socio-cognitive capacities, transitive inference (TI) is the ability to indirectly infer the relationship between two individuals, by knowing their respective relationship with a third individual. For example, if individual A is dominant over individual B, and individual B is dominant over individual C, then individual C can infer that individual A is dominant over themself. TI is thus particularly useful in hierarchical-group living. This complex ability requires a long-term memory about the relationships (relational memory) and the ability to use this knowledge to infer the indirect relationships. The common task to test TI in animals is the five-terms series task (Piaget 1928 ; Bryant and Trabasso 1971 ; McGonigle and Chalmers 1977 ). In this task, the individuals have to learn dyadic relationships between successive items in a five-term ordered sequence A > B > C > D > E (where the letters stand for the different items). That is, for example, when faced with the items A and B, individuals are reinforced when choosing A; when faced with B and C, they are reinforced when choosing B, and so on. Then, individuals are tested on the unanticipated, nonadjacent pair BD. Choosing B over D is interpreted as a demonstration of TI, as B and D are indirectly linked through their relationship with the third item C (learned relationships: B > C and C > D). However, the use of the five-terms series task has been questioned in the literature (Allen 2006 ). Since in a five-terms series, B is directly linked to A, the most positive item in terms of probability of reinforcement, and D is directly linked to E, the most negative item, this configuration impedes discussions about the cognitive process at stake both during the relational learning and during TI, that is, either a relational representation of the ordinal series, or a more associative-based representation. In the present study, we adapted a six-term series task to adult hens (Gallus gallus domesticus), with a twofold objective. The first one was to use a larger sequence of items (three non-adjacent and non-endpoint pairs: BD, BE and CE) allowing for the drawing of more solid conclusions about TI abilities, making it more difficult for animals to respond to items based solely on associative processes. This approach has already been successfully used in other avian species, such as pigeons and corvids (von Fersen et al. 1991 ; Bond et al. 2003 ; Daniels et al. 2014 ). The second one was to broaden our knowledge of the cognitive abilities of domesticated birds compared to non-domesticated species, since the former are generally expected to be less capable due to artificial selection. (Ferreira et al 2023 ). Hens were trained following a procedure adapted from the hybrid training procedure in pigeons, as proposed by Daniels et al. ( 2014 ). We ran a six-terms series of arbitrary items of different shape and color. Pairs were trained successively (AB, BC, CD, DE and EF), and mixed sessions, containing each previously trained pair, were included at each step. Such a hybrid training procedure, by gradually increasing complexity, allows to maximize and equalize the performance for every pair trained and to avoid different delays between training each pair and test sessions (Vasconcelos 2008 ; Daniels et al. 2014 ). The configuration of the task allowed us to test three different inference tests: BD, BE and CE. As already mentioned, instead of relying solely on a single pairwise comparison, these three tests allow for complementary analyses and more robust results (Vasconcelos 2008 ; Bond et al. 2010 ). We analyzed the learning performance, the TI performance and the choice latency for each adjacent pair (premise pairs) and for each nonadjacent pair (control and inference trials, respectively), at the group level, and at the individual level to measure eventual inter-individual variations. We put an emphasis on the learning accuracy between the premise pairs as it might give some indication about the cognitive resolution of the TI task (Bond et al. 2010 ). Finally, we tested for alternative cues that could have driven the choice behaviour of individuals through the configuration of the sessions. METHOD ETHICS APPROVAL This experimental procedure was approved by the Val de Loire Ethics Committee (approval n° CE19–2021-0211-1, CEEA VdL, France). Animal care and experimental treatments complied with the French and European guidelines for housing and care of animals used for scientific purposes (European Union Directives 2010/63/AU). SUBJECTS AND HOUSING Six adult laying hens (Isa Brown strain), aged 2 to 3 years old, were included in the procedure. The hens were maintained at the Pôle d’Expérimentation Avicole de Tours, where the experiment took place (UE PEAT, INRAE, 2018. Experimental Poultry Facility, doi: 10.15454/1.5572326250887292E12 ). They had access to a wood-chip littered barn (25m 2 ) equipped with nesting boxes and perches, and had access to an outside enclosure (approx. 30m 2 ) enriched with perches. Water was provided ad libitum, and food was delivered at will once the experiments of each testing day were completed. Birds were kept in a stable social group of 21 individuals on a 6am to 6pm daylight cycle. All experiments took place between 9am and 3pm and the testing order of the individuals was counterbalanced each day. STIMULI Each individual was attributed a unique sequence of 6 items, composed of the same items in a different order. The items (4x4cm) were printed in the center of white cards (10x8cm). Each item had a different color (yellow, red, black, green, blue and pink) and a different shape (square, circle, cross, diamond, triangle, star) (Fig. 1 ). The colors and the shapes were chosen in order to maximize the difference between the items. With the same purpose, the blank rate and the structural information rate were not controlled. Each individual had a unique sequence order of these items, thus, for clarity, we will name the items depending on their sequence position: A, B, C, D, E and F. APPARATUS AND GENERAL OPERATIVE 1. Apparatus The apparatus consisted in a starting box and a separation between the box and the experimenter, both made in condensed wood. The starting box (69cm long x 55cm large x 80cm high) was closed with a wire-meshed door (Fig. 2 a). A separation wall with a one-way mirror stood 20cm away from the starting box. This separation wall provided a 10cm high x 40cm large opening allowing the experimenter to present the cards to the hen without her being able to see the experimenter (Fig. 2 b). The pairs of cards were presented on a solid cardboard display (Fig. 2 c). 2. General procedure Before training, each individual was habituated individually to the new testing environment and to be carried out from the social group and placed in the starting box in a separate, adjacent room. Then each individual was accustomed to peck on a neutral card (black figure printed on a white card with the same dimensions as the test cards) to get mealworms. This habituation was carried out using a clicker-training procedure (Feng et al. 2016 ). After this procedure, the stimuli habituation and the training sessions started for each individual. During habituation, the experimenter randomly presented each item of the series, as a prior habituation, and each hen pecked directly on each item regardless of the shapes or the colors. Then, through the training sessions, the hens would learn to peck preferentially (i.e., to choose) on one item in the pair presented to get the reward (i.e., the prior item among the two in the series: for example, in ABC, peck at B when faced with B-C). For each trial, the pair was prepared out of the sight of the individual. Then, the wire-meshed door was opened and the experimenter presented the pair on the display. A choice was considered when the hen pecked on one card. If her choice was incorrect, the display was immediately taken back and the hen was brought back in the box for a 10-second wait. Then, a maximum of 3 correction trials was given with the exact same trial. If her choice was correct, the experimenter clicked and waited for a second peck on the same card to give her the reward (3 mealworms, either for a successful standard trial or a correction trial). After each trial, the hen was brought back in the starting box and the inter-trial interval was of 15 to 20 seconds. If the hen did not respond within 2 minutes, the experimenter closed the door and waited 10 seconds before trying again. If the hen did not respond for a total of 3 minutes, she was considered unmotivated and the session was postponed. Sessions could last from 20 to 45 minutes depending on the individual performance (number of correction trials needed) and depending on the number of trials in the session. Each individual had one session per day. For each training or testing session, the prior item for each pair was equally presented right and left among the trials (e.g., in the pair CD, the prior item is C), and a prior item was presented no more than twice in a row on the same side. A learning criterion was required at each training step: both in single-pair training and in mixed sessions (detailed below), the hen had to reach a minimum of 75% of success in 2 consecutive sessions with more than 50% of success for each side (right or left) of presentation of the prior item for each pair. However, if an individual reached 90% successful trials or more for both sides within one session, then the learning criterion was considered to be reached. The learning criteria stood to avoid any potential side bias, i.e., systematically pecking on the left or on the right card, independently of the card. In case of a side bias, a maximum of 8 correction sessions were run, that were adapted to the bias and the strength of the bias on a case-by-case basis. PROCEDURE 1. Training sessions First, during the single-pair training of the premise pairs, each pair of adjacent items was learned in the sequence order (e.g., AB, then BC, etc.). These training sessions comprised 20 trials, with an equal amount of left and right presentations of the prior item. We measured the number of sessions individuals needed to reach the learning criterion for each pair. Then, to ensure that hens were able to remember the correct choice for the items that were involved in two different situations (e.g., C is the prior item the hen had to peck on in CD but not in BC), mixed sessions, where all pairs previously learned were retrained, were added to our training schedule. These mixed sessions were included after the training of BC, CD, DE and EF, respectively (Fig. 3 ). This hybrid training procedure (successive pair training intermixed in mixed sessions) has been shown to lead to a faster significant performance for the premise pairs (Daniels et al. 2014 ). For a consistency of the number of trials per pair per session, and considering the conditions of prior items presentations, we allocated a total number of trials per session accordingly. Mixed sessions after BC training sessions had a total of 20 trials; those after CD had 18 trials; those after DE had 32 trials; those after EF had 40 trials. The number of presentations of each pair was equal in each session. We controlled afterwards that the number of presentations of each pair did not influence the performance in inference trials. We measured the number of sessions the individuals needed to reach the learning criterion for each mixed step. For the last three mixed sessions (which included the 5 premise pairs), a baseline performance (percentage of success) and the response latency was measured for each premise pair. 2. Test sessions Each individual had 12 test sessions with 46 trials each: 40 trials of adjacent pairs (8 trials per pair), 3 control trials and 3 inference trials. Control and inference trials were non-rewarded and with no correction trials. Extremity items are not used in inference trials as they are either always reinforced (A) or never reinforced (F). Thus, nonadjacent pairs involving an extremity item (e.g., A C or C F ) do not test for TI but allow for the basic understanding of associative values according to the reinforcement of each item during training (end-anchor effet, von Fersen et al. 1991 ). Control trials consisted in nonadjacent pairs with either the very first item A or the very last item F (AC, AD, AE, BF, CF or DF). These A and F control trials were pseudo-randomly distributed and equally presented among the 12 test sessions. Inference trials were trials of nonadjacent items (but not extremity items) for which transitive inference was needed to give the correct response. The pairs involved items that were both equally reinforced (for example, in BD: B + C- and D + E-) and non-reinforced (for example, in BD: A + B- and C + D-) during the training. The six-term series enabled to test TI through three inference trials (BD, BE and CE). Each one of the possible inference trials was presented once per test session. The sides of presentation of the prior item for each possible inference trial, were pseudo-randomized and equalized within two sessions. During the test sessions, the performance for the premise pairs was tracked. In case of a side- or an item-bias, we integrated optional correction sessions with, in main proportion, the difficult pair, and the two adjacent pairs, to ensure an equalized level of performance for every premise pair throughout test sessions. A maximum of two correction sessions was added each two test sessions. We measured the number of successful trials for the premise pair trials, for control trials, and for inference trials. We measured the response latency for control trials and for inference trials. STATISTICAL ANALYSIS All statistical analyses were performed using R version 4.2.1 (R Core Team 2022 ). We considered p-values below 0.05 to be significant, and p-values below 0.1 as a statistical tendency, for all statistical analyses. Chance level was considered at 50% of success. A non-parametric approach was preferred due to the small sample size (6 to 4 individuals). The response variables were either the number of sessions to criteria, or the percentage of success, or the response latency. The fixed variables were the side for the presentation of the prior item, the trial type, the pair and the number of the session. Anova permutational models were performed when several explanatory variables were to be considered, with individuals as random effect (aovperm function, package permuco , Frossard and Renaud 2021 ; np = 10000, type = permutation). Symmetry tests or one-way permutation tests were run to analyze the effect of a fixed variable with more than 2 modalities on the response, when the data were paired or not, respectively (package coin , Hothorn et al. 2006 ; two-tailed comparisons, np = 10000). Pairwise permutation t-tests were run for Posthoc analyses to detail the comparison between more than two modalities (pairwise.perm.t.test function, package RVAideMemoire , Hervé 2022 ; np = 10000, Holm correction). Two-tailed Wilcoxon rank-sum tests were used to assess the statistical significance of a comparison against the chance level, with Holm correction when appropriate (manual correction). Two-tailed exact binomial tests were used to test the statistical significance of individual performances (binom.test function). Homogeneity of variances was assessed with Levene tests before model fitting and before running multiple comparison analyses. Graphics were performed with the package ggplot2 (Wickham et al. 2019 ). For statistical analyses including less than 10 responses, the results are reported as the median and the first and third quartiles (MED [Q1:Q3]). For other statistical analyses, the results are reported as the mean and the standard deviation (MEAN +/- SD). As training and testing parameters can cause the use of associative cues in test trials, we ran supplementary analyses afterwards to control for some eventual response biases because of the sessions’ configuration (Guez and Audley 2013 ). We controlled for an effect of the reinforcement history and the reinforcement ratio of the items during training (Daisley et al. 2010 , 2021 ; Hotta et al. 2020 ; Okouchi and Lattal 2006 ; Lazareva et al. 2004 ) and of the configuration of premise pair trials previous to the test trials in test sessions (Russell et al. 1996 ). These analyses are detailed in Supplementary Materials ( Supplementary Material S1 ). RESULTS TRAINING SESSIONS The details of the number of sessions to reach the learning criterion at each step for each individual are presented in Supplementary Materials S1 , Table S3 . 1. Single-pair training of the premise pairs (i.e., pairs of successive items At the group level, the mean number of sessions needed to reach the learning criteria did not differ between the different pairs (variance homogeneity, one-way permutation test, chi2 = 4.5469, df = 4, p = 0.337; global median number of sessions = 3 [3:4] sessions to reach the criterion). 2. Mixed sessions At the group level, the mean number of sessions needed to reach the learning criterion was significantly different depending on the number of pairs included in the session (no variance homogeneity, one-way permutation test, chi2 = 12.203, df = 3, p = 0.007). More precisely, the learning speed was higher in the mixed sessions including pairs [AB to BC] compared to other stages of mixed sessions (significant with pairs [AB to CD]: p = 0.046; statistical tendency with [AB to DE]: p = 0.082 and with [AB to EF]: p = 0.074; pairwise permutation t-test). 3. Baseline performance for the premise pairs Four out of six individuals successfully finished the training by reaching the learning criteria ( Supplementary Materials , Table S2 ). Two individuals have not progressed to the required steps and were not included in the following analyses. In the last 3 mixed sessions, the mean performance for the premise pairs was of 80.42 +/- 17.88% of success, which is higher than chance level (two-tailed Wilcoxon rank-sum test, V = 1526.5, p < 0.001; Supplementary Materials , Table S3 ). We used a permutational anova model analysis to estimate the effect of the pair presented and of the side of the prior item, on the percentage of success. We found an effect of the side (F = 8.69, p = 0.011), with a global side bias for left over right (mean performance when prior item on the left = 85.69 +/- 19.29%; on the right = 74.81 +/- 25.59%) that was not different between the pairs (F = 1.464, p = 0.260). Moreover, we found an effect of the pair presented on the performance (F = 3.343, p = 0.038; Fig. 4 ). A multiple pairwise analysis revealed significant performance differences between pairs AB and DE (p = 0.035), CD and EF (p = 0.007) and DE and EF (p = 0.002) and a statistical tendency for the difference between pairs BC and EF (p = 0.053; pairwise permutation t-test). The performance for the premise pairs at the end of training for each individual is detailed in Fig. 5 . 4. Response latency for the premise pairs in the 3 last mixed sessions. The mean response latency for the premise pairs was 3.097 +/- 3.335 seconds. We found no difference of response latency between the premise pairs (one-way permutation test, variance homogeneity; chi2 = 4.9606, df = 4, p = 0.291). At the individual level, there was a significant difference in the response latency between the pairs for two individuals with a slower response latency for DE and EF (for Dion: between AB and DE: p = 0.020; between BC and DE: p = 0.014; between AB and EF: p = 0.048; between BC and EF: p = 0.048; for Daenerys: between AB and EF, p = 0.018; pairwise permutation t-test). The response latency for each premise pairs at the end of the training for each individual is detailed in Fig. 6 . TEST SESSIONS Four individuals successfully finished the training and were tested for transitive inference. The test sessions had 40 premise trials, 3 inference trials (BD, BE and CE) and 3 control trials per session. Each individual had 12 test sessions. The performance in control trials validated the basic understanding of the associative rule of the (non-)reinforced extremity items, as every hen performed better than chance level (total of 36 control trials; group median performance = 98.61 [95.14:100] %; binomial tests, p < 0.05; performance for A- trials = 97.22 [94.44:100] %; for F- trials = 100 [95.83:100] %). TI performance. All hens performed better than the chance level (total in 36 inference trials; 80.56% for Octo, 83.33% for Starr, 94.44% for Dion and 72.22% for Daenerys; binomial tests, p < 0.05). The median performance at the group level was 81.94%. Starr, Dion and Daenerys were successful in their very first inference trial. Within the 12 first inference trials, Dion and Daenerys performed significantly better than the chance level (12/12 and 10/12, respectively; binomial tests, p < 0.05; 7/12 for Octo and Starr). We found no side bias for test trials (anova permutation model analysis; F = 1.19, p = 0.314) either for control or inference trials (no interaction; F = 0.629, p = 0.432). Comparison of the different inference trials. At the group level, we found a statistical tendency for the difference in the performance between the 3 trial types at the group level with a higher performance for BE trials compared to BD and CE trials (variance homogeneity, symmetry test, maxT = 2.250, p = 0.063; median performance for BE = 100 [97.9:100] %; for BD = 83.30 [70.83:91.60] %; for CE = 66 [64.08:70.47] %). At the individual level, each hen performed better than 50% of success for each trial type (BD, BE and CE). All individuals performed significantly better than chance level when tested with BE, two individuals performed significantly better than chance level when tested with BD, and one individual when tested with CE (statistical significance for more than 10 successful trials among 12 for each inference trial possibility; two-tailed binomial test, p < 0.05; Fig. 7 ). The performance in inference trials was not explained by the differential ratio reinforcement of the items during training, except for Daenerys (Fig. 7 ). Response latency. We found a significant difference in the response latency between premise pairs, control trials and inference trials (One-way permutation test, chi2 = 7.553, df = 2, p = 0.023) with a significant difference between premise pairs and control trials (pairwise permutation t-test, p < 0.001; mean response latency for premises pairs trials = 3.0965 +/-3.33 seconds, for control trials = 2.043 +/-1.106 seconds, for inference trials = 2.519 +/- 2.189 seconds). We found no significant difference in the response latency between the different pairs presented in control and in inference trials (homogeneity of variances, One-way permutation test, chi2 = 7.89, df = 8, p = 0.444). In inference trials, the mean response latency for BD trials was of 2.271 +/- 1.201 seconds; for BE trials = 2.308 +/- 1.207 seconds; and for CE trials = 2.933 +/- 3.438 seconds. CONTROL FOR CONFOUNDING FACTORS ON CHOICE BEHAVIOR AND TI PERFORMANCE We analyzed whether the hens could have relied on other cues to be successful in test trials. These analyses are detailed in the Supplementary Materials S2 . The results show that the reinforcement ratios of the items and the configuration of the test sessions could not have allowed most hens to perform higher than the chance level in test trials, as (1) neither the reinforcement history nor the reinforcement ratio of the items resulting from the training stage influenced the performance in test trials, excepted for Daenerys, and (2) the configuration of the previous pair trial influenced the choice behavior in the following test trial. Interestingly, the fact that test trials were not rewarded might have affected the performance of the hens in premise pairs during test sessions. A permutation anova analysis showed that the choice to peck at an item in a premise pair significantly increased when the individual had pecked on this item in the previous non-rewarded test trial (control or inference trial) but only if it was presented on the other side (anova permutation model analysis; interaction: F = 7.108, p = 0.008). In parallel, whether hens gave the correct (C T ) or the incorrect (I T ) response at test trials did not significantly impact their performance (correct C H or incorrect I H ) at the next premise pair trial which presented one of the same items (two-tailed Wilcoxon tests with Holm correction; mean occurrence of I T I H =0, I T C H =10 +/- 4.97, C T I H =0, C T C H =42.5 +/- 9.85). DISCUSSION Our results show that the hens were capable of transitive inference when confronted with the 6-term series task. This confirms what was observed in previous studies in poultry that used a set up with a five-item series (in chicks: Daisley et al. 2010 , 2021 ; in geese: Weiß et al. 2010 ). We found an inter-individual variability in the resolution of the task, and notably, the performance of one hen suggested a resolution of the task through a mental representation of the series. The TI performance of most hens was not impacted neither by the reinforcement ratios of the items, nor the configuration of the sessions, supporting the use of transitive inference in this task, and making it harder to support a purely associative-based resolution of the task (except for one individual, Daenerys). Overall, the present study expands our knowledge on how chickens learn and solve a relational task. LEARNING PERFORMANCE Overall, the hens needed a mean of 51.25 +/- 13.43 sessions to complete the training stage. Hens were faster to learn the six-term series than greylag geese confronted with a 5-terms series with a similar hybrid training procedure (mean of 83.4 +/-17.1 sessions; Weiss et al. 2010). Two main reasons might explain this difference. Firstly, the learning criteria may have been more demanding in Weiss et al.’s study. Secondly, our hens were tested in a controlled environment without further distractions, which was not the case for the greylag geese. As in greylag geese (Weiß et al. 2010 ) and in pigeons (Clement and Zentall 2003 ), the hens learned the first two premise pairs at a similar speed ( Supplementary material S2 , Table S2 ). While these authors considered this performance to be part of a natural tendency of the animals to choose a familiar stimulus over a novel one, we observed a different pattern. We observed that the hens tended to apply the rule previously learned at the start of the training sessions for each premise pair (5 to 10 first trials), that is, they tended to first avoid the item they had not to peck at for the previous pair (e.g., do not peck at B in BC because that was the item to not peck at in AB). A difficulty to step to a new pair has already been found in other species (Treichler and Van Tilburg 1996 in macaques; Bond et al. 2003 in corvids). Thus, we hypothesize that, first, hens tended to apply a previous rule when facing new situations, which could be referenced as rule generalization, and then, that the learning speed of the new pair could rather be attributed to a high behavioral flexibility to learn the new rule (see Degrande et al. 2022 ). The comparison of the performance of the premise pairs has been shown to bring information about the cognitive process that could be at stake in the n-term series task (Bond et al. 2010 ). When comparing the performance between the premise pairs at the end of the training, each hen showed a better performance for AB and EF compared to the other pairs. This performance can be associated with what is known as the end-anchor effect. In simple terms, this serial position effect (i.e., a performance that depends on the serial position of the items in the ordinal series) implies better performance when the pair includes an extremity item (here, A or F; Allen 2006 ). This effect can be explained through the strong conditioning (positive and negative, respectively) that comes with the very first item of the series, that is always reinforced, and the very last, that is never reinforced. Concerning the response latency, two out of four individuals showed a choice speed that was dependent on the serial position of the pairs, with a slower response for the pairs at a greater distance from A. Some authors argue that this effect suggest a mental representation of the ordinal series (Terrace 2005 ): the item A being the reference point of the spatial representation of the series, the response latency for the premise pairs increases with the distance from A. This result was significant for Daenerys (AB < EF) and for Dion (AB < DE and EF, and BC < DE and EF). However, associative hypotheses can account for this effect. For example, the first-item effect hypothesis (von Fersen et al. 1991 ) claims that the first item of the series (A) has a strong positive value and is thus associated with a shorter response latency, and that the response latency for the premise pairs involving the following items in the series will decrease gradually with the distance to A. Another hypothesis is that a higher number of presentations of the pairs during training might generate a shorter response latency for these pairs. This needs to be further investigated as the result is not significant at the group level in our study. At the individual level, two out of our six individuals did not manage to end the training stage, by not being able to mix the first two premise pairs (Elizabeth) or not being able to mix the five premise pairs together (Savana). The fact that some individuals never reach the learning criteria has been found in other species (e.g., in pigeons: von Fersen et al. 1991 ). The n-term task, commonly used for TI testing, involves some individual parameters that are prerequisites to the task and that can strongly impact the performance, independent of the ability to perform TI. Independently of inferential and relational memory abilities, this task requires a long-term memory, a retrieval capacity, a certain level of behavioral flexibility, and the ability to apply a different rule for the same item depending on the context of presentation (i.e., depending on the pair). Personality and social structure issues have been shown to impact these parameters in birds (chicks: Daisley et al. 2021 ; geese: Weiss et al. 2010). This observation calls for precautions when testing for TI in animals with the common n-term series task. Finally, we found a significant side bias for left over right in the last mixed sessions of the relational training. In the same way, Daisley et al. ( 2010 ) found that chicks that could use their left eye only had a better TI performance in comparison to chicks that could use their right eye only. As the bird’s visual pathway have been shown to be contralateral (Deng and Rogers 1998 ), the authors conclude that the right hemisphere may be more implicated in TI, which is consistent with a development of TI capacities through relational representations in mammals (see for example Dusek and Eichenbaum 1997 ). However, we did not find a significant side bias for the TI performance, which could mean that the cognitive process implied in relational learning is different than the one implied in relational retrieval in birds. TRANSITIVE INFERENCE PERFORMANCE The hens showed transitive inference in the 6-term series task. Each hen performed significantly better than chance in 36 inference trials (non-rewarded test trials), and two hens did it in 12 trials. Notably, one individual (Dion) performed significantly better than chance level for each of the three different trials (BD, BE and CE). Further analyses showed that the hens could not have used some alternative strategy other than transitive inference to perform better than chance in inference trials, as we controlled for a strategy based on the reinforcement ratio of the items and for a performance based on the configuration of the sessions (see Supplementary Materials S2 ). TI performance was not dependent on the reinforcement ratio of the items in three out of the four hens, which is an argument against a performance relying on associative responses. The six-term series enabled us to study further the cognitive resolution of the task through the three inference trials. Hens performed better in inference trials with more distant items, i.e., in the nonadjacent pair BE (performance significantly better than chance for the four hens), compared to pairs BD and CE (sign. for two hens and for one hen, respectively). Such a performance, that is related to the distance between two items, is referred to as the symbolic distance effect (Moyers et al. 2018 ). It has been found in other species as pigeons (von Fersen et al. 1991 ; Daniels et al. 2014 ), corvids (Bond et al. 2003 ) or even in humans (for example: Bryant and Trabasso 1971 ) and is consistent with linear models of TI (von Fersen et al. 1991 ; Daniels et al. 2014 ). However, the premise pairs were trained in a temporal order that was consistent with their ordinal rank, thus, we cannot state from our results whether the hens mentally organized the items of the series along a linear representation (Couvillon and Bitterman 1992 ; MacLean et al. 2008 ). This symbolic distance effect is also explainable by the Value Transfer Theory (VTT) from von Fersen et al. ( 1991 ) or derived associative-based models (see for example Zentall et al. 1996 ; discussed in: Allen 2006 ; Vasconcelos 2008 ; Guez and Audley 2013 ). In these models, each item is being transferred a partial associative value from its adjacent items in the series, depending on the reward contingency that occurs in the pairs presented. This way, A transfers its positive value along the adjacent items, and F transfers its non-rewarding, negative value along its adjacent items. This transfer mechanism could thus make it easier to respond to BE through associative values than BD or CE. According to the symbolic distance between the items, the performance for the nonadjacent pairs BD and CE should have been similar. However, we found that the performance for BD was better than that for CE for two out of four individuals (not significant). The decreasing TI performance from BE to BD, and from BD to CE has also been found in pigeons in Daniels et al. ( 2014 ) at the group level. Overall, it is possible that the worse performance in CE is related to a recency effect for the learning of the pair EF (with E rewarded), which hypothesis must be validated through further studies (Bond et al. 2003 ). Focusing on the performance of each individual separately, we show that the hens might have used different cognitive strategies to solve the TI task. At one extreme, Dion probably solved the task through a relational representation of the series: she showed a high performance for the premise pairs and showed no first- or last-item effect, and an almost perfect performance for the three different inference trials. Her performance highlights the relevance of studying the response latency to observe potential response differences between the pairs at a high level of performance. At the other extreme, Daenerys used a more associative strategy, as her performance could be predicted by the values of the differential ratio of reinforcement at the end of her training. Her learning performances and her inference performances match an associative model including a transfer of associative values through the items. More broadly, the relational learning involved in this task might be multiply encoded. The TI task, as it is designed in most experiments, might simultaneously involve both cognitive- (through a relational representation) and associative- (through associative-based models) strategies (Jacobs 2006 ; Lazareva and Wasserman 2006 ; Bond et al. 2010 ). CONCLUSION AND PERSPECTIVES To our knowledge, this is the first attempt to further study the underlying process of transitive inference in the domestic chicken, and more generally in Galloanserae . Our results confirm that adult chickens are capable of transitive inference, and support that their transitive response cannot be explained just in terms of an associative-based model. Moreover, we found that some hens might be able to solve the transitive task through a mental representation of the series. Since our results resemble those found in Neoaves (pigeons and corvids), our hypothesis is that TI is based on the same cognitive processes in both phylogenetic groups, and that the cognitive strategy to solve the transitive task might be driven mainly by individual parameters within species. Our study brings additional information about the cognitive capacities in chickens, which might be more complex than often assumed. Declarations ETHICS APPROVAL This experimental procedure was approved by the Val de Loire Ethics Committee (approval n° CE19 – 2021-0211-1, CEEA VdL, France). Animal care and experimental treatments complied with the French and European guidelines for housing and care of animals used for scientific purposes (European Union Directives 2010/63/AU). SUPPLEMENTARY MATERIALS Supplementary Material is available online in the INRAE data repository at https://doi.org/10.57745/TOXZ8L . DATA AND MODEL AVAILABILITY STATEMENT The data that support the study findings is hosted in the INRAE data repository at https://doi.org/10.57745/TOXZ8L and is available on reasonable request from the corresponding author at [email protected] . AUTHORS CONTRIBUTION Rachel Degrande, Fabien Cornilleau, Ludovic Calandreau, Léa Lansade and Océane Amichaud contributed to the study design. The habituation and the first training stages of the hens with clicker training were performed by Océane Amichaud. Material preparation, data collection and analysis were performed by Rachel Degrande. Plotine Jardat helped with data analysis. Benoit Piegu guided the analyses presented in the Supplementary Materials . The first draft of the manuscript was written by Rachel Degrande and all authors commented on previous versions of the manuscript. Ludovic Calandreau supervised the research. All authors read and approved the final manuscript. DECLARATION OF INTEREST The authors have no conflict of interest to disclose. ACKNOWLEDGEMENTS The authors acknowledge the experimental unit Pôle d’Expérimentation Avicole de Tours (UE PEAT, INRAE, 2018. Experimental Poultry Facility, DOI: 10.15454/1.5572326250887292E12) where hens were maintained and where the experiments took place. FINANCIAL SUPPORT STATEMENT This research received the funding support of the PHASE research Department (INRAE) and a PhD fellowship INRAE-Région Centre Val de Loire (France). References Abeyesinghe SM, McLeman MA, Owen RC, et al. (2009) Investigating social discrimination of group members by laying hens. Behav Processes 81:1–13. https://doi.org/10.1016/j.beproc.2008.11.017 Allen C (2006) Transitive inference in animals: Reasoning or conditioned associations? In: Hurley S, Nudds M (eds) Rational Animals? Oxford University Press, pp 175–186 Aust U, Range F, Steurer M, Huber L (2008) Inferential reasoning by exclusion in pigeons, dogs, and humans. Anim Cogn 11:587–597. https://doi.org/10.1007/s10071-008-0149-0 Avarguès-Weber A, Portelli G, Benard J, et al. (2010) Configural processing enables discrimination and categorization of face-like stimuli in honeybees. J Exp Biol 213:593–601. https://doi.org/10.1242/jeb.039263 Bond AB, Kamil AC, Balda RP (2003) Social complexity and transitive inference in corvids. Anim Behav 65:479–487. https://doi.org/10.1006/anbe.2003.2101 Bond AB, Wei CA, Kamil AC (2010) Cognitive representation in transitive inference: A comparison of four corvid species. Behav Processes 85:283–292. https://doi.org/10.1016/j.beproc.2010.08.003 Bryant PE, Trabasso T (1971) Transitive inferences and memory in young children. Nature 232:456–458. https://doi.org/10.1038/232456a0 Camarena HO, García-Leal O, Burgos JE, et al. (2018) Transitive Inference Remains Despite Overtraining on Premise Pair C+D-. Front Psychol 9:1791. https://doi.org/10.3389/fpsyg.2018.01791 Clement TS, Zentall TR (2003) Choice based on exclusion in pigeons. Psychon B Rev 10:959–964. https://doi.org/10.3758/BF03196558 Couvillon PA, Bitterman ME (1992) A conventional conditioning analysis of “transitive inference” in pigeons. Journal of Experimental Psychology: Anim Behav Processes 18:308–310. https://doi.org/10.1037/0097-7403.18.3.308 Craig JV (1978) Aggressive Behavior of Chickens: Some Effects of Social and Physical Environments. Technical Report, KansasnState University. Presented at the 27th Annual National Breeder’s Roundtable, Kansas City, May 11, 1978. Croney CC, Newberry RC (2007) Group size and cognitive processes. Appl Anim Behav Sci 103:215–228. https://doi.org/10.1016/j.applanim.2006.05.023 Daisley JN, Vallortigara G, Regolin L (2010) Logic in an asymmetrical (social) brain: Transitive inference in the young domestic chick. Soc Neurosci 5:309–319. https://doi.org/10.1080/17470910903529795 Daisley JN, Vallortigara G, Regolin L (2021) Low-rank Gallus gallus domesticus chicks are better at transitive inference reasoning. Commun Biol 4:1344. https://doi.org/10.1038/s42003-021-02855-y Daniels CW, Laude JR, Zentall TR (2014) Six-term transitive inference with pigeons: Successive-pair training followed by mixed-pair training: Six-term transitive inference with pigeons. J Exp Anal Behav 101:26–37. https://doi.org/10.1002/jeab.65 Degrande R, Cornilleau F, Jardat P, Ferreira VHB, Lansade L, Calandreau L (2024) A cognitive approach to better understand foraging strategies of the adult domestic hen. Preprint access at https://doi.org/10.21203/rs.3.rs-4121447/v1 Degrande R, Cornilleau F, Lansade L, Jardat P, Colson V, Calandreau L (2022) Domestic hens succeed at serial reversal learning and perceptual concept generalisation using a new automated touchscreen device. Animal 16:16(8), 100607. https://doi.org/10.1016/j.animal.2022.100607 Deng C, Rogers LJ (1998) Bilaterally projecting neurons in the two visual pathways of chicks. Brain Res 794:281–290. https://doi.org/10.1016/S0006-8993(98)00237-6 Dusek JA, Eichenbaum H (1997) The hippocampus and memory for orderly stimulus relations. Proc Natl Acad Sci USA 94:7109–7114. https://doi.org/10.1073/pnas.94.13.7109 Feng LC, Howell TJ, Bennett PC (2016) How clicker training works: Comparing Reinforcing, Marking, and Bridging Hypotheses. Appl Anim Behav Sci 181:34–40. https://doi.org/10.1016/j.applanim.2016.05.012 Ferreira, V. H. B., Lansade, L., Calandreau, L., Cunha, F., & Jensen, P. (2023). Are domesticated animals dumber than their wild relatives? A comprehensive review on the domestication effects on animal cognitive performance. Neurosci Biobehav Rev, 154, Article 105407. https://doi.org/10.1016/j.neubiorev.2023.105407 Frossard J, Renaud O (2021) “Permutation Tests for Regression, ANOVA, and Comparison of Signals: The permuco Package.” J Stat Softw 99:1–32. https://doi.org/10.18637/jss.v099.i15 Galizio A, Doughty AH, Williams DC, Saunders KJ (2017) Understanding behavior under nonverbal transitive-inference procedures: Stimulus-control-topography analyses. Behav Processes 140:202–215. https://doi.org/10.1016/j.beproc.2017.05.010 Greene AJ, Spellman BA, Levy WB, et al. (2001) Relational learning with and without awareness: Transitive inference using nonverbal stimuli in humans. Mem Cogn 29:893–902. https://doi.org/10.3758/BF03196418 Guez D, Audley C (2013) Transitive or Not: A Critical Appraisal of Transitive Inference in Animals. Ethology 119:703–726. https://doi.org/10.1111/eth.12124 Hervé M (2022) RVAideMemoire: Testing and Plotting Procedures for Biostatistics. R package version 0.9-81-2. https://www.R-project.org Hogue M-E, Beaugrand JP, Laguë PC (1996) Coherent use of information by hens observing their former dominant defeating or being defeated by a stranger. Behav Processes 38:241–252. https://doi.org/10.1016/S0376-6357(96)00035-6 Hothorn T, Hornik K, van de Wiel M, Zeileis A (2006) “A Lego system for conditional inference.” Am Stat 60:257–263. https://doi.org/10.1198/000313006X118430 Hotta T, Ueno K, Hataji Y, et al (2020) Transitive inference in cleaner wrasses (Labroides dimidiatus). PLoS ONE 15:e0237817. https://doi.org/10.1371/journal.pone.0237817 Jacobs L (2006) From Movement to Transitivity: The Role of Hippocampal Parallel Maps in Configural Learning. Rev Neurosci, 17(1-2) 99-109. https://www.doi.org/ 10.1515/revneuro.2006.17.1-2.99 Lazareva OF, Paxton Gazes R, Elkins Z, Hampton R (2020) Associative models fail to characterize transitive inference performance in rhesus monkeys (Macaca mulatta). Learn Behav 48:135–148. https://doi.org/10.3758/s13420-020-00417-6 Lazareva OF, Smirnova AA, Bagozkaja MS, et al (2004) Transitive responding in hooded crows requires linearly ordered stimuli. J Exp Anal Behav 82:1–19. https://doi.org/10.1901/jeab.2004.82-1 Lazareva OF, Wasserman EA (2006) Effect of stimulus orderability and reinforcement history on transitive responding in pigeons. Behav Processes 72:161–172. https://doi.org/10.1016/j.beproc.2006.01.008 MacLean EL, Merritt DJ, Brannon EM (2008) Social complexity predicts transitive reasoning in prosimian primates. Anim Behav 76:479–486. https://doi.org/10.1016/j.anbehav.2008.01.025 McGonigle BO, Chalmers M (1977) Are monkeys logical? Nature, 267(5613), 694–696. https://doi.org/10.1038/267694a0 Mikolasch S, Kotrschal K, Schloegl C (2013) Transitive inference in jackdaws (Corvus monedula). Behav Processes 92:113–117. https://doi.org/10.1016/j.beproc.2012.10.017 Moyers SC, Adelman JS, Farine DR, et al. (2018) Exploratory behavior is linked to stress physiology and social network centrality in free-living house finches (Haemorhous mexicanus). Horm Behav 102:105–113. https://doi.org/10.1016/j.yhbeh.2018.05.005 Okouchi H, Lattal KA (2006) An analysis of reinforcement history effects. J Exp Anal Behav 86:31–42. https://doi.org/10.1901/jeab.2006.75-05 Piaget J (1928) Judgment and reasoning in the child. Humana Mente 3(12):551-554 R Core Team (2022) R: A language and environment for statistical Computing. https://www.R-project.org Russell J, McCormack T, Robinson J, Lillis G (1996) Logical versus Associative Performance on Transitive Reasoning Tasks by Children: Implications for the Status of Animals Performance. Qu J Exp Psychol 49B:231–244 Terrace HS (2005) The simultaneous chain: a new approach to serial learning. Trends Cogn Sci 9:202–210. https://doi.org/10.1016/j.tics.2005.02.003 Treichler FR, Van Tilburg D (1996) Concurrent conditional discrimination tests of transitive inference by macaque monkeys: List linking. Journal of Experimental Psychology: Anim Behav Processes 22:105–117. https://doi.org/10.1037/0097-7403.22.1.105 Vasconcelos M (2008) Transitive inference in non-human animals: An empirical and theoretical analysis. Behav Processes 78:313–334. https://doi.org/10.1016/j.beproc.2008.02.017 von Fersen L, Wynne CD, Delius JD, Staddon JE (1991) Transitive inference formation in pigeons. Journal of Experimental Psychology: Anim Behav Processes 17:334–341. https://doi.org/10.1037/0097-7403.17.3.334 Weiß BM, Kehmeier S, Schloegl C (2010) Transitive inference in free-living greylag geese, Anser anser. Anim Behav 79:1277–1283. https://doi.org/10.1016/j.anbehav.2010.02.029 Werchan DM, Gómez RL (2013) Generalizing memories over time: Sleep and reinforcement facilitate transitive inference. Neurobiol Learn Mem 100:70–76. https://doi.org/10.1016/j.nlm.2012.12.006 Wickham H, Averick M, Bryan J, et al. (2019) Welcome to the Tidyverse. JOSS 4:1686. https://doi.org/10.21105/joss.01686 Zentall TR, Peng D, Miles L (2019) Transitive inference in pigeons may result from differential tendencies to reject the test stimuli acquired during training. Anim Cogn 22:619–624. https://doi.org/10.1007/s10071-019-01257-2 Zentall TR, Sherburne LM, Roper KL, Kraemer PJ (1996) Value Transfer in a Simultaneous Discrimination Appears to Result From Within-Event Pavlovian Conditioning. J Exp Psychol: Anim Behav Processes, 22(1), 68-75. https://doi.org/10.1037/0097-7403.22.1.68 Zuberbühler K, Byrne RW (2006) Social cognition. Curr Biol, 16:R786–R790. https://doi.org/ 10.1016/j.cub.2006.08.046 Additional Declarations No competing interests reported. Supplementary Files DegrandeetalTransitivereasoninghensSupplementaryMaterials.pdf Cite Share Download PDF Status: Published Journal Publication published 19 Nov, 2024 Read the published version in Animal Cognition → Version 1 posted Editorial decision: Revision requested 06 Aug, 2024 Reviews received at journal 12 Jul, 2024 Reviews received at journal 21 Jun, 2024 Reviewers agreed at journal 24 May, 2024 Reviewers agreed at journal 23 May, 2024 Reviewers invited by journal 22 May, 2024 Editor assigned by journal 22 May, 2024 Submission checks completed at journal 17 May, 2024 First submitted to journal 16 May, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4431359","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":306470434,"identity":"e37ae2bc-dc19-4505-9dad-94ecf08e6439","order_by":0,"name":"Rachel Degrande","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDElEQVRIiWNgGAWjYHACNgaGAgbGBihbDkQdeEBQiwFCizFYSwIpWhLBDHxazNsPP3vwwYBBtp9/7cPPBWU26fPDDj8E2mInp9uAXYvMmTRzwxkGDMYzZzw3lp5xLi134+00A6CWZGOzA9i1SDDksEnzGDAkbrhxjEGat+1w7sbZCSAtBxK34dLC/4ZN+g9Qy/4bx5h/87b9Tzecnf4BvxYJoC0MIFv429iAthxIkJfOIWCLxDMzyR4DCeMZN9jYrHnOJRtukM4pOJBggMcv/MnPJH5U2Mj29x9jvs1TZicvPzt984cPFXZyuLTAQ4FBIgHCNACrNMCrHAr4oYbKNxCjehSMglEwCkYSAACPkFxcM6A67wAAAABJRU5ErkJggg==","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":true,"prefix":"","firstName":"Rachel","middleName":"","lastName":"Degrande","suffix":""},{"id":306470436,"identity":"9753ac03-fb2a-4bde-930b-17f73ab06ff6","order_by":1,"name":"Océane Amichaud","email":"","orcid":"","institution":"IFIP - Institut du Porc, 9 Boulevard du Trieux, 35740 Pacé","correspondingAuthor":false,"prefix":"","firstName":"Océane","middleName":"","lastName":"Amichaud","suffix":""},{"id":306470437,"identity":"e86a4864-604b-4702-bd9f-54d75a5b3eeb","order_by":2,"name":"Benoît Piégu","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Benoît","middleName":"","lastName":"Piégu","suffix":""},{"id":306470440,"identity":"68e5dcd7-c4d2-41de-93a1-78b3263b63e6","order_by":3,"name":"Fabien Cornilleau","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Fabien","middleName":"","lastName":"Cornilleau","suffix":""},{"id":306470442,"identity":"88dde421-f3bd-4ebd-83cd-15a952f351c3","order_by":4,"name":"Plotine Jardat","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Plotine","middleName":"","lastName":"Jardat","suffix":""},{"id":306470443,"identity":"e98c9909-e8cb-4182-b273-c4356e921e97","order_by":5,"name":"Vitor Hugo Bessa Ferreira","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Vitor","middleName":"Hugo Bessa","lastName":"Ferreira","suffix":""},{"id":306470444,"identity":"94f11f6d-624d-4f62-b4a3-3a5eae0365cb","order_by":6,"name":"Violaine Colson","email":"","orcid":"","institution":"Laboratoire de Physiologie et Génomique des Poissons, INRAE, Rennes","correspondingAuthor":false,"prefix":"","firstName":"Violaine","middleName":"","lastName":"Colson","suffix":""},{"id":306470446,"identity":"21356377-3f29-41aa-8b3c-6b71cdfbe220","order_by":7,"name":"Léa Lansade","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Léa","middleName":"","lastName":"Lansade","suffix":""},{"id":306470447,"identity":"c6d1800e-9d98-4007-82d5-18de169e50c9","order_by":8,"name":"Ludovic Calandreau","email":"","orcid":"","institution":"CNRS, INRAE, Université de Tours, PRC (Physiologie de la Reproduction et des Comportements), F-37380, Nouzilly, Indre-et-Loire","correspondingAuthor":false,"prefix":"","firstName":"Ludovic","middleName":"","lastName":"Calandreau","suffix":""}],"badges":[],"createdAt":"2024-05-16 13:29:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4431359/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4431359/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10071-024-01914-1","type":"published","date":"2024-11-19T15:57:19+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":57705996,"identity":"86f6d7ca-12dd-4500-a40b-460efaad4777","added_by":"auto","created_at":"2024-06-04 14:52:11","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":20567,"visible":true,"origin":"","legend":"\u003cp\u003eExample of the priority sequence of the items for one individual. The colors and shapes correspond to those that have been printed on the cards for the procedure (yellow square, red circle, black cross, green diamond, blue triangle, pink star). The priority order is from item A to item F.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/0c4033109f4f46b8158a1388.png"},{"id":57706000,"identity":"b8925931-d847-4cf7-8dd5-d1ad23101933","added_by":"auto","created_at":"2024-06-04 14:52:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":907420,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a)\u003c/strong\u003e Apparatus, with the starting box and the separation wall. \u003cstrong\u003e(b) \u003c/strong\u003eIn each trial, the experimenter showed a pair of cards through the opening on the separation wall, and the tested hen was allowed to choose, by pecking, one of the two items presented. (\u003cstrong\u003ec\u003c/strong\u003e) The cards were presented on the display.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/47b62c9832476b75beb3e999.png"},{"id":57706575,"identity":"690f1a4d-35b5-4253-8041-e610d15763ad","added_by":"auto","created_at":"2024-06-04 15:00:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":65380,"visible":true,"origin":"","legend":"\u003cp\u003eSequence of the pair training of adjacent items and of the mixed sessions. Individuals had to pass a learning criterion at each step. The outlined items are the prior items in the pair.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/c3dc95f7a5dee7a72e3798ef.png"},{"id":57705995,"identity":"fb7bb5fb-8075-4853-b872-fdf475dca10c","added_by":"auto","created_at":"2024-06-04 14:52:11","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":30670,"visible":true,"origin":"","legend":"\u003cp\u003ePercentage of success at the group level for each premise pair at the end of the training (3 last mixed sessions – 8 trials per pair per session). *: p\u0026lt;0.05; **: p\u0026lt;0.01; +: p\u0026lt;0.1.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/c68a43de4035cbddef95fa0a.png"},{"id":57706574,"identity":"5e32bceb-2623-4c53-b1c8-5214e2bf2acf","added_by":"auto","created_at":"2024-06-04 15:00:11","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":70287,"visible":true,"origin":"","legend":"\u003cp\u003eMean percentage of success at the individual level for each premise pair at the end of the training (3 last mixed sessions), for the four individuals that passed the learning criteria. At the individual level, there was no significant difference in the performance of the premise pairs.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/36d777173e7eab5cd3f0a795.png"},{"id":57706576,"identity":"d606f54a-7fdd-40c0-b31d-73a533cad27b","added_by":"auto","created_at":"2024-06-04 15:00:11","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":61988,"visible":true,"origin":"","legend":"\u003cp\u003eMean response latency in seconds for each premise pair for each individual, at the end of the training (3 last mixed sessions). *: p\u0026lt;0.05 (pairwise permutation t-test).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/14d20a404f1a2f4eb9b6cc1c.png"},{"id":57706002,"identity":"b8bdb8cc-b56b-4918-8d57-7adf9fa7c9ab","added_by":"auto","created_at":"2024-06-04 14:52:11","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":114220,"visible":true,"origin":"","legend":"\u003cp\u003eDetailed performance for each type of inference trial for each individual. Left y-axis: percentage of success in 12 inference trials. Right y-axis: differential ratio of reinforcement in percentage (see \u003cem\u003eSupplementary Materials S2\u003c/em\u003e). For example, the higher D\u003csub\u003eBD\u003c/sub\u003e is, the more B has been reinforced in comparison to D. The black horizontal line indicates the performance to reach to be significantly higher than the chance level is 83,33% (10 trials out of 12).\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/5193b6efcd54ebd66f2b8bbb.png"},{"id":69834860,"identity":"6b5f0456-03b8-484b-a63e-6d3009450c28","added_by":"auto","created_at":"2024-11-25 16:09:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1718983,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/c06549a3-509d-441c-a23b-0af3645347cd.pdf"},{"id":57705997,"identity":"44508fca-0f03-46c3-b224-fdda64dc3042","added_by":"auto","created_at":"2024-06-04 14:52:11","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":297345,"visible":true,"origin":"","legend":"","description":"","filename":"DegrandeetalTransitivereasoninghensSupplementaryMaterials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4431359/v1/13e01e245bea39f8d3fef371.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Transitive reasoning in the adult domestic hen in a six-term series task","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eTo adapt efficiently to their living environment, animals generally use specific cognitive capacities, which may depend on species-specific physical or social constraints. In large groups, the ability to recognize conspecifics and to position themselves in the hierarchy enables individuals to avoid potential injuries due to predictable conflicts (Zuberb\u0026uuml;hler and Byrne \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Croney and Newberry \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). In free-living conditions, hens form social groups organized according to a strict linear hierarchy, maintained by displays of dominance, also known as \u0026ldquo;pecking order\u0026rdquo; (Craig \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1978\u003c/span\u003e). They are known to use a wide range of socio-cognitive skills to navigate such a complex social structure. For example, hens are able to recognize their peers based on several individual features (Abeyesinghe et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), and to adapt their behavior towards an unknown individual, depending on whether this individual won or lost the interaction with a known, dominant individual (Hogue et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1996\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAmong these socio-cognitive capacities, transitive inference (TI) is the ability to indirectly infer the relationship between two individuals, by knowing their respective relationship with a third individual. For example, if individual A is dominant over individual B, and individual B is dominant over individual C, then individual C can infer that individual A is dominant over themself. TI is thus particularly useful in hierarchical-group living. This complex ability requires a long-term memory about the relationships (relational memory) and the ability to use this knowledge to infer the indirect relationships.\u003c/p\u003e \u003cp\u003eThe common task to test TI in animals is the five-terms series task (Piaget \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e1928\u003c/span\u003e; Bryant and Trabasso \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1971\u003c/span\u003e; McGonigle and Chalmers \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e1977\u003c/span\u003e). In this task, the individuals have to learn dyadic relationships between successive items in a five-term ordered sequence A\u0026thinsp;\u0026gt;\u0026thinsp;B\u0026thinsp;\u0026gt;\u0026thinsp;C\u0026thinsp;\u0026gt;\u0026thinsp;D\u0026thinsp;\u0026gt;\u0026thinsp;E (where the letters stand for the different items). That is, for example, when faced with the items A and B, individuals are reinforced when choosing A; when faced with B and C, they are reinforced when choosing B, and so on. Then, individuals are tested on the unanticipated, nonadjacent pair BD. Choosing B over D is interpreted as a demonstration of TI, as B and D are indirectly linked through their relationship with the third item C (learned relationships: B\u0026thinsp;\u0026gt;\u0026thinsp;C and C\u0026thinsp;\u0026gt;\u0026thinsp;D). However, the use of the five-terms series task has been questioned in the literature (Allen \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Since in a five-terms series, B is directly linked to A, the most positive item in terms of probability of reinforcement, and D is directly linked to E, the most negative item, this configuration impedes discussions about the cognitive process at stake both during the relational learning and during TI, that is, either a relational representation of the ordinal series, or a more associative-based representation.\u003c/p\u003e \u003cp\u003eIn the present study, we adapted a six-term series task to adult hens (Gallus gallus domesticus), with a twofold objective. The first one was to use a larger sequence of items (three non-adjacent and non-endpoint pairs: BD, BE and CE) allowing for the drawing of more solid conclusions about TI abilities, making it more difficult for animals to respond to items based solely on associative processes. This approach has already been successfully used in other avian species, such as pigeons and corvids (von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Bond et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Daniels et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The second one was to broaden our knowledge of the cognitive abilities of domesticated birds compared to non-domesticated species, since the former are generally expected to be less capable due to artificial selection. (Ferreira et al \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHens were trained following a procedure adapted from the hybrid training procedure in pigeons, as proposed by Daniels et al. (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). We ran a six-terms series of arbitrary items of different shape and color. Pairs were trained successively (AB, BC, CD, DE and EF), and mixed sessions, containing each previously trained pair, were included at each step. Such a hybrid training procedure, by gradually increasing complexity, allows to maximize and equalize the performance for every pair trained and to avoid different delays between training each pair and test sessions (Vasconcelos \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Daniels et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The configuration of the task allowed us to test three different inference tests: BD, BE and CE. As already mentioned, instead of relying solely on a single pairwise comparison, these three tests allow for complementary analyses and more robust results (Vasconcelos \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Bond et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWe analyzed the learning performance, the TI performance and the choice latency for each adjacent pair (premise pairs) and for each nonadjacent pair (control and inference trials, respectively), at the group level, and at the individual level to measure eventual inter-individual variations. We put an emphasis on the learning accuracy between the premise pairs as it might give some indication about the cognitive resolution of the TI task (Bond et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Finally, we tested for alternative cues that could have driven the choice behaviour of individuals through the configuration of the sessions.\u003c/p\u003e"},{"header":"METHOD","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eETHICS APPROVAL\u003c/h2\u003e \u003cp\u003e This experimental procedure was approved by the Val de Loire Ethics Committee (approval n\u0026deg; CE19\u0026ndash;2021-0211-1, CEEA VdL, France). Animal care and experimental treatments complied with the French and European guidelines for housing and care of animals used for scientific purposes (European Union Directives 2010/63/AU).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eSUBJECTS AND HOUSING\u003c/h2\u003e \u003cp\u003eSix adult laying hens (Isa Brown strain), aged 2 to 3 years old, were included in the procedure. The hens were maintained at the P\u0026ocirc;le d\u0026rsquo;Exp\u0026eacute;rimentation Avicole de Tours, where the experiment took place (UE PEAT, INRAE, 2018. Experimental Poultry Facility, doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.15454/1.5572326250887292E12\u003c/span\u003e\u003cspan address=\"10.15454/1.5572326250887292E12\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). They had access to a wood-chip littered barn (25m\u003csup\u003e2\u003c/sup\u003e) equipped with nesting boxes and perches, and had access to an outside enclosure (approx. 30m\u003csup\u003e2\u003c/sup\u003e) enriched with perches. Water was provided ad libitum, and food was delivered at will once the experiments of each testing day were completed. Birds were kept in a stable social group of 21 individuals on a 6am to 6pm daylight cycle. All experiments took place between 9am and 3pm and the testing order of the individuals was counterbalanced each day.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eSTIMULI\u003c/h2\u003e \u003cp\u003eEach individual was attributed a unique sequence of 6 items, composed of the same items in a different order. The items (4x4cm) were printed in the center of white cards (10x8cm). Each item had a different color (yellow, red, black, green, blue and pink) and a different shape (square, circle, cross, diamond, triangle, star) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The colors and the shapes were chosen in order to maximize the difference between the items. With the same purpose, the blank rate and the structural information rate were not controlled. Each individual had a unique sequence order of these items, thus, for clarity, we will name the items depending on their sequence position: A, B, C, D, E and F.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eAPPARATUS AND GENERAL OPERATIVE\u003c/span\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e1. \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003eApparatus\u003c/span\u003e\u003c/h2\u003e \u003cp\u003eThe apparatus consisted in a starting box and a separation between the box and the experimenter, both made in condensed wood. The starting box (69cm long x 55cm large x 80cm high) was closed with a wire-meshed door (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). A separation wall with a one-way mirror stood 20cm away from the starting box. This separation wall provided a 10cm high x 40cm large opening allowing the experimenter to present the cards to the hen without her being able to see the experimenter (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The pairs of cards were presented on a solid cardboard display (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2. \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003eGeneral\u003c/span\u003e procedure\u003c/h2\u003e \u003cp\u003eBefore training, each individual was habituated individually to the new testing environment and to be carried out from the social group and placed in the starting box in a separate, adjacent room. Then each individual was accustomed to peck on a neutral card (black figure printed on a white card with the same dimensions as the test cards) to get mealworms. This habituation was carried out using a clicker-training procedure (Feng et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAfter this procedure, the stimuli habituation and the training sessions started for each individual. During habituation, the experimenter randomly presented each item of the series, as a prior habituation, and each hen pecked directly on each item regardless of the shapes or the colors. Then, through the training sessions, the hens would learn to peck preferentially (i.e., to choose) on one item in the pair presented to get the reward (i.e., the prior item among the two in the series: for example, in ABC, peck at B when faced with B-C).\u003c/p\u003e \u003cp\u003eFor each trial, the pair was prepared out of the sight of the individual. Then, the wire-meshed door was opened and the experimenter presented the pair on the display. A choice was considered when the hen pecked on one card. If her choice was incorrect, the display was immediately taken back and the hen was brought back in the box for a 10-second wait. Then, a maximum of 3 correction trials was given with the exact same trial. If her choice was correct, the experimenter clicked and waited for a second peck on the same card to give her the reward (3 mealworms, either for a successful standard trial or a correction trial). After each trial, the hen was brought back in the starting box and the inter-trial interval was of 15 to 20 seconds. If the hen did not respond within 2 minutes, the experimenter closed the door and waited 10 seconds before trying again. If the hen did not respond for a total of 3 minutes, she was considered unmotivated and the session was postponed.\u003c/p\u003e \u003cp\u003eSessions could last from 20 to 45 minutes depending on the individual performance (number of correction trials needed) and depending on the number of trials in the session. Each individual had one session per day. For each training or testing session, the prior item for each pair was equally presented right and left among the trials (e.g., in the pair CD, the prior item is C), and a prior item was presented no more than twice in a row on the same side. A learning criterion was required at each training step: both in single-pair training and in mixed sessions (detailed below), the hen had to reach a minimum of 75% of success in 2 consecutive sessions with more than 50% of success for each side (right or left) of presentation of the prior item for each pair. However, if an individual reached 90% successful trials or more for both sides within one session, then the learning criterion was considered to be reached. The learning criteria stood to avoid any potential side bias, i.e., systematically pecking on the left or on the right card, independently of the card. In case of a side bias, a maximum of 8 correction sessions were run, that were adapted to the bias and the strength of the bias on a case-by-case basis.\u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ePROCEDURE\u003c/span\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e1. \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003eTraining sessions\u003c/span\u003e\u003c/h2\u003e \u003cp\u003eFirst, during the single-pair training of the premise pairs, each pair of adjacent items was learned in the sequence order (e.g., AB, then BC, etc.). These training sessions comprised 20 trials, with an equal amount of left and right presentations of the prior item. We measured the number of sessions individuals needed to reach the learning criterion for each pair.\u003c/p\u003e \u003cp\u003eThen, to ensure that hens were able to remember the correct choice for the items that were involved in two different situations (e.g., C is the prior item the hen had to peck on in CD but not in BC), mixed sessions, where all pairs previously learned were retrained, were added to our training schedule. These mixed sessions were included after the training of BC, CD, DE and EF, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). This hybrid training procedure (successive pair training intermixed in mixed sessions) has been shown to lead to a faster significant performance for the premise pairs (Daniels et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). For a consistency of the number of trials per pair per session, and considering the conditions of prior items presentations, we allocated a total number of trials per session accordingly. Mixed sessions after BC training sessions had a total of 20 trials; those after CD had 18 trials; those after DE had 32 trials; those after EF had 40 trials. The number of presentations of each pair was equal in each session. We controlled afterwards that the number of presentations of each pair did not influence the performance in inference trials.\u003c/p\u003e \u003cp\u003eWe measured the number of sessions the individuals needed to reach the learning criterion for each mixed step. For the last three mixed sessions (which included the 5 premise pairs), a baseline performance (percentage of success) and the response latency was measured for each premise pair.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2. \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003eTest sessions\u003c/span\u003e\u003c/h2\u003e \u003cp\u003eEach individual had 12 test sessions with 46 trials each: 40 trials of adjacent pairs (8 trials per pair), 3 control trials and 3 inference trials. Control and inference trials were non-rewarded and with no correction trials.\u003c/p\u003e \u003cp\u003eExtremity items are not used in inference trials as they are either always reinforced (A) or never reinforced (F). Thus, nonadjacent pairs involving an extremity item (e.g., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eA\u003c/span\u003eC or C\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eF\u003c/span\u003e) do not test for TI but allow for the basic understanding of associative values according to the reinforcement of each item during training (end-anchor effet, von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e). Control trials consisted in nonadjacent pairs with either the very first item A or the very last item F (AC, AD, AE, BF, CF or DF). These A and F control trials were pseudo-randomly distributed and equally presented among the 12 test sessions.\u003c/p\u003e \u003cp\u003eInference trials were trials of nonadjacent items (but not extremity items) for which transitive inference was needed to give the correct response. The pairs involved items that were both equally reinforced (for example, in BD: B\u0026thinsp;+\u0026thinsp;C- and D\u0026thinsp;+\u0026thinsp;E-) and non-reinforced (for example, in BD: A\u0026thinsp;+\u0026thinsp;B- and C\u0026thinsp;+\u0026thinsp;D-) during the training. The six-term series enabled to test TI through three inference trials (BD, BE and CE). Each one of the possible inference trials was presented once per test session. The sides of presentation of the prior item for each possible inference trial, were pseudo-randomized and equalized within two sessions.\u003c/p\u003e \u003cp\u003eDuring the test sessions, the performance for the premise pairs was tracked. In case of a side- or an item-bias, we integrated optional correction sessions with, in main proportion, the difficult pair, and the two adjacent pairs, to ensure an equalized level of performance for every premise pair throughout test sessions. A maximum of two correction sessions was added each two test sessions.\u003c/p\u003e \u003cp\u003eWe measured the number of successful trials for the premise pair trials, for control trials, and for inference trials. We measured the response latency for control trials and for inference trials.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eSTATISTICAL ANALYSIS\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed using R version 4.2.1 (R Core Team \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). We considered p-values below 0.05 to be significant, and p-values below 0.1 as a statistical tendency, for all statistical analyses. Chance level was considered at 50% of success. A non-parametric approach was preferred due to the small sample size (6 to 4 individuals).\u003c/p\u003e \u003cp\u003eThe response variables were either the number of sessions to criteria, or the percentage of success, or the response latency. The fixed variables were the side for the presentation of the prior item, the trial type, the pair and the number of the session. Anova permutational models were performed when several explanatory variables were to be considered, with individuals as random effect (aovperm function, package \u003cem\u003epermuco\u003c/em\u003e, Frossard and Renaud \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; np\u0026thinsp;=\u0026thinsp;10000, type\u0026thinsp;=\u0026thinsp;permutation). Symmetry tests or one-way permutation tests were run to analyze the effect of a fixed variable with more than 2 modalities on the response, when the data were paired or not, respectively (package \u003cem\u003ecoin\u003c/em\u003e, Hothorn et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; two-tailed comparisons, np\u0026thinsp;=\u0026thinsp;10000). Pairwise permutation t-tests were run for Posthoc analyses to detail the comparison between more than two modalities (pairwise.perm.t.test function, package \u003cem\u003eRVAideMemoire\u003c/em\u003e, Herv\u0026eacute; \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; np\u0026thinsp;=\u0026thinsp;10000, Holm correction). Two-tailed Wilcoxon rank-sum tests were used to assess the statistical significance of a comparison against the chance level, with Holm correction when appropriate (manual correction). Two-tailed exact binomial tests were used to test the statistical significance of individual performances (binom.test function). Homogeneity of variances was assessed with Levene tests before model fitting and before running multiple comparison analyses. Graphics were performed with the package ggplot2 (Wickham et al. \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor statistical analyses including less than 10 responses, the results are reported as the median and the first and third quartiles (MED [Q1:Q3]). For other statistical analyses, the results are reported as the mean and the standard deviation (MEAN +/- SD).\u003c/p\u003e \u003cp\u003eAs training and testing parameters can cause the use of associative cues in test trials, we ran supplementary analyses afterwards to control for some eventual response biases because of the sessions\u0026rsquo; configuration (Guez and Audley \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). We controlled for an effect of the reinforcement history and the reinforcement ratio of the items during training (Daisley et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2010\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Hotta et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Okouchi and Lattal \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Lazareva et al. \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) and of the configuration of premise pair trials previous to the test trials in test sessions (Russell et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e1996\u003c/span\u003e). These analyses are detailed in \u003cem\u003eSupplementary Materials\u003c/em\u003e (\u003cb\u003eSupplementary Material S1\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eTRAINING SESSIONS\u003c/h2\u003e\n \u003cp\u003eThe details of the number of sessions to reach the learning criterion at each step for each individual are presented in \u003cem\u003eSupplementary Materials S1\u003c/em\u003e, \u003cstrong\u003eTable S3\u003c/strong\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e1. Single-pair training of the premise pairs (i.e., pairs of successive items\u003c/h2\u003e\n \u003cp\u003eAt the group level, the mean number of sessions needed to reach the learning criteria did not differ between the different pairs (variance homogeneity, one-way permutation test, chi2\u0026thinsp;=\u0026thinsp;4.5469, df\u0026thinsp;=\u0026thinsp;4, p\u0026thinsp;=\u0026thinsp;0.337; global median number of sessions\u0026thinsp;=\u0026thinsp;3 [3:4] sessions to reach the criterion).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e2. Mixed sessions\u003c/h2\u003e\n \u003cp\u003eAt the group level, the mean number of sessions needed to reach the learning criterion was significantly different depending on the number of pairs included in the session (no variance homogeneity, one-way permutation test, chi2\u0026thinsp;=\u0026thinsp;12.203, df\u0026thinsp;=\u0026thinsp;3, p\u0026thinsp;=\u0026thinsp;0.007). More precisely, the learning speed was higher in the mixed sessions including pairs [AB to BC] compared to other stages of mixed sessions (significant with pairs [AB to CD]: p\u0026thinsp;=\u0026thinsp;0.046; statistical tendency with [AB to DE]: p\u0026thinsp;=\u0026thinsp;0.082 and with [AB to EF]: p\u0026thinsp;=\u0026thinsp;0.074; pairwise permutation t-test).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e3. Baseline performance for the premise pairs\u003c/h2\u003e\n \u003cp\u003eFour out of six individuals successfully finished the training by reaching the learning criteria (\u003cem\u003eSupplementary Materials\u003c/em\u003e, \u003cstrong\u003eTable S2\u003c/strong\u003e). Two individuals have not progressed to the required steps and were not included in the following analyses. In the last 3 mixed sessions, the mean performance for the premise pairs was of 80.42 +/- 17.88% of success, which is higher than chance level (two-tailed Wilcoxon rank-sum test, V\u0026thinsp;=\u0026thinsp;1526.5, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; \u003cem\u003eSupplementary Materials\u003c/em\u003e, \u003cstrong\u003eTable S3\u003c/strong\u003e). We used a permutational anova model analysis to estimate the effect of the pair presented and of the side of the prior item, on the percentage of success. We found an effect of the side (F\u0026thinsp;=\u0026thinsp;8.69, p\u0026thinsp;=\u0026thinsp;0.011), with a global side bias for left over right (mean performance when prior item on the left\u0026thinsp;=\u0026thinsp;85.69 +/- 19.29%; on the right\u0026thinsp;=\u0026thinsp;74.81 +/- 25.59%) that was not different between the pairs (F\u0026thinsp;=\u0026thinsp;1.464, p\u0026thinsp;=\u0026thinsp;0.260). Moreover, we found an effect of the pair presented on the performance (F\u0026thinsp;=\u0026thinsp;3.343, p\u0026thinsp;=\u0026thinsp;0.038; Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). A multiple pairwise analysis revealed significant performance differences between pairs AB and DE (p\u0026thinsp;=\u0026thinsp;0.035), CD and EF (p\u0026thinsp;=\u0026thinsp;0.007) and DE and EF (p\u0026thinsp;=\u0026thinsp;0.002) and a statistical tendency for the difference between pairs BC and EF (p\u0026thinsp;=\u0026thinsp;0.053; pairwise permutation t-test). The performance for the premise pairs at the end of training for each individual is detailed in Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\u003cspan\u003e\n \u003cp\u003e\u003cstrong\u003e4. Response latency for the premise pairs in the 3 last mixed sessions.\u003c/strong\u003e\u003c/p\u003e\n \u003c/span\u003e\n \u003cp\u003eThe mean response latency for the premise pairs was 3.097 +/- 3.335 seconds. We found no difference of response latency between the premise pairs (one-way permutation test, variance homogeneity; chi2\u0026thinsp;=\u0026thinsp;4.9606, df\u0026thinsp;=\u0026thinsp;4, p\u0026thinsp;=\u0026thinsp;0.291). At the individual level, there was a significant difference in the response latency between the pairs for two individuals with a slower response latency for DE and EF (for Dion: between AB and DE: p\u0026thinsp;=\u0026thinsp;0.020; between BC and DE: p\u0026thinsp;=\u0026thinsp;0.014; between AB and EF: p\u0026thinsp;=\u0026thinsp;0.048; between BC and EF: p\u0026thinsp;=\u0026thinsp;0.048; for Daenerys: between AB and EF, p\u0026thinsp;=\u0026thinsp;0.018; pairwise permutation t-test). The response latency for each premise pairs at the end of the training for each individual is detailed in Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eTEST SESSIONS\u003c/h2\u003e\n \u003cp\u003eFour individuals successfully finished the training and were tested for transitive inference. The test sessions had 40 premise trials, 3 inference trials (BD, BE and CE) and 3 control trials per session. Each individual had 12 test sessions. The performance in control trials validated the basic understanding of the associative rule of the (non-)reinforced extremity items, as every hen performed better than chance level (total of 36 control trials; group median performance\u0026thinsp;=\u0026thinsp;98.61 [95.14:100] %; binomial tests, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; performance for A- trials\u0026thinsp;=\u0026thinsp;97.22 [94.44:100] %; for F- trials\u0026thinsp;=\u0026thinsp;100 [95.83:100] %).\u003c/p\u003e\n \u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003eTI performance.\u003c/span\u003e All hens performed better than the chance level (total in 36 inference trials; 80.56% for Octo, 83.33% for Starr, 94.44% for Dion and 72.22% for Daenerys; binomial tests, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The median performance at the group level was 81.94%. Starr, Dion and Daenerys were successful in their very first inference trial. Within the 12 first inference trials, Dion and Daenerys performed significantly better than the chance level (12/12 and 10/12, respectively; binomial tests, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; 7/12 for Octo and Starr). We found no side bias for test trials (anova permutation model analysis; F\u0026thinsp;=\u0026thinsp;1.19, p\u0026thinsp;=\u0026thinsp;0.314) either for control or inference trials (no interaction; F\u0026thinsp;=\u0026thinsp;0.629, p\u0026thinsp;=\u0026thinsp;0.432).\u003c/p\u003e\n \u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003eComparison of the different inference trials.\u003c/span\u003e At the group level, we found a statistical tendency for the difference in the performance between the 3 trial types at the group level with a higher performance for BE trials compared to BD and CE trials (variance homogeneity, symmetry test, maxT\u0026thinsp;=\u0026thinsp;2.250, p\u0026thinsp;=\u0026thinsp;0.063; median performance for BE\u0026thinsp;=\u0026thinsp;100 [97.9:100] %; for BD\u0026thinsp;=\u0026thinsp;83.30 [70.83:91.60] %; for CE\u0026thinsp;=\u0026thinsp;66 [64.08:70.47] %). At the individual level, each hen performed better than 50% of success for each trial type (BD, BE and CE). All individuals performed significantly better than chance level when tested with BE, two individuals performed significantly better than chance level when tested with BD, and one individual when tested with CE (statistical significance for more than 10 successful trials among 12 for each inference trial possibility; two-tailed binomial test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e). The performance in inference trials was not explained by the differential ratio reinforcement of the items during training, except for Daenerys (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003eResponse latency.\u003c/span\u003e We found a significant difference in the response latency between premise pairs, control trials and inference trials (One-way permutation test, chi2\u0026thinsp;=\u0026thinsp;7.553, df\u0026thinsp;=\u0026thinsp;2, p\u0026thinsp;=\u0026thinsp;0.023) with a significant difference between premise pairs and control trials (pairwise permutation t-test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; mean response latency for premises pairs trials\u0026thinsp;=\u0026thinsp;3.0965 +/-3.33 seconds, for control trials\u0026thinsp;=\u0026thinsp;2.043 +/-1.106 seconds, for inference trials\u0026thinsp;=\u0026thinsp;2.519 +/- 2.189 seconds). We found no significant difference in the response latency between the different pairs presented in control and in inference trials (homogeneity of variances, One-way permutation test, chi2\u0026thinsp;=\u0026thinsp;7.89, df\u0026thinsp;=\u0026thinsp;8, p\u0026thinsp;=\u0026thinsp;0.444). In inference trials, the mean response latency for BD trials was of 2.271 +/- 1.201 seconds; for BE trials\u0026thinsp;=\u0026thinsp;2.308 +/- 1.207 seconds; and for CE trials\u0026thinsp;=\u0026thinsp;2.933 +/- 3.438 seconds.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eCONTROL FOR CONFOUNDING FACTORS ON CHOICE BEHAVIOR AND TI PERFORMANCE\u003c/h2\u003e\n \u003cp\u003eWe analyzed whether the hens could have relied on other cues to be successful in test trials. These analyses are detailed in the \u003cem\u003eSupplementary Materials S2\u003c/em\u003e. The results show that the reinforcement ratios of the items and the configuration of the test sessions could not have allowed most hens to perform higher than the chance level in test trials, as (1) neither the reinforcement history nor the reinforcement ratio of the items resulting from the training stage influenced the performance in test trials, excepted for Daenerys, and (2) the configuration of the previous pair trial influenced the choice behavior in the following test trial.\u003c/p\u003e\n \u003cp\u003eInterestingly, the fact that test trials were not rewarded might have affected the performance of the hens in premise pairs during test sessions. A permutation anova analysis showed that the choice to peck at an item in a premise pair significantly increased when the individual had pecked on this item in the previous non-rewarded test trial (control or inference trial) but only if it was presented on the other side (anova permutation model analysis; interaction: F\u0026thinsp;=\u0026thinsp;7.108, p\u0026thinsp;=\u0026thinsp;0.008). In parallel, whether hens gave the correct (C\u003csub\u003eT\u003c/sub\u003e) or the incorrect (I\u003csub\u003eT\u003c/sub\u003e) response at test trials did not significantly impact their performance (correct C\u003csub\u003eH\u003c/sub\u003e or incorrect I\u003csub\u003eH\u003c/sub\u003e) at the next premise pair trial which presented one of the same items (two-tailed Wilcoxon tests with Holm correction; mean occurrence of I\u003csub\u003eT\u003c/sub\u003eI\u003csub\u003eH\u003c/sub\u003e=0, I\u003csub\u003eT\u003c/sub\u003eC\u003csub\u003eH\u003c/sub\u003e =10 +/- 4.97, C\u003csub\u003eT\u003c/sub\u003eI\u003csub\u003eH\u003c/sub\u003e =0, C\u003csub\u003eT\u003c/sub\u003eC\u003csub\u003eH\u003c/sub\u003e =42.5 +/- 9.85).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eOur results show that the hens were capable of transitive inference when confronted with the 6-term series task. This confirms what was observed in previous studies in poultry that used a set up with a five-item series (in chicks: Daisley et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2010\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; in geese: Weiß et al. \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). We found an inter-individual variability in the resolution of the task, and notably, the performance of one hen suggested a resolution of the task through a mental representation of the series. The TI performance of most hens was not impacted neither by the reinforcement ratios of the items, nor the configuration of the sessions, supporting the use of transitive inference in this task, and making it harder to support a purely associative-based resolution of the task (except for one individual, Daenerys). Overall, the present study expands our knowledge on how chickens learn and solve a relational task.\u003c/p\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eLEARNING PERFORMANCE\u003c/h2\u003e \u003cp\u003eOverall, the hens needed a mean of 51.25 +/- 13.43 sessions to complete the training stage. Hens were faster to learn the six-term series than greylag geese confronted with a 5-terms series with a similar hybrid training procedure (mean of 83.4 +/-17.1 sessions; Weiss et al. 2010). Two main reasons might explain this difference. Firstly, the learning criteria may have been more demanding in Weiss et al.’s study. Secondly, our hens were tested in a controlled environment without further distractions, which was not the case for the greylag geese.\u003c/p\u003e \u003cp\u003eAs in greylag geese (Weiß et al. \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) and in pigeons (Clement and Zentall \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), the hens learned the first two premise pairs at a similar speed (\u003cem\u003eSupplementary material S2\u003c/em\u003e, \u003cb\u003eTable S2\u003c/b\u003e). While these authors considered this performance to be part of a natural tendency of the animals to choose a familiar stimulus over a novel one, we observed a different pattern. We observed that the hens tended to apply the rule previously learned at the start of the training sessions for each premise pair (5 to 10 first trials), that is, they tended to first avoid the item they had not to peck at for the previous pair (e.g., do not peck at B in BC because that was the item to not peck at in AB). A difficulty to step to a new pair has already been found in other species (Treichler and Van Tilburg \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e1996\u003c/span\u003e in macaques; Bond et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2003\u003c/span\u003e in corvids). Thus, we hypothesize that, first, hens tended to apply a previous rule when facing new situations, which could be referenced as rule generalization, and then, that the learning speed of the new pair could rather be attributed to a high behavioral flexibility to learn the new rule (see Degrande et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe comparison of the performance of the premise pairs has been shown to bring information about the cognitive process that could be at stake in the n-term series task (Bond et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). When comparing the performance between the premise pairs at the end of the training, each hen showed a better performance for AB and EF compared to the other pairs. This performance can be associated with what is known as the end-anchor effect. In simple terms, this serial position effect (i.e., a performance that depends on the serial position of the items in the ordinal series) implies better performance when the pair includes an extremity item (here, A or F; Allen \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). This effect can be explained through the strong conditioning (positive and negative, respectively) that comes with the very first item of the series, that is always reinforced, and the very last, that is never reinforced.\u003c/p\u003e \u003cp\u003eConcerning the response latency, two out of four individuals showed a choice speed that was dependent on the serial position of the pairs, with a slower response for the pairs at a greater distance from A. Some authors argue that this effect suggest a mental representation of the ordinal series (Terrace \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2005\u003c/span\u003e): the item A being the reference point of the spatial representation of the series, the response latency for the premise pairs increases with the distance from A. This result was significant for Daenerys (AB \u0026lt; EF) and for Dion (AB \u0026lt; DE and EF, and BC \u0026lt; DE and EF). However, associative hypotheses can account for this effect. For example, the \u003cem\u003efirst-item effect\u003c/em\u003e hypothesis (von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) claims that the first item of the series (A) has a strong positive value and is thus associated with a shorter response latency, and that the response latency for the premise pairs involving the following items in the series will decrease gradually with the distance to A. Another hypothesis is that a higher number of presentations of the pairs during training might generate a shorter response latency for these pairs. This needs to be further investigated as the result is not significant at the group level in our study.\u003c/p\u003e \u003cp\u003eAt the individual level, two out of our six individuals did not manage to end the training stage, by not being able to mix the first two premise pairs (Elizabeth) or not being able to mix the five premise pairs together (Savana). The fact that some individuals never reach the learning criteria has been found in other species (e.g., in pigeons: von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e). The n-term task, commonly used for TI testing, involves some individual parameters that are prerequisites to the task and that can strongly impact the performance, independent of the ability to perform TI. Independently of inferential and relational memory abilities, this task requires a long-term memory, a retrieval capacity, a certain level of behavioral flexibility, and the ability to apply a different rule for the same item depending on the context of presentation (i.e., depending on the pair). Personality and social structure issues have been shown to impact these parameters in birds (chicks: Daisley et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; geese: Weiss et al. 2010). This observation calls for precautions when testing for TI in animals with the common n-term series task.\u003c/p\u003e \u003cp\u003eFinally, we found a significant side bias for left over right in the last mixed sessions of the relational training. In the same way, Daisley et al. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) found that chicks that could use their left eye only had a better TI performance in comparison to chicks that could use their right eye only. As the bird’s visual pathway have been shown to be contralateral (Deng and Rogers \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e1998\u003c/span\u003e), the authors conclude that the right hemisphere may be more implicated in TI, which is consistent with a development of TI capacities through relational representations in mammals (see for example Dusek and Eichenbaum \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). However, we did not find a significant side bias for the TI performance, which could mean that the cognitive process implied in relational learning is different than the one implied in relational retrieval in birds.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eTRANSITIVE INFERENCE PERFORMANCE\u003c/h2\u003e \u003cp\u003eThe hens showed transitive inference in the 6-term series task. Each hen performed significantly better than chance in 36 inference trials (non-rewarded test trials), and two hens did it in 12 trials. Notably, one individual (Dion) performed significantly better than chance level for each of the three different trials (BD, BE and CE). Further analyses showed that the hens could not have used some alternative strategy other than transitive inference to perform better than chance in inference trials, as we controlled for a strategy based on the reinforcement ratio of the items and for a performance based on the configuration of the sessions (see \u003cem\u003eSupplementary Materials S2\u003c/em\u003e). TI performance was not dependent on the reinforcement ratio of the items in three out of the four hens, which is an argument against a performance relying on associative responses.\u003c/p\u003e \u003cp\u003eThe six-term series enabled us to study further the cognitive resolution of the task through the three inference trials. Hens performed better in inference trials with more distant items, i.e., in the nonadjacent pair BE (performance significantly better than chance for the four hens), compared to pairs BD and CE (sign. for two hens and for one hen, respectively). Such a performance, that is related to the distance between two items, is referred to as the symbolic distance effect (Moyers et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). It has been found in other species as pigeons (von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Daniels et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), corvids (Bond et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) or even in humans (for example: Bryant and Trabasso \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1971\u003c/span\u003e) and is consistent with linear models of TI (von Fersen et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Daniels et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). However, the premise pairs were trained in a temporal order that was consistent with their ordinal rank, thus, we cannot state from our results whether the hens mentally organized the items of the series along a linear representation (Couvillon and Bitterman \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e1992\u003c/span\u003e; MacLean et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). This symbolic distance effect is also explainable by the Value Transfer Theory (VTT) from von Fersen et al. (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) or derived associative-based models (see for example Zentall et al. \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e1996\u003c/span\u003e; discussed in: Allen \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Vasconcelos \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Guez and Audley \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). In these models, each item is being transferred a partial associative value from its adjacent items in the series, depending on the reward contingency that occurs in the pairs presented. This way, A transfers its positive value along the adjacent items, and F transfers its non-rewarding, negative value along its adjacent items. This transfer mechanism could thus make it easier to respond to BE through associative values than BD or CE.\u003c/p\u003e \u003cp\u003eAccording to the symbolic distance between the items, the performance for the nonadjacent pairs BD and CE should have been similar. However, we found that the performance for BD was better than that for CE for two out of four individuals (not significant). The decreasing TI performance from BE to BD, and from BD to CE has also been found in pigeons in Daniels et al. (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) at the group level. Overall, it is possible that the worse performance in CE is related to a recency effect for the learning of the pair EF (with E rewarded), which hypothesis must be validated through further studies (Bond et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFocusing on the performance of each individual separately, we show that the hens might have used different cognitive strategies to solve the TI task. At one extreme, Dion probably solved the task through a relational representation of the series: she showed a high performance for the premise pairs and showed no first- or last-item effect, and an almost perfect performance for the three different inference trials. Her performance highlights the relevance of studying the response latency to observe potential response differences between the pairs at a high level of performance. At the other extreme, Daenerys used a more associative strategy, as her performance could be predicted by the values of the differential ratio of reinforcement at the end of her training. Her learning performances and her inference performances match an associative model including a transfer of associative values through the items. More broadly, the relational learning involved in this task might be multiply encoded. The TI task, as it is designed in most experiments, might simultaneously involve both cognitive- (through a relational representation) and associative- (through associative-based models) strategies (Jacobs \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Lazareva and Wasserman \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Bond et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e "},{"header":"CONCLUSION AND PERSPECTIVES","content":"\u003cp\u003eTo our knowledge, this is the first attempt to further study the underlying process of transitive inference in the domestic chicken, and more generally in \u003cem\u003eGalloanserae\u003c/em\u003e. Our results confirm that adult chickens are capable of transitive inference, and support that their transitive response cannot be explained just in terms of an associative-based model. Moreover, we found that some hens might be able to solve the transitive task through a mental representation of the series. Since our results resemble those found in \u003cem\u003eNeoaves\u003c/em\u003e (pigeons and corvids), our hypothesis is that TI is based on the same cognitive processes in both phylogenetic groups, and that the cognitive strategy to solve the transitive task might be driven mainly by individual parameters within species. Our study brings additional information about the cognitive capacities in chickens, which might be more complex than often assumed.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eETHICS APPROVAL\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis experimental procedure was approved by the Val de Loire Ethics Committee (approval n\u0026deg; CE19 \u0026ndash; 2021-0211-1, CEEA VdL, France). Animal care and experimental treatments complied with the French and European guidelines for housing and care of animals used for scientific purposes (European Union Directives 2010/63/AU).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSUPPLEMENTARY MATERIALS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary Material is available online in the INRAE data repository at https://doi.org/10.57745/TOXZ8L\u003cem\u003e.\u003c/em\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDATA AND MODEL AVAILABILITY STATEMENT\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the study findings is hosted in the INRAE data repository at https://doi.org/10.57745/TOXZ8L and is available on reasonable request from the corresponding author at [email protected]. \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHORS CONTRIBUTION\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRachel Degrande, Fabien Cornilleau, Ludovic Calandreau, L\u0026eacute;a Lansade and Oc\u0026eacute;ane Amichaud contributed to the study design. The habituation and the first training stages of the hens with clicker training were performed by Oc\u0026eacute;ane Amichaud. Material preparation, data collection and analysis were performed by Rachel Degrande. Plotine Jardat helped with data analysis. Benoit Piegu guided the analyses presented in the \u003cem\u003eSupplementary Materials\u003c/em\u003e. The first draft of the manuscript was written by Rachel Degrande and all authors commented on previous versions of the manuscript. Ludovic Calandreau supervised the research. All authors read and approved the final manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDECLARATION OF INTEREST\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no conflict of interest to disclose.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eACKNOWLEDGEMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge the experimental unit P\u0026ocirc;le d\u0026rsquo;Exp\u0026eacute;rimentation Avicole de Tours (UE PEAT, INRAE, 2018. Experimental Poultry Facility, DOI: 10.15454/1.5572326250887292E12) where hens were maintained and where the experiments took place. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFINANCIAL SUPPORT STATEMENT\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received the funding support of the PHASE research Department (INRAE) and a PhD fellowship INRAE-R\u0026eacute;gion Centre Val de Loire (France).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAbeyesinghe SM, McLeman MA, Owen RC, et al. (2009) Investigating social discrimination of group members by laying hens. Behav Processes 81:1\u0026ndash;13. https://doi.org/10.1016/j.beproc.2008.11.017\u003c/li\u003e\n \u003cli\u003eAllen C (2006) Transitive inference in animals: Reasoning or conditioned associations? In: Hurley S, Nudds M (eds) Rational Animals? Oxford University Press, pp 175\u0026ndash;186\u003c/li\u003e\n \u003cli\u003eAust U, Range F, Steurer M, Huber L (2008) Inferential reasoning by exclusion in pigeons, dogs, and humans. Anim Cogn 11:587\u0026ndash;597. https://doi.org/10.1007/s10071-008-0149-0\u003c/li\u003e\n \u003cli\u003eAvargu\u0026egrave;s-Weber A, Portelli G, Benard J, et al. (2010) Configural processing enables discrimination and categorization of face-like stimuli in honeybees. J Exp Biol 213:593\u0026ndash;601. https://doi.org/10.1242/jeb.039263\u003c/li\u003e\n \u003cli\u003eBond AB, Kamil AC, Balda RP (2003) Social complexity and transitive inference in corvids. Anim Behav 65:479\u0026ndash;487. https://doi.org/10.1006/anbe.2003.2101\u003c/li\u003e\n \u003cli\u003eBond AB, Wei CA, Kamil AC (2010) Cognitive representation in transitive inference: A comparison of four corvid species. Behav Processes 85:283\u0026ndash;292. https://doi.org/10.1016/j.beproc.2010.08.003\u003c/li\u003e\n \u003cli\u003eBryant PE, Trabasso T (1971) Transitive inferences and memory in young children. Nature 232:456\u0026ndash;458. https://doi.org/10.1038/232456a0\u003c/li\u003e\n \u003cli\u003eCamarena HO, Garc\u0026iacute;a-Leal O, Burgos JE, et al. (2018) Transitive Inference Remains Despite Overtraining on Premise Pair C+D-. Front Psychol 9:1791. https://doi.org/10.3389/fpsyg.2018.01791\u003c/li\u003e\n \u003cli\u003eClement TS, Zentall TR (2003) Choice based on exclusion in pigeons. Psychon B Rev 10:959\u0026ndash;964. https://doi.org/10.3758/BF03196558\u003c/li\u003e\n \u003cli\u003eCouvillon PA, Bitterman ME (1992) A conventional conditioning analysis of \u0026ldquo;transitive inference\u0026rdquo; in pigeons. Journal of Experimental Psychology: Anim Behav Processes 18:308\u0026ndash;310. https://doi.org/10.1037/0097-7403.18.3.308\u003c/li\u003e\n \u003cli\u003eCraig JV (1978) Aggressive Behavior of Chickens: Some Effects of Social and Physical Environments. Technical Report, KansasnState University. Presented at the 27th Annual National Breeder\u0026rsquo;s Roundtable, Kansas City, May 11, 1978.\u003c/li\u003e\n \u003cli\u003eCroney CC, Newberry RC (2007) Group size and cognitive processes. Appl Anim Behav Sci 103:215\u0026ndash;228. https://doi.org/10.1016/j.applanim.2006.05.023\u003c/li\u003e\n \u003cli\u003eDaisley JN, Vallortigara G, Regolin L (2010) Logic in an asymmetrical (social) brain: Transitive inference in the young domestic chick. Soc Neurosci 5:309\u0026ndash;319. https://doi.org/10.1080/17470910903529795\u003c/li\u003e\n \u003cli\u003eDaisley JN, Vallortigara G, Regolin L (2021) Low-rank Gallus gallus domesticus chicks are better at transitive inference reasoning. Commun Biol 4:1344. https://doi.org/10.1038/s42003-021-02855-y\u003c/li\u003e\n \u003cli\u003eDaniels CW, Laude JR, Zentall TR (2014) Six-term transitive inference with pigeons: Successive-pair training followed by mixed-pair training: Six-term transitive inference with pigeons. J Exp Anal Behav 101:26\u0026ndash;37. https://doi.org/10.1002/jeab.65\u003c/li\u003e\n \u003cli\u003eDegrande R, Cornilleau F, Jardat P, Ferreira VHB, Lansade L, Calandreau L (2024) A cognitive approach to better understand foraging strategies of the adult domestic hen. Preprint access at https://doi.org/10.21203/rs.3.rs-4121447/v1\u003c/li\u003e\n \u003cli\u003eDegrande R, Cornilleau F, Lansade L, Jardat P, Colson V, Calandreau L (2022) Domestic hens succeed at serial reversal learning and perceptual concept generalisation using a new automated touchscreen device. Animal 16:16(8), 100607. https://doi.org/10.1016/j.animal.2022.100607\u003c/li\u003e\n \u003cli\u003eDeng C, Rogers LJ (1998) Bilaterally projecting neurons in the two visual pathways of chicks. Brain Res 794:281\u0026ndash;290. https://doi.org/10.1016/S0006-8993(98)00237-6\u003c/li\u003e\n \u003cli\u003eDusek JA, Eichenbaum H (1997) The hippocampus and memory for orderly stimulus relations. Proc Natl Acad Sci USA 94:7109\u0026ndash;7114. https://doi.org/10.1073/pnas.94.13.7109\u003c/li\u003e\n \u003cli\u003eFeng LC, Howell TJ, Bennett PC (2016) How clicker training works: Comparing Reinforcing, Marking, and Bridging Hypotheses. Appl Anim Behav Sci 181:34\u0026ndash;40. https://doi.org/10.1016/j.applanim.2016.05.012\u003c/li\u003e\n \u003cli\u003eFerreira, V. H. B., Lansade, L., Calandreau, L., Cunha, F., \u0026amp; Jensen, P. (2023). Are domesticated animals dumber than their wild relatives? A comprehensive review on the domestication effects on animal cognitive performance. Neurosci Biobehav Rev, 154, Article 105407. https://doi.org/10.1016/j.neubiorev.2023.105407\u003c/li\u003e\n \u003cli\u003eFrossard J, Renaud O (2021) \u0026ldquo;Permutation Tests for Regression, ANOVA, and Comparison of Signals: The permuco Package.\u0026rdquo; J Stat Softw 99:1\u0026ndash;32. https://doi.org/10.18637/jss.v099.i15\u003c/li\u003e\n \u003cli\u003eGalizio A, Doughty AH, Williams DC, Saunders KJ (2017) Understanding behavior under nonverbal transitive-inference procedures: Stimulus-control-topography analyses. Behav Processes 140:202\u0026ndash;215. https://doi.org/10.1016/j.beproc.2017.05.010\u003c/li\u003e\n \u003cli\u003eGreene AJ, Spellman BA, Levy WB, et al. (2001) Relational learning with and without awareness: Transitive inference using nonverbal stimuli in humans. Mem Cogn 29:893\u0026ndash;902. https://doi.org/10.3758/BF03196418\u003c/li\u003e\n \u003cli\u003eGuez D, Audley C (2013) Transitive or Not: A Critical Appraisal of Transitive Inference in Animals. Ethology 119:703\u0026ndash;726. https://doi.org/10.1111/eth.12124\u003c/li\u003e\n \u003cli\u003eHerv\u0026eacute; M (2022) RVAideMemoire: Testing and Plotting Procedures for Biostatistics. R package version 0.9-81-2. \u003cu\u003ehttps://www.R-project.org\u003c/u\u003e\u003c/li\u003e\n \u003cli\u003eHogue M-E, Beaugrand JP, Lagu\u0026euml; PC (1996) Coherent use of information by hens observing their former dominant defeating or being defeated by a stranger. Behav Processes 38:241\u0026ndash;252. https://doi.org/10.1016/S0376-6357(96)00035-6\u003c/li\u003e\n \u003cli\u003eHothorn T, Hornik K, van de Wiel M, Zeileis A (2006) \u0026ldquo;A Lego system for conditional inference.\u0026rdquo; Am Stat 60:257\u0026ndash;263. https://doi.org/10.1198/000313006X118430\u003c/li\u003e\n \u003cli\u003eHotta T, Ueno K, Hataji Y, et al (2020) Transitive inference in cleaner wrasses (Labroides dimidiatus). PLoS ONE 15:e0237817. https://doi.org/10.1371/journal.pone.0237817\u003c/li\u003e\n \u003cli\u003eJacobs L (2006) From Movement to Transitivity: The Role of Hippocampal Parallel Maps in Configural Learning. Rev Neurosci, 17(1-2) 99-109. \u003cu\u003ehttps://www.doi.org/\u003c/u\u003e10.1515/revneuro.2006.17.1-2.99\u003c/li\u003e\n \u003cli\u003eLazareva OF, Paxton Gazes R, Elkins Z, Hampton R (2020) Associative models fail to characterize transitive inference performance in rhesus monkeys (Macaca mulatta). Learn Behav 48:135\u0026ndash;148. https://doi.org/10.3758/s13420-020-00417-6\u003c/li\u003e\n \u003cli\u003eLazareva OF, Smirnova AA, Bagozkaja MS, et al (2004) Transitive responding in hooded crows requires linearly ordered stimuli. J Exp Anal Behav 82:1\u0026ndash;19. https://doi.org/10.1901/jeab.2004.82-1\u003c/li\u003e\n \u003cli\u003eLazareva OF, Wasserman EA (2006) Effect of stimulus orderability and reinforcement history on transitive responding in pigeons. Behav Processes 72:161\u0026ndash;172. https://doi.org/10.1016/j.beproc.2006.01.008\u003c/li\u003e\n \u003cli\u003eMacLean EL, Merritt DJ, Brannon EM (2008) Social complexity predicts transitive reasoning in prosimian primates. Anim Behav 76:479\u0026ndash;486. https://doi.org/10.1016/j.anbehav.2008.01.025\u003cu\u003e\u0026nbsp;\u003c/u\u003e\u003c/li\u003e\n \u003cli\u003eMcGonigle BO, Chalmers M (1977) Are monkeys logical? Nature, 267(5613), 694\u0026ndash;696. https://doi.org/10.1038/267694a0\u003cu\u003e\u0026nbsp;\u003c/u\u003e\u003c/li\u003e\n \u003cli\u003eMikolasch S, Kotrschal K, Schloegl C (2013) Transitive inference in jackdaws (Corvus monedula). Behav Processes 92:113\u0026ndash;117. https://doi.org/10.1016/j.beproc.2012.10.017\u003c/li\u003e\n \u003cli\u003eMoyers SC, Adelman JS, Farine DR, et al. (2018) Exploratory behavior is linked to stress physiology and social network centrality in free-living house finches (Haemorhous mexicanus). Horm Behav 102:105\u0026ndash;113. https://doi.org/10.1016/j.yhbeh.2018.05.005\u003c/li\u003e\n \u003cli\u003eOkouchi H, Lattal KA (2006) An analysis of reinforcement history effects. J Exp Anal Behav 86:31\u0026ndash;42. https://doi.org/10.1901/jeab.2006.75-05\u003c/li\u003e\n \u003cli\u003ePiaget J (1928) Judgment and reasoning in the child. Humana Mente 3(12):551-554\u003c/li\u003e\n \u003cli\u003eR Core Team (2022) R: A language and environment for statistical Computing. \u003cu\u003ehttps://www.R-project.org\u003c/u\u003e\u003c/li\u003e\n \u003cli\u003eRussell J, McCormack T, Robinson J, Lillis G (1996) Logical versus Associative Performance on Transitive Reasoning Tasks by Children: Implications for the Status of Animals Performance. Qu J Exp Psychol 49B:231\u0026ndash;244\u003c/li\u003e\n \u003cli\u003eTerrace HS (2005) The simultaneous chain: a new approach to serial learning. Trends Cogn Sci 9:202\u0026ndash;210. https://doi.org/10.1016/j.tics.2005.02.003\u003c/li\u003e\n \u003cli\u003eTreichler FR, Van Tilburg D (1996) Concurrent conditional discrimination tests of transitive inference by macaque monkeys: List linking. Journal of Experimental Psychology: Anim Behav Processes 22:105\u0026ndash;117. https://doi.org/10.1037/0097-7403.22.1.105\u003c/li\u003e\n \u003cli\u003eVasconcelos M (2008) Transitive inference in non-human animals: An empirical and theoretical analysis. Behav Processes 78:313\u0026ndash;334. https://doi.org/10.1016/j.beproc.2008.02.017\u003c/li\u003e\n \u003cli\u003evon Fersen L, Wynne CD, Delius JD, Staddon JE (1991) Transitive inference formation in pigeons. Journal of Experimental Psychology: Anim Behav Processes 17:334\u0026ndash;341. https://doi.org/10.1037/0097-7403.17.3.334\u003c/li\u003e\n \u003cli\u003eWei\u0026szlig; BM, Kehmeier S, Schloegl C (2010) Transitive inference in free-living greylag geese, Anser anser. Anim Behav 79:1277\u0026ndash;1283. https://doi.org/10.1016/j.anbehav.2010.02.029\u003c/li\u003e\n \u003cli\u003eWerchan DM, G\u0026oacute;mez RL (2013) Generalizing memories over time: Sleep and reinforcement facilitate transitive inference. Neurobiol Learn Mem 100:70\u0026ndash;76. https://doi.org/10.1016/j.nlm.2012.12.006\u003c/li\u003e\n \u003cli\u003eWickham H, Averick M, Bryan J, et al. (2019) Welcome to the Tidyverse. JOSS 4:1686. https://doi.org/10.21105/joss.01686\u003c/li\u003e\n \u003cli\u003eZentall TR, Peng D, Miles L (2019) Transitive inference in pigeons may result from differential tendencies to reject the test stimuli acquired during training. Anim Cogn 22:619\u0026ndash;624. https://doi.org/10.1007/s10071-019-01257-2\u003c/li\u003e\n \u003cli\u003eZentall TR, Sherburne LM, Roper KL, Kraemer PJ (1996) Value Transfer in a Simultaneous Discrimination Appears to Result From Within-Event Pavlovian Conditioning. J Exp Psychol: Anim Behav Processes, 22(1), 68-75. https://doi.org/10.1037/0097-7403.22.1.68\u003cu\u003e\u0026nbsp;\u003c/u\u003e\u003c/li\u003e\n \u003cli\u003eZuberb\u0026uuml;hler K, Byrne RW (2006) Social cognition. Curr Biol, 16:R786\u0026ndash;R790. \u003cu\u003ehttps://doi.org/\u003c/u\u003e10.1016/j.cub.2006.08.046\u003cu\u003e\u0026nbsp;\u003c/u\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"animal-cognition","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"anco","sideBox":"Learn more about [Animal Cognition](http://link.springer.com/journal/10071)","snPcode":"10071","submissionUrl":"https://submission.nature.com/new-submission/10071/3","title":"Animal Cognition","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"transitive inference, relational memory, cognition, chicken, Gallus gallus domesticus","lastPublishedDoi":"10.21203/rs.3.rs-4431359/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4431359/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTransitive inference (TI) is a disjunctive syllogism that allows an individual to indirectly infer a relationship between two components, by knowing their respective relationship to a third component (if A\u0026thinsp;\u0026gt;\u0026thinsp;B and B\u0026thinsp;\u0026gt;\u0026thinsp;C then A\u0026thinsp;\u0026gt;\u0026thinsp;C). The common procedure is the 5-terms series task, in which individuals are tested on indirect, unlearned relations. Few bird species have been tested for TI to date, which limits our knowledge of the phylogenetic spread of such reasoning ability. Here we tested TI in adult laying hens using a more solid methodology, the 6-terms series task, which has not been tested in poultry so far. Six hens were trained to learn direct relationships in a sequence of six arbitrary items (A\u0026thinsp;\u0026gt;\u0026thinsp;B\u0026thinsp;\u0026gt;\u0026thinsp;C\u0026thinsp;\u0026gt;\u0026thinsp;D\u0026thinsp;\u0026gt;\u0026thinsp;E\u0026thinsp;\u0026gt;\u0026thinsp;F) in a hybrid training procedure. Then, 12 testing sessions were run, comprising 3 non-rewarded inference trials each: BD, BE, and CE. All subjects showed TI within 12 inference trials and were capable of TI whatever the relative distance between the items in the series. We found that TI performance was not impacted by the reinforcement ratios of the items for most individuals; thus, making it harder to support a purely associative-based resolution of the task. We suggest that TI is based on the same cognitive processes in poultry (\u003cem\u003eGalloanserae\u003c/em\u003e) than in modern flying birds (\u003cem\u003eNeoaves\u003c/em\u003e), and that the cognitive strategy to solve the task might be driven mainly by individual parameters within species. These results contribute to a better understanding of transitive reasoning in birds.\u003c/p\u003e","manuscriptTitle":"Transitive reasoning in the adult domestic hen in a six-term series task","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-04 14:52:06","doi":"10.21203/rs.3.rs-4431359/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-08-06T11:06:43+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-07-12T21:31:38+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-06-21T10:50:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211336058749961368017967072569420943540","date":"2024-05-24T16:28:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"24820324426339235839326333472845265545","date":"2024-05-23T11:01:33+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-05-23T02:59:12+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-05-22T20:34:08+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-05-17T10:41:38+00:00","index":"","fulltext":""},{"type":"submitted","content":"Animal Cognition","date":"2024-05-16T13:27:17+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"animal-cognition","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"anco","sideBox":"Learn more about [Animal Cognition](http://link.springer.com/journal/10071)","snPcode":"10071","submissionUrl":"https://submission.nature.com/new-submission/10071/3","title":"Animal Cognition","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"7eb5f69f-eb6d-4497-94e4-6862c92803f8","owner":[],"postedDate":"June 4th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-11-25T16:01:06+00:00","versionOfRecord":{"articleIdentity":"rs-4431359","link":"https://doi.org/10.1007/s10071-024-01914-1","journal":{"identity":"animal-cognition","isVorOnly":false,"title":"Animal Cognition"},"publishedOn":"2024-11-19 15:57:19","publishedOnDateReadable":"November 19th, 2024"},"versionCreatedAt":"2024-06-04 14:52:06","video":"","vorDoi":"10.1007/s10071-024-01914-1","vorDoiUrl":"https://doi.org/10.1007/s10071-024-01914-1","workflowStages":[]},"version":"v1","identity":"rs-4431359","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4431359","identity":"rs-4431359","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}