Behavioural Processes 81 (2009) 1–13
Contents lists available at ScienceDirect
Behavioural Processes journal homepage: www.elsevier.com/locate/behavproc
Investigating social discrimination of group members by laying hens Siobhan M. Abeyesinghe ∗ , Morven A. McLeman, Rachael C. Owen, Claire E. McMahon, Christopher M. Wathes Centre for Animal Welfare, Department of Clinical Veterinary Sciences, Royal Veterinary College, Hawkshead Lane, North Mymms, Hatﬁeld AL9 7TA, United Kingdom
a r t i c l e
i n f o
Article history: Received 28 July 2008 Received in revised form 20 November 2008 Accepted 25 November 2008 Keywords: Social discrimination Hens Operant learning
a b s t r a c t Social relationships in domestic fowl are commonly assumed to rely on social recognition and its pre-requisite, discrimination of group-mates. If this is true, then the unnatural physical and social environments in which commercial laying hens are typically housed, when compared with those in which their progenitor species evolved, may compromise social function with consequent implications for welfare. Our aims were to determine whether adult hens can discriminate between unique pairs of familiar conspeciﬁcs, and to establish the most appropriate method for assessing this social discrimination. We investigated group-mate discrimination using two learning tasks in which there was bi-directional exchange of visual, auditory and olfactory information. Learning occurred in a Y-maze task (p < 0.003; n = 7/8) but not in an operant key-pecking task (p = 0.001; n = 1/10). A further experiment with the operanttrained hens examined whether failure was speciﬁc to the group-mate social discrimination or to the response task. Learning also failed to occur in this familiar/unfamiliar social discrimination task (p = 0.001; n = 1/10). Our ﬁndings demonstrate unequivocally that adult laying hens kept in small groups, under environmental conditions more consistent with those in which sensory capacities evolved, can discriminate group members: however, appropriate methods to demonstrate discrimination are crucial. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The progenitor of the domestic fowl (Gallus gallus) naturally forms small polygynous harems or bachelor groups of up to 30 birds, adapted to communication in a forest habitat (Collias and Collias, 1996; Mench and Keeling, 2001). The artiﬁcial social and physical environments in which poultry are housed commercially comprise dim light intensities, restricted space, elevated noise and poor atmospheres. These may interfere with the effectiveness of social signalling (e.g. Hughes et al., 1974; Algers and Jensen, 1985; D’Eath and Stone, 1999; Jones et al., 2001) and alter the ability of hens to recognise conspeciﬁc identity and/or intent (D’Eath and Keeling, 2003), thereby disrupting social function, with potential implications for welfare (Craig et al., 1969; Grigor et al., 1995; Freire et al., 1997; D’Eath and Stone, 1999). To understand the importance of the social environment to hens, it is necessary to gauge their cognitive capacity for establishing stable social relationships under environmental conditions which do not compromise the transmission and perception of social signals, thereby providing a baseline against which commercial environmental conditions can be compared. This requires a robust and repeatable method and the use of live stimuli.
∗ Corresponding author. E-mail address: [email protected]
(S.M. Abeyesinghe). 0376-6357/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.beproc.2008.11.017
Social relationships in fowl are commonly assumed to rely upon individual recognition (Guhl and Ortman, 1953; Bradshaw, 1991; Hauser and Huber-Eicher, 2004), although this has never been demonstrated explicitly and hens rely primarily upon visual displays, postures and vocalisations for communication (Wood-Gush, 1971). Use of status cues, such as body size and comb size and colour, to maintain social order without memory for individuals is also plausible (Wood-Gush, 1971; Maynard-Smith and Harper, 1988; Pagel and Dawkins, 1997; Estevez et al., 1997), although correlations between such features and rank do not always appear to be reliable (Guhl and Ortman, 1953; Bradshaw, 1992c; Cloutier and Newberry, 2000). For example, ﬂocks of dubbed hens still form peck orders, though this alternate social strategy may be constrained by group size, when individual recognition becomes impracticable due to memory limitation and hens therefore appear to be more socially tolerant than those in small ﬂocks (D’Eath and Keeling, 2003; Estevez et al., 2003). Many studies have investigated discrimination of familiars from strangers (e.g. Hughes, 1977; Bradshaw, 1992a; Grigor et al., 1995; Jones et al., 1996; D’Eath and Dawkins, 1996; D’Eath and Stone, 1999; Marin et al., 2001; Deng and Rogers, 2002; D’Eath and Keeling, 2003; Hauser and Huber-Eicher, 2004; Porter et al., 2006; Guzman and Marin, 2008), and hens appear to attend to the task behaviour of high ranking birds over subordinates, regardless of skill (Nicol and Pope, 1999), but the pre-requisite to true individual recognition based on idiosyncratic cues, as deﬁned by Zayan (1994), is
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
the discrimination between familiar conspeciﬁcs irrespective of relative rank and un-weighted by other social information. Concerns have been raised over the ﬁdelity of artiﬁcial cues (Bradshaw and Dawkins, 1993; Dawkins, 1996; D’Eath and Dawkins, 1996; Patterson-Kane et al., 1997) and only one study has reported such a discrimination of live stimuli. Bradshaw (1991) used a Y-maze learning task to demonstrate social discrimination in an extremely small number of laying hens viewing stimuli at 75 cm. However, Dawkins’ subsequent assertion (Dawkins, 1995, 1996) that social recognition in hens may only occur at distances less than 30 cm, a constraint imposed by the binocular visual ﬁeld thought to be used for social discrimination, raises the possibility that Bradshaw’s (1991) birds, although discriminating more than one pair, may have been forced to do so on the basis of cues which are not used socially but still are apparent at a distance, such as gross differences in size or plumage colouration. Thus his subjects may not have demonstrated true group-mate discrimination. In this study, the ability of two small groups of hens, kept under environmental conditions considered optimal for perception of social signals, to discriminate between individual group members in a Y-maze choice task (Experiment Y-FF; FF—familiar/familiar) and an operant pecking task (Experiment Op-FF), respectively, was investigated. The aims were to determine whether group-mate discrimination could occur at short distances, and to determine which task was most appropriate for studies of social cognitive capacity in laying hens. The Y-maze task was similar to that used previously with pigs (McLeman et al., 2005), but the operant pecking task offered the opportunity to minimise subject handling and labour and to increase the number of trials per training session in which subjects could learn. Experiments Y-FF and OpFF were run simultaneously. A third discrimination of familiar vs. unfamiliar hens (Experiment Op-FU: FU—familiar/unfamiliar)
was subsequently conducted with the operant-trained birds using the operant pecking task to examine further the appropriateness of that method using a different social category for discrimination. 2. Methods 2.1. Subjects and housing Two groups, of 15 and 16 experimentally naïve Hy-line Brown pullets, respectively, were obtained from a commercial producer at point-of-lay and housed in separate, naturally ventilated 3 m × 3 m pens bedded with wood-shavings. A bell-drinker, a tube feeder, perches, nest-boxes containing a small amount of straw and a dish of layer-grit were provided in each pen. Birds were fed ad libitum on commercial layer pellets and were exposed to natural daylight, with a mean (±S.D.) illuminance of 153 ± 49 lx. In addition, their photoperiod was gradually extended from that experienced during rearing (11 h) to 16 h over 11 weeks using a supplementary ﬂuorescent strip light. All guidelines and requirements set out in the Principles of Laboratory Animal Care (National Institutes of Health, U.S.A., Publication No. 86-23, revised 1985) and the UK Animals (Scientiﬁc Procedures) Act 1986 were followed. 2.2. Apparatus 2.2.1. Y-maze For the Y-maze choice task (experiment Y-FF), a Y-maze was constructed from three stainless steel rectangular chambers joined by a non-slip, ridged, stainless steel triangular, ﬂoor piece. Each chamber comprised an opaque PVC hinged door, a transparent acrylic roof to admit light and a non-slip ridged stainless steel ﬂoor, and
Fig. 1. Experiment Y-FF: plan view of the Y-maze apparatus.
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
rested on castors 0.15 m from the ﬂoor (Fig. 1). The roof of the central zone was transparent polypropylene. Inside each arm of the Y-maze was placed a small stimulus box (0.4 m × 0.4 m × 0.6 m) on castors that could hold a stimulus hen; allowing her free rotation of movement and adequate head height for standing and stretching during a test. The stimulus boxes comprised wooden side and back walls, a 0.01 m gauge steel mesh ﬂoor and roof, and a transparent UV-transmitting acrylic front panel (Altuglas UVX solarium and sun-bed grade acrylic). The wooden surfaces were painted with white gloss-paint to facilitate cleaning and uniform reﬂection of light. Circular apertures (0.1 m diameter) were cut into the centre of the front panels at hen head height (0.34 m from the mesh ﬂoor) to provide ‘social exchange ports’, ensuring close visual inspection of stimuli (considered important for social recognition in hens; Dawkins, 1995, 1996) and bi-directional exchange of odour cues. Each stimulus box also contained a central feeder at the front. A small plastic pot ﬁxed to the ﬂoor at the end of each arm of the maze, behind the stimulus box, was used for placement of feed rewards for the test subject. The drop door between the Y-maze start-box and the triangular central zone was opaque to prevent test subjects viewing stimulus placement. Doors to both arms were made of 0.04 m gauge steel mesh, preventing immediate access whilst allowing stimuli to be viewed during a pre-choice period. The experimenter operated all doors from a central position using pulleys. The apparatus was housed in a naturally ventilated barn, lit by ﬂuorescent strip lights and daylight through windows to achieve a mean illuminance of 137 ± 38 lx at the test bird’s head height in the central zone and 139 ± 10 lx at the stimulus birds’ head height within the stimulus boxes. A system of four overhead cameras linked via a quad-splitter, VCR and monitor was used to observe and record hen behaviour in each area of the maze.
An opaque pulley-operated drop door between the operant box and stimulus boxes was raised and lowered manually to control the timing of stimulus presentation. The pecking keys consisted of metal-framed, circular (4 cm diameter) Perspex disks which could be illuminated with white LEDs through a diffuser to produce a uniform light ﬁeld. Key lights were calibrated using a luminance meter to ensure equal luminance (737 ± 2.5 cd m−2 ). The keys were calibrated to ensure that equal energy was required to operate them (≥0.2 mJ) whereupon a solid-state Hall-effect sensor registered a peck. To obtain an appropriate light environment, the test room’s windows were obscured and the test area was lit by ﬂuorescent strip lights. The operant chamber was lit with a dimmable white halogen (150 W) ﬂood-light to achieve a mean illuminance of 218 ± 31 lx at the test bird’s head height during a trial. Two dimmable halogen spot-lights (50 W) were also suspended over the positions of the stimulus boxes to achieve a mean illuminance of 233 ± 13 lx at the stimulus birds’ head height. The food reward (maggots) was delivered to the operant chamber via an enclosed conveyor feeder at ﬂoor height so that the test bird’s view of the stimuli was not obscured. The keys, halogen lights and feed conveyer were connected to a control box and PC which allowed each item to be operated manually, disabled, or automated via purpose-built software (Solutions for Research Ltd., Silsoe) to run a series of consecutive trials. Correct choice and food delivery was signalled by illumination of feeder with a further halogen spotlight (50 W) and a ‘reward sound’ generated by the PC and played back through PC speakers adjacent to the apparatus. Incorrect choices resulted in an ‘incorrect sound’. An overhead camera linked to a VCR and monitor system was used to observe and record hen behaviour.
2.2.2. Operant chamber The apparatus used in the operant pecking task (Experiments Op-FF and Op-FU) consisted of three boxes, i.e. an operant chamber and two stimulus boxes (Fig. 2). Each stimulus box measured 0.27 m × 0.40 m × 0.60 m and, barring the slightly smaller size (still allowing free rotation of movement) to facilitate simultaneous stimulus presentation, was constructed in the same manner as the stimulus boxes of the Y-maze, also possessing a social exchange port. The operant chamber measured 0.60 m × 0.60 m × 0.60 m and comprised wooden side and back walls, a 0.01 m gauge steel mesh ﬂoor and roof, and a UV-transmitting acrylic front panel. A section was cut out of each edge of the acrylic panel allowing (i) access to the two operant pecking keys ﬁxed to the walls on either side of the panel symmetrically above a central food trough; and (ii) alignment with the social exchange ports on each stimulus box.
One group of hens was assigned exclusively to each learning task (operant, n = 15; Y-maze, n = 16). To minimise the number of birds used, test birds were also used as stimuli for two of their ﬂock-mates; thus hens were allocated to within-group triads (5 per pen), based on similarity of live-weight. Birds within each triad were tagged with leg bands of the same colour, so that this colour could not be used as a discrimination cue during testing; however, a different number of small holes in each band allowed identiﬁcation by the experimenter. Stimulus pairs were assigned within a triad, such that each bird played the role of test subject, positive stimulus (rewarded choice) for one ﬂock-mate and negative stimulus (unrewarded choice) for the other, once in each of three daily test sessions. This design resulted in 15 unique stimulus pairings within each pen. The positive and negative stimuli assigned to each test bird remained the same throughout the experiment. The three sessions per triad were evenly spaced and in the same order throughout the day to minimise carry over effects, the likelihood of satiation and frustration. Birds did not remain within the test apparatus for more than 30 min per session. Hens were acclimatised initially to the apparatus by placing the whole triad in the test area, allowing them to ﬁnd pre-placed food (maggots) in the feeder(s) and to explore whilst the experimenter accustomed them to the apparatus operation. Once conﬁdent, test birds were placed individually in the test apparatus for response training with the appropriate positive stimulus in one stimulus box; the second stimulus box remaining empty (after Bradshaw, 1991 and McLeman et al., 2005) to reduce uncertainty during the period of task acquisition. Correct selection of the positive stimulus was rewarded with 3 maggots; selection of the negative stimulus was unrewarded. The stimulus boxes were swapped between sessions to ensure equal wear and avoid attention to the apparatus. Once the appropriate response had been learnt, subjects were tested daily; each session comprised a number of consecutive choices or
Fig. 2. Experiment Op-FF: plan view of the operant apparatus.
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
trials. A portion of (20%) forced trials (Abeyesinghe et al., 2005) for the positive stimulus were allocated randomly within the trial sequence to remind subjects of choice contingencies and ensure receipt of some rewards. Forced trials did not count towards the performance score. To separate positional from stimulus choice and avoid acquisition of side bias, stimulus positions were swapped quasi-randomly in a balanced manner between trials so that the positive stimulus appeared an equal number of times on each side, but occurred no more than three consecutive trials on the same side within a session. Stimulus boxes (obscured from test bird view) were moved between trials, regardless of whether a side-change was required, to eliminate the possibility of cueing test bird choice. Stimuli were fed maggots by the operator during the inter-trial intervals to encourage them to remain facing the front of the stimulus boxes. Although physical access to other individuals was limited to the portholes, the following criteria were introduced to safeguard the subjects’ welfare in the event of aggression. First, if a bird was pecked through the porthole the drop door was partially lowered to prevent contact. Secondly, if consistent attempts to peck aggressively occurred, or a bird made aggressive physical contact on more than one occasion, the session was terminated and all birds were returned to their home pen(s). Boxes were cleaned with disinfectant and biological odourremoving solution (Odalim: Vétoquinol UK Ltd. or Odourfree: Alstoe) and feeders were cleaned with non-tainting disinfectant wipes between sessions to minimise residual odours. If subjects showed a side bias (≥8/10 or ≥17/24 choices to in the Y-maze and operant tasks, respectively) on 4 consecutive sessions a bias-correction procedure was introduced whereby the positive stimulus was presented on the neglected side for 80% of the free choices until ≥50% choices per session for the previously neglected side were made, whereupon the normal training procedure resumed. Performance data were recorded as the proportion of correct free choices per session. Bias correction trials did not count towards the success criterion. When the success criterion (≥8/10 or ≥17/24 choices for the positive stimulus on two consecutive sessions for the Y-maze and operant tasks, respectively) on the positive stimulus vs. empty box (F+ E; familiar vs. empty) discrimination was met by a test bird, she immediately progressed to discriminating between her positive and negative stimuli (F+ F− ; familiar vs. familiar), otherwise conducted in exactly the same manner. Once the success criterion was achieved on the group-mate discrimination, the experiment was completed for that subject. For the ﬁrst month, training for each subject (n = 15) was conducted on alternate days, thereafter those subjects performing most consistently and achieving the highest proportion of correct responses per session were selected to continue daily training (Ymaze, n = 8; operant, n = 10). The other hens were withdrawn from the experiment as test subjects. 2.3.1. Group-member discrimination in the Y-maze (Y-FF) In the Y-maze, discrimination training was conducted in sessions of 10 free trials plus two forced trials. In each trial, the test subject was placed in the start-box, the positive stimulus was placed in one arm, the empty box or negative stimulus was placed in the other arm and the correct choice feeder was baited. A large container of maggots was kept at ﬂoor level at the central point between the Y-maze arms to counter the unlikely possibility of birds locating rewards by scent. At the trial’s start, the door to the central zone was raised and the test subject was allowed to enter, whereupon the door was immediately lowered behind her and an enforced, pre-choice inspection period of 5 s commenced. Following this, the mesh door(s)s to the arm(s) were raised: one or both depending on whether the trial was forced or free choice. Test subjects were required to walk into the chosen arm and past the stimulus box to
locate the feeder concealed behind it, whereupon the doors were lowered. If the incorrect arm was chosen, the subject found an empty feeder. After a choice, the test subject was gently lifted out of the apparatus and returned to the start box ready for the next trial, which began immediately. If subjects did not exit the start box or choose an arm within 5 min, the trial was skipped. If this occurred in two or more consecutive trials, then the session was terminated. Two alterations to the apparatus were introduced during discrimination training. Following approximately 200 free-choice trials per subject with little indication of task learning (only one bird had progressed to F+ F− discrimination), the mesh doors to the Y-maze arms were replaced with hardboard drop doors, each cut with a rectangular hole appropriate to the size and position of the stimulus box front panel. This facilitated interaction between stimuli and test subjects and obscured the potentially distracting view of the routes into each Y-maze arm prior to choice. However, a number of subjects continued to move past the stimulus boxes directly to the apex of the central triangular zone and wait for the doors to rise, neglecting to inspect their stimuli. This was addressed after a further 80 free-choice trials per subject by relocating the stimulus box’s position within each arm from the foot of the triangle to the apex, to re-engage the test birds’ attention. The stimulus boxes were swapped to the apex of the central triangle and then back to the foot on two consecutive sessions (remaining in one position within sessions) after which they remained at the apex position for the remainder of the experiment. One test subject (R2y) avoided approaching her positive stimulus (R1y), which behaved aggressively towards her. After 420 trials on the F+ E discrimination she was allocated a replacement; a previously unused group-mate of similar weight to the original positive stimulus. This new stimulus bird was given a blank red leg band, her original leg band being removed to avoid discrimination on the basis of leg band colour. 2.3.2. Group-member discrimination in the operant chamber (Op-FF) During shaping of the key-peck response only one key at a time was illuminated, following the position of the positive stimulus, although subjects were trained equally on both sides. The experimenter manually operated the conveyer to deliver a reward (3 maggots) only when the test subject made progressively more accurate movements towards the illuminated, operational key until ﬁnally only a ﬁrm peck (which would register a response on the computer) was required. After each reward delivery, the drop door was lowered and the key-light switched off for a few seconds. The door was then raised, the appropriate key illuminated and then the process was repeated. Once test subjects had learnt that only a ﬁrm peck produced a reward, and that only lit keys produced food, they were progressed rapidly to discrimination training with both operant keys lit and operational during free choices. Sessions consisted of 24 free-choice trials plus 6 forced choices (three per side) for the positive stimulus. The sequence of events for each operant pecking trial is shown in Fig. 3. A stimulus inspection period of 3 s before the keys became operational was imposed to encourage informed choices and a timeout punishment for incorrect choices was introduced to aid discrimination. Subjects failing to complete 30 trials in a session were trained for a maximum of 20 min, regardless of the number of trials completed. As with the Y-maze task some modiﬁcations were made during the F+ F− discrimination phase to try and facilitate learning. After approximately 570 free-choice training trials per subject, the apparatus was modiﬁed to allow slightly closer visual inspection of stimuli by removing the Perspex from the front of only the test chamber; this was potentially important when stimulus birds stayed near the back of the stimulus boxes. The test light environment was also altered to introduce daylight, in correspondence with
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
Fig. 3. Experiment Op-FF: trial sequences for free and forced choice trials during discrimination training. The diagram illustrates the view from the test bird’s perspective and events occurring, according to test bird response, during a trial. Stimuli were positioned prior to trial start which was signalled by a commencement tone and raising the drop door. A 3 s inspection period was allowed before the stimulus key(s) were illuminated. During forced choice trials pecks on the unlit incorrect key had no consequence. A peck on the correct illuminated key resulted in it darkening, accompanied by a ‘correct choice’ tone, a light illuminating the feeder, reward delivery and the drop door. The test bird was given 2 s to consume the reward before commencement of the ITI (15 s). A peck on the incorrect key resulted in a ﬁxed ‘time-out’ punishment interval of 5 s prior to commencement of the inter-trial interval. If neither key was pecked the trial continued for a maximum of 30 s whereupon a ﬁxed trial-end-to-trial-start inter-trial interval (ITI) of 15 s commenced automatically.
the home pen environment, ensuring the presence of certain wavelengths available in this spectrum (potentially including UV), which could potentially affect recognition. 2.3.3. Discrimination of familiarity in the operant chamber (Op-FU) The test subject performance in the operant task group-member discrimination (Op-FF see Section 3.2) was very poor. To examine further whether failure was generic to any social discrimination using this conditioned task, or speciﬁc to the group-member discrimination, we investigated the performance of the same test subjects on a familiar vs. unfamiliar discrimination, which hens have previously performed in functional tasks (e.g. Hughes, 1977; Bradshaw, 1992a; Dawkins, 1995; Grigor et al., 1995; D’Eath and Keeling, 2003; D’Eath and Stone, 1999). Thus we expected to see clear behavioural evidence of functional discrimination and, if the operant task was appropriate, rapid acquisition of the conditioned response would be likely.
Procedures followed those in the experiment Op-FF group-mate discrimination stage (F+ F− ) except that the familiar negative stimulus (F− ) was replaced with an unfamiliar negative stimulus bird (U− ) from one of two ﬂocks of the same size, kept under identical conditions. The positive stimulus remained familiar so that test birds were not required to approach a potentially aversive unfamiliar bird to make a correct response. The unfamiliar stimulus pens were within potential olfactory and auditory range of the operant test birds’ pen, but prior to the ﬁrst test session, the test subjects and unfamiliar stimuli had never seen each other. The unfamiliar individual used was rotated in order from a panel of ﬁve birds, so that each test subject saw each unfamiliar stimulus only once per week, but each unfamiliar stimulus was used twice a day. This maximised the duration for which they would remain essentially ‘unfamiliar’ (Bradshaw, 1992a; Porter et al., 2006), whilst minimising animal use. Only 10 sessions were conducted because of an expectation of rapid improvement in learning if the task was appropriate to social discrimination, a requirement to avoid familiarisation with
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
the unfamiliar stimuli and to minimise exposure of birds to potentially stressful social challenge.
chance can be calculated from the Geometric distribution, N=
2.3.4. Behaviour To aid interpretation of performance, behavioural observations were made on video recordings from three sessions per subject at key stages throughout training: (a) unlearned—early on in discrimination training before the stimulus associations had been learnt or the equivalent ‘early’ stage for ultimately unsuccessful subjects; (b) pre-success—immediately prior to a successful session, during acquisition of stimulus associations or the equivalent ‘middle’ stage for ultimately unsuccessful subjects; (c) at criterion—the session during which the success criterion was met or the equivalent ‘late’ stage for ultimately unsuccessful subjects. Given the different performances in the F+ E, F+ F− and F+ U− tasks, this was done for each of these stages separately with the operant method. The following data were recorded for both methods: (i) stimulus bird head orientation(s) towards the Y-maze central area or operant test chamber during the inspection period (facing forwards, sideways or backwards); (ii) test subject pre-choice body position (location within the inspection area) and (iii) head-orientation at the end of the inspection period; (iv) number of stimuli ‘inspected’ (test subject’s head turned in the direction of the stimulus box); and (v) mean latency (s) to make a choice (Y-maze duration from raising the pulley doors until the subject’s whole body entered a arm and operant latency to ﬁrst key-peck following the keys becoming operational). Although we accept that head orientation does not guarantee that the stimulus was in fact being viewed, given the apparent lateralisation of vision in chicks (Deng and Rogers, 2002), Dawkins’ (1995, 1996) clearly describes discrimination at close range using a stereoscopic assessment in unconstrained adult fowl. Therefore, following this reported use of the binocular visual ﬁeld, this is a reasonable assumption. In addition, for the operant task we recorded social, exploratory and displacement and behaviour of test subjects pre-choice and during the inter-trial interval: threats and aggressive pecks given or received and submissions performed (Wood-Gush, 1971); pecking of the operant chamber ﬂoor, wall or Perspex (bouts separated by ≥2 s); scratching (raking feet across chamber ﬂoor in a paddling motion associated with foraging behaviour); preening; head-ﬂicking (Zimmerman et al., 2003) and circling (walking a lap clockwise or anticlockwise around the operant box commencing at the front, moving to the back and terminating at the front, without pause or performance of an interrupting behaviour). 2.4. Analysis The Binomial Law was used to determine the likelihood of a test subject achieving the discrimination criterion by chance alone (McLeman et al., 2005). Assuming the probability, p, of a test hen choosing the positive stimulus by chance is 0.5 in a single trial and using the Y-maze experiment as an example, then the exact probability, q, of eight or more similar choices within a session of 10 free-choice trials is calculated from the Binomial distribution, q = (0.5)10 + 10(0.5)9 (0.5) + 45(0.5)8 (0.5)2 = 0.0547.
Assuming no learning and therefore, independence of trials, the likelihood of a non-discriminator choosing the positive stimulus eight or more times in two consecutive sessions is, by the Multiplication rule: q2 = 0.00299. If the probability of each outcome (either + or −) is 0.5 (i.e. no learning has occurred), then the expected number of sessions, N, before a single session of 8/10 successful discriminations occurs by
(1 − q) = 17 sessions. q
By the same formula, the expected number of sessions before a run of two consecutive successful discriminations is achieved is N=
2(1 − q2 ) = 666 sessions. q2
Similarly for the operant tasks the likelihood of a nondiscriminator reaching criterion (≥17/24 free choices) on two consecutive sessions is q2 = 0.0322 = 0.00102,
and the expected number sessions, N, before a single successful session and two consecutive successful sessions occur by chance are ≈30 and ≈1950, respectively. Given the limited sampling of sessions observed, descriptive statistics of the behaviour data were calculated. 3. Results 3.1. Discrimination in the Y-maze (Y-FF) Fig. 4 displays the performance of individual test subjects over time. For most subjects a clear learning curve, or gradual increase in the proportion of successful choices over successive sessions, was not seen until approximately 2/3 of the way into training. 3.1.1. Y-maze positive stimulus vs. empty box discrimination (F+ E) All 8 subjects selected to continue training were successful in signiﬁcant discrimination between their positive stimulus and an empty box (p < 0.003). Although one bird (P1y) learnt relatively quickly, on average each subject took a total of 315 ± 108 free-choice trials to meet the success criterion (mean ± S.D.). This may not be a true representation of capability as learning progressed very rapidly (49 ± 42 free-choice trials) in subjects which had not already learnt (n = 7) following modiﬁcations to the apparatus (Fig. 4. indicated by ‘’ and labelled with lower case letters); in particular with the relocation of the stimulus box from the foot to the apex. The subject in receipt of agonistic behaviour from her positive stimulus (R2y) learnt within 50 trials after her stimulus was changed (as indicated by ; Fig. 4). Across all test subjects, 18 single signiﬁcant sessions were achieved, in addition to the 16 successive criterion sessions (Fig. 4). From the calculations conducted in Section 2.4, we would expect a total of only 15 signiﬁcant sessions occurring by chance during the 252 sessions across subjects, indicating subjects may have learned before consistency of performance was obtained. Left bias was most often seen in this training phase (all sessions: left bias 26.2%; right bias 10.3%) although only two subjects (B1y and R2y) required correction. 3.1.2. Y-maze group-member discrimination (F+ F− ) Seven of the eight test subjects successfully discriminated between their group-mate pairs. This generally took fewer trials per subject than the F+ E discrimination (F+ F− : mean ± S.D. trials to success = 83 ± 72, n = 7) and for some subjects occurred very quickly (Fig. 4). The bird (R2y) which took the longest to learn the F+ E discrimination did not achieve success on the F+ F− discrimination within the remainder of the experimental time available, i.e. 130 free-choice trials. Although there were fewer single signiﬁcant sessions in this stage across all subjects compared with the F+ E discrimination (7 sessions), this reﬂected the faster learning of most test subjects (Fig. 4); only 4 signiﬁcant sessions in an overall total of 71
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
Fig. 4. Experiment Y-FF: performance scores of birds (out of 10 free-choice trials) across training sessions for both familiar positive stimulus vs. empty stimulus box (F+ E ) and familiar positive stimulus vs. familiar negative stimulus (F+ F− ) discrimination tasks in the Y-maze. Success criterion for each session (≥8/10 choices) is indicated by an unbroken horizontal line. Points  labelled with letters refer to the introduction of methodological changes: a = daily training, b = replacement of mesh doors with hardboard doors, c = alternating stimulus box position, d = stimulus boxes remaining at apex. The asterisk (for bird R2y) denotes the session during which the positive stimulus was ﬁrst replaced for another less aggressive positive.
would have been expected by chance, again indicating some learning had occurred before criterion was met. During the group-mate discrimination 29.6% were signiﬁcant compared with 13.5% in the F+ E discrimination. Overall fewer bias sessions were recorded (all sessions left bias: 12.1%; right bias 10.3%) and bias was recorded equally often for both sides. Only the subject (R2y), which ulti-
mately did not achieve the F+ F− discrimination, needed a bias correction. 3.1.3. Behaviour in the Y-maze The orientation of the stimulus bird’s heads did not affect the test subject’s success; stimulus birds’ heads faced either forwards
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
or sideways during the majority (≥84%) of trials in all observed periods, regardless of choice outcome. Before the test subjects had learned, when the stimulus boxes were at the foot of the central triangle, they waited next to the doors to the Y-maze arms (at the triangle apex) in 79.8% of the trials observed. In this location, subjects’ heads were unobservable or were oriented towards the apex and/or exits just prior to the doors being raised and they tended to ‘look’ towards both stimuli on only 31% of trials as they moved past. As subjects began to learn the stimulus associations (pre-success), their inspection positions altered more frequently to locations allowing close stimulus examination and ‘inspection’ of both stimuli prior to making a choice was recorded in 88% of trials observed. Success (at criterion) was associated with close proximity to the stimuli, in a central position during the inspection period. At this stage, test subjects faced the positive stimulus as the doors were raised (60 out of 84 observed trials, 71%) and had ‘inspected’ both stimuli prior to choice (96% trials). Average decision times tended to be slightly longer just prior to success than at any other time (unlearned: 3.6 ± 2.7 s vs. presuccess: 4.6 ± 2.3 s vs. at criterion: 3.8 ± 1.4 s), perhaps indicating that more attention was being paid to the stimuli. Decisions during incorrect trials tended to be completed slightly more quickly than correct ones (pooled across stages; 3.6 ± 1.3 s vs. 4.3 ± 2.0 s, respectively), perhaps reﬂecting impulsive choices. 3.2. Discrimination in the operant chamber (Op-FF and Op-FU) Fig. 5 displays the performance of individual test subjects over time in the operant task. 3.2.1. Operant positive stimulus vs. empty box discrimination (F+ E) All 10 subjects selected to continue operant training were successful in signiﬁcantly discriminating their positive stimulus from an empty box within, on average 226 ± 98 trials, although the fastest subject (S3o) took only 48 trials. Prior to all subjects achieving success, 11 single signiﬁcant sessions (in addition to the 20 criterion sessions) were performed across subjects during the course of this training stage, whilst only 3 would be expected by chance (Section 2.4) in a total of 94. For most of the subjects a clear learning curve was seen for this discrimination (Fig. 5). No bias was recorded more often to one side than the other (all sessions: left bias 16.0 %; right bias 12.8%) and only two subjects required bias correction (Y2o and B3o). 3.2.2. Operant group-member discrimination (F+ F− ) Only one subject (B2o) of the 10 tested was successful at the group-mate discrimination (after 432 trials or 18 sessions; Fig. 5). Despite the apparatus’ modiﬁcations (Fig. 5 ‘’, labelled with lower case letters) no consistent improvement in bird performance was seen. After an average (excluding the successful subject) of 573 ± 63 trials on this discrimination, the experiment was terminated for the remaining subjects. Eight signiﬁcant sessions would have been expected by chance within the total across subjects of 233 sessions and 14 were obtained. This suggests some learning occurred but consistency of performance was not achieved, however, the majority of these single successful sessions we associated with only two birds (including B2o). The percentage of signiﬁcant sessions (including the single criterion achieved) was much smaller (6%) than for the F+ E discrimination (33%), reﬂecting the apparent general failure to learn, which is also clearly seen in the lack of performance curves in Fig. 5. More bias was seen towards the left key relative to the right during this phase (all sessions: left bias 15.0%; right bias 6.4%) and 4 subjects required correction sessions.
Fig. 5. Experiments Op-FF and Op-FU: performance scores of birds (proportion correct out of 24 free-choice trials) across training sessions for Op-FF familiar positive stimulus vs. empty stimulus box (F+ E ) and familiar positive stimulus vs. familiar negative stimulus (F+ F− ) discrimination tasks and the Op-FU familiar positive stimulus vs. unfamiliar negative stimulus (F+ U− [䊉]) discrimination task in the operant chamber. Success criterion for each session (≥17/24 choices) is indicated by an unbroken horizontal line. Points  labelled with letters refer to the introduction of methodological changes: a = daily training, b = removal of Perspex and concordance of light environment with home pen.
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
3.2.3. Operant familiarity discrimination (F+ U− ) After 10 sessions on the familiar vs. unfamiliar (F+ U− ) discrimination, only one subject (S2o) achieved success criterion (Fig. 5) and although four further non-consecutive signiﬁcant sessions were recorded (4% of sessions), 3 would be expected by chance within the 100 session total. Successful performances did not appear to be associated with particular negative stimuli. Three subjects were corrected for bias; bias was recorded on 24% all sessions and left bias was seen most often (63% bias recorded). 3.2.4. Behaviour in the operant chamber 184.108.40.206. Pre-choice behaviour. As with the Y-maze, operant stimuli spent the majority of test trials facing the test birds or facing sideways in all discrimination tasks, regardless of choice outcome. However, as the F+ U− task progressed, the positive stimuli began to face backwards or to the side more often (% trials stimulus birds were recorded facing forwards: F+ stimulus unlearned/early 84%, pre-success/middle 64%, at criterion/late 59%; U− stimulus unlearned/early 81%, pre-success/middle 80%, at criterion/late 83%). Overall, test subjects faced towards the front of the chamber and avoided the back of the chamber during the inspection period in all trials observed and there was a tendency for them to stay to the left or right of the front area rather than centrally (<25% trials observed). Unlike in the Y-maze task, generally only one stimulus was inspected prior to choice regardless of ultimate success or learning stage and there did not appear to be a relationship between latency to ﬁrst key-peck and choice outcome. Whilst low frequencies of social and exploratory behaviour were recorded overall, more social interactions per subject, particularly test bird submission and receipt of threat from stimuli, were observed during the F+ E discrimination (30.6 total interactions pooled across all sampled trials) compared with the F+ F− (15.2) and F+ U− (15.7) discriminations. Overall test subjects received more aggressive behaviour from stimuli than they performed towards stimuli in all operant tasks (total counts aggressive threats and pecks, respectively, performed by stimuli, F+ E = 40 and 3, F+ F− = 25 and 7, F+ U− = 45 and 9, respectively, vs. performed by test subjects F+ E = 35 and 2, F+ F− = 13 and 2, F+ U− = 10 and 4, respectively; data pooled across all subjects and trials sampled). Exploratory or displacement behaviour was rarely seen prior to choice. Unsurprisingly, F+ E test subjects were located substantially more often next to (60–70% all trials sampled) and oriented their head towards the positive stimulus more than anywhere else in the chamber immediately prior to choice, but tended to do so more at success (test subject head orientation towards F+ : unlearned, 39% trials; pre-success, 46% trials; at criterion, 50% trials sampled). The performance of threats by stimuli and submissions performed by the test subjects tended to increase with success (F+ E total threats received per bird; unlearned 0.1, pre-success 0.1, at criterion 0.2; F+ E submissions performed per bird unlearned 5.1, pre-success 6.1, at criterion 6.5). A preference for standing next to the positive stimulus (cf. centrally or next to the negative stimulus) was evident before task learning in the single successful F+ F− group-mate discriminating subject (B2o: unlearned, F+ 59% vs. F− 33% trials; pre-success, F+ 57% vs. F− 31% trials; at criterion, F+ 46% vs. F− 38% trials sampled). Though this subject was far more frequently found oriented towards the keys before task learning (+ve key 55%, −ve key 29% trials), at criterion she was almost equally oriented towards the positive stimulus, or either key (21–29%) and ‘inspected’ at least one stimulus in all trials. In contrast, unsuccessful F+ F− subjects remained consistently oriented to the keys throughout (21–29% trials sampled at all stages) and omitted to ‘inspect’ stimuli during up to 15% of trials throughout. Following the pattern of the Y-maze subjects, B2o tended to take slightly longer to chose just prior to success
during F+ F− than at any other time (B2o: unlearned, 4.3 ± 0.6 s; pre-success, 5.6 ± 1.9 s; at criterion, 4.7 ± 0.8 s; unsuccessful subjects: early, 4.6 ± 1.1 s; middle, 4.9 ± 2.0 s; late, 5.4 ± 1.9 s). Social behaviour in F+ F− discriminating B2o did not differ substantially from the average unsuccessful subject except that she tended to perform more (F+ F− total counts threats and pecks performed per subject; B2o = 3 and 1, respectively, vs. unsuccessful = 1.4 and 0.1, respectively; data pooled across all trials sampled) and receive less (F+ F− total counts threats and pecks received per subject; B2o = 0 and 0, respectively, vs. unsuccessful = 2.8 and 0.8, respectively) threats and aggressive pecks overall, suggesting high rank. Exploratory behaviour was recorded for only F+ F− discriminating B2o (total count of 4 occurrences of ﬂoor-pecking across all trials sampled) which did not perform head-ﬂicks, in contrast to her unsuccessful companions (F+ F− total counts per unsuccessful subject across all trials = 1.3). Although additionally tending towards the central position just prior to success, the single successful F+ U− discriminating test subject (S2o) was primarily recorded on the left. At criterion she was recorded on the left in 83% trials and more often next to (S2o: unlearned, F+ 50% vs. U− 42% trials; pre-success, F+ 32% vs. U− 32% trials; at criterion, F+ 57% vs. U− 39% trials sampled) and oriented towards (S2o: unlearned, F+ 38% vs. U− 17% trials; pre-success, F+ 32% vs. U− 24% trials; at criterion, F+ 65% vs. U− 35% trials sampled) her positive familiar stimulus than her unfamiliar negative, having more frequently ‘inspected’ both stimuli before choice (S2o unlearned, 8%; pre-success, 12%; at criterion, 30% trials sampled). Overall F+ U− discriminating S2o received more threats than the average unsuccessful test subject (total counts per bird, data pooled across all stages S2o = 8.0 vs. non-discriminating birds = 4.1) but did not engage in or receive any interaction at success. In addition she performed more head ﬂicks pre-choice than the average unsuccessful test subject (total counts per bird, data pooled across all stages S2o = 2.0 vs. non-discriminating birds = 0.2). The remaining unsuccessful F+ U− subjects, regardless of stage in training were positioned more uniformly across the front of the test chamber pre-choice, than their successful companion or during previous operant discriminations. As the experiment progressed, they were also recorded more equally next to both stimuli (early, F+ 45% vs. U− 29% trials; middle, F+ 38% vs. U− 36% trials; late, F+ 33% vs. U− 39% trials sampled) but, regardless of stage, oriented equally towards all features except the back of the chamber (% trials recorded oriented towards: F+ 16% trials; U− 16% trials; front central 26% trials; +ve key 19% trials; −ve key 19% trials; back of the chamber 0% trials; data pooled across all trials sampled). Mean latency to ﬁrst key-peck did not really differ with experimental progression for unsuccessful F+ U− subjects (early, 5.0 ± 2.2 s; middle, 4.8 ± 1.6 s; late, 5.1 ± 1.4 s) and again there was no relationship between choice outcome and decision speed. Aggressive behaviour performed by test subjects during this F+ U− discrimination was due to a only two birds (B2o and S3o) and threats performed by these subjects towards the unfamiliar stimuli were more overt, involving elevation of neck feathers and greater neck extension, than seen towards familiar stimuli. 220.127.116.11. Inter-trial interval behaviour (drop door lowered). During the inter-trial interval, test subjects were most frequently recorded at the front of the chamber regardless of discrimination or learning stage, but they spent more time at the back of the box during the F+ U− discrimination inter-trial interval (frequency of recording birds at the back of the chamber during the ITI [counts per trial]: F+ E, 0.1; F+ F− , 0.1; F+ U− , 0.2; data pooled across all learning stages and test subjects). During the F+ E discrimination successful test subjects most frequently pecked the chamber ﬂoor and circled (0.3 and 2.8 counts per trial, respectively; data pooled across birds) during their
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
wait. The successful group-mate discriminating (B2o) behaved and more frequently than her unsuccessful F+ F− non-discriminating companions (F+ F− counts of ﬂoor pecking and circling per trial: B2o 0.8 and 0.7, respectively, vs. non-discriminating birds 0.1 and 0.5, respectively) and also frequently performed bouts of scratching the ﬂoor (0.5 counts per trial vs. non-discriminating F+ F− = 0). Unsuccessful F+ F− subjects were further recorded less frequently in all areas, suggesting less movement around the chamber (frequency of subjects found in location, counts per trial: F+ F− successful front-left 1.4, front-right 1.3, centre 0.6, back of chamber 0.2; F+ F− unsuccessful front-left 0.9, front-right 0.7, centre 0.6, back of chamber 0.0) and whilst F+ F− discriminating subject B2o never showed head-ﬂicks during the ITI, her unsuccessful F+ F− non-discriminating companions, F+ U− discriminating S2o and unsuccessful F+ U− non-discriminating subjects all displayed the same frequencies (0.3 counts per trial, respectively, across all stages). The F+ U− discriminating S2o and unsuccessful F+ U− nondiscriminating subjects performed similar levels of circling behaviour (0.4 counts per trial, respectively), but again at lower frequencies than F+ F− discriminating B2o. Overall F+ U− discriminating S2o tended relatively to avoid the front-central area and more frequently visit the back of the chamber compared with the unsuccessful F+ U− subjects, which tended to be more frequently recorded at the front-left of the chamber (frequency of birds found in location, counts per trial: successful: front-left, 0.6; front-right, 0.8; front-centre, 0.4; back of test chamber, 0.4; unsuccessful: frontleft, 0.9; front-right, 0.7; front-centre, 0.7; back of test chamber, 0.2; data pooled across all stages sampled).
4.1.1. Discrimination between a live stimulus and empty stimulus box (F+ E) Dawkins (1995) predicted that social discrimination in close proximity should be learnt much more quickly than from a distance. It was therefore surprising that our test subjects were substantially slower to learn the initial discrimination between a hen and an empty box than Bradshaw’s (1991) subjects, although some performed comparably. For the Y-maze this was probably because our initial apparatus conﬁguration was not conducive to tasklearning. Requiring birds to walk past the stimuli to obtain rewards potentially promoted this route, rather than the reward cue, as the primary focus. Most subjects successfully discriminated very quickly after the apparatus was modiﬁed, at rates (49 ± 42 trials) similar to Bradshaw’s (1991) subjects (67 ± 29 trials). The operant pecking response may simply have been harder or taken longer to learn than a Y-maze task since operant subjects were ultimately successful in this stage. However, the F+ E discrimination task may actually have been facilitated in Bradshaw’s (1991) subjects by viewing stimuli at a distance, depending on the ‘decision rule’ subjects were using. If this rule was simply either ‘choose the occupied box’ or ‘don’t choose the empty box’ then there would have been no necessity to identify the stimulus occupant and discrimination could have occurred using only very coarse cues, such as colour or shape. In closer proximity, motivation to interact may have initially distracted test subjects from the discrimination task. Finally, a mild food deprivation (4 h) and a smaller sample size (n = 3) also enabled Bradshaw (1991) to train potentially more motivated subjects, more frequently and for longer within a session than in this study, providing greater opportunity for birds to learn quickly.
4.1.2. Discrimination between live familiar stimuli (F+ F− ) In accordance with Dawkins’ (1995) prediction about proximity, our Y-maze test subjects learnt the second, ﬂock-mate discrimination far more quickly than Bradshaw’s (1991) birds, even though their training was approached in a similar manner. Only one bird did not learn the Y-maze task within the constraints of experimental time, however, stress interference (Mendl, 1999) associated with the agonistic behaviour of her initial positive stimulus may have contributed to this failure. The general speed of learning in the Y-maze suggests that once the nature of the task had been grasped, the way in which the stimuli and rewards were presented in this study was more amenable to task performance than in Bradshaw’s method. The rapid success and reduction in bias during this Y-maze discrimination stage implies further that at least some of our subjects were using cues associated with their speciﬁc positive stimulus in the preceding discrimination between a hen and an empty box and some transfer of learning between discriminations occurred. In contrast the majority of our operant subjects failed to demonstrate group-mate discrimination and the only successful subject still took over twice as many trials as Bradshaw’s (1991) hens and over 20 times as many trials as our fastest Y-maze subjects. The number of trials performed (maximum 720) implies that the failure to learn was unlikely to have been due to insufﬁcient opportunity and that the apparatus’ alterations were ineffective, at least at the stage at which they were introduced. Operant success on the F+ E but not F+ F− discrimination, suggests that subjects were using simple decision rules which allowed successful performance without social recognition; e.g. using perception of low frequency visual details, such as the shape and form of the positive stimulus whilst still focussing on the foreground. It is possible that success of the single F+ F− discriminating operant subject may have been an artefact of her relationships with her stimuli. Although this could only have been conﬁrmed by presenting her with another stimulus pair, which was not possible within this experiment, similarities between her behaviour and that of successful Y-maze subjects sup-
The aims of this study were two-fold, i.e., to determine whether domestic hens are capable of discriminating between group-mates and to determine the best method of conducting a group-mate discrimination task in terms of learning speed, reliability, and proportion of subjects achieving criterion. Although Bradshaw (1991) demonstrated ﬂock-mate discrimination with a small sample, later investigations (Dawkins, 1995, 1996) raised doubts, regarding the social nature of these ﬁndings, which we attempted to address in our study. Given the adjustment of viewing distance, the larger sample size and the use of unique stimulus pairs for each test subject – to increase generality of the response in our study compared with Bradshaw’s (1991) – we have unequivocally demonstrated that it is possible for adult hens to learn a ﬂock-mate discrimination task using live stimuli. However, the method used is crucial. 4.1. Method comparison The majority of the literature on social discrimination in domestic fowl reports studies using discrimination tasks based on observation of functional behaviour, e.g. either preference or aggression between unfamiliar birds; however, an overt, predictable functional response between familiars, even with a clear dominance relationship, has been difﬁcult to identify (Bradshaw, 1992b; D’Eath, 1998). To circumvent this problem, we used a conditional learning task to investigate the cognitive ability of hens to discriminate ﬂock-mates, after Bradshaw (1991) and McLeman et al. (2005). Despite the abstract nature of this type of task, its relevance is supported by the speculation that individual recognition is likely to rely on learned discriminations between familiar conspeciﬁcs (Zayan and Lamberty, 1988). However, there were clear differences in performance in the two tasks.
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
port a true discrimination. The greater activity of this bird (B2o; e.g. ground-pecking and circling) during the inter-trial interval compared with her unsuccessful counterparts could also reﬂect strategies to reduce impulsive behaviour (Grosch and Neuringer, 1981). The head-ﬂicking behaviour observed during the inter-trial interval in unsuccessful group-mate discriminating subjects is suggestive of frustration (Zimmerman et al., 2003) which may lead to more impulsive responding. 4.1.3. Discrimination between live familiar and unfamiliar stimuli (F+ U− ) Given the strong evidence of discrimination between ﬂockmates and strangers from previous studies (e.g. Hughes, 1977; Bradshaw, 1992a; Dawkins, 1995; Grigor et al., 1995; D’Eath and Keeling, 2003; D’Eath and Stone, 1999), failure in both operant discrimination tasks (F+ F− and F+ U− ) was unexpected. Behavioural analysis indicated that two problems occurred. The majority of the test subjects appeared to ‘ignore’ the stimuli; they did not orient towards them any more than other features at the front of the box and, contrary to our predictions, did not even appear to behave differently towards the unfamiliar, despite agonistic behaviour performed by the stimuli. It could be argued that since birds had previously been in atmospheric and auditory contact then they may not have been perceived as unfamiliar; however, this is extremely unlikely as previous studies (D’Eath and Stone, 1999; Hauser and Huber-Eicher, 2004) found visual cues to be crucial in discrimination. The two birds which did appear to discriminate behaviourally (but not via the conditioned response) included the previously successful subject from Experiment O-FF (B2o and S3o). These test birds were so distracted by social interaction with the unfamiliar stimulus that they did not peck the operant key within the choice period and thus failed to perform the operant task. Dukas (2002) reviews how successful task performance is constrained by the amount of attention simultaneously focussed on other activities, which, in a learning task context, may include natural behaviour that the subjects are additionally motivated to perform, e.g. nesting behaviour or social interaction. Indeed the agonistic behaviour received by the successful bird, her tendency to position herself away from the unfamiliar stimulus and the frustration indicated by head-ﬂicking, suggests she may have been avoiding aggression from the unfamiliar stimulus rather than demonstrating a learned response. Bradshaw (1991) conducted a familiar vs. stranger discrimination with three naïve birds using a Y-maze learning task and reported discrimination by only one subject, after more than 500 trials. This is noteworthy because we might assume, if cues used to determine familiarity are also used in determining identity, they should have performed similarly to his ﬂock-mate discriminating birds, even if they were simply using simple shape and colour at this distance. Bradshaw (1991) cites his ﬁndings to be ‘an artefact of a poor training procedure’ which did not include discrete stages in training, but it is also possible that they may be due to the stress of approaching an unfamiliar conspeciﬁc (Mendl, 1999), an artefact of the small sample size, or a conﬂict between functional and conditioned behaviour as we observed in a minority of our subjects. If the latter consistently occurs then it potentially renders any learning task inappropriate for testing discrimination between familiar and unfamiliar birds. Further investigation is required to establish this. 4.2. Factors affecting task performance Given these ﬁndings regarding both group-mate and familiarity discriminations made with the current apparatus conﬁguration and test procedure, the operant method – unlike the Y-maze – appears to be an inappropriate method to demonstrate social discrimina-
tion in hens. There are several factors involved in task performance which might explain this. 4.2.1. Cue consistency The ease with which a reward association is formed will be affected by the consistency of cue appearance as well as its spatial and temporal presentation. Using live unrestrained stimuli, although presumed to be more socially relevant, introduced the risks that the stimuli could behave differently during different sessions and that the test subject could not always view them from the same perspective. However, this was not an issue in our study as, for the most part, the stimuli remained facing the test subject in both experiments (Y-FF and Op-FF). This implies that those features that are necessary for discrimination were available either from the frontal viewing position, consistent with previous reports that the head and neck need to be visible (Guhl and Ortman, 1953; Dawkins, 1996) or were irrelevant to the stimulus position altogether. During the Op-FU experiment, the positive (familiar) stimuli, tended to face away from the test subjects more often as the experiment progressed, potentially increasing the difﬁculty of the discrimination (Bradshaw, 1992a; Porter et al., 2006); this may have reﬂected an avoidance of agonistic test subject behaviour which was actually directed to the unseen (by the positive stimulus) unfamiliar. 4.2.2. Cue attendance Attendance to the stimuli as cues to signal reward was also an important feature of both discrimination tasks. Initially, test birds in both tasks appeared to be distracted by the response-reward association and did not appear to regard the stimulus birds as relevant discriminating cues. However, following the apparatus’ modiﬁcations, stimulus attendance improved for the Y-maze subjects, such that a clear difference in stimulus inspection by test subjects could be seen at success. In contrast, the majority of operant subjects continued to direct their attention to the keys at the moment of choice, supporting a more impulsive strategy as opposed to informed decision making, although choices did not differ temporally. The majority tendency to disregard stimuli in the operant tasks could have been exacerbated by the requirements: (i) to train birds to key-peck for food before testing commenced and (ii) to position the keys to the far left and right of the chamber to avoid obstructing the view of each stimulus. Thus test subjects may have attended to the foreground and not the stimuli beyond and the key-peck-food association potentially overshadowed the conspeciﬁc cue. This is supported by the failure to obtain even behavioural discrimination by the majority of operant subjects in the F+ U− discrimination suggesting that they were not attending to the stimuli as task relevant at all. It is possible that this may have been a carry over from their previous experience of operant discrimination (during Op-FF), however, this only supports the exclusion of the operant task for group-mate discrimination and still begs the question of whether naive test birds might show more behavioural discrimination which, as demonstrated by our minority, may conﬂict with the conditioned response. 4.2.3. Appropriateness of the method and response requirement There are a number of reasons why the operant test subjects might have disregarded the stimuli as discriminating cues. In contrast to Y-maze subjects, which approached from a distance to make a choice, the continuous close proximity to the operant keys at all times appeared to encourage impulsive behaviour in the operant subjects by reducing the incentive to move location and peck the alternative key. For reasons outlined earlier, the nature of our experiments by necessity employed conspeciﬁcs as cues to signal food rewards, however, there are arguably limited contexts when this might naturally occur. The operant test hens may have had difﬁculty learning that a pecking a key in front of particular bird
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13
leads to food as it is not a natural outcome associated with a social stimulus. One possible explanation of the test subjects’ behaviour refers to the respective relationships of the Y-maze and operant tasks to different phases of the natural foraging process (Timberlake, 1988). The Y-maze locomotor response to obtain food is representative of the search phase of foraging, where a conspeciﬁc may naturally play a role in locating a food source, e.g. through food calling or through being observed whilst eating. In this case, discrimination between conspeciﬁcs may be useful to determine the more successful locators of food. The operant task is more representative of the food acquisition and handling phase, where the bird is directly engaged in obtaining food, i.e. by pecking. Here there is no natural role for a conspeciﬁc to signal food; rather her presence is more likely to be associated with competition. In addition, social discrimination in this circumstance is more relevant to the probability of having to cede the resource. In this case, the identity of the conspeciﬁc potentially became irrelevant if test subjects learnt that their food source could not be accessed by the stimuli. As a result the use of stimulus birds as reward cues in the operant pecking method may have been so irrelevant to the test bird’s natural behaviour that they could not learn the discrimination task. 4.2.4. Intimidation/distraction Even where stimulus inspection did occur, this does not necessarily imply that test subjects perceived stimuli as cues to reward; rather they may have represented potential competition for food or sources of social conﬂict. At least one test bird in each method appeared to avoid her positive stimulus as a result of threatening behaviour and aggression by the latter, although avoidance was not signiﬁcant. Stimuli, either observing or hearing food that they were unable to access being consumed by test subjects, may have been frustrated and responded agonistically towards the test birds. In fact, with the operant task, pecking towards another bird at head height could be construed as a threat by the stimulus, leading to antagonism and test subject reluctance to perform the task. Test subjects, particularly of low social rank could conceivably have been in conﬂict between motivation to obtain the food rewards and to avoid inciting agonistic behaviour in their stimuli. Although overt agonistic behaviour was rarely seen, the observed test birds’ relatively frequent performance of submission and receipt of threat from stimuli in the operant task, particularly during the F+ E discrimination, suggests that the social conﬂict as described was plausible. With the F+ U− operant discrimination, aggression was more frequently performed by a small number of test subjects and the stimuli only, often resulting in fearfulness or retaliation which distracted the test subject from conditional responding as discussed earlier. 4.3. Implications of ﬁndings The ability of hens to discriminate between group-mates supports the possibility that hens may use individual recognition (as deﬁned by Zayan, 1994) to maintain social hierarchies, with consequent implications for disruption of social cognition under commercial conditions. Whilst this ﬁnding under experimental conditions does not necessarily imply that hens naturally discriminate group-mates in a social context, an inability to do so would preclude the ability (D’Eath and Keeling, 2003) and the energetic costs and cognitive demands involved argue against evolutionary development of an unnecessary and unused cognitive ability. This does not, however, preclude an ability to alter social strategy under conditions which render such discrimination either ineffective or inefﬁcient, such as large ﬂock size (see Mench and Keeling, 2001). The paradigm of simultaneous stimulus comparison provides a useful method to investigate one of the mechanisms underly-
ing individual recognition; however, it is unlikely that recognition would normally involve direct comparison of individuals against one another. It may be possible to investigate recognition and memory on a individual basis further using successive presentation in a symmetrically reinforced task, such as described by Gheusi et al. (1994), modiﬁed to use a locomotor response, as in the Y-maze method. Despite these limitations, the Y-maze ﬂock-mate discrimination provides a promising model for further investigation of the sensory modalities used in social recognition and their precedence (after McLeman et al., 2005 with pigs) as well as the impact of environmental factors on social cognition. Although in this study a familiar vs. unfamiliar discrimination was not conducted using the Y-maze approach, the lack of success with the operant task, together with Bradshaw’s (1991) ﬁndings raises the possibility that learning tasks may be inappropriate for investigating functional social discriminations where behavioural responses may conﬂict with performance of conditioned responses. This conclusion requires further investigation, but if it proves to be the case, then it may preclude the possibility of comparing abilities of domestic fowl to discriminate different levels of social category and the cues involved using a single method; i.e. establishing the relative ease of each cognitive task. Our study does, however, provide support for the use of live stimuli as relevant social cues given an appropriate context for task performance. Acknowledgements This study was funded by the Biotechnology and Biological Sciences Research Council. We are grateful to Mr. John Lowe for his technical aid, staff of the RVC Biological Services Unit for animal husbandry and an anonymous reviewer for constructive comments on the manuscript. References Abeyesinghe, S.M., Nicol, C.J., Hartnell, S.J., Wathes, C.M., 2005. Can domestic fowl, Gallus gallus domesticus, show self-control? Anim. Behav. 70, 1–11. Algers, B., Jensen, P., 1985. Communication during suckling in the domestic pig—effects of continuous noise. Appl. Anim. Behav. Sci. 14, 49–61. Bradshaw, R.H., 1991. Discrimination of group members by laying hens Gallusdomesticus. Behav. Process. 24, 143–151. Bradshaw, R.H., 1992a. Conspeciﬁc discrimination and social preference in the laying hen. Appl. Anim. Behav. Sci. 33, 69–75. Bradshaw, R.H., 1992b. The expression of agonistic behavior in groups of laying hens—a regression-analysis. Appl. Anim. Behav. Sci. 33, 63–68. Bradshaw, R.H., 1992c. Individual attributes as predictors of social-status in smallgroups of laying hens. Appl. Anim. Behav. Sci. 34, 359–363. Bradshaw, R.H., Dawkins, M.S., 1993. Slides of conspeciﬁcs as representatives of real animals in laying hens (Gallus-domesticus). Behav. Process. 28, 165–172. Cloutier, S., Newberry, R.C., 2000. Recent social experience, body weight and initial patterns of attack predict the social status attained by unfamiliar hens in a new group. Behaviour 137, 705–726. Collias, N.E., Collias, E.C., 1996. Social organization of a red junglefowl, Gallus gallus, population related to evolution theory. Anim. Behav. 51, 1337–1354. Craig, J.V., Biswas, D.K., Guhl, A.M., 1969. Agonistic behaviour inﬂuenced by strangeness, crowding and heredity in female domestic fowl (Gallus gallus). Anim. Behav. 17, 498–506. D’Eath, R.B., 1998. Can video images imitate real stimuli in animal behaviour experiments? Biol. Rev. 73, 267–292. D’Eath, R.B., Dawkins, M.S., 1996. Laying hens do not discriminate between video images of conspeciﬁcs. Anim. Behav. 52, 903–912. D’Eath, R.B., Keeling, L.J., 2003. Social discrimination and aggression by laying hens in large groups: from peck orders to social tolerance. Appl. Anim. Behav. Sci. 84, 197–212. D’Eath, R.B., Stone, R.J., 1999. Chickens use visual cues in social discrimination: an experiment with coloured lighting. Appl. Anim. Behav. Sci. 62, 233–242. Dawkins, M.S., 1995. How do hens view other hens—the use of lateral and binocular visual-ﬁelds in social recognition. Behaviour 132, 591–606. Dawkins, M.S., 1996. Distance and social recognition in hens: implications for the use of photographs as social stimuli. Behaviour 133, 663–680. Deng, C., Rogers, L.J., 2002. Social recognition and approach in the chick: lateralization and effect of visual experience. Anim. Behav. 63, 697–706. Dukas, R., 2002. Behavioural and ecological consequences of limited attention. Philos. Trans. R. Soc. B Biol. Sci. 357, 1539–1547. Estevez, I., Keeling, L.J., Newberry, R.C., 2003. Decreasing aggressions with increasing group size in young domestic fowl. Appl. Anim. Behav. Sci. 84, 213–218.
S.M. Abeyesinghe et al. / Behavioural Processes 81 (2009) 1–13 Estevez, I., Newberry, R.C., de Reyna, L.A., 1997. Broiler chickens: a tolerant social system? Etologia 5, 19–29. Freire, R., Appleby, M.C., Hughes, B.O., 1997. Assessment of pre-laying motivation in the domestic hen using social interaction. Anim. Behav. 54, 313–319. Gheusi, G., Bluthe, R.M., Goodall, G., Dantzer, R., 1994. Social and individual recognition in rodents—methodological aspects and neurobiological bases. Behav. Process. 33, 59–87. Grigor, P.N., Hughes, B.O., Appleby, M.C., 1995. Social inhibition of movement in domestic hens. Anim. Behav. 49, 1381–1388. Grosch, J., Neuringer, A., 1981. Self control in pigeons under the Mischel paradigm. J. Exp. Anal. Behav. 35, 3–21. Guhl, A.M., Ortman, L.L., 1953. Visual patterns in the recognition of individuals among chickens. Condor 55, 287–298. Guzman, D.A., Marin, R.H., 2008. Social reinstatement responses of meat-type chickens to familiar and unfamiliar conspeciﬁcs after exposure to an acute stressor. Appl. Anim. Behav. Sci. 110, 282–293. Hauser, J., Huber-Eicher, B., 2004. Do domestic hens discriminate between familiar and unfamiliar conspeciﬁcs in the absence of visual cues? Appl. Anim. Behav. Sci. 85, 65–76. Hughes, B.O., 1977. Selection of group size by individual laying hens. Br. Poult. Sci. 18, 9–18. Hughes, B.O., Wood-Gush, D.G.M., Jones, R.M., 1974. Spatial organization in ﬂocks of domestic fowls. Anim. Behav. 22, 438–445. Jones, J.B., Wathes, C.M., Persaud, K.C., White, R.P., Jones, R.B., 2001. Acute and chronic exposure to ammonia and olfactory acuity for n-butanol in the pig. Appl. Anim. Behav. Sci. 71, 13–28. Jones, R.B., Mills, A.D., Faure, J.M., 1996. Social discrimination in Japanese quail Coturnix japonica chicks genetically selected for low or high social reinstatement motivation. Behav. Process. 36, 117–124. Marin, R.H., Freytes, P., Guzman, D., Jones, R.B., 2001. Effects of an acute stressor on fear and on the social reinstatement responses of domestic chicks to cagemates and strangers. Appl. Anim. Behav. Sci. 71, 57–66.
Maynard-Smith, J., Harper, D.G.C., 1988. The evolution of aggression: can selection generate variability? Philos. Trans. R. Soc. B 319, 557–570. McLeman, M.A., Mendl, M., Jones, R.B., White, R., Wathes, C.M., 2005. Discrimination of conspeciﬁcs by juvenile domestic pigs, Sus scrofa. Anim. Behav. 70, 451–461. Mench, J., Keeling, L.J., 2001. The social behaviour of domestic birds. In: Keeling, L.J., Gonyou, H.W. (Eds.), Social Behaviour in Farm Animals. CABI Publishing, New York, pp. 117–209. Mendl, M., 1999. Performing under pressure: stress and cognitive function. Appl. Anim. Behav. Sci. 65, 221–244. Nicol, C.J., Pope, S.J., 1999. The effects of demonstrator social status and prior foraging success on social learning in laying hens. Anim. Behav. 57, 163–171. Pagel, M., Dawkins, M.S., 1997. Peck orders and group size in laying hens: ‘future contracts’ for non-aggression. Behav. Process. 40, 13–25. Patterson-Kane, E., Nicol, C.J., Foster, T.M., Temple, W., 1997. Limited perception of video images by domestic hens. Anim. Behav. 53, 951–963. Porter, R.H., Arnould, C., Simac, L., Hild, S., 2006. Retention of individual recognition in chicks and the effects of social experience. Anim. Behav. 72, 707–712. Timberlake, W., 1988. Evolution, behaviour systems and ‘self control’: the ﬁt between organism and test environment. Behav. Brain Sci. 11, 694–695. Wood-Gush, D.G.M., 1971. The Behaviour of the Domestic Fowl. Heinemann Educational Books Ltd., London. Zayan, R., 1994. Mental representations in the recognition of conspeciﬁc individuals. Behav. Process. 33, 233–246. Zayan, R., Lamberty, Y., 1988. Experimental evidence of social recognition in domestic fowl: learning or habituation? In: Blanchard, R.J., Brain, P.F., Blanchard, D.C., Parmigiani, S. (Eds.), Ethoexperimental Approaches to the Study of Behavior. Kluwer Academic Publishers, Dordrecht, Netherlands; Boston, pp. 691–702. Zimmerman, P.H., Lundberg, A., Keeling, L.J., Koene, P., 2003. The effect of an audience on the gakel-call and other frustration behaviours in the laying hen (Gallus gallus domesticus). Anim. Welf. 12, 315–326.