Stability and individual variability of social attachment in imprinting

Filial imprinting has become a model for understanding memory, learning and social behaviour in neonate animals. This mechanism allows the youngs of precocial bird species to learn the characteristics of conspicuous visual stimuli and display affiliative response to them. Although longer exposures to an object produce stronger preferences for it afterwards, this relation is not linear. Sometimes, chicks even prefer to approach novel rather than familiar objects. To date, little is known about how filial preferences develop across time. This study aimed to investigate filial preferences for familiar and novel imprinting objects over time. After hatching, chicks were individually placed in an arena where stimuli were displayed on two opposite screens. Using an automated setup, the duration of exposure and the type of stimuli were manipulated while the time spent at the imprinting stimulus was monitored across 6 days. We showed that prolonged exposure (3 days vs 1 day) to a stimulus produced robust filial imprinting preferences. Interestingly, with a shorter exposure (1 day), animals re-evaluated their filial preferences in functions of their spontaneous preferences and past experiences. Our study suggests that predispositions influence learning when the imprinting memories are not fully consolidated, driving animal preferences toward more predisposed stimuli.

www.nature.com/scientificreports/ and blue can be considered as predisposed imprinting stimuli. In our study, we used objects of different colours to investigate whether spontaneous preferences are steady or can change in time. Filial imprinting preferences have been well described 3,6 . However, how these preferences develop in time and vary depending on the animals' predispositions has been poorly documented. We know that longer exposure (from a couple of minutes to a few hours) produces stronger preferences for the imprinting stimulus (familiar stimulus) 4,20,43 . However, after imprinting, the preference for approaching familiar objects and avoiding novel objects is not merely steady nor incremental. On the contrary, in some situations, chicks prefer to approach novel rather than familiar objects, an unexpected behaviour. For instance, Bateson et al. 20,44 have observed that in the initial stage of imprinting-i.e. 15 and 30 min after the beginning of the imprinting phase but not after 60 min-chicks are motivated to be exposed to novel objects. More recently, the early shift from the first object to the exploration of alternative stimuli has been observed in different breeds of chicks that were tested on their spontaneous preferences to approach a stuffed hen versus a scrambled version of it. Versace et al. have shown that while in the first 5 min of visual experience, all breeds had a preference for the stuffed hen, 5 min later, one breed started to explore the other stimulus 45 . Interestingly, preferences for novel stimuli in imprinting appear also at much later stages 16,17 . In this paper, we focus on imprinting responses up to 6 days after hatching.
The longitudinal aspect of our study enables us to investigate the paradoxical phenomenon of the preference for unfamiliar imprinting objects. While it has been shown that exploration of novelty takes place at different stages of imprinting, how and why this counterintuitive phenomenon appears remains an open question. To date, the transient preference for unfamiliar stimuli, named 'slight-novelty preference' by Bateson, has been described and modelled as a phenomenon driven by the need to explore different points of view of the imprinting stimulus to build a full representation of it 46 . According to this hypothesis, the preference for exploring objects slightly different from the imprinting stimulus would help recognise different points of view of the mother hen and build a complete representation of it 6,44,47 . This hypothesis is supported by other studies showing that when two stimuli are presented in close temporality, they became "blended" as a unique stimulus for the animals 48,49 . However, the hypothesis that (only) in the first hour of exposure chicks explore novel stimuli to improve the imprinting object's representation has been confuted by behavioural 16,17 and physiological/biochemical studies 12,[50][51][52] . For instance, a novelty preference has been observed after the first day of imprinting. After 3 days of imprinting, chicks prefer novel visual patterns presented as a sequence of stimuli 16 or as simultaneous multimodal pattern 17 . Interestingly, sex differences have been observed 17 , with males preferring unfamiliar stimuli and females preferring familiar stimuli 53,54 .
Little is known about individual differences in imprinting behaviour. Templeton and Smith described that chicks' response to an effective stimulus varied across a wide range of performance and was not affected by genetics 55 . Gribosvkiy et al. 56 developed a quantitative methodology to study the inter-individual variability among chicks in imprinting and showed high variability between individuals and behavioural types. Chicks with higher behavioural flexibility had a stronger preference for novelty in a generalisation task after conditioning 57 . Little is known whether individual differences apply to imprinting, mainly due to the difficulties in tracking and analysing imprinting behaviour at the individual level across time. To overcome this difficulty, we use automated behavioural tracking techniques [58][59][60][61] , studying individual imprinting preferences for multiple consecutive days.
We built an automated setup to continuously track chicks' behaviour from the first exposure to the imprinting stimuli for six consecutive days. Chicks were individually housed in an arena with two opposite monitors. The imprinting and test stimuli position was counterbalanced between monitors while we measured the distance of chicks from the stimuli. Imprinting duration and testing duration were manipulated. Objects of different colours were used as imprinting objects to investigate the role of colour on chicks preferences. In Experiment 1, chicks were imprinted for 1 day with one stimulus and tested for 5 days with two stimuli. In Experiment 2, the imprinting duration was increased to 3 days and chicks were tested for 3 days. In Experiment 3, chicks were imprinted with one object for 1 day, then with another one for 2 days and tested for 3 days. In Experiment 4, we replicated a similar procedure than Experiment 3, but this time assessed the animal preference between their primary or secondary imprinting object. In such settings, with prolonged and continuous behavioural monitoring, we investigated how filial preferences developed in time at the group and individual level.

Results
Experiment 1. Imprinting. There were non-significant effects of Condition (imprinted with a green hourglass or imprinted with a blue cube; F(1, 28) = 0.57, p = 0.46), Sex (F(1, 28) = 0.18, p = 0.67) or interaction Sex × Condition, F(1, 28) = 0.14, p = 0.71) on the time spent close to the imprinting stimulus. The chicks significantly remained closer to the imprinting stimulus (t(31) = 83.25, p < 0.001, Cohen's d = 14.72), spending 96% of their time (± 0.56 SEM) close to the imprinting stimulus. All chicks (32) remained significantly more on the side of the arena in which the imprinting stimulus was displayed (Table S1 in the Supplementary Material). In the blue condition, 10 chicks (63%) had a significant preference for the imprinting stimulus, 5 (31%) had no preference, and 1 (6%) significantly preferred the unfamiliar stimulus. In the green condition, 7 chicks (44%) had a significant preference for the imprinting stimulus, 6 (37%) had no preference, and 3 (19%) had a significant preference for the unfamiliar stimulus (Table S1 in   In the blue condition, 14 chicks (87.5%) had a significant preference for the imprinting stimulus, 2 (12.5%) had no preference, and none significantly preferred the unfamiliar stimulus. In the green condition, 10 chicks (62.5%) had a significant preference for the imprinting stimulus, 4 (25%) had no preference, and 2 (12.5%) had a significant preference for the unfamiliar stimulus (Table S2 in the Supplementary Material). Levene's test showed that the variances of the two conditions were significantly different (F(1, 30) = 6.14, p < 0.05). Chicks imprinted with the green stimulus showed higher variability in their preferences for the imprinting stimulus during testing (σ 2 = 380.85) than chicks imprinted with the blue stimulus (σ 2 = 129.91). All chicks (33) remained significantly more on the side of the arena, where the primary imprinting stimulus was displayed (Table S3 in   In experiment 1, the preference for the imprinting stimulus was stable stable across days for the chicks imprinted with the blue stimulus (blue line) but not with the green stimulus (green line) (p < 0.05, *; p < 0.01, **; p < 0.001, ***). Green asterisks represent the statistical significance of the group of chicks imprinted with the green stimulus. The blue asterisks represent the statistical significance of the group of chicks imprinted with the blue stimulus.  The preference for the primary imprinting stimulus was significantly different from chance level for the chicks imprinted with the blue stimulus (t(16) = 12.27, p < 0.001, Cohen's d = 2.98, Bonferroni correction) with an average time spent close to the primary imprinting stimulus of 83% (± 2.66 SEM). The Preference score was non-significantly different from chance level for the chicks imprinted with the green stimulus (t(15) = − 1.94, Figure 2. In experiment 2, the preference for the imprinting stimulus was stable across days for both conditions (p < 0.001, ***). The blue dots represent the preference score of the chicks imprinted with the blue stimulus. The green dots represent the preference score of the chicks imprinted with the green stimulus. Filled dots show the individuals having a significant preference while empty dots show the individuals having no preference. The asterisks represent the overall significance of both conditions pooled, against no preference. In experiment 3, the object used during the primary imprinting phase strongly influenced chicks' preference for it when exposed to a novel stimulus (p < 0.001, ***). The blue boxplot represents the preference score of the chicks imprinted with the blue stimulus. The green boxplot represents the preference score of the chicks imprinted with the green stimulus. Filled dots show the individuals having a significant preference while empty dots show the individuals having no preference. The asterisks represent the overall significance of each condition, against no preference and between conditions.  (17) had a significant preference for the imprinting stimulus while primary imprinted with the blue stimulus (Table S3 in the Supplementary Material). Whereas for the chicks primarily imprinted with the green stimulus, 2 (13%) had a significant preference for their primary imprinting stimulus, 6 (37%) had no preference and 8 (50%) had a preference for the unfamiliar stimulus (Table S3 in   Individual preferences were calculated and showed that 32 (97%) chicks remained significantly more on the side of the arena, where the primary imprinting stimulus was displayed, and 1 (3%) did not (Table S4 in  In the blue condition, 1 chick (7%) had a significant preference for the primary imprinting stimulus, 9 (60%) had no preference, and 5 (33%) significantly preferred the unfamiliar stimulus. In the green condition, 2 chicks (13%) had a significant preference for the primary imprinting stimulus, 9 (56%) had no preference, and 5 (31%) had a significant preference for the unfamiliar stimulus (Table S4 in   www.nature.com/scientificreports/

Discussion
Due to the difficulties in assessing animals behaviours over prolonged durations, the temporal stability and individual variability of social attachment in filial imprinting have remained unexplored. To understand more about it, we used an automated behavioural tracking method and followed the animals' preferences for familiar and novel stimuli for 6 consecutive days. The temporal stability of the imprinting preferences was investigated by manipulating the duration of the imprinting and the stimuli used. When imprinted for 14 h over 1 day (Experiment 1), the chicks exhibited an unsteady preference for their imprinting stimulus compared to when exposed for 42 h over 3 days (Experiment 2). In fact, after 1 day of imprinting, the filial preferences were disparate between conditions. While the chicks of the blue condition always had a preference for their imprinting stimulus at testing, the chicks of the green condition lost their significant preference for the imprinting stimulus on the fourth testing day. They started to explore more the unfamiliar stimulus (blue stimulus). Since we know that chicks mainly rely on colour to recognise their artificial imprinting objects 38 , this difference confirms previous reports of an advantage of blue over green imprinting stimuli [40][41][42] . In contrast, after 3 days of imprinting, chicks of both conditions had a robust and stable preference for their imprinting objects. Moreover, we excluded the possibility that the difference observed was affected by the time spent close to the imprinting stimuli by showing that bot conditions spent the same amount of time close to their respective stimulus during the imprinting phase. The preference observed in Experiment 1 for the imprinting stimulus across days in the blue condition and on the first testing day of the green condition indicated that chicks imprinted on their respective stimuli. Nevertheless, 14 h of imprinting are insufficient to produce a robust and stable imprinting preference for artificial stimuli. The unlearned preferences are influencing the animals' filial preferences. Therefore, the decrease of preference for the imprinting stimulus in the green condition suggests that the blue stimulus is more attractive to the chicks. Due to repeated testing, secondary imprinting with the blue stimulus might take place in the green condition. This would explain why the animals spend more time close to it rather than a general lack of memory (the preference is stabilised at chance-level in the green condition and blue-imprinted chicks steadily remembered and prefered the imprinting stimulus). The difference between blue and green-imprinted chicks is apparent also looking at the individual performances. More than half of the chicks had a preference for the imprinting stimulus, and only 6% had a preference for the novel stimulus in the blue-imprinted chicks. In contrast, only less than half of the chicks had a preference for the imprinting stimulus, and 19% preferred the novel stimulus in the green-imprinted chicks.
Several biochemicals changes associated with imprinting have been described later than 15 h after the start of the imprinting process, confirming the idea that imprinting might not be fully consolidated on the first day of exposure 12 . Furthermore, the mechanisms responsible for the spontaneous preferences observed in chicks strongly influence the imprinting memory 25,26 . In Experiment 1, it seems that after 14 h of exposure to a stimulus, the imprinting memories are available but not fully consolidated yet. The preferences also seem more plastic after imprinting with less predisposed stimuli.
Hence, because the same experience produces different learning outcomes, it appears that predispositions affect both learning and the between-subjects variability in learning, with faster and stronger learning and less variability when subjects are exposed to predisposed stimuli.
The analysis of individual behaviours revealed that some chicks had consistent preferences for unfamiliar stimuli not only at the very beginning of imprinting, as hypothesised by Bateson's model 46 . By increasing the chicks' exposure to their imprinting objects to 42 h over 3 days, we observed more robust and stable filial preferences with time for both stimuli (Experiment 2) but still a higher inter-individual variability within the green-imprinted chicks. These results are in line with previous experiments in which preferences for unfamiliar objects have been observed even after 3 days of imprinting in males 16,17 . As stated in the introduction, males are more inclined to approach unfamiliar conspecifics than females, usually showing a strong attachment to their conspecifics 53,54 . In this study, the filial preference was similar in both sexes.
More prolonged imprinting exposure has been associated with stronger preference scores for the imprinting stimulus 4,20 . Furthermore, our study suggests that the imprinting duration strongly influences the filial preference steadiness. After 42 h over 3 days of exposure to an object, the imprinting memory appears to be consolidated for both artificial stimuli (green and blue). Nonetheless, animals' spontaneous preferences for specific stimuli are still, to a lower degree, influencing chicks' filial preferences. The variability within the green condition (less predisposed colour) was three-time higher than in the blue condition. While almost all chicks showed a strong preference for their imprinting objects in the blue condition, more than a third did not prefer their imprinting stimulus in the green condition.
The evidence that prolonged exposure to an object leads to more stable preferences in time is convincing and in line with previous evidence 20,43 . Nevertheless, the ontogenetic stage at which the preferences were tested could have influenced filial preferences. In the third experiment, we assessed whether this was the case. As in the first experiment, both conditions (blue and green) were exposed to their respective objects for 14 h (day 1), but this time, their filial preference was tested from day 4 to day 6, after exposure to a novel object on day 2 and 3 (this prevented a complete 'social' deprivation). Similarly to what observed in the first experiment (short imprinting duration), the filial preferences observed differed between conditions. In the blue condition, all the individuals preferred their imprinting object, showing that the memory of the imprinting stimulus lasted although chicks had been detached by the initial stimulus for days. At the same time, preferences among individuals of the green conditions were disparate with 13% of the individuals preferring the imprinting object, 37% showing no preferences and even 50% showing a preference for the novel object. Interestingly, the preferences observed here were not wholly similar to the first experiment. The preferences observed in both conditions were stable in time. Then, one could argue that the filial preferences observed resulted from a lack of memory, but the different patterns of preference between conditions and the literature suggest otherwise. In the case of a memory loss, chicks would have either approached the more attractive stimulus (blue object) or not chosen any. However, the www.nature.com/scientificreports/ results showed both patterns depending on the primary imprinting stimulus used. Moreover, studies exploring successive imprinting always described a recall of the primary imprinting object 21,62 .
In Experiment 4, we assessed whether chicks had a preference for their primary imprinting stimulus compared to their secondary imprinting stimulus during the testing phase. Both conditions showed a similar preference for the secondary imprinting stimulus. As previously shown, chicks can imprint on multiple objects 22,23 . Furthermore, a preference for a primary imprinting stimulus can be reversed after prolonged exposure with a secondary imprinting object 63 , which is in line with the experimental settings used here (1 day of primary imprinting and 2 days of secondary imprinting). It is then very likely that the filial bond formed with the secondary imprinting object has influenced the chicks' filial preferences toward their primary imprinting stimulus.
In all experiments, the filial imprinting preferences were all pointing in the same direction. Overall, chicks of the blue condition (where blue is a more predisposed colour) had a more robust and stable preference in time for their imprinting stimulus than the chicks of the green condition (where green is a less predisposed colour). The differences between conditions were not the result of the time spent close to their respective objects during imprinting, given that chicks engaged with the imprinting stimuli for the same amount of time. This strongly suggests that some features of the objects (e.g. colour) are more efficient for forming filial imprinting preferences. Further studies should be performed to understand the influence of colour in comparison to shape.
Altogether, our results indicate that the temporal stability of filial imprinting preferences is influenced by the amount of experience (exposure duration and successive imprinting) and spontaneous preferences (predispositions). Moreover, using automated tracking for assessing chicks' behaviour for several days, we show that chicks with similar experiences can have steady and robust idiosyncratic differences in their preferences for familiar vs novel stimuli. Some chicks consistently preferred to approach their imprinting stimulus, while others preferred the unfamiliar stimulus, even if they had the same experience. Moreover, this consistent inter-individual variability (a phenomenon already documented in other animal species, such as fruit flies) 59,64-68 was modulated by the animals' spontaneous preferences. Further studies should clarify whether these differences stem from genetic variability and/or derive from stochasticity in the course of development 69 , as well as their neurobiological basis.

Materials and methods
Ethical note. This study was carried out in compliance with the European Union and the Italian law on the treatment of animals. The experimental procedures were performed in accordance with the ARRIVE guidelines and approved by the Ethical Committee of the University of Trento and licenced by the Italian Health Ministry (permit number 53/2020).

Subjects.
We used 128 domestic chicks (Gallus gallus) of the strain Ross 308 (a strain selected to be sexually dimorphic at birth, based on the feathers). The eggs were coming from a commercial hatchery (Azienda Agricola Crescenti) and were incubated at the University of Trento under standards controlled conditions (37.7 °C and 40% of humidity). Three days before hatching eggs were transferred into opaque individual boxes within a hatching chamber (37.7 °C and 60% of humidity).
Setup. Several apparatuses were used simultaneously. Each apparatus had a rectangular shape (90 cm × 60 cm × 60 cm, Fig. 5). A high-frequency computer screen (ASUS MG248QR, 120 Hz) was located on each smaller wall and used to display stimuli. A Microsoft life camera was located on the top of the apparatus at 105 cm from the ground to record the behaviours of the animal. Food and water were located in the middle of the apparatus and available ad libitum.
Stimuli. Three-dimensional virtual visual stimuli were created (Fig. 5) and animated on Blender (v2.79). The objects were different in term of colours and shapes (green hourglass, hex: 30B619; blue cube, hex: 2EBAFF; pink cylinder, hex: C33CDB) but had similar sizes (5 cm × 5 cm, Fig. 5). The stimuli were animated (linear movement) in a 3D environment and were crossing the screen in 4.5 s (from left to right). The video displaying the stimuli was exported with a high frame frequency (120 frames per second, fps).
General procedure. After hatching, chicks were sexed in darkness and were transported in another room and individually placed in their apparatus for 6 days in a day-night cycle (14:10 h). During the day, the chicks were exposed to the stimuli displayed on the screens. The displaying of the stimuli was divided into different sessions depending on the experimental phase (from 2 h to 30 min). The position of the stimuli on the screens was counterbalanced across sessions. During the night, dark screens were displayed. Four different experiments were performed. Each experiment was divided into 2 or 3 distinct phases (primary imprinting, secondary imprinting and testing) and conditions (blue and green). The duration of each phase was manipulated from one experiment to another. Chicks were donated to local farms at the end of the experiment.
Primary imprinting. This phase was the first one of each experiment. The chicks were exposed to a single imprinting stimulus (the blue or the green depending on the condition). The imprinting sessions lasted 2 h (7 sessions) on the first primary imprinting day and 1 h on the following days (13 sessions interrupted by 5 min period of dark screens).
Secondary imprinting. In Experiments 3 and 4, this phase followed the primary imprinting phase and lasted 2 days. The chicks were exposed to a new stimulus (a pink cylinder). The sessions were lasting 1 h (13 sessions interrupted by 5 min period of dark screens). www.nature.com/scientificreports/ Testing. Depending on the experiment, the testing phase was either following the primary (Experiment 1 and 2) or secondary imprinting phase (Experiment 3 and 4). The chicks were exposed to two stimuli (primary imprinting stimulus vs novel stimulus or primary vs secondary imprinting stimulus), and their preferences were monitored. The sessions lasted thirty minutes (24 sessions interrupted by 5 min period of dark screens between each session). Experiment 1. Chicks were exposed to an imprinting stimulus for 1 day (blue or green stimulus depending on the condition) and then tested with two stimuli (imprinting stimulus vs unfamiliar stimulus) for 5 days (Fig. 6A).
Subjects. We imprinted 16 animals (8 females, 8 males) with the green hourglass (green condition) and 16 animals (8 females, 8 males) with the blue cube (blue condition). Experiment 2. Chicks were exposed to an imprinting stimulus for 3 days (blue or green stimulus depending on the condition) and then tested with two stimuli (imprinting stimulus vs unfamiliar stimulus) for 3 days (Fig. 6B).
Subjects. We imprinted 16 animals (8 females, 8 males) with the green hourglass (green condition) and 16 animals (8 females, 8 males) with the blue cube (blue condition). Experiment 3. Chicks were exposed to a primary imprinting stimulus for 1 day (blue or green stimulus depending on the condition), secondary imprinting stimulus (pink stimulus) for 2 days and then tested with two stimuli (primary imprinting stimulus vs unfamiliar stimulus) for 3 days (Fig. 6C).
Subjects. We imprinted 16 animals (8 females, 8 males) with the green hourglass (green condition) and 16 animals (8 females, 8 males) with the blue cube (blue condition). Experiment 4. Chicks were exposed to a primary imprinting stimulus for 1 day (blue or green stimulus depending on the condition), secondary imprinting stimulus (pink stimulus) for 2 days and then tested with two stimuli (primary imprinting stimulus vs secondary imprinting stimulus) for 3 days (Fig. 6D). deep-learning toolbox made to track efficiently animal behaviours 61 . The preference for a stimulus was assessed using the time spent inside the closest zone to it (30 cm wide). The apparatus had been virtually divided into three equal zones corresponding to the left, centre and right side of each arena (Fig. 5).
Imprinting phases. During these phases (primary and secondary), the number of seconds [s] spent close to the stimulus (in the 30 cm zone close to the screen) was analysed to check for the amount of time spent attending the imprinting object.
Testing phase. For this phase, the Preference for the imprinting stimulus [%] was calculated using the following formula: Preference for the imprinting stimulus = Time spent close primary imprinting stimulus Time spent close both screens ×100. www.nature.com/scientificreports/