Novel object trials are commonly used to assess aversion to novelty (neophobia), and previous work has shown neophobia can be influenced by the social environment, but whether the altered behaviour persists afterwards (social learning) is largely unknown in wild animals. We assessed house sparrow (Passer domesticus) novel object responses before, during, and after being paired with a conspecific of either similar or different behavioral phenotype. During paired trials, animals housed with a similar or more neophobic partner demonstrated an increased aversion to novel objects. This change did not persist a week after unpairing, but neophobia decreased after unpairing in birds previously housed with a less neophobic partner. We also compared novel object responses to non-object control trials to validate our experimental procedure. Our results provide evidence of social learning in a highly successful invasive species, and an interesting asymmetry in the effects of social environment on neophobia behavior depending on the animal’s initial behavioral phenotype.

2.1 Subject capture & housing

            We captured 24 house sparrows (12 males, 12 females) in Iberville and Lafourche Parishes, Louisiana, USA, using mist nets near bird feeders in November 2019. We individually housed sparrows in cages in a vivarium at Louisiana State University with unlimited access to mixed seeds, grit, a vitamin-rich food supplement (Purina Lab Diet for small birds), and water. Light cycle was synchronized with capture date (10.5L:13.5D) and maintained at this level for the duration of the experiment. Sparrows were visually but not acoustically isolated from one another to prevent observations of their neighbours’ novel object trials. Sparrows received an aluminum band with a unique number and were acclimated to captivity for 8-10 weeks until novel object trials began in mid-January.

2.2 Study design

            This study took place over five weeks. Week 1, test individual phenotypes: We exposed individual house sparrows to three randomly selected novel objects and two control days (no object) in a randomized order. Week 2, assess individual phenotypes and create pairs: We observed recordings for each sparrow’s time to feed from its food dish. Based on average feeding time, we created house sparrow pairs that were similar in approach time (matched phenotype controls; 5 pairs; Table S1 in Supplementary Material). The remaining birds were paired with a sparrow with a different phenotype (mixed phenotype pairing; 7 pairs; Table S1). We introduced pairs to one another in a new cage to reduce the possibility of territoriality. One individual from each pair received a second aluminum band to distinguish individuals in recordings. We acclimated pairs together for three days before continuing novel object trials. Week 3, paired trials: We exposed pairs to three randomly selected objects, novel to both individuals, and two control conditions in a random order. Week 4, individual housing acclimation: We separated pairs and returned individuals to their original cage during the fourth week to allow acclimation before beginning the final novel object exposures. Week 5, re-test individual phenotypes: We exposed sparrows to three new novel objects and two control conditions in a random order to test whether having a cage mate had lasting effects on neophobia behaviour (evidence of social learning).

2.3 Novel object exposure procedure

Over the five weeks, we exposed sparrows to a total of 15 trials, six controls and nine unique novel objects: a blinking light, a white cover over part of the dish, yellow pipe cleaners, a purple plastic egg, a red-painted dish, a tinfoil hood, three gold bells, pink puffs, and an opened blue cocktail umbrella (images in Supplementary Material). Objects were chosen to maximize the diversity of textures, colours, and shapes of novel objects, and all objects were placed on, in, or immediately above the food dish. Sparrows saw each object once. These objects were based on objects shown to elicit delayed approach and feeding in another songbird species [32], but altered slightly to share fewer common features (e.g., red color) that might incidentally target ecologically-relevant cognitive biases [33].

The night before a trial, we removed food dishes from cages 30 min before lights off and fasted sparrows overnight. We replenished seed in the dishes and outfitted them with an object (or no object, for controls). The next morning, 30 min after lights on, researchers entered the room, began a video recording of all cages (12 cameras, 2 birds per camera), replaced food dishes according to individual treatments, and left for 1 h. Researchers were present in the room < 4 min. Sparrows were therefore fasted for a total of 15 h (1 h light, 14 h dark) before neophobia or control treatments. This fasting period ensured that sparrows would be motivated to approach the food dish. Upon re-entering the room, the video recording was stopped, and objects removed from food dishes. Videos were scored for the time each sparrow took to eat from the food dish. Two different observers scored all videos. To ensure intercoder reliability, these observers re-watched 14 videos from paired trials (n = 28 feeding events) one month later, and ANOVA tests showed no significant differences within (F_2,81<0.001; p = 0.99) or between (F_1,54<0.001, p=0.99) observers.

2.4 Statistical approach

We split sparrows into three categories: (1) paired with an individual of similar neophobia response (control; n = 10), (2) paired with a less neophobic individual (partner faster to feed in the presence of novel objects; n = 7), (3) paired with a more neophobic individual (partner slower to feed in the presence of novel objects; n = 7; Table S1). See Supplementary Material for datapoint losses. We used Cox proportional hazard models to investigate the mere effect of having a cage mate (1) as well as having a cage mate with a different neophobia phenotype (2 and 3) separately using the ‘coxme’ package [34] in R Studio version 3.6.3 [35]. Using a survival analysis approach [36] avoids creating arbitrary threshold values when a subject does not perform the expected behaviour during the allotted time period, which may bias alternative statistical approaches – i.e., giving subjects a time of 3600 s if they do not feed during a 60 min trial. Each model included subject as a random effect, week as a fixed effect and were interpreted as full models. Analyses were performed for control trials (to detect effects of the experimental procedure) and novel object trials (to detect neophobia effects) separately, resulting in six models. Novel object trials also included object as a fixed effect. Because effects of neophobia were detected in models run using novel object trials only, the contrasts among objects were arbitrarily against the object coded as ‘1’, so we do not report these contrasts. Instead, to validate that each object significantly increased latency to feed, we ran a seventh model that included all trials (all individuals, control, and object trials). This seventh model included subject as a random effect and object (or not) as a fixed effect.

References:
[32] de Bruijn R, Romero LM. 2019 Prior restraint stress inhibits habituation to novel objects in the European starlings (Sturnus vulgaris). J. Exp. Zool. Part A Ecol. Integr. Physiol. , doi:10.1002/jez.2327. (doi:10.1002/jez.2327)
[33] Greggor AL, Thornton A, Clayton NS. 2015 Neophobia is not only avoidance: improving neophobia tests by combining cognition and ecology. Curr. Opin. Behav. Sci. 6, 82–89. (doi:10.1016/j.cobeha.2015.10.007)
[34] Therneau T. 2020 coxme: mixed effect cox models. , URL: https://CRAN.R-project.org/package=coxme. (doi:10.1017/CBO9781107415324.004)
[35] R Core Team. 2020 R: A language and environment for statistical computing. R Found. Stat. Comput. , Vienna, Austria. , URL https://www.R-project.org/.[36] Stankowich T, Coss RG. 2007 The re-emergence of felid camouflage with the decay of predator recognition in deer under relaxed selection. Proc. R. Soc. B Biol. Sci. 274, 175–182. (doi:10.1098/rspb.2006.3716)

Additional information about animal husbandry

Sparrows were provided with a variety of perches and dust baths as enrichment. On the rare occasions that animals escaped from cages in the bird room when cage doors were opened during animal husbandry or neophobia tests, they were immediately and safely recaptured using a handheld net and returned to their housing. Animals were solo housed for part of this study, but all birds were housed in the same room and birds could hear each other. Because we were interested in individual animal personalities and how personality can change based on social learning, it was necessary to first test the animals individually before testing them in a group (paired) context. We also wished to understand if changes in personality persisted outside of the group context. Solo housing was therefore absolutely essential for the design of our study. Because this was only a behavioral study and did not involve any physiological manipulations, we requested and received permission from our university’s Institutional Animal Care And Use Committee (IACUC) to transfer the animals to another protocol after the published project was finished, so as to minimize animal numbers required for our research. Animals were therefore not euthanized at the end of this study. However, one control bird died of natural causes during week 4 (see Datapoint losses below). The age of wild sparrows brought into captivity is not known beyond “juvenile” and “adult,” nor is it known whether sparrows have any chronic illnesses or parasites, which are common in wildlife. Therefore, the occasional death due to natural causes in wild birds is expected. 

Datapoint Losses:

Control Pairs: One individual escaped during object exposure during week 3 paired trials. We removed these datapoints for both birds (control pairing; week 3 object trials n = 28). Week 5 (unpaired) novel object videos were lost (three trials) for one control bird (control pairing; week 5 object trials n = 27, week 5 control trials n = 18).

Mixed Pairs: Due to a camera failure, one trial was lost during week 3 (paired trials; n = 20).

Guide to datafiles:

Files are split by pairing type and by control or object trials.

Pairing Effects:

Control Trials (no object presented)
R Code:                       “R Code week comparisons by pairing_control trials only.R”
Data:   [control pair]   “Approach times_controls all weeks_control trials only.csv”
            [more neo]      “Approach times_more neophobic all weeks_control trials only.csv”
            [less neo]        “Approach times_less neophobic all weeks_control trials only.csv”

Object Trials
R Code:                       “R Code week comparisons by pairing_no control trials.R”
Data:   [control pair]   “Approach times_controls all weeks_no control trials.csv”
            [more neo]      “Approach times_more neophobic all weeks_no control trials.csv”
            [less neo]        “Approach times_less neophobic all weeks_no control trials.csv”

Object Effects (seventh model):

R Code:                       “R Code all weeks all phenotypes for object effects.R”
Data:                           “Approach times_all phenotypes all weeks.csv”

Hazard Ratio 95% CI calculations:

“Calculating Hazard Ratio 95CIs.xls”