Generalization refers to the process by which animals assign different stimuli into cognitive categories based on their similarity/dissimilarity to previously experienced stimuli. Stimuli which strongly deviate from previous experience may be avoided due to their novelty (i.e., neophobia), while stimuli which are sufficiently similar to a known/recognized cue may elicit the same response as that cue (i.e., generalization). While generalization has been widely researched, few studies have examined among-individual variance in its expression. Our study quantified among-individual variation in neophobia and generalization in house sparrows (Passer domesticus) by placing a series of different objects next to each subject’s food source on successive days and measuring the subject’s latency to approach the food source in the presence vs. absence of each object. Similarly to previous work, we found that house sparrows are neophobic on average and exhibit significant among-individual variance in neophobia. More interestingly, we found that the neophobic response declined across presentations of different novel objects. This implies that the sparrows generalized some aspect(s) of the objects. We also found significant among-individual variance in the rate at which approach latency changed across this series, possibly reflecting individual differences in propensity to generalize. These results raise new questions about how neophobia, habituation, and generalization are linked and about the potential for selection to act on these traits under different ecological conditions.

Subjects and acclimation

We conducted experiments on 10 female and 8 male house sparrows (Passer domesticus) at the University of Kentucky’s Ecological Research and Education Center (38°4'52", -84°28'14”). We captured these subjects using mist nets from two sites in Lexington, KY: a backyard in a suburban housing development (38°0'28", -84°30'39") and the seed warehouse of a farm store in an industrial/commercial area (38°0'52", -84°26'29"). Eleven of the subjects (5 female, 6 male) we captured between January and March 2017 and assayed between May and September 2017, while the other seven (5 female, 2 male) we captured and assayed during the same time frames in 2018. All subjects were developmentally adults (i.e. after hatch year), but their exact ages were unknown. The choice of 18 subjects was based on practical constraints including the amount of available aviary space and the time needed to conduct a full set of behavioral assays; no a priori sample size calculation was conducted. No captured subjects were excluded from the experiment, nor did we have any a priori criteria for doing so. One captured subject (a sixth female captured in 2017) escaped from the experimental aviary before the completion of the full set of behavioral assays and was thus excluded from the final dataset/analysis.

Following capture, subjects were tagged with a unique combination of colored plastic leg bands for visual ID; these are standard for birds and do not impede normal behavior. Between capture and commencement of behavioral assays, we held the subjects in mixed-sex flocks in outdoor aviaries (3.7 x 2.4 x 2.4m), each having a wooden frame with 1 x 1 cm wire mesh enclosing it on all sides except for the ends, which were solid plywood. Each aviary contained a tree and two elevated platforms (attached to the back wall) on which we placed dishes of water and food (a mix of 3 parts white millet and 1 part chicken starter mix), both provided ad libitum. For the behavioral assays, we moved each subject to one of four “experimental aviaries,” identical to the housing aviaries, where it stayed for the duration of the assays. Experimental aviaries had six platforms, four on one wall and two on the opposite wall. Each of these four platforms contained a specialized feeding board on an aluminum tray; under one board was a balance used to obtain subject mass. For symmetry, we placed wooden spacers that matched the height of the scale under the other three boards. The two platforms on the opposite wall contained a water dish (bottom platform) and a camera tripod for filming experimental trials (top platform).

After moving each subject from the group holding aviary to the experimental aviary, we gave it a variable period of two to five days to acclimate to its new surroundings with minimal disturbance (with the exception of the first two subjects tested, which had 33 acclimation days). The number of acclimation days is included in the data file. After acclimation, we trained each subject for the behavioral assays as described below.

Training protocol

The novel object presentations which comprise this data set were embedded in a set of trials designed to assess other behaviors, addressing other questions unrelated to neophobia and generalization. Our full trial set required the birds to be trained to find food on the feeding boards, each of which contained eight wells. A fully trained subject would find food by flipping a lightly weighted paper lid off of a well and digging through sand to find millet seed that was buried there. Subjects experienced sand and lids from the moment they entered the aviary, but food was slowly buried in sand, and the sand was covered by the lids over a 4-day period. We began the full sequence of behavioral assays on the 5th day after training started.

Behavioral assays

The first two days of behavioral assays involved assessing subjects’ preference for particular boards given a difference in the distribution of seed among wells on each board. This distribution was continued through all trials, including those analyzed here. In addition, subjects’ feeding boards were refilled four times per day, at about 8am, 12pm, 3pm, and 6pm. The third day marked the beginning of our neophobia trials, for which we modified the feeding schedule to include a 1hr deprivation period (in which all four boards were completely removed from the aviary) prior to the 12pm and 3pm feedings. This deprivation functioned to increase and roughly standardize the subjects’ feeding motivation during the subsequent trials. At the 12pm (“control”) trial, we reintroduced the (refilled) food boards as usual and filmed the subject’s response over the following hour. At the 3pm (“novel object”) trial, we reintroduced the boards with a novel object placed immediately next to the same corner (front right from camera view) of each board. We then filmed the subject’s response over the following hour. The timing/order of the control and novel object trials were the same for each subject because our goal was to uncover pre-existing individual differences in the processing of cues, and this requires reducing differences in stimuli received during the experiment. 

We repeated this schedule over the following three days, using the first four (different) novel objects from a sequence of eight. We constructed these novel objects using Mega Bloks® such that each object was of a similar size but differed in shape and/or color. Each subject experienced one of three predetermined orders of objects. To enable the assessment of individual differences (as with trial order) it would have been ideal for each subject to experience the same object presentation order. However, practical constraints on time and materials prevented this, so we account for this potential confound in our data with the variable “object order.”

After the four neophobia assays, we devoted the following three days to further assays of board preference which included two days of reduced food across all three boards with food. Following this was a recovery day during which birds had access to their boards as before but were disturbed only when the boards were changed. We then repeated the novel object trials as before but with the remaining four objects in the sequence. Each subject therefore experienced a total of eight neophobia trials, with a six-day interval between the fourth and fifth trials. Upon completion of this second round of trials, we released our subjects back into the wild.

Video scoring

We scored video sessions by recording each subject’s “latency to land,” defined as the time in seconds between (a) the door being closed behind the last board placed in the aviary and (b) the first time that the bird had both feet touching down on any part of a feeding platform, board, or aluminum tray. All video recordings were started before the boards were placed so that this could be timed precisely. If a bird did not land on any part of a platform, board, or tray during the approximately 1-hour session, its latency to land was recorded as the time between door closure and the end of the video (N = 13 of 288 trials, 3600–3900 seconds). All videos were scored by A.L. McLaughlin. Because latency to land (as defined) was relatively objective and easy to observe, intercoder reliability was not measured, and scoring was not blind with respect to subject ID or novel object presence/ID (which would be difficult to accomplish given that both are readily visible in the videos). 

Data processing

Latency to land was log-transformed for all statistical analyses to reduce its skewness and treated as a Gaussian variable. Details of our statistical analysis (conducted in R version 3.5.1 and SAS 9.4®) may be found in the associated manuscript (McLaughlin and Westneat in press, Ethology manuscript number ETH-22-0172). 

Microsoft Excel is required to open our main data file (“Novelty_Data_DRYAD_version.xlsx”); open-source alternatives are Google Sheets and Microsoft Office Excel Online. Our ReadMe file (“ReadMe.txt”) may be opened in any text editor.