Evidence suggesting that desire-state attribution may govern food sharing in Eurasian jays

Subjects.

Data from seven male-female Eurasian jay pairs were collected in the breeding seasons 2010–2012 (March–June). The experiments were approved by the University of Cambridge and conducted under the UK Home Office Project Licenses PPL 89/1975 and PPL 80/2519. Pairs included 14 Eurasian jays (7 males, 7 females) from two colonies (colony 1: n = 8, 4 y old; colony 2: n = 6, 3 y old). One male had different female partners between 2010 and 2011. The second female partner participated solely in conditions in which data were only collected for males. His partner from 2010 repaired with another, previously untested, male and this pair was tested in 2011. For the purpose of the analyses, data from the males in these two pairs are considered independent, whereas an average across 2010 and 2011 was calculated for all data concerning the female.

Colonies 1 and 2 were housed in two separate outdoor aviaries (20 × 6 × 3m) with indoor testing compartments (2 × 1 × 2 m) at one end that the birds could access from the aviary via opaque flap doors (0.5 × 0.5 m), which were opened and closed by the experimenter. Birds were fed a maintenance diet of soaked dog biscuits, cheese, seeds, nuts, and fruit and had ad libitum access to water.

Procedure.

Testing took place in the compartments attached to the aviary. Each of our three experiments consisted of a prefeeding phase and a test phase (Fig. 1 A–C) and was preceded by ∼2 h during which birds had no access to their maintenance diet. During testing, females and males were placed in separate, adjacent compartments that were joined by a wire mesh window. All pairs were tested only once in a day.

Specific Satiety Experiment.

To ensure that jays had a specific satiety for our test foods, we investigated whether birds would respond to the prefeeding and therefore devaluation of a particular food by subsequently eating a smaller amount of that food relative to when this food was not devalued. We prefed birds on one food type—either MD, W, or M—and then gave them access to two bowls containing W and M (Fig. 1A). All birds started with being prefed MD (baseline), whereas the order of prefeeding the two test foods was counterbalanced across birds. The mesh window was covered by transparent Perspex to prevent food sharing from occurring during prefeeding and by opaque material during the test phase, such that birds could not observe the choices made by their partner.

We calculated the difference of the proportion of W eaten between the baseline (prefed MD) and the conditions in which birds were prefed one of the two test foods (prefed W and prefed M). If birds learned about their specific satiety to the two test foods during testing (17), we predicted that when the differences were compared, a decrease should be shown from the prefed M to the prefed W trials.

Food-Sharing Experiment.

In this experiment, we investigated whether males would respond to the females’ specific satiety by sharing fewer of the food items that had been devalued relative to the amount of that food shared when it had not been devalued for the female. In the prefeeding phase of the food-sharing experiment, females ate either MD, W, or M, whereas males ate MD in all trials (Fig. 1B). During prefeeding, the mesh window was either covered by a transparent Perspex screen (seen condition; Fig. 1 B, i), or with opaque material (unseen condition; Fig. 1 B, ii). The order of the seen and unseen conditions was counterbalanced between pairs. All pairs started with the female being prefed MD (baseline), whereas the order of prefeeding the female the two test foods was counterbalanced across pairs. During prefeeding, in this and all experiments described, both M and W were present in the testing area to control for odor cues and were visible to the male in the seen condition of the food-sharing experiment. In the test phase of the food-sharing experiment, males were given 20 choices between W or M (Movie S1). The position of the foods was pseudorandomized with no food appearing on the same side on more than two consecutive trials. If the males did not make a choice within 30 s, the foods were removed. Each opportunity to make a choice was followed by a 40-s delay in which males could share with the female through the open mesh window (Fig. 1B).

We calculated the difference in the proportion of W shared by the males between the baseline (prefed MD) and the condition in which the females were prefed one of the two test foods (prefed W and prefed M). If males responded to the females’ specific satiety to the two test foods, we predicted that when the above differences were compared, a decrease in the proportion of W shared should be shown from the prefed M to the prefed W trials.

In the seen condition, males could rely on both the information gained in the prefeeding phase and the information from the females’ behavior during the test phase to inform their decision on what to feed her. In contrast, in the unseen condition, the only information available to the males was the females’ behavior during the test phase. The procedure was identical to the seen condition in all other respects. Therefore, if an effect was found in the unseen condition and the males’ sharing behavior did not differ between the seen and the unseen conditions, this would indicate that males are able to respond to the females’ specific satiety by reacting to her behavioral cues [stimulus-bound behavior reading (26)]. If, however, the males’ sharing behavior differed between the seen and the unseen conditions, this would indicate that using stimulus-bound behavior reading alone cannot account for the males’ response to the females’ specific satiety. Thus, seeing the female being prefed would be crucial for any effect found in the seen condition.

Observational Specific Satiety Experiment.

Finally, we investigated the possibility that the males could have developed observational specific satiety by watching the females interact with one food type. Evidence that males did not develop specific satiety themselves through watching the females being prefed on a particular food is important for showing that the males’ subsequent sharing behavior reflected their own, rather than the females’ specific satiety (self–other differentiation). Two males in colony 1, which had been tested on specific satiety and food sharing in 2010, were tested on this experiment in 2011 (with an additional test of specific satiety to compare the data across years). All other pairs were tested on the observational specific satiety immediately after the food-sharing experiment. Here, birds were prefed in the same manner as the seen condition of the food-sharing experiment and the all pairs started with the female being prefed MD (baseline), whereas the order of prefeeding the female the two test foods was counterbalanced across pairs. Subsequently, the females were released into the aviary and the males were given access to two bowls containing W and M (homologous to the test phase of the specific satiety experiment; Fig. 1C).

We calculated the difference of the proportion of W eaten by the males between the baseline (females prefed MD) and the conditions in which females were prefed one of the two test foods (prefed W and prefed M). In accordance with the prediction that males did not develop observational specific satiety through watching the females being prefed the test foods, we expected the males’ eating pattern in this experiment to differ to their eating pattern in the specific satiety experiment.

Analysis.

All graphs show the difference in the proportion of W eaten or shared between the baseline (prefed MD) and trials in which one of the test foods had been prefed (prefed W and prefed M). Relating the results from the test trials to the performance during the baseline ensured that interindividual variation both in the amount of food eaten or shared as well as in general food preferences were taken into account. Because of the birds’ general preference for the two test foods (both W and M) over MD, they are likely to eat more of the test foods than of the MD during prefeeding. Therefore, they might experience a higher general satiety in the test trials than in the baseline. This might lead to a decrease in the amount of food eaten or shared during the test phase. However, if they developed specific satiety (specific satiety and observational specific satiety experiments) or take the females’ specific satiety into account (food-sharing experiment), this decrease was predicted to be larger for the prefed and therefore devalued food than for the other food type.

The data were live scored by experimenter 1 (L.O.) for colony 1 and experimenter 2 (R.C.S.) for colony 2. A naive rater coded 51% of sharing bouts for which the recorded video allowed them to clearly see the sharing event. The same bouts were then coded by experimenter 1 who was blind to the condition in which the birds were tested at the time of rating. Interobserver reliability for the type of larvae shared by the male was high (n = 94, Cohen’s κ = 0.87).

Because of the small sample sizes, data were analyzed using exact nonparametric tests (31). The main analyses were based upon clear predictions—specific satiety experiment: a decrease in the proportion of W eaten relative to the baseline (prefed MD) after birds had been prefed W (W devalued) compared with when birds had been prefed M (M devalued); and seen condition: a decrease in the proportion of W shared by the males relative to the baseline (females prefed MD) after females had been prefed W compared with when females had been prefed M. Our two controls for testing the behavioral criteria for state attribution required the comparison between test conditions. In accordance with the state-attribution hypothesis, the decrease in the proportion of W eaten from prefed M to prefed W (relative to the baseline) was predicted to be higher in the specific satiety than in the observational specific satiety experiment and predicted to be higher in the seen than in the unseen condition of the food-sharing experiment. Thus, all of the comparisons were calculated using directional exact Wilcoxon signed-rank tests (one-tailed P values). For all other analyses, for which there were no clear predictions tests were nondirectional (two-tailed P values). Alpha for all tests was set at 0.05.