Subjects
Our subjects were juvenile pigs (N = 10, 6 males and 4 females, Xage ± SD = 7.0 ± 1.24 months, Minnesota and mixed miniature variants) and dogs (N = 12 family dogs passed the criteria out of 19 tested, see below), 7 males and 5 females, Xage ± SD = 6.91 ± 1.92 months, from 8 different breeds). All animals were living in human families from ~ 8 weeks of age (more details in Supplementary Table S1). Subjects from both species were tested intermixed, there was no fixed species order.
Procedure
The study was carried out in the laboratory (4.45 × 3.86 m room) of the Department of Ethology (Eötvös Loránd University, Budapest). A transparent plastic container (the apparatus, 15 × 15 cm) was placed equidistant from the two longer sides of the test room (Fig. 1), upside down (over a few titbits of sausages for dogs and apple/dog food for pigs—based on preparatory owner reports and pilot trials we assumed that these food types were of similarly high value for the individuals, since all of them willingly ate the offered titbits) on a wooden platform (40 × 60 cm), with the base permanently fixed to the platform. The upper cover part—with holes on it—could be moved off the platform easily by manipulation (solvable phase), but it could also be securely attached so the food was still visible but not accessible (unsolvable phase, adapted from Passalacqua et al. 2013).
Each individual test was done in the same day. We followed the method applied by Passalacqua et al. (2013) with some modifications; we increased the time of the unsolvable phase by 30 s to allow more time for the investigated behaviours to evolve, and we added an additional Baseline phase as well for observing any human- or apparatus-oriented behaviour in the absence of food, and familiarizing the subject (S) with the apparatus. The whole procedure consisted of a baseline phase (60 s), a solvable phase (5 solvable trials) and an unsolvable phase (90 s) in fixed order. Before the test session began, the S, the owner (O) and the experimenter (E) entered the test room and the S was allowed to walk around off leash and explore the room for 60 s.
During all phases O and E kneeled down by two different sides of the apparatus at 50 cm distance facing towards it. In the beginning of each phase O kept S in between his/her legs with S facing the apparatus. In the baseline phase, E manipulated the apparatus without covering it, placing the upper part next to the fixed bottom part. O let S free when E signalled with her hand. E and O, with the hands behind their backs stayed passive, following S with their gaze.
After a short (− 30 s) break, the test continued with the solvable phase. E showed a piece of food to S, placed it on the bottom of the apparatus without covering and O let S free. This served for testing S’s motivation for eating and informing the S where the food would be placed. E then showed another piece of food to S, placed it on the bottom of the apparatus and covered it with the upper part without closing it securely. O let S free. E and O did not move. The trial ended when S obtained the food or after a maximum 60 s. Only subjects succeeding (accessing the food) a minimum 3 out of 5 times (10/10 pigs, 12/19 dogs) were tested in the unsolvable phase that followed immediately, and included in analyses.
The unsolvable phase was identical to the solvable phase, except that the cover part of the apparatus was securely closed (i.e. food inaccessible for S).
Behaviour coding
All tests were video recorded for behavioural analysis by Solomon Coder (v. 090913; © András Péter https://solomoncoder.com). Starting from O releasing S, during the baseline and unsolvable phases we measured latency and duration of orientation to humans, frequency of orientation-alternation between the apparatus and the humans, and durations of vocalization, human-oriented vocalization and apparatus-interaction. During the solvable phase we measured the latency to success (i.e. solving the task), and apparatus-interaction and human-orientation in the first trial (see Supplementary Table S2 for behavioural variables definitions).
The recordings were coded by one main coder, and twenty percent of them was also coded by a secondary coder. Interrater agreement was near perfect for ‘Success latency’ (ICC = 0.99 for both species). We used the raw coding sheets to calculate the agreement between the two raters for ‘orientation to human and to apparatus’ and ‘apparatus-interaction’, where the occurrence (yes/no) of any of these behaviours was marked every 0.2 s. The agreement was near perfect for ‘orientation’ (Cohen’s Kappa, ĸpigs = 0.89, ĸdogs = 0.97, Ps < 0.001) and also for ‘apparatus-interaction’ (ĸpigs = 0.93, ĸdogs = 0.98, Ps < 0.001). We therefore used the coding of the main coder only to extract the variables of interest.
Data analysis
We used the R statistical environment (v.3.5.0. R Development Core Team) with the following packages: lme4, emmeans and ggplot2. We used Shapiro–Wilk test and data visualization (normal Q-Q plots) to check for the distribution of the response variables and residuals, and applied Box-Cox power transformations with optimal lambda parameters where it was necessary to fulfil normality criteria. We used non-parametric tests where neither transformation method resulted in a normal distribution. We built a linear mixed-effects model (LMM) with ‘Success latency’ as the response variable, trial (within-subject) and species (between-subject) as explanatory factors and individual subjects as a random factor for the solvable phase analysis. We used Mann–Whitney–Wilcoxon tests to compare the total times pigs and dogs spent orienting at the two humans and manipulating the apparatus in the first trial of the solvable phase—to test for factors that could possibly explain success latency differences between the two species (see “Results”).
Since we did not aim to test for the familiarity effect of the humans on the subjects’ interspecific behaviour and the owner-experimenter contrast was not well controlled to make it clearly interpretable, we did not consider orientation to owner and experimenter as separate measures throughout the main analyses, but used a combined ‘orientation to human’ variable instead. To test for main effects of phase (baseline vs. unsolvable, within-subject factor) and species, as well as for their interaction, we built LMMs with either ‘orientation to human’ (duration, s), ‘latency of orientation to human’ (s), and ‘apparatus-interaction’ (duration, s) as response variables. For testing the same main and interaction effects on ‘orientation-alternation’ (Poisson-distributed count data) we built a generalized mixed-effects model (GLMM) fit by maximum likelihood using Laplace approximation. We included individual subjects as a random factor in all the models and obtained corrected multiple post-hoc comparisons for the fixed factors. In all the above models we used the data from the first 60 s of the unsolvable phase (and all data from the 60 s long baseline phase) to make a fair comparison between the two phases. However, to further explore species differences during the total duration of the unsolvable phase in the above mentioned response variables, we used Mann–Whitney–Wilcoxon tests and two-sample t test (according to data distribution). To show how the human-oriented behaviours in the unsolvable phase were divided across the experimenter and the owner, we built LMMs for ‘orientation’, ‘latency of orientation’, a GLMM for ‘orientation-alternation’, and tested for main effects of orientation target (experimenter and owner), species and their interaction (see Supplementary material).
We compared the number of pigs and dogs that vocalized in the baseline and unsolvable phase by Chi-square test (with Yates' continuity correction), and used Mann–Whitney–Wilcoxon test to compare the duration of their vocalization. Because of dogs’ overall less vocalization (see “Results”) we further analysed pigs’ vocal behaviour only. To see whether there was any difference between the proportions of time pigs vocalized while exhibiting different orientation behaviours, we calculated the ratios of ‘human- and apparatus-oriented vocalization’ and ‘orientation to human or apparatus’ in both conditions (to make a fair comparison between the proportions of times—expressed as the ratio of the total duration of the session—pigs spent vocalizing while being oriented either to the humans or to the apparatus), and compared them by Wilcoxon signed-rank tests.