Concept learning is considered a high-level adaptive ability. Thus far, it has been studied in laboratory via asocial trial and error learning. Yet, social information use is common among animals but it remains unknown whether concept learning by observing others occurs. We tested whether pied flycatchers (Ficedula hypoleuca) form conceptual relationships from the apparent choices of nest-site characteristics (geometric symbol attached to the nest box) of great tits (Parus major). Each wild flycatcher female (n = 124) observed one tit pair that exhibited an apparent preference for either a large or a small symbol and was then allowed to choose between two nest boxes with a large and a small symbol, but the symbol shape was different to that on the tit nest. Older flycatcher females were more likely to copy the symbol size preference of tits than yearling flycatcher females when there was a high number of visible eggs or a few partially visible eggs in the tit nest. However, this depended on the phenotype; copying switched to rejection as a function of increasing body size. Possibly the quality of and overlap in resource use with the tits affected flycatchers’ decisions. Hence, our results suggest that conceptual preferences can be horizontally transmitted across co-existing animals, which may increase the performance of individuals that utilize concept learning abilities in their decision-making.

Setup of the experiment

The experiment was conducted in mixed and coniferous forests near the city of Oulu, in northern Finland, between 2012-2014. Nest boxes were situated along small roads and great tits were allowed to settle and breed freely. The basic sampling unit in the study was a set of two pairs of closely situated nest boxes inhabited by a pair of great tits and a pair of pied flycatchers. Prior to the arrival of flycatchers, we randomized one of the symbol shape (triangle/circle) – symbol size (large/small) associations and attached a white plastic symbol around the entrance hole of an active great tit nest. An empty nest box was then erected near (3-6 m) the great tit’s nest box on a tree of the same species, at the same height and facing same direction as the occupied box, and the same symbol shape but opposite symbol size compared to that on the tit nest box was attached on the empty box. Two empty boxes, entrance holes facing towards the tit box pair, were erected 20-25 m from the great tit’s nest box and 3-6 m apart from each other, with both symbol sizes, but the opposite symbol shape compared to that on the great tit’s nest box, randomized on the boxes. These two vacant nest boxes were meant for flycatchers to choose between. We put one litre of sawdust into both vacant nest boxes because flycatchers highly prefer it as a nest platform over empty nest boxes or great tit nests. This reduces the risk that pied flycatchers take over the great tit nest or choose the empty nest box near the great tit nest. These setups were situated one km apart to minimize the risk that flycatchers perceive more than one symbol-size association apparently preferred by great tits.

This design was intended to inform arriving flycatchers that the resident great tit pair had apparently preferred one of the sizes of the symbol. Flycatchers were then given a choice to either copy or reject the size concept preference of tits with symbols whose shape was different from that on tit nest. In this way, the only cues provided by the tits that the flycatchers could use were the nesting success (clutch size) and its association with the symbol size (large/small). The sizes of the symbols were: small and large triangle: 17.7 and 83.6 cm², respectively, and small and large circle: 19.5 and 40.5 cm², respectively. The size differences between small triangles and small circles as well as large triangles and large circles were about 10% and 50%, respectively, and were chosen to avoid the possibility that flycatchers would use the surface area of the small and large symbols in making the choice and not the relative size difference. To avoid the possibility that the previous year’s symbol shape and size preferences of tits affect current choices, the study area was changed annually. Great tits and pied flycatchers were breeding in natural cavities in the study area in low numbers but there were no nest boxes other than ours in the area.

Measurements

We visited the nests at least every second day and determined the choice of each flycatcher female upon the appearance of nest material in the nest box. During egg-laying, great tits usually cover eggs totally or partially but sometimes leave them totally exposed. We therefore monitored how many eggs were visible and the clutch size in great tit nests on the day of flycatcher choice. Once the flycatcher had made a choice, all symbols and the two empty nest boxes were removed. Flycatcher females were captured during incubation and their age was estimated and their tarsus length was measured. Age of the females was estimated from tail and wing feathers as young (born in the previous year and therefore breeding for the first time) or adult (at least two years old and probably has bred earlier). All birds were handled and measured by JTF under his ringing license (2975) and following Finnish legislature.

Statistical analyses

Usually most tits have initiated egg-laying prior to the arrival of flycatchers. However, in our data, 21 flycatchers (out of 124 made a choice and settled before great tits had laid their first egg. We therefore analysed the data in two parts: 1) data including choices made during egg-laying or incubation of great tits and 2) data including all choices irrespective of the state of great tit reproductive cycle. Great tit nests where egg-laying was not initiated before flycatcher choice mostly consisted of undeveloped moss nests or nearly/fully finished nests with hair cover but where egg-laying was not yet initiated, or were deserted nests and did not result in a nesting attempt (but at the moment of flycatcher choice it was not known whether the great tit territory was active or not). This division is biologically meaningful because our previous studies have demonstrated that the perceivable clutch size of the tits provides important information for flycatchers and strongly affects their choices. The possibility to perceive eggs in great tit nests also confirms that the great tit territory is active because tits often start nest building in several cavities, one of which is chosen. The level of testosterone also decreases from nest-building to egg-laying state, which plausibly reflects to the behaviour of great tits. In the main text, we present the results of the data including choices made during egg-laying or incubation of great tits, while the results of the whole data are shown in the supplementary material.

In both data sets, we tested how the phenotype of the nest of the demonstrator tit and the phenotype of the flycatcher female explain flycatcher choices. We used generalized linear models (R function ‘glm’, binomial error distribution and logistic link function) to determine if tit clutch size (6.1 ± 0.28 S.E. eggs), the proportion of visible eggs on the day of flycatcher choice (0.56 ± 0.046), the phenotype of the flycatcher female in terms of morphology (length of tarsus; 19 ± 0.047 mm) and age (adult/born in the previous year; 59 adults and 44 yearlings) explain the symbol size choice of the flycatcher female (binary variable: matching [44 times; coded as 1] vs. opposing [59 times; coded as 0] symbol size choice than that on the tit’s nest). The full model included all main effects, the three-way interaction among tit clutch size at the time of choice, the proportion of visible clutch size and flycatcher female age, as well as all two-way interactions among these three variables. Visual evaluation of residual plots indicated that the full model fitted the data well, and residual deviance (114 on 97 degrees of freedom) did not indicate overdispersion.

All analyses were conducted using R (version 4.1.2). We derived all meaningful sub-models of the full model with the function ‘dredge’, ranked the models using Akaike’s Information Criterion adjusted for small sample sizes (AICc), and derived the 95% confidence set of models (i.e. the set of models over which the cumulative sum of Akaike weights is ≤ 0.95) for further consideration. We further removed all models that were more complex extensions of the model with the lowest AICc value from the model set and all models that were >6 AICc units from the best model. We used ∆AICc, Akaike weights of the models and their evidence ratios to determine if there was a single superior model explaining the data or a set of nearly equal models. If there was no single best model (∆AICc< 2 and evidence ratio < 5) we used model averaging; function ‘model.avg’ in model parameter estimation and inference. Otherwise, we based our inferences on the single best model.

Meanings of abbreviations used in the FC_choices.csv-file:

GT = Great tit

FC = Flycatcher

F = Female