Insects may acquire social information by active communication and through inadvertent social cues. In a foraging setting, the latter may indicate the presence and quality of resources. Although social learning in foraging contexts is prevalent in eusocial species, this behaviour has been hypothesised to also exist between conspecifics in non-social species with sophisticated behaviours, including Heliconius butterflies. Heliconius are the only butterfly genus with active pollen feeding, a dietary innovation associated with a specialised, spatially faithful foraging behaviour known as trap-lining. Long-standing hypotheses suggest that Heliconius may acquire trap-line information by following experienced individuals. Indeed, Heliconius often aggregate in social roosts, which could act as ‘information centres’, and present conspecific following behaviour, enhancing opportunities for social learning. Here, we provide a direct test of social learning ability in Heliconius using an associative learning task in which naïve individuals completed a colour preference test in the presence of demonstrators trained to feed randomly or with a strong colour preference. We found no evidence that Heliconius erato, which roost socially, used social information in this task. Combined with existing field studies our results add to data which contradict the hypothesised role of social learning in Heliconius foraging behaviour.
Experimental subjects and arena
Experimental subjects originated from first-generation insectary-reared stock populations of Heliconius erato phyllis, descended from multiple wild-caught females collected in Mata do Jiqui, Natal, Brazil (5°55'39"S, 35°10'59"W). We maintained stock populations in large outdoor cages (3 x 3 x 2.5 m) in which free-flying butterflies were able to engage in natural social and flight behaviours, including chasing, mating and following. At night, individuals were also observed to form roosts of 2-22 individuals. Stock butterflies had access to hostplants (Passiflora misera and P. galbana) and rewarding artificial white flowers. All butterflies were individually labelled with unique IDs. The test arena was composed of purple and yellow artificial flowers. These colours were chosen given that they are, on average, both relatively unpreferred . Rewarding flowers contained a ~20% sugar solution mixed with bee-pollen supplement while unrewarding flowers were empty. Twelve flowers of each colour were placed on a grid of 24, with randomised positions (Supplementary Information Figure S1).
Selection of demonstrators
Butterflies were randomly assigned to one of two demonstrator groups, each collectively subjected to four days of training (Supplementary Information Figure S2). For one group, only purple flowers were positively reinforced to strengthen preference for purple. For the other, to minimise long-term colour preference, the colour of the rewarding flower was randomly determined for each trial. Training was run between 08:00-16:00. In the following day, a final 5-minute test was conducted to determine demonstrators' colour preferences. Preference was calculated as the proportion of landings on purple and yellow flowers out of 20 landing events. Demonstrators with a ≤ 60% purple preference were assigned to the control group, whereas demonstrators with a ≥ 80% purple preference were assigned to the knowledgeable group, creating two demonstrator groups with unbiased and biased preferences respectively (Figure 1).
Social learning experiment
The social learning experiment consisted of two phases (Figure S3). (1) During pre-training, naïve butterflies, hereafter ‘observers’, fed on artificial white flowers between 08:00-16:00 to get accustomed to artificial flowers. (2) In the trialling phase, which lasted for four days, butterflies were randomly assigned to either the control or the knowledgeable demonstrator groups. Observers were released in pairs along with 10 demonstrators to ensure that most of individual-level feeding attempts were by the demonstrators. During trials, we scored the number of feeding attempts made by each observer on purple rewarding flowers and yellow unrewarding flowers for 15 minutes. A choice was scored when the butterfly landed on a flower.
Data were analysed using generalised linear mixed models in R using the lme4 package . First, we asked whether demonstrators from different groups preferred the rewarding colour, using a binomial GLMM with response variable 'preference for rewarding colour' (proportion of landings on purple flowers) and fixed factor 'group' (control and knowledgeable). Then, for observers, we examined whether there were intergroup differences in preference for rewarding colour over time, using a binomial GLMM with response variable 'preference for rewarding colour' and fixed factors 'group' (control and knowledgeable) and 'trial day' (1 to 4). In the latter, identity was set as a random effect due to repeated measures. Finally, we analysed whether observers preferred flowers occupied by demonstrators using a binomial GLMM with response variable 'local preference' (0 = no; 1 = yes) and fixed factors 'group' (control and knowledgeable) and 'trial day' (1 to 4).