Inferring the decision rules that drive co-foraging affiliations in wild mixed-species parrot groups

Abstract

Animals gathered around a specific location or resource may represent mixed-species aggregations or mixed-species groups. Patterns of individuals choosing to join these groups can provide insight into the information processing underlying these decisions. However, we still have little understanding of how much information these decisions are based upon. We used data on 12 parrot species to test what kind of information each species may use about others to make decisions about which mixed-species aggregations to participate in. We used co-presence and joining patterns with categorization and model fitting methods to test how these species could be making grouping decisions. Species generally used a simpler lower-category method to choose which other individuals to associate with, rather than basing these decisions on species-level information. We also found that the best-fit models for decision-making differed across the 12 species and included different kinds of information. We found that not only does this approach provide a framework to test hypotheses about why individuals join or leave mixed-species aggregations, it also provides insight into what features each parrot could have been using to make their decisions. While not exhaustive, this approach provides a novel examination of the potential features that species could use to make grouping decisions and could provide a link to the perceptive and cognitive abilities of the animals making these minute-by-minute decisions. This article is part of the theme issue 'Mixed-species groups and aggregations: shaping ecological and behavioural patterns and processes'.

Keywords: categorization; coarse-graining; mixed-species groups; social decision-making; social interactions; species co-presence.

Figure 1.

The 12 parrot species used in the analysis. The drawings depict species differences in size (shown from smallest on the left to largest on the right), coloration and body shape. For size comparison, mean species mass is shown in grams, rounded to nearest whole number [15]. Abbreviations show the two-letter codes used in the text and other figures. Species artwork by V. Darby Moore. (Online version in colour.)

Figure 2.

Example joining events extracted from dataset of binary presence or absence of species. In the left example, species A is absent at t₁, but joined species B at t₂. The right example shows a joining event from the main data partition, highlighted with the blue box. RB was absent at 5:35 in zone 1A while other species were present. RB joined BH (and SC) in zone 1A at 5:40. (Online version in colour.)

Figure 3.

Summaries of the extent of mixed-species aggregations. Panel (a) shows the numbers of scans with groups comprised of different numbers of species within zones (with the number of unique sampling days shown above each bar). Panel (b) summarizes how often each species participated in mixed-species groupings within zones. For each focal species, columns show: (1) the median number of species present in a zone during a single scan when each focal species was present, (2) the median number of focal individuals present during each scan at each zone, (3) the median total number of individuals present during each scan at each zone (across both the focal species and all other species), (4) the median species diversity in groupings during each scan at each zone that each focal species was present in, and (5) the proportion of scans where the focal was the only species present on the zone. Cell colours indicate each value’s proportion compared to the maximum value per measure for each column, with red shades highlighting the maximum values for each column. (Online version in colour.)

Figure 4.

Raw co-presence counts between each pair of species. Each cell shows the number of observations in which the two species were present on the same zone at the same time. The two species that occurred together the most were RB and CF (1343 scans) and the least were DH and RG (seven scans). Axes show species ordered by body size, from DH (smallest) to RG (largest). (Online version in colour.)

Figure 5.

Networks of co-presence and joining among parrot species, showing affiliative (blue) and avoidant (red) edges among species pairs. The top two panels show the network inferred from the controlled co-presence data: (a) affiliative edges indicate species that occurred together on the same zone of the clay lick significantly more than expected by chance, while (b) avoidant edges indicate species that occurred together significantly less than expected by chance. The bottom two panels show the network inferred from the joining data: (c) affiliative edges, where A → B shows species A was significantly more likely to join a zone if B was already present and (d) avoidant edges, where A → B shows species A was significantly less likely to join a zone if B was already present. Dashed lines in panels (a) and (c) show shared community membership, identified from community detection on the two types of affiliative networks. Nodes in all four networks were arranged by body side, running clockwise from RG (largest) to DH (smallest). (Online version in colour.)

Figure 6.

Body mass differences explained part of the variation in controlled co-presence association types. Negative associations were more likely to occur between species with a larger difference in body mass and positive associations were more likely to occur between species with similar body mass. Bars show the difference in body mass between all possible pairs of species, ordered from largest to smallest: blue bars show edges that were significantly affiliative in the controlled co-presence network (figure 5a) and red bars show edges that were significantly avoidant (figure 5b). We found a significant association between body mass differences and the type of association (affiliative/avoidant) in the main data partition (a), then tested for and found the same pattern in our validation data partition (b). The mean body mass of each species is shown in figure 1. (Online version in colour.)

Figure 7.

Best-fit categorization schema models by species for dynamic joining decisions. Each focal species is labelled in the first column, followed by the best-fit model inferred from that species’ joining patterns. All best-fit categorization schemas, with the exception of those for WB and WE, had ΔAICs > 2, making them strongly differentiated from the next-best model (see electronic supplementary material, table S1 for AIC results). Across the table, species that the focal significantly prefers are shown in colour with a white category label, while species they significantly avoid are shown in greyscale with a black category label. Cells containing only text indicate species with which the focal did not have a significantly positive or negative association. All cells are labelled with a text code that indicates the category that each species falls into under the best-fit categorization model (size: S, small; M, medium; L, large; large macaw: N = is not a large macaw, Y = is a large macaw; head colour: G, green; O, orange; K, black; B, blue; R, red; clade: PK, parakeet; SP, small parrot; LP, large parrot; SM, small macaw; LM, large macaw; face colour: GG, green−grey; O, orange; Y, yellow; W, white; B, blue). See table 1 for species assignments to categories. Species artwork by V. Darby Moore. (Online version in colour.)

Figure 8.

The relationship between usage complexity of the best-fit schemas for each species and centrality in the co-presence network. The y-axis is ‘usage complexity’, the number of statistically significant categories within the best-fit categorization schema that each species used when joining other birds on the clay lick. The x-axis shows each species’ degree centrality (a,b) and betweenness centrality (c,d) in the inferred co-presence network. We found degree and betweenness were both significantly predictive of usage complexity in the main data partition (a,c), but only betweenness remained significant when tested in our validation data partition (b,d).