Network theory has provided a general way to understand mutualistic plant–animal interactions at the community level. However, the mechanisms responsible for interaction patterns remain controversial. In this study we use a combination of statistical models and probability matrices to evaluate the relative importance of species morphological and nutritional (phenotypic) traits and species abundance in determining interactions between fleshy-fruited plants and birds that disperse their seeds. The models included variables associated with species abundance, a suite of variables associated with phenotypic traits (fruit diameter, bird bill width, fruit nutrient compounds), and the species identity of the avian disperser. Results show that both phenotypic traits and species abundance are important determinants of pairwise interactions. However, when considered separately, fruit diameter and bill width were more important in determining seed dispersal interactions. The effect of fruit compounds was less substantial and only important when considered together with abundance-related variables and/or the factor ‘animal species’.

Summary: 1. Secondary seed dispersal is a multistep process with two or more phases, which involve different dispersers that usually extend the distance from the seed’s parent plant. This ecological process has been recorded in some subtropical oceanic islands, where predatory birds commonly consume frugivorous lizards and disperse seeds already consumed by the lizards.2. We evaluated the overall importance of this type of secondary dispersal in the Canary Islands, the only place worldwide where it has been studied in depth. From an examination of all the islands and their suitable habitats, we found seeds from 78 plant species inside 2098 shrike pellets and 5304 kestrel pellets. A greater number of species were secondarily dispersed by kestrels (76; 97%) than by shrikes (26; 34%).3. Forty‐four (56%) of the total species detected in pellets were identified at the species level, comprising 73% native and 27% introduced species. Seventy per cent of these identified species were fleshy fruit‐bearing plants and 84% of the interactions took place in open habitats, close to coastal areas.4. Germination experiments showed that seeds of at least 32 plant species were viable after being removed from the bird pellets. A similar pattern of seed germination was detected for seeds from the droppings of lizards and pellets of shrikes, showing both to be effective dispersers. However, the seeds dispersed by kestrels had different levels of success depending on the number of gut passages experienced. Seeds that had undergone double gut treatment (lizard and secondary ingestion by kestrel) had reduced germination rates of many small‐ and medium‐sized seeds compared with seeds ingested by lizards and discarded inside the lizard guts by kestrels.5. We also studied the relationship between body length and gape width of lizards in order to assess limitations on the sizes and quantities of seeds available for secondary dispersal. Kestrels can disperse a greater number and variety of seeds because they predate larger lizards that potentially carry greater seed loads.6. Synthesis. The current results show how these non‐standard long‐distance dispersal events produced by predatory birds can be considered as a regular and generalized process on all islands of the Canary archipelago.

Network models of frugivory and seed dispersal are usually static. To date, most studies on mutualistic networks assert that interaction properties such as species' degree (k) and strength (s) are strongly influenced by species abundances. We evaluated how species' degree and strength change as a function of temporal variation not only in species abundance, but also in species persistence (i.e., phenology length). In a two-year study, we collected community-wide data on seed dispersal by birds and examined the seasonal dynamics of the above-mentioned interaction properties. Our analyses revealed that species abundance is an important predictor for plant strength within a given sub-network. However, our analyses also reveal that species' degree can often be best explained by the length of fruiting phenology (for plants degree) or by the number of fruiting species (for dispersers degree), which are factors that can be decoupled from the relative abundance of the species participating in the network. Moreover, our results suggest that generalist dispersers (when total study period is considered) act as temporal generalists, with degree constrained by the number of plant species displaying fruits in each span. Along with species identity, our findings underscore the need for a temporal perspective, given that seasonality is an inherent property of many mutualistic networks.