Biotic interactions (e.g. predation and mutualism) occur between individuals and accumulate across a community to shape species-level interaction structure. Increasingly intraspecific variation in interactions is recognized as an important component of ecological processes, with the potential to have substantial repercussions for the outcomes of plant-frugivore seed dispersal mutualisms. Skewed interaction structures, where a few individuals are highly connected and most have few interactions are increasingly identified at the individual-level. However, the way individual-level interactions accumulate across a whole plant-frugivore to species interactions and shape interaction outcomes remains unclear. To address this gap in our knowledge, we studied the interactions of three sympatric frugivorous lemur species (Varecia variegata editorum, Eulemur rufifrons, and Eulemur rubriventer) with individual plants within a defined study area. We studied the consistency of skewed patterns across time, lemur species and plant species; compared individual and species-level network structure; studied the intrinsics (DBH, fruit crop size, species) and extrinsic (fruit availability and richness) factors affecting interaction structure; and tested interaction structure impact on seed movement. We found a substantial and consistent skew in the interactions of individual plants with the lemurs, such that 70% of all plant individuals received a single visit. This abundance of single-visit plants was consistent among lemur species, across time and between plant species. Highly visited plant individuals (with > 20 visits per lemur species) occurred infrequently and only for visitation by V. variegata editorum and E. rubriventer. These differences in lemur visitation were significant at the individual-level, though missing in their interactions with plant species. Individual interaction networks were smaller, more specialized and had lower connectivity than their species counterparts. These differences demonstrate the importance of individual-level interactions, which contain substantial interaction variation lost at the species scale. Individual-level interactions were positively influenced only by DBH and fruit crop size and conspecific fruit availability. This skewed interaction structure impacted seed dispersal patterns, so that seeds were most likely to be deposited within 15m of highly visited plants. Our results highlight the importance of individual-level interaction variation for seed dispersal mutualisms and call for further study on the consistency of such patterns across ecosystems.

Fieldwork took place in the Valohoaka research area within Ranomafana National Park (RNP) in southeastern Madagascar. Ranomafana National Park is composed of protected montane rainforest with an elevation ranging from 600 – 1500 m (Wright and Andriamihaja 2002) and yearly rainfall averaging 2835mm (monthly averages 10-1200mm) (Dunham, Erhart, and Wright 2011). Our field site was a 2,000 X 2,000 m (4 km²) forest section where previous observations of all three focal lemur species (Eulemur rufifrons, Eulemur rubriventer and Varecia Variegata editorum) facilitated access to habituated populations (Razafindratsima, Jones, and Dunham 2014; Tonos et al. 2021). These frugivorous lemur species are known to play a substantial role as seed dispersers for the native plant community, which is composed of at least 330 plant species (Razafindratsima and Dunham 2015).

Data collection took place from June 2022 through December 2023, and was organized in expeditions of 11 days each. During these expeditions, we undertook two days of phenology observations and nine days of lemur foraging observations. On average, we undertook two expeditions per month, though local holidays, logistical constraints, and inclement weather occasionally intervened with this schedule. We completed 31 expeditions during which we conducted 6-11 primate observation days, for a total of 280 days.

1.1        Primate observations

For each lemur species, a dedicated team of three observers conducted nine continuous days of direct observations per expedition. Each observation day consisted of 8 hours of continuous data collection on a group of the team’s target species. At the beginning of each observation day, each team entered the study area, located a group or individual of their target species, and then began the 8-hour observation period. Shorter observation periods occurred when lemur groups were found late enough that observations were cut off by nightfall or when lemur groups were lost due to faster-than-usual travel speeds, terrain, or weather. In the case that a lemur group was lost, the team would search for the same, or another group, of the same species within the study area and resume observations. When a lemur group was found, an individual was selected as a focal for feeding and seed-dispersal observations. As lemur groups of the Eulemur species tend to forage and travel together cohesively (Pyritz, Kappeler, and Fichtel 2011; Tecot and Romine 2012), a randomly selected individual is a reasonable proxy for the broad group foraging and seed-dispersal behaviors. For V. variegata observations focused on a single individual (rather than a group) was necessary due to their flexible fission-fusion dynamics in which group composition changes readily and individuals are often alone (Baden, Webster, and Kamilar 2016). At any instance of feeding (frugivory)(Appendix S1: Figure S1) or seed dispersal (defecation containing seeds) by the focal, we collected the time of the event, the geographic location using a GPS device, and the species and individual identifier of the tree where the lemur was located. For feeding events, we additionally noted the plant part and species consumed (may differ from the tree species in the case of vines and epiphytes), the diameter at breast height (DBH), and the fruiting phenology status of the plant consumed. For epiphytes and vines, the identifying tag was placed on the host tree, so that a particular plant is indexed by a combination of tree tag and species name. Fruiting phenology was graded on a 5-category system in which a zero is given for no fruits, 1 corresponded to 1-25% of the crown containing fruit, 2 for 25-50%, 3 for 50-75% and 5 for >75% as utilized in similar studies (Dunham et al. 2018).

1.2        Phenology

We established 27 transects (100 x 4 meters) within our 4 km² study area in May 2023. The transects were placed by selecting 30 random start points (>100m apart) and then randomly assigning a direction (selecting a number between 0 and 360 degrees). We established the transects by navigating to the start point and extending the transect line 100 meters in the selected bearing, using a compass and measuring tape. Three of the proposed transects were removed during this process due to the steepness of the terrain and the presence of large boulders. Along each transect (within 2 meters of the established center line at each side), we marked (with a unique numbered aluminum tag) and measured the DBH and height of every tree with a DBH >10 cm. During the first two days of every expedition after transect establishment, we survey the fruiting phenology of every plant along our transects (tagged or not), recording the species, individual tag (as relevant), DBH, height, transect section (10 sections per transect of 10 m each). For each plant, we record the presence and extent of each phenological event observed (new leaves, budding, flowering, and fruiting). The extent of the phenology event is rated on the same five-category system as above.

Using these phenology data, we estimated the total fruit availability index (FAI) for each expedition, as well as the FAI for each plant species within each expedition. The FAI was calculated according to the methods of (Sengupta and Radhakrishna 2016) by multiplying the mean density across all transects to the mean basal area and mean fruiting score across all individuals of a species within each expedition. For total expedition FAI, species specific values were summed for each expedition.