Morphological evolution and the behavioral organization of agricultural division of labor in the leafcutter ant Atta cephalotes

WOLO: Wilson Only Looks Once – Estimating ant body mass from reference-free images using deep convolutional neural networks

“…Body weight estimation from images is a challenging task, and usually only feasible if reference lengths or other cues are provided. In Atta leaf-cutter ants workers, body size variation is accompanied by changes in body shape [24,[37][38][39][40][41][42][43]; it is thus in principle possible to learn how to estimate body mass or other size proxies without external cues, and independent of image magnification [14]. To this end, we trained a variety of deep convolutional neural networks to estimate leaf-cutter ant body mass directly from cropped or full-frame samples.…”

Large deformation diffeomorphic mapping of 3D shape variation reveals two distinct mandible and head capsule morphs in Atta vollenweideri leaf‐cutter worker ants

“…Males were not investigated as part of this work, but may well represent an additional morph. The three shape categories seemingly coincide with functional specialisation: the queen is the only reproductively active female; minims are rarely found outside the nest, but are by far the dominant size‐class inside the fungal gardens (Figure S3; Muratore et al., 2023; Wetterer, 1999; Wilson, 1980); medias of all sizes actively partake in foraging (Clark, 2006; Muratore et al., 2023; Rudolph & Loudon, 1986; note that task preference can further vary with size within the medias, e.g. Camargo et al., 2007; Püffel, Meyer, et al., 2023).…”

“…Indeed, larger workers are more likely to cut tougher leaves (Clark, 2006; Evison & Ratnieks, 2007; Nichols‐Orians, 1991), illustrating how size‐differences can be functionally advantageous. However, whether, and if so how, leaf‐cutter ant workers also differ in shape is controversial; the number of morphs reported in the literature lies anywhere between one and five (Cherrett, 1972; Feener et al., 1988; Franks & Norris, 1987; Hernandez & Caetano, 1995; Muratore et al., 2023; Rudolph & Loudon, 1986; Silva et al., 2016; Wetterer, 1991, 1999). The key difficulty is that shape variation can be subtle, and is thus generally harder to quantify than differences in body size.…”

“…Indeed, classic ergonomic theory suggests that ant colonies maximised their fitness if each essential task was performed by a morphologically specialised worker (Oster & Wilson, 1978). However, this prediction stands in sharp contrast to empirical observations: ant workers may perform more than 40 tasks (Wilson, 1976), but no more than four to five "worker castes" have been reported for any ant species (Franks, 1985;Muratore et al, 2023;Silva et al, 2016). One possible explanation for this seeming discrepancy may be that task specialisation can be achieved with isometric variations in body size alone, with the putative advantage that no complex changes in the developmental programme of workers are needed.…”