Population density and starvation as key drivers of cannibalism in the hydrothermal vent crab Xenograpsus testudinatus

Cannibalism is reported here for the first time in the hydrothermal vent crab Xenograpsus testudinatus, an endemic species inhabiting sulfur-rich shallow vent ecosystems. We examined how food deprivation and population density influence cannibalistic behavior under controlled laboratory conditions. Size-matched crabs were maintained at three initial densities (low: 15; medium: 45 (control group); high: 90 individuals per tank) and subjected to a 28-day starvation period. Cannibalism and mortality were recorded daily. Weekly cumulative counts were summarized descriptively and analyzed using Poisson generalized linear mixed models (GLMMs), with tank included as a random effect and effective density (accounting for mortality over time) modeled as a continuous covariate. GLMMs showed significant effects of initial density, week, and effective density on cannibalism (all p < 0.001). Relative to medium density, low-density tanks showed higher baseline cannibalism (log coefficient = 0.883 ± 0.332; incidence rate ratio [IRR] ≈ 2.42), whereas high-density tanks showed lower baseline rates (−0.870 ± 0.106; IRR ≈ 0.42). Cannibalism increased markedly over time, becoming significant in Weeks 2–4 and peaking in Week 4 (log coefficient = 1.978 ± 0.154; IRR ≈ 7.23). Effective density independently amplified realized cannibalism (log coefficient = 1.776 ± 0.080; IRR ≈ 5.91 per 1 SD increase), indicating that encounter-driven processes intensified as remaining individuals became concentrated. Consistent with model predictions, survival declined as cannibalism increased; patterns suggest that early removal of weaker individuals may transiently reduce mortality risk, followed by intensified competition that promotes further cannibalism during prolonged starvation. Overall, our results suggest that cannibalism in X. testudinatus is influenced by both population density and food limitation. However, because no fed control group was included, the independent contributions of starvation and density cannot be fully disentangled. Future studies should address this limitation for a clearer understanding of the underlying drivers.