- The paper finds that larger ant workers have lower mortality risk and longer lifespans, with a hazard ratio of 0.67 per tenfold increase in body mass, while colony size and temperature do not significantly modify this relationship.
- Senescence is largely independent of body size, occurring in 14 of 18 species and varying instead with circadian activity, while field conditions generally produce steeper aging trajectories than laboratory conditions.
- Thermal mortality increases sharply below 20 °C before reaching a plateau, but matinal Rhytidoponera ants show 66% excess mortality risk above 30 °C, indicating that extreme-heat frequency and lineage may outperform body size in vulnerability assessments.
Overview and motivation
Body size is the canonical predictor of worker longevity in social insects, and it is routinely deployed as a one-dimensional proxy for mortality risk in comparative and applied work. This study, built on a paired field–laboratory dataset of 18 Australian ant species (2,363 cohort-day observations; 1,826 mortality events; 106 cohorts across 39 colonies; 1,148 laboratory workers), argues that "mortality risk" is in fact a bundle of three components that need not share a common predictor: duration (average lifespan), senescence trajectory (the Weibull shape parameter ρ), and thermal vulnerability (the sensitivity of mortality hazard to temperature). The paper tests all three axes simultaneously and finds that no single predictor — body size included — is detectable across all three at this sample size.
The study system is a temperate assemblage at the La Trobe Wildlife Sanctuary (Melbourne, Victoria), spanning five subfamilies and the full range of body sizes and circadian activity regimes (matinal, diurnal, crepuscular). The work extends a prior mark–recapture study of the same assemblage by partitioning mortality risk, testing whether size buffers thermal mortality, and characterising senescence trajectories.
Duration: body size predicts lifespan, and the effect appears intrinsic
The primary Cox proportional-hazards model, with cluster-robust standard errors and subfamily fixed effects, yields a hazard ratio of HR = 0.67 per log₁₀ unit of body mass (95% CI [0.52, 0.87], p=0.002): each tenfold increase in worker mass reduces daily mortality hazard by roughly a third. PGLS confirms negligible phylogenetic structure in this signal (Pagel's λ^=0.06, p=0.90).
The paper's mechanistic argument proceeds by exclusion. Three candidate moderators of the size–longevity effect were tested:
- Colony size (colony-dilution hypothesis): null interaction (OLS p=0.60).
- Temperature (extrinsic-buffering hypothesis): null interaction (HR = 1.00, 95% CI [0.98, 1.03], p=0.72). The authors report ~97% power to detect an interaction as small as HR = 1.05, so this null constitutes positive evidence against ecologically meaningful thermal buffering by size.
- Foraging rate: a weak but significant interaction (LRT p=0.014, ΔAIC = −4.1), which qualifies the intrinsic reading; the authors caution it should be interpreted cautiously given moderate collinearity (VIF ≈ 4).
With the two extrinsic moderators null, intrinsic physiology — lower mass-specific metabolic rate and plausibly reduced oxidative stress — remains the most parsimonious account. The size benefit is also directionally stronger in the field than in the laboratory (1.28× on the log-hazard scale), consistent with extrinsic mortality filtering acting on an intrinsic size advantage. One important caveat, which the paper states plainly: the HR = 0.67 reflects a between-species pattern. When species identity is absorbed as fixed effects or Cox strata, the coefficient remains directionally consistent (HR = 0.71–0.77) but non-significant, reflecting narrow intraspecific size variation rather than absence of effect.
Trajectory: senescence tracks circadian niche, not body size
Actuarial senescence (ρ>1) is the dominant pattern, present in 14 of 18 species (mean ρ=1.37). Critically, ρ is uncorrelated with body mass (Spearman p=0.0020, p=0.0021) — the trait predicting duration does not predict the shape of ageing. Instead, p=0.0022 differs significantly among circadian niches (Kruskal–Wallis p=0.0023, p=0.0024), with matinal species senescing most steeply and the matinal-vs-crepuscular contrast significant (p=0.0025). The matinal–diurnal contrast is not individually significant, partly due to a single diurnal outlier (Iridomyrmex notialis, p=0.0026), so the signal is best described as a graded matinal-to-crepuscular continuum.
Field conditions steepen senescence nearly universally: p=0.0027 in 17 of 18 species (Wilcoxon p=0.0028), and the result holds (14 of 15) when three species with pooled-fallback field estimates are excluded. The authors interpret this as field amplification of a species-typical intrinsic trajectory, while noting that laboratory and field environments differ enough that Δp=0.0029 should be read as a qualitative direction, not a precise magnitude.
Thermal risk: a plateau at 20 °C and a taxon-specific excess
Temperature raises hazard by 3.6% per °C in the main-effects model (HR = 1.036, λ^=0.060), but the relationship is strongly nonlinear. A piecewise spline with breakpoint λ^=0.061 °C outperforms the linear model decisively (ΔAIC = −38; Davies permutation λ^=0.062). Below 20 °C, hazard rises steeply (HR = 1.216/°C); above it, the slope collapses to a near-plateau (HR = 1.016/°C). Because 65.7% of observations fall above the breakpoint and the assemblage mean maximum temperature (22.8 °C) already sits past it, most realised thermal variation lies in the attenuated regime. The authors attribute the plateau, cautiously, to behavioural buffering — nest retreat once surface temperatures exceed foraging optima — supported by correspondence between λ^=0.063 and published foraging limits for these genera, but not demonstrated by direct behavioural measurement.
This buffer is not shared equally. The joint subfamily × temperature interaction is significant (LRT λ^=0.064, λ^=0.065), and only the Ectatomminae contrast — represented entirely by Rhytidoponera — is individually significant: an excess thermal sensitivity of 5% per °C (HR = 1.05, λ^=0.066) above the assemblage response. During extreme heat events (maxt > 30 °C), Rhytidoponera experience a 66% excess hazard (HR = 1.66, 95% CI [1.11, 2.48], λ^=0.067). The species-level PGLS is directionally consistent but non-significant (λ^=0.068/°C, λ^=0.069, p=0.900), and Pagel's p=0.901 indicates no phylogenetic signal in thermal sensitivity — so this is a genus-level pattern, not a conserved clade-wide one.
Synthesis: three axes, three predictors, and a collinearity problem
The central claim is a three-axis reorganisation: body size predicts duration only; senescence trajectory is associated with circadian regime; thermal vulnerability is lineage-patterned. The three signals converge on matinal Rhytidoponera (Ectatomminae) as the candidate most-vulnerable lineage — steep senescence, directionally elevated thermal sensitivity, and matinal activity. Because the plateau sits near the assemblage mean temperature, the operative climate variable is the frequency of extreme-heat days (> 30 °C) rather than mean warming; vulnerability projections built on mean-temperature anomalies would understate risk for heat-sensitive lineages, and size-based vulnerability indices would systematically misrank the most exposed taxa.
The paper is appropriately explicit that this attribution rests on a serious confound: circadian niche and subfamily identity are strongly collinear in this assemblage (Cramér's p=0.902; variance-partitioning shared fraction −0.45). All three Rhytidoponera are matinal and both Myrmecia are crepuscular, so Claims 2 and 3 cannot be fully disentangled at p=0.903; the attribution to circadian niche versus lineage identity is presented as two complementary framings of the same signal, with cross-phylogenetic sampling identified as the key open empirical question.
Limitations
The paper distinguishes confirmatory from hypothesis-generating results. Claim 1 (size–duration, HR = 0.67) and the null size × temperature interaction are robust to sensitivity analyses and adequately powered (87% and ~97%, respectively). Claims 2 and 3 are provisional: the thermal signal rests on three Rhytidoponera species; the PGLS does not reach p=0.904 (retrospective power 41% at p=0.905; minimum detectable effect p=0.906); the Spearman null on p=0.907–mass cannot exclude true correlations as large as p=0.908 at 80% power; and temperature is measured as daily station maxima ~1 km away, which can diverge from microhabitat temperatures by 10–20 °C, potentially confounding thermal-hazard estimates with foraging guild. Three species have field p=0.909 from pooled fallback models, though sensitivity analyses confirm the field-amplification result is not driven by them. The foraging-rate interaction (VIF ≈ 4) and unmeasured confounders at p=0.600 further temper the intrinsic-physiology inference.
Conclusion
This study shows that the size–longevity rule, however robust, captures only one of three components of worker mortality risk in ants. Duration scales with body size in a pattern consistent with intrinsic physiology, independent of colony size and temperature; senescence trajectory is associated with circadian activity regime rather than size; and thermal hazard follows a nonlinear function plateauing above 20 °C, with a genus-specific excess in matinal Rhytidoponera that intensifies during extreme heat. The practical consequence is concrete and testable: predicting ant responses to warming requires tracking extreme-heat-day frequency and weighting circadian regime and lineage identity — not body size — when ranking vulnerable taxa, with cross-phylogenetic replication needed to separate the collinear circadian and lineage signals.
