Why does this matter? Climate change can alter where species occur and when plants flower or pollinators are active. We did not test future changes directly, but our results suggest that pollination networks will not all respond in the same way. 8/8

The strongest common thread was climate. Temperature and rainfall were generally better-supported predictors of specialisation than latitude, biodiversity or ecosystem productivity. Yet their effects were nonlinear and differed greatly among pollinator groups. 7/8

The classic expectations were not supported. Specialisation neither consistently increase towards the Equator, nor towards the poles. Network-level and pollinator specialisation often peaked around 20–30° N, near the tropical–subtropical boundary. 5/8

Besides the data size, we treated limits of previous studies: controlled sampling completeness, separated community-wide networks from group-focused ones, and analysed plants and pollinator groups separately. This distinguished broad trends from group-specific responses. 4/8

To test this, we compiled 3,415 quantitative networks with >110,000 links among 5,343 pollinator and 6,126 plant species across all major terrestrial biomes. The study brought together 142 researchers from 32 countries. 3/8

Specialisation is an evolutionary bargain. A close match can make feeding and pollen transfer highly efficient, but it also creates dependence on particular partners. Ecologists have long expected such close relationships to be especially common in the tropics. 2/8

New study published in @NatureEcoEvo! We assembled the largest global dataset of plant–pollinator networks to ask a simple question: do interactions become more specialised towards the tropics? doi.org/10.1038/s41559-026-0… Study led by @SaileeSakhalkar @RobertTropek 1/8

Empirical elevational networks partly follow these MDE predictions, but also show clear deviations. 👉 Mid-domain effect matters, although biology leaves a major fingerprint. Ignoring MDE risks misinterpreting null patterns as biological signals.

Our null-model simulations show that MDE alone can generate strong structure: - unimodal peaks in realised links, generality and vulnerability - U-shaped patterns in connectance and often nestedness All emerging from geometry, not biology...

If species overlap most in the middle of a gradient, interactions among overlapping species should too. Under MDE alone, network metrics are predicted to show unimodal or U-shaped patterns, even without any niche differences or environmental filtering.