Despite a large consensus on increasing facilitation among plants with increasing stress in alpine regions, a number of different outcomes of interaction have been observed, which impedes the generalisation of the ‘stress-gradient hypothesis’ (SGH). With the aim to reconcile the different viewpoints on the stress-interaction relationship in alpine environments we hypothesized that fine nurse variations within a single life form (cushion) may explain this pattern variability To test this hypothesis, we compared the magnitude of the stress-interaction relationship in a single study area with that observed in existing studies involving cushions, worldwide. We characterized the nurse effects of cushions on the whole plant community at inter-specific, intra-specific and intra-individual levels along a stress gradient in the dry, alpine tropics of Bolivia (4400 m, 4700 m and 4900 m a.s.l). Using a relative index of interaction (RII) we included our data in a meta-analysis on the nurse effects of cushions along alpine gradients, worldwide. At inter-specific level, the loose cushion Pycnophyllum was a better nurse than the compact Azorella compacta. However, at intra-individual level facilitation was higher at the periphery than at the centre of cushions, exceeding in magnitude the variation observed at inter-specific level. This pattern was associated with higher minimum temperature and lower mortality at the periphery of cushions. The net effects of cushions on plant communities became more positive at higher elevation, corroborating the SGH. Within our single site in Bolivia, fine morphological nurse variations captured a similar variability in the stress-interaction relationship as that observed in a subset of studies on cushions on a worldwide scale. This suggests that fine variations in nurse traits, in general those not considered in protocols dealing with facilitation or in restoration/conservation management plans, explain in part the current discrepancies among SGH studies in alpine regions.