Environmental enrichment (EE) reduces stereotypic behaviour (SB), but typically only partially. Using American mink (n = 17) as models, we tested the hypotheses that the effectiveness of EE reflects the degree to which subjects utilise it, and also the SB's degree of ‘establishment’ (its frequency and within-bout predictability). In Non-Enriched cages, our subjects performed Carnivora-typical Locomotor SBs; some also stereotypically scrabbled against the cage walls. Each mink was then enriched: re-housed in a cage connected by an overhead tunnel to a compartment enriched with manipulable objects and running water. Here, both types of SB declined (Locomotor SB: from 5.8% to 0.9% of observations; Scrabbling: from 1.6% to 0.2%). Individual differences in these lowered SB levels were stable, correlating positively between observation periods (ρ ≥ +0.66). Similar stability was evident in all measures of enrichment use (partial r ≥ + 0.74). However, predictions that EE would be least effective for ‘established’, i.e. highly frequent or predictable, SBs were not supported. Furthermore, nor did low enrichment use predict small reductions in SB: instead, unexpectedly, frequent interactions with enrichment items predicted the smallest absolute reductions in Locomotor SB (partial r = + 0.55). This was because the smallest reductions in this SB occurred in mink with the lowest pre-enrichment levels (partial r = −0.60), who also interacted most with the enrichments (partial r = −0.52). Mink with high pre-enrichment levels of Locomotor SB, in contrast, rarely interacted with the enrichments, instead becoming more inactive (partial r = + 0.52). These individuals were also perseverative (displaying more inappropriate response repetition in a previous food-rewarded guessing task), thence potentially behaviourally inflexible. However, their perseveration did not predict less interaction with enrichments. Furthermore, these animals did markedly alter their behaviour when enriched, reducing their SB and shifting their preferred resting locations from home cages to the overhead tunnels (partial r = + 0.70): individual differences that were, again, stable over time (co-variance across observation periods for inactivity in different areas: partial r ≥ +0.65). Individuals thus used enriched environments in consistent, qualitatively different ways that were predicted by pre-enrichment levels of Locomotor SB: low-SB mink became more active when enriched, frequently engaging with manipulable objects and running water, while high-SB subjects instead shifted resting locations and became more inactive. These divergent response styles are unexpected: not predicted by any current understanding of SB. The underlying mechanisms and generality to other species therefore now need investigation.