If we want to understand or predict future ecosystem composition, it is important to validate whether well-known traits of animals remain ecologically relevant. Highly specialised species, for example, may die or be forced to relocate, as habitats change. But what about more nuanced shifts or adjustments to new circumstances? Obligate coral-dwelling damselfishes on coral reefs are known to depend on live branching coral to survive. However, in recent studies, they were observed to survive the localized extinction of their primary habitat following coral bleaching. To address this apparent paradox in the trait of coral dependence, we documented the fine-scale spatial behaviour of obligate coral-dwellers in relation to habitat quality. We found very little effect of habitat quality on space use. Thus, obligate coral-dwelling fishes may prefer branching live coral, but their ‘obligate’ dependence may be more flexible and context dependent. As ecosystems reconfigure, plasticity in behavioural traits may be critical for the persistence of fish populations. Likewise, plasticity may be required of us – attempting to quantify ecology using traits. Rather than relying on pre-existing measurements and retrofitting ecological relevance to them, the key may be to invert the process, i.e. to identify context-specific ecological processes first and then identify traits as quantifiable proxies.