Psychopathy, a multi-faceted construct characterised by antisocial behaviours and callous-unemotional traits, is associated with atypical learning. However, the nature of this atypical learning is unclear due to inconsistencies in the literature. Attentional and outcome processing theories of psychopathy do not provide an adequate explanation of these inconsistencies, but I argue that an attentional-associative account can. For instance, Mackintosh's (1975) theory can be used to interpret learning differences in psychopathy through disruption in establishing a cue's associability: a representation of the relative validity of a cue used to determine attentional allocation, and so how easily the cue is learnt about. The discrepancy between previous studies can be attributed to methodological differences and how these differentially influence the associability of a stimulus. In a series of experiments I test this hypothesis and the conditions under which learning differences are seen in psychopathy. I report that psychopathic traits predict altered learning about inhibitory cues, but not excitatory cues, and show that these effects are sensitive to manipulation of predictiveness and cognitive load. By incorporating principles of reduced processing capacity in psychopathic individuals into an attentional-associative model, I am able to explain the results of my experiments, and provide an account for previous discrepancies in the literature.