Emotional learning is a pivotal process enabling organisms to predict motivationally significant stimuli in the environment. Previous research has suggested that only specific stimuli that have threatened survival across evolution are learned preferentially. Here, we challenge this view by testing an alternative model, which holds that stimuli detected as relevant to the organism's concerns benefit from preferential emotional learning beyond their valence and evolutionary status. Across several empirical studies, we provide evidence showing that (a) similar to threat-relevant stimuli, positive relevant stimuli are likewise preferentially learned during Pavlovian aversive conditioning, (b) initially neutral stimuli can induce facilitated Pavlovian aversive conditioning after being temporarily associated with goal-relevance, and (c) such preferential learning depends on inter-individual differences in motivation. These findings suggest that learning biases in Pavlovian conditioning are driven by a relevance detection mechanism that is not specific to threat, and foster new insights into the basic mechanisms underlying emotional learning.