Audio-visual (AV) associations are central to many aspects of behavior, including the initial steps of learning to read. The acquisition of AV pairings has been explored in individuals with varying literacy skills, including children with developmental dyslexia. Most previous studies examined performance in AV associative tasks looking at the pairings between linguistic auditory material and visual stimuli, thus confounding AV learning with phonological and/or verbal abilities. In the present study, we introduce an AV learning paradigm relying on non-linguistic auditory stimuli and novel visual shapes. We fit trial-by-trial performance and compare the response patterns of 52 Italian-speaking children with developmental dyslexia (DD) with those of age-matched (N = 54) and of younger, reading-matched (N = 51) typically-developing children. All groups showed increasing accuracy across trials, but children with DD learned less efficiently than their peers. These findings suggest that difficulties in forming AV associations through repeated exposure may underlie dyslexia, even when linguistic demands are minimized.