The ability to learn visual-phonological associations is a unique predictor of word reading, and individuals with developmental dyslexia show impaired ability in learning these associations. In this study, we compared developmentally dyslexic and nondyslexic adults on their ability to form cross-modal associations (or "bindings") based on a single exposure to pairs of visual and phonological features. Reading groups were therefore compared on the very early stages of associative learning. We used a working memory framework--including experimental designs used to investigate cross-modal binding. Two change-detection experiments showed a group discrepancy in binding that was dependent on spatial location encoding: Whereas group performance was similar when location was an inconsistent cue (Experiment 1), nondyslexic readers showed higher accuracy in binding than dyslexics when location was a consistent cue (Experiment 2). A cued-recall task confirmed that location information discriminates binding ability between reading groups in a more explicit memory recall task (Experiment 3). Our results show that recall for ephemeral cross-modal bindings is supported by location information in nondyslexics, but this information cannot be used to similar effect in dyslexic readers. Our findings support previous demonstrations of cross-modal association difficulty in dyslexia and show that a group discrepancy exists even in a single, initial presentation of visual-phonological pairs. Effective use of location information as a retrieval cue is one mechanism that discriminates reading groups, which may contribute to the longer term cross-modal association problems characteristic of dyslexia.