A myriad of different deficits have been evidenced in dyslexic subjects, but the underlying causes of dyslexia are unclear. We hypothesised that reading difficulties could be partly due to difficulty forming grapheme-phoneme correspondences, as part of a more global impairment forming appropriate associations between related information in working memory, which may also affect within-modal binding of visual feature information. In two experiments, we assessed the efficacy with which dyslexic readers can integrate individual feature information into unified objects and maintain these integrated representations in visual working memory, compared to a non-dyslexic group. In Experiment 1, we employed a recognition (change-detection) test to directly assess memory for feature-conjunctions compared to memory for individual features. In Experiment 2 we varied the amount of feature information to be remembered in a cued-recall test to examine whether integrated object representations or individual features constituted the unit of visual working memory capacity. Results demonstrated that both dyslexic and non-dyslexic groups stored feature information as integrated object representations. Dyslexic readers’ overall accuracy was not less than non-dyslexic readers in either experiment, but qualitative differences in memory across different conditions in cued-recall were revealed. These results are applied to existing memory literature and causal hypotheses of dyslexia. Most importantly, this study provides evidence that any binding deficit in dyslexic readers may be limited to cross-modal associations and binding in the visual domain is not likely to be affected.