The exploration of a familiar object by hand can benefit its identification by eye. What is unclear is how much this multisensory cross-talk reflects shared shape representations versus generic semantic associations. Here, we compare several simultaneous priming conditions to isolate the potential contributions of shape and semantics in haptic-to-visual priming. Participants explored a familiar object manually (haptic prime) while trying to name a visual object that was gradually revealed in increments of spatial resolution. Shape priming was isolated in a comparison of identity priming (shared semantic category and shape) with category priming (same category, but different shapes). Semantic priming was indexed by the comparisons of category priming with unrelated haptic primes. The results showed that both factors mediated priming, but that their relative weights depended on the reliability of the visual information. Semantic priming dominated in Experiment 1, when participants were free to use high-resolution visual information, but shape priming played a stronger role in Experiment 2, when participants were forced to respond with less reliable visual information. These results support the structural description hypothesis of haptic-visual priming (Reales and Ballesteros in J Exp Psychol Learn Mem Cogn 25:644-663, 1999) and are also consistent with the optimal integration theory (Ernst and Banks in Nature 415:429-433, 2002), which proposes a close coupling between the reliability of sensory signals and their weight in decision making.