Our research is a first attempt to study phonological representations in postlingually deaf subjects using a multichannel cochlear implant. Before deafness, these subjects had developed normal language. The present study investigated how phonological representations are assessed through cochlear implant inputs by comparing priming within the auditory modality, and priming in a visual-auditory, cross-modal, condition. Two postlingually deafened adults participated in two lexical decision experiments where word primes were phonologically paired with a word target (e.g. vedette/dette), or a pseudoword target (e.g. banane/nane). The same word primes were also paired with non phonologically related target words (e.g. vedette/chat) and pseudowords (e.g. banane/repe). In addition, the same targets were used in phonologically-related pairs and in phonologically-unrelated ones. Results showed different priming effects for each patient. In one patient, priming was observed for word targets in the unimodal condition only. In the other patient, priming was observed for word targets and interference was observed for pseudoword targets in the cross-modal condition, whereas no effect was observed in the unimodal condition. In addition, this last patient made more errors for pseudowords than for words in the cross-modal condition. These results were interpreted as suggesting that lexical phonological representations participated to priming effects. Moreover, our results suggest that phonological word forms can be activated by visual primes via cochlear implants.