In a recent study using a masked priming lexical decision task, New, Araujo, and Nazzi found priming of targets by primes sharing consonants (jalu-JOLI) but not by primes sharing vowels (vobi-JOLI). To examine the orthographic or phonological/lexical nature of this consonantal bias, and to determine whether vocalic priming can be obtained under different presentation conditions, we manipulated the duration of prime presentation. Experiment 1 examined masked priming effects for consonant- or vowel-related primes with 33 ms prime durations, while Experiments 2 and 3 examined masked priming effects with longer prime processing times (66 ms or 50 + 16 ms). Results replicated the relative advantage of consonant over vocalic priming, and established that the consonant bias was not at the orthographic level but rather at the phonological and lexical levels. Furthermore, primes sharing vowels with the target showed inhibition for longer processing times. The implications of these findings for models of visual word recognition are discussed.