A lexical decision task was used with a masked priming procedure to investigate whether and to what extent neighborhood distribution influences the effect of prime duration in masked orthographic priming. French word targets had two higher frequency neighbors that were either distributed over two letter positions (e.g., "LOBE/robe-loge") or concentrated on a single letter position (e.g., "FARD/tard-lard"). Word targets were preceded by their highest frequency neighbor or by a control prime. Four prime durations were compared (27, 39, 53, and 67 ms). Results showed that the inhibitory priming effect found for words with distributed neighbors at a 67-ms stimulus onset asynchrony (SOA) was canceled when prime duration decreased. In contrast, no priming effect was found in any of the four prime durations for words with concentrated neighbors. Simulations run on the word materials revealed that the interactive activation model (McClelland & Rumelhart, 1981) captured the increasing inhibitory priming effect in the distributed neighbor condition but failed to capture the loss of priming in the concentrated neighbor condition. (Contains 2 notes, 3 tables, and 1 figure.)