Three experiments compared the qualitative pattern of participants' word-naming performance in a syntactic priming task with the qualitative pattern of performance generated by a recurrent network model. Experiments 1 and 2 demonstrated that when participants had a 600-ms response deadline, the appropriateness of a syntactic prime affected their naming times for high-frequency words but not low-frequency words. However, Experiment 2 also demonstrated that participants made the most pronunciation errors when naming inconsistent low-frequency words (e.g., pint) that were preceded by an inappropriate prime. The results of Experiment 3 suggest that participants' naming times to both high- and low-frequency words are affected by syntactic primes when there is no response deadline. The implication of these findings for the study of syntactic priming in English and other languages is discussed.