Semantic priming in a recurrent network: Comment on Dalrymple-Alford and Marmurek (1999)

Simulations reported by E. C. Dalrymple-Alfard, and H. C. Marmurek (1999a), using the distributed memory model of semantic priming (M. E. J. Masson, 1 995), suggested that the model simulates priming in a word-naming task by p roducing interference from unrelated primes rather than facilitation from r elated primes, relative to a target-alone condition, and that fully recurre nt networks may not produce priming through indirect pathways (i.e., semant ic units to phonological units). In this article, new simulations show (a) why the model did not produce facilitation relative to a target-alone condi tion in previous simulations, (b) why a priming effect relative to an unrel ated-prime condition nevertheless occurs, and (c) that the model does produ ce facilitation relative to a target-alone condition if he the network is n ot reset to a random state after each trial. Simulation results reported by M. E. J. Masson (1995) are replicated using this method. A postscript resp onds to E. C. Dalrymple-Alford and H. H. C. Marmurek's (1999b) response.