A neural network model of backward priming
Semantic priming refers to facilitation in the recognition or naming of a word (the target, e.g., KING) following a semantically related word (the prime, e.g. CROWN). Backward priming demonstrates similar effects for related words presented in the reverse order. The first goal of this study was to examine priming effects in both directions for asymmetrically associated words. It also investigated the contributions of automatic and strategic processing mechanisms, lexical frequency, semantic relatedness, and association strength to priming in a lexical decision task. The second goal was to develop a neural network model capable of simulating those priming effects. Results showed that both the human participants and the neural network model displayed both forward and backward priming effects. The results of the human participants and model were then compared in order to evaluate the neural network as a model of human lexical processing.