Psychological theories of positive and negative repetition priming have relied on mechanisms such as residual activity/active suppression or episodic match/mismatch. Here we propose an alternative theory developed from a cognitive neuroscience perspective in which positive and negative priming reflect neural learning mechanisms that serve to develop and shape long-term knowledge representations in the brain. We present a biologically constrained connectionist model that incorporates "biased-competition" mechanisms of attentional selection and a synaptic plasticity rule derived from neuroscience studies. In the model, we show that positive priming results from high neural activity levels and overall synaptic strengthening, and negative priming results from lower activity levels and overall synaptic weakening. The competitive interactions that are claimed to occur during attentional selection and the way that these interact with the plasticity rule also make novel behavioral predictions that we are evaluating through complementary empirical studies by manipulating the relative brightness and familiarity of stimuli.