The work of connectionist researchers is examined in order to understand better the implications for modeling second language learning processes. Connectionism is a biologically-oriented framework for understanding complex behavior, and provides a modeling tool (computer simulation) that behaves and learns without rules being explicitly wired into it. Its origins in cognitive science are traced to the 1950s, and its evolution within the field of artificial intelligence is reviewed briefly. It is noted that while there has been some discussion about parallel distributed processing (PDP) and its potential for understanding cognitive processes in the literature of second language acquisition (SLA), there has been little empirical work involving computer simulation of SLA. Several studies have addressed the utility of computer modeling for explaining some discrete SLA phenomena. Some arguments against connectionism are also found. It is concluded that if the more advanced PDP models can overcome some of the current problems and can allow predictions to be made about real second language learners, connectionism can be useful to SLA researchers. (Contains 26 references.) (MSE)