This paper argues that explanations for the origins of lexical and non lexical (recurring utterances) aphasic speech automatisms must take into account: (a) both neurophysiological as well as environmental factors; (b) initial and subsequent productions of the utterance. A preliminary model is proposed which accounts for initial and subsequent productions and provides the basis for further investigation into the neurogenesis of different subtypes of speech automatism. The model acknowledges the lack of straight linguistic input, it reflects the apparent holistic preparation and the invariance of production and it distinguishes between the two major types (lexical and non lexical).