ABSTRACT
The models of language most widely used to analyze functional damage are constructed in terms of discrete processing components that operate independently of each other. From this assumption, it follows that the stimulus dimensions affecting performance should point directly to the general location of the defective component within the language mechanism: Sensitivity to attributes involving only word-specific dimensions (e.g., the status of the target as a real word) would indicate the disturbance occurs after the level where the structural aspects of input and output codes are established. If variables related to structure (e.g., word length) do exert a major influence, the conclusion is that damage has taken place to encoding procedures which operate before the categorization of perceptual information as higher level units. We present evidence from three brain-damage patients that creates some difficulty for this general approach. Analysis of the cases suggests that relatively central components of the language mechanism are influencing ongoing events at more peripheral stages that compute a structural description. Such interactivity qualifies the notion of a completely transparent relationship between a disordered functional component and performance.
