Abstract

The feature and domain-specific models of semantic memory were explored in three experiments involving ten patients with mild to moderate dementia of the Alzheimer type (DAT), and eleven age- and education-matched normal elderly controls. In Experiment 1, the living/nonliving things dissociation was examined using a confrontation-naming task comprising Snodgrass and Vanderwarts' (1980) line drawings of living (e.g., animals), and nonliving (e.g., tools) things. In Experiment 2, verb deficits in general, object/action naming dissociation, and past and present-progressive verb dissociation were explored using data from Experiment 1, and Fiez and Tranel's (1997) standardized Action Naming Test. This test comprised single and paired color pictures of actions, eliciting gerundial (e.g., baking ) and regularly inflected past-tense verbs (e.g., baked ), respectively. In Experiment 3, verb category-specific deficits were explored in a confrontation-naming task that consisted of short-movies of actions that depicted two verb categories: verbs similar in semantic and argument structure (lexical causatives; e.g., peeling ), and verbs similar in semantic content (movement verbs; e.g., crawling , and perception verbs; e.g., watching ). Four main hypotheses were examined: (1) The patients would exhibit a living/nonliving things dissociation with a selective living things deficit; (2) If the patients did exhibit an object/action naming dissociation with a selective verb deficit (e.g., Cappa et al., 1998), this dissociation would hold across two different sets of verb stimuli depicting actions: static (e.g., pictures), and dynamic (e.g., short-movies); (3) Similar to putative category specificity involving noun-labeled categories (e.g., the living/nonliving things dissociation), patients' verb deficits would also be characterized by category specificity; and (4) Following the domain-specific model, the patients' verb deficits would be characterized by a category dissociation specifically affecting the classes of movement and perception verbs; and following the feature-specific models, the patients' verb deficits would be characterized by a category dissociation, specifically affecting the class of lexical causatives. The results revealed that the patients showed a living/nonliving things dissociation with a selective living things deficit substantiating the localization assumption inherent in both domain and feature specific models. Further, when static images were used to depict verbs, both the patients and the controls had significantly more errors naming actions than objects. In contrast, when short-movies were used to depict verbs, no object/action naming dissociation was observed. These results suggested that the object/action naming dissociation observed in both the patient and control groups could have been a function of the verb stimuli (e.g., static images) used. The results further revealed that similar to their noun deficits, patients' verb deficits were also characterized by category dissociation, specifically affecting perception verbs. Patients had significantly more errors naming perception verbs than movement verbs. This dissociation was to some extent attributed to the abstract and context-dependent semantic underpinnings of perception verbs used in a referential processing task and presented to patients with global cognitive deficits. The pattern of verb category-specificity observed suggested that both semantic feature and semantic content are reflected in the organization of semantic memory with the implication that domain and feature specific models are fundamentally similar.