Abstract

In this dissertation an attempt was made to uncover the source of verbal associations. The investigation focused on establishing the locus of representation for associative relationships in the cognitive system and whether this locus is different from that for semantic relationships. A picture naming task and an object decision task were used within the standard priming paradigm, in which the target is preceded by a prime. A dual-level model was proposed in which associative relatedness is represented at a lemma level that connects the lexical form representation of a word to its semantic information. According to this model an interaction between associative and categorical relatedness should occur in picture naming, but not in object decision, when primes and targets share both relationships, and this is what was observed. To investigate the mechanisms of associative priming, asymmetrically associated prime-target pairs were used to create two situations. In the forward priming condition the target was an associate of the prime (e.g., brick-house), and in the backward priming condition the prime was an associate of the target (e.g., babyrattle). Unexpectedly, facilitation was observed for backward priming at the short SOA in picture naming. Because no effect was observed for this condition in the object decision task, and given that forward priming produced facilitation in both tasks spreading activation was upheld as the mechanism for associative priming. In order to investigate whether the source of the relationship between associates might be in their latent semantic content, the impact of instrument relationships (e.g., grinder-coffee), script relationships (e.g., zoo-tiger), and proximity in multidimensional semantic space were also investigated in the picture naming task. Items that were close in semantic space, but did not share any semantic relationships, produced the same priming pattern as category co-ordinates in picture naming (i.e., interference), while instrumental and script relationships did not produce a priming pattern that matched either that observed for associative or categorical relatedness. These results were taken to indicate that the source of associative relationships is in the co-occurrence of words in the language, which further supported the main claim of a dual-level model where information about verbal associations is stored outside semantic memory