Academic journal article Memory & Cognition

An RT Distribution Analysis of Relatedness Proportion Effects in Lexical Decision and Semantic Categorization Reveals Different Mechanisms

Academic journal article Memory & Cognition

An RT Distribution Analysis of Relatedness Proportion Effects in Lexical Decision and Semantic Categorization Reveals Different Mechanisms

Article excerpt

Published online: 24 My 2014

© Psychonomic Society, Inc. 2014

Abstract The magnitude of the semantic priming effect is known to increase as the proportion of related prime-target pairs in an experiment increases. This relatedness proportion (RP) effect was studied in a lexical decision task at a short prime-target stimulus onset asynchrony (240 ms), which is widely assumed to preclude strategic prospective usage of the prime. The analysis of the reaction time (RT) distribution suggested that the observed RP effect reflected a modulation of a retrospective semantic matching process. The pattern of the RP effect on the RT distribution found here is contrasted to that reported in De Wit and Kinoshita's (2014) semantic categorization study, and it is concluded that the RP effect is driven by different underlying mechanisms in lexical decision and semantic categorization.

Keywords Semantic priming . Relatedness proportion . RT distribution analysis . Lexical decision . Retrospective semantic matching

The size of the semantic priming effect increases when the proportion of related prime-target pairs in the experimental list increases. This finding, first reported by Tweedy, Lapinski, and Schvaneveldt (1977), is referred to as the relatedness proportion (RP) effect and is generally taken as an indicator of strategic use of the prime. The standard explanation of the RP effect is in terms of the expectancy generation strategy. Specifically, participants generate expectancies regarding the identity of the target on the basis of the semantic properties of the prime (Becker, 1980; Neely, 1977), and they are more likely to do so when the proportion of related prime-target pairs is high and, hence, the expectation is likely to be correct on most trials. Consistent with the assumption that it takes time to generate expectancies, reliable RP effects are generally found only at long prime-target stimulus onset asynchronies (SOAs; see Hutchison, 2007, for a review). As was noted by Hutchison, Neely, and Johnson (2001), "when the SOA is short (under 300 ms), RP often has little or no effect on semantic priming" (p. 1451). On the basis of the two-process model of semantic priming (Neely, 1977, 1991; Posner & Snyder, 1975), which explains semantic priming effects in terms of a fast-acting, automatic spreading activation process and a slow-acting controlled process, it has been widely assumed that semantic priming effects obtained at short SOAs of 250 ms or less reflect the former. There is a growing recognition, however, that this assumption may be incorrect.

One line of evidence comes from studies that found RP effects at short SOAs (e.g., De Groot, 1984; De Wit & Kinoshita, 2014). De Groot used the lexical decision task and manipulated the proportion of related prime-target pairs (RPs of .25, .50, .75, and 1.00) at different SOAs (240, 540, and 1,040 ms). She found robust RP effects at all levels of SOAs, and even at the shortest SOA of 240 ms, which is generally considered too short for the expectancy generation strategy to be used, the size of semantic effect was greater (by 16 ms) in the .75 RP condition than in the .25 RP condition. De Wit and Kinoshita also manipulated RP (.25 vs. .75 related pairs) at a short SOA of 240 ms but used a semantic categorization task ("Is it an animal?") instead of a lexical decision task. They too found that the semantic priming effect was larger (by 47 ms) in the high RP than in the low RP condition.

The present study investigates the mechanism responsible for the RP effect at a short SOA in the lexical decision task, using RT distribution analyses, and contrasts it with the effect recently reported with the semantic categorization task (De Wit & Kinoshita, 2014). As will be discussed shortly, an analysis of the RT distribution can be more informative than the analysis of mean RT alone. De Wit and Kinoshita found that in their semantic categorization task, the semantic priming effect was reflected asa shift in the RT distribution and that the RP manipulation further increased the amount of shift (the theoretical implication of this pattern of RT distribution will be described later in the Discussion section). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.