Academic journal article Psychonomic Bulletin & Review

The Turple Effect Is Modulated by Base Word Frequency: Implications for Models of Lexical and Semantic Access

Academic journal article Psychonomic Bulletin & Review

The Turple Effect Is Modulated by Base Word Frequency: Implications for Models of Lexical and Semantic Access

Article excerpt

People who are asked to classify whether words presented visually belong to the category of animals respond to nonwords derived from animal names more slowly than they do to nonwords derived from nonanimal names. This is known as the turple effect (Forster, 2006; Forster & Hector, 2002). In the present article, we show that the turple effect is modulated by the frequency of the animal names from which the nonwords are derived: In particular, we show that nonwords derived from high-frequency animal names are rejected faster than those derived from low-frequency animal names. We discuss the implications of this result for two approaches to lexical and semantic access modeling.

The aim of the present study is to evaluate two models of lexical and semantic access from printed words: Pecher, Zeelenberg, and Wagenmakers's (2005) model (PZW) and the links model (Forster, 2006; Forster & Hector, 2002). These models can be seen as representing two classes: cascaded models (e.g., Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) and staged models (e.g., Morton, 1969). In cascaded models, activation in one module flows on continuously to successive ones without intermodule thresholds. In staged models, processing at each level must be completed before activation passes to the next one. Which approach to modeling cognition better fits the empirical evidence is a crucial issue in theories of word processing (e.g., Navarrete & Costa, 2005).

To achieve our goal, we made use of the turple effect (Forster, 2006; Forster & Hector, 2002). In a semantic decision task that uses animals as exemplars, nonwords derived from animal names are rejected more slowly than are nonwords derived from nonexemplar names and nonwords without neighbors. Responses to the latter two types of stimuli do not differ.

How Cascaded Models Account for the Turple Effect

Carreiras, Perea, and Grainger (1997) proposed a cascaded model of semantic decision (CPG) as an extension to the multiple read-out model of visual word recognition (Grainger & Jacobs, 1996). Within the CPG model, decisions about whether the stimuli belong to the category of animals are based on the activation level in an "animalness" feature unit. If the activation level reaches a given criterion, a "yes" decision is generated; "no" decisions are produced when a "yes" decision is not reached within a given amount of time (i.e., when a deadline expires). However, such a model cannot account for the turple effect, because its "no" responses are based on a deadline, irrespective of the type of stimulus (nonexemplar words, nonwords derived from animal names, or nonwords derived from nonanimal names). Such an approach runs against the empirical evidence showing that the neighbors of nonwords are relevant to the decision. The CPG model comprises a mechanism that solves this problem: If low levels of activation (i.e., below the criterion) are detected in the animalness unit, the deadline is extended, because the stimulus could be a lowfrequency animal name. Through their base words, nonwords such as turple activate the animalness units enough to cause an extension of the deadline, although not of the criterion. This extension leads to slower responses.

A shortcoming of this model is that it does not account for a well-known effect: High-frequency nonexemplar words are rejected faster than low-frequency nonexemplar words (Monsell, Doyle, & Haggard, 1989). If "no" decisions are generated when a deadline expires, as posited by the CPG model, the decision time would be the same for all nonexemplar words; that is, the frequency of the nonexemplar words should be irrelevant. Pecher et al. (2005) proposed a cascaded model (PZW) that accounts for the frequency effect of nonexemplar words. Although it has no deadline, the PZW model has one unit that collects evidence for a "yes" decision and another that collects evidence for a "no" decision.1 If activation in those two units rises faster for high-frequency words, then decisions are generated earlier for high- than for low-frequency words, regardless of whether the response is "yes" or "no. …

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


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.