Repetition priming and frequency attenuation in lexical access.

TLDR

Repetition priming in lexical decision tasks is stronger for low‑frequency words, creating a frequency‑attenuation effect that challenges frequency‑ordered search models and has attracted attention due to the marked speed and accuracy gains for repeatedly presented words. The study proposes that frequency attenuation arises from episodic memory involvement in lexical decision processing. Masked primes produced constant repetition effects across frequencies, while long‑term repetition effects were unreliable without a lexical‑decision response and returned when a response was required, indicating that repetition effects comprise a brief, frequency‑independent lexical component and a longer‑term, frequency‑sensitive episodic component.

Abstract

Repetition priming effects in lexical decision tasks are stronger for low-frequency words than for high-frequency words. This frequency attenuation effect creates problems for frequency-ordered search models that assume a relatively stable frequency effect. The suggestion is made that frequency attenuation is a product of the involvement of the episodic memory system in the lexical decision process. This hypothesis is supported by the demonstration of constant repetition effects for high- and low-frequency words when the priming stimulus is masked; the masking is assumed to minimize the influence of any possible episodic trace of the prime. It is further shown that long-term repetition effects are much less reliable when the subject is not required to make a lexical decision response to the prime. When a response is required, the expected frequency attenuation effect is restored. It is concluded that normal repetition effects consist of two components: a very brief lexical effect that is independent of frequency and a long-term episodic effect that is sensitive to frequency. There has been much recent interest in the fact that in a lexical decision experiment, where subjects are required to classify letter strings as words or nonwords, there is a substantial increase in both the speed and the accuracy of classificatio n for words that are presented more than once during the experiment, even though considerable time may have elapsed between successive presen