Psychonomic Bulletin and Review (in press)
黄瓜火腿肠
Repetition and form priming interact with neighborhood density at a brief stimulus-
ont asynchrony
蒸菜Manuel Perea and Eva Rosa
Universitat de València
RUNNING HEAD: Repetition Priming and Neighborhood
Correspondence:
Manuel Perea
Departament de Metodologia
秦王墓
Facultat de Psicologia
Av. Blasco Ibáñez, 21
46010-València (Spain)
Fax: (34) 96 3864668月考之后
e-mail: mperea@uv.es
Abstract
The relationships between repetition/form priming effects and neighborhood density were analyzed in two masked priming experiments with the lexical decision task. Given that form priming effects appear to be influenced by a word's orthographic neighborhood, it is theoretically important to find out whether or not repetition priming also differs as a function of the word’s orthographic neighborhood. Within an activation framework, repetition and form priming effects are just quantitatively different phenomena, whereas the two effects are qualitatively different in a rial-ordered model of lexical access (the “entry-opening” model). The results show that repetition and form priming effects were stronger for “hermit” words than for words with many neighbors. The results po some problems for both activation and rial-ordered models. The implications of the results for determining how neighbors affect the identification of a word are discusd.
名词作定语Repetition and form priming interact with neighborhood density at a brief stimulus-
ont asynchrony
Investigations of the effects of lexical similarity provide valuable information into the process underlying word recognition and a number of papers reporting the results of such investigations have appeared recently (e.g., Andrews, 1989, 1992, 1996; Carreiras, Perea, & Grainger, 1997; Forster & Shen, 1996; Grainger, 1990; Grainger & Jacobs, 1996; Grainger, O'Regan, Jacobs, & Seguí, 1989, 1992; Huntsman & Lima, 1996; Paap & Johann, 1994; Perea & Pollatk, 1998; Pollatk, Perea, & Binder, 1999; Sears, Hino, & Lupker, 1995; Snodgrass & Mintzer, 1993). The data indicate that, upon the visual prentation of a word, similarly spelled words (the so-called “neighbors”) become partially activated and affect the speed of lexical access. Virtually all of the experiments have adopted Coltheart, Davelaar, Jonasson, and Besner's (1977) definition of orthographic neighbor —any word that can be created by changing one letter of the stimulus word, prerving letter positions (e.g., peace, poach, and beach are orthographic neighbors of peach)— and have defined the neighborhood of a word to be the t of neighbors of that word (or N).加拿大国花
One interesting way of investigating neighborhood structure is by prenting a prime followed by a neighboring target (or the identical target). In this context, the masked priming technique (Forster, 1987, 1998; Forster & Davis, 1984; Forster, Davis, Schoknecht, & Carter, 1987) has been the most fr
uitful paradigm to study competition process at the earliest stages of word recognition. The priming stimulus is orthographically and/or phonologically related to the target and is prented briefly (30-66 ms) just prior to the target. The prime is preceded by a forward pattern mask and, under the conditions, the trace of the prime is relatively inaccessible to conscious report. (Obviously, the fact that the prime is replaced by the target at a very short stimulus-ont asynchrony —SOA— does not necessarily imply that the prime is no longer procesd once the target replaces the prime.)
Masked priming effects with orthographic neighbors
Prior rearch with the masked priming technique has found that target words are primed by orthographically similar nonword primes (relative to an unrelated control condition), although the effects are restricted to target words extracted from small neighborhoods (e.g., album) in both lexical decision (e.g., Forster, 1987; Forster et al., 1987; Forster & Taft, 1994) and naming tasks (Forster & Davis, 1991). In addition, inhibitory relatedness effects have been obtained with orthographically related word primes which are more frequent than the word target (e.g., blue-BLUR; Bijeljac-Babic, Biardeau, & Grainger, 1997; Ferrand & Grainger, 1994; Grainger, Colé, & Seguí, 1991; Perea & Rosa, 1998; Seguí & Grainger, 1990), although this issue remains controversial (e Forster, 1987; 恒星英语网
Forster & Veres, 1998) (NOTE 1). Furthermore, associative, morphological and translation priming effects have also been obtained with the masked priming technique (e.g., de Groot & Nas, 1991; Gollan, Forster, & Frost, 1997; Lukatela & Turvey, 1994; Perea & Gotor, 1997; Williams, 1994). Taken together, the results strongly suggest that masked priming effects occur at the lexical level rather than at a sublexical level.
The form priming effects from the above-cited studies can be readily explained in terms of an interactive activation model (McClelland & Rumelhart, 1981; e also Ferrand & Grainger, 1992; Grainger, 1992). In this model, the lexical entries corresponding to the more frequent words have higher resting levels than the units corresponding to less frequent words. In addition, there is mutual inhibition among the candidates at the lexical level, and a lexical unit is recognized when its level of activation reaches a pre-specified decision criterion or when its level of activation ris significantly above the activation level of other candidates. The interactive activation model captures the existence of inhibition when the orthographically related prime is of higher frequency than the target (e Jacobs & Grainger, 1992; Simulation 2): When the prime (e.g. blue)
is a higher frequency neighbor of the target (blur), during the processing of the target blur the node of blue is even more activated than that of blur, thus increasing the inhibition on the node for blur (co
mpared with an unrelated control condition).
In addition, the interactive activation model can also capture the facilitative priming effects with nonword primes (via sublexical activation). Activation from the nonword primes at sublexical levels (letters, letter clusters, phonological units, syllabic units) feeds forward to the lexical level, with the conquent top-down feedback. As nonword primes activate lexical reprentations more and more compared to target activation, then lexical level competition will appear, canceling out sublexical facilitation (e Ferrand & Grainger, 1994). In the ca of target words with many neighbors, the activation from the related nonword primes may have spread along the lexical level so that inhibition between word units will cancel out sublexical facilitation (e Forster et al., 1987; Experiment 5). In the ca of words with no neighbors (“hermit” words; e.g., typhus), the amount of lexical inhibition from the target's similar words will be negligible and only facilitation will be obtained. That is, the density constraint for form primes appears to fall out as a natural conquence of the recognition process in the interactive activation model (e General Discussion for the simulations on the interactive activation model).
Given that form priming effects appear to be influenced by a word's orthographic neighborhood, it is theoretically important to find out whether or not repetition priming also differs as a function of a word
’s orthographic neighborhood. In an activation framework, repetition priming and form priming effects are just quantitatively different phenomena: If the inhibitory effects of masked primes are indeed preactivation effects, as suggested by Seguí and Grainger (1990), it might be expected that words with many neighbors would show less masked repetition priming effects than words with few or no neighbors. The masked prentation of a word with many neighbors (e.g., peach) should produce a significant ri in the activation level of its (high-frequency) neighbors (e.g.,
beach, peace, reach, etc.). The activation levels of the reprentations will continue to be (partially) supported by information from the target word. So, the reprentations that remain in a heightened state of activation during target processing will influence target recognition (via lateral inhibition between word nodes). In contrast, if only one word node is activated (e.g., hermit words), no significant inhibition from the other word nodes will be produced. That result would strengthen the argument that competition among candidates (via lateral inhibition) plays a role in visual word recognition. Of cour, specific predictions can only be achieved by running computer simulations (e General Discussion for the simulations on the interactive activation model).
In the framework of a rial-ordered model of lexical access (“entry-opening” model), Forster and Davis (1984) suggested that immediately after an entry has been accesd, it is left in a moderately 爆笑喜剧小品大全
excited state so that information can be extracted from it more rapidly. As a result, the time for other processing systems to extract information from the entry might decrea (e Forster & Davis, 1984). In this model, neighborhood density is relevant only to form priming effects. Specifically, form priming is considered a special ca of repetition priming in which the entry for the target has been previously opened. Given that clo matches are ignored in a den neighborhood (in order to keep the number of candidates to a reasonable number), the model predicts an interaction between neighborhood density and form priming. In contrast, the model clearly predicts additive effects of neighborhood density and repetition priming, since an exact match will always open the entry for the target word. (NOTE 2). Forster and Taft (1994) claimed that “there is no sign that repetition priming is affected by neighborhood density, as shown by Forster and Davis (1984) and Forster et al. (1987)” (p. 858). However, Forster and colleagues did not systematically investigate that question. It may be important to note that Forster et al. (1987; Experiment 3) reported a 38 ms repetition effect for four-letter words and a 56 ms repetition effect for eight-letter words. Since length is highly correlated with neighborhood density, words with few