|
COPYRIGHT 2001 Massachusetts Institute of Technology
Nicolas Dumay [1,2]
Abdelrhani Benraiss [3]
Brian Barriol [3]
Cecile Colin [2]
Monique Radeau [1,2]
Mireille Besson [3]
Abstract
* Phonological priming between bisyllabic (CV.CVC) spoken items was examined using both behavioral (reaction times, RTs) and electrophysiological (event-related potentials, ERPs) measures. Word and pseudoword targets were preceded by pseudoword primes. Different types of final phonological overlap between prime and target were compared. Critical pairs shared the last syllable, the rime or the coda, while unrelated pairs were used as controls. Participants performed a target shadowing task in Experiment 1 and a delayed lexical decision task in Experiment 2. RTs were measured in the first experiment and ERPs were recorded in the second experiment. The RT experiment was carried out under two presentation conditions, In Condition 1 both primes and targets were presented auditorily, while in Condition 2 the primes were presented visually and the targets auditorily. Priming effects were found in the unimodal condition only. RTs were fastest for syllable overlap, intermediate for rime overlap, and slowest for coda o verlap and controls that did not differ from one another. ERPs were recorded under unimodal auditory presentation. ERP results showed that the amplitude of the auditory N400 component was smallest for syllable overlap, intermediate for rime overlap, and largest for coda overlap and controls that did not differ from one another. In both experiments, the priming effects were larger for word than for pseudoword targets. These results are best explained by the combined influences of nonlexical and lexical processes, and a comparison of the reported effects with those found in monosyllables suggests the involvement of rime and syllable representations.
INTRODUCTION
The assignment of a meaningful interpretation to a stretch of speech requires the mapping between the incoming signal and stored word forms. This mapping process, which leads to spoken word recognition, depends upon the structural properties of word representations, the organization of the mental lexicon, and the prelexical input representations used in lexical access (see Frauenfelder & Floccia, 1999 for an introduction). One tool psycholinguists have developed to study spoken word recognition is phonological priming (for an overview: Zwitserlood, 1996). The logic of this paradigm is that processing a stimulus--the prime--should affect the processing of a subsequent stimulus--the target-- when prime and target share some formal information that matches a representation involved in lexical access. The priming situation is rather far from naturalistic conditions of speech processing. However, as priming enables to bias the state of the lexical processing system prior to a critical stimulation, it appears an at tractive indirect way to approach analytically how words are recognized in a situation of real communication.
The large phonological priming literature shows a remarkable dissociation depending on overlap location. Studies that focused on the effect of initial overlap between two successively presented items (e.g., sad- sack) have obtained rather messy results. In tasks as lexical decision and identification in noise, both facilitation and inhibition, or more often no effect at all, have been reported (see the review by Radeau, Morais, & Segui, 1995, Appendix A; and additional results by Cutler and Chen, 1995; Goldinger, 1998; Praamstra, Meyer, & Levelt, 1994; Praamstra & Stegeman, 1993). And, in the target shadowing task (Radeau, Morais, & Dewier, 1989), what underlies the often reported inhibitory effect between monosyllabic words (Brown, 1990; Radeau, 1995; Radeau et al., 1995; Slowiaczek & Hamburger, 1992) is a matter of controversy (see Goldinger, 1999; Hamburger & Slowiaczek, 1996, 1999; Radeau & Colin, 1996).
In contrast to initial overlap, final overlap has systematically produed facilitatory effects, independently of the task performed: identification in noise (Slowiaczek, Nusbaum, & Pisoni, 1987), lexical decision (Slowiaczek, McQueen, Soltano, & Lynch, 2000; Radeau, Segui, & Morais, 1994, 1995; Praamstra et al., 1994; Praamstra & Stegeman, 1993; Corina, 1992) and shadowing (Slowiaczek et al., 2000; Radeau, 1995; Radeau et al., 1995). The final overlap facilitation is transient: the effect was shown to decrease when the interstimulus interval (ISI) was lengthened (Radeau et al., 1995). Studies using monosyllabic words have clearly shown that final overlap facilitation resulted from the two items sharing at least the phonological rime [1] (e.g., back-lack): that is, the vowel and the following optional consonants in a syllable (e.g., e in be; ack in lack). In fact, using items with a CCVC consonant-vowel phonological structure that shared one, two, or three phonemes from the end, Radeau (1995) observed facilita tion of rime overlap [VC] (flamme-tram) but no increase in this effect when the overlap included a consonant of the onset [2] in addition to the rime [CVC] (gramme-tram). When the two items shared only the last consonant--the coda--(flemme--tram), no effect was found, Further evidence that rime overlap is required to obtain final facilitation has recently been provided by Slowiaczek et al. (2000). These authors showed that facilitation was no longer observed when prime and target shared the vowel but none of the following consonants (e.g., shade-bake) so that rime overlap was not fulfilled.
The rime effect seems to occur rather early in the course of target processing. Indeed, it has been shown to be modality-dependent: the effect requires that prime and target be presented auditorily, and dissipates under crossmodal presentation (Radeau et al., 1994; Radeau, 1995). Furthermore, the failure to obtain any modulation of the effect by prime-target relative frequency (Radeau et al., 1995) or prime lexicality (words vs. pseudowords; Slowiaczek et al., 2000; Radeau et al., 1994; Burton, 1992) is also consistent with such a prelexical view.
Bisyllabic items have only been used in studies that have primarily focused on the effect of final syllable overlap (Titone & Connine, 1997; Cutler & Chen, 1995; Cutler, van Ooijen, & Norris, 1999; Burton, 1992; Corina, 1992; Emmorey, 1989). In these experiments, facilitation was found using both lexical decision and shadowing, and for both word and pseudoword primes. One exception is a study by Emmorey (1989, Exp. 1) using words with a weak-STRONG syllabic stress pattern that is uncommon in English (strong syllables are reported in uppercase letters). Final syllable overlap gave rise to facilitation when the strong syllable corresponded to the morphemic root (subMIT-perMIT), but produced no effect when it had no morphological status (duRESS-caRESS).
As regards the level of representation that gives rise to final overlap facilitation between bisyllabic items, the data from overlap length manipulations are inconsistent. Using words with the STRONG-weak common stress pattern in English, Emmorey (1989, Exps. 2 and 3) found a large facilitatory effect of syllable overlap (TANgo- CARgo), which was independent of the morphological status of the syllable, but no effect of a rime overlap, which in this case always corresponded to an inflectional suffix (DANcing-GRAzing). By contrast, Burton (1992, Exp. 2) obtained facilitatory effects of similar magnitude for syllable overlap (FOrage-COUrage) and rime overlap (BONdage-COUrage), where in this case the rime unit was part of the root, or a derivation, rather than an inflection.
The results of a study by Titone and Connine (1997) shed further light on the representations involved, These authors used phonological priming to examine whether any syllabification principle is applied during on-line spoken word recognition, and if so, what principle is used. Pairs of CVCCVC items were presented in which a word or pseudoword target (e.g., MARKet or TARKet, respectively) was primed by a pseudoword sharing all but the first phoneme with the target, in one of three conditions. In two conditions, primes had been artificially syllabified by the insertion of silence in accordance either with the Maximal Onset Principle (e.g., Pulgram, 1970) or with the Stress Principle (Bailey, 1978). The Maximal Onset Principle states that the maximum number of consonants attach to a syllable onset provided this sequence exists at the beginning of a word (LAR.ket; the "." indicates a syllable boundary). The Stress Principle (Bailey, 1978) states that consonants preferentially attach to stressed syllables 3 rega rdless of phonotactics (LARK.et). The third condition, in which primes had not been artificially syllabified, served as control (FARKet). The largest facilitation was found in the control condition that used natural items. In the artificial conditions, primes syllabified in agreement with the Maximal Onset Principle produced a far greater facilitation than those syllabified according to the Stress Principle. In the case of word targets, primes segmented according to the Maximal Onset Principle were equivalent to the unmanipulated controls. These results suggest both that phonological priming is sensitive to some syllabification cues, and that the Maximal Onset Principle of syllabification has a perceptual reality in spoken word recognition, at least for English.
Leading psycholinguistic models of spoken word recognition fail to provide an account completely satisfactory of (1) the phonological facilitation that (2) is restricted to final overlap. The Cohort model predicts priming effects indeed, but only for initial overlap. Whereas the original version predicted facilitatory effects (Marslen-Wilson & Welsh, 1978), the current version that incorporates transient bottom-up active inhibition predicts inhibitory effects of initial overlap (Gaskell & Marslen-Wilson, 1996, 1998; Marslen-Wilson, 1993). The bottom-up inhibition mechanism reduces the activation level of prime competitors that no longer match the speech input, including the target. In NAM, the Neighborhood Activation Model (Luce, 1986; see also Luce, Pisoni, & Goldinger, 1990; Luce & Pisoni, 1998') as in PARSYN (Luce, Goldinger, Auer, & Vitevitch, 2000), a connectionist instantiation of NAM able to deal with priming, target identification is negatively affected by the density and the frequency of the target neighbors. Presentation of a phonologically related prime should thus have two opposite effects. On the one hand, at the word-level, as a target neighbor of which the frequency is enhanced, the phonologically related prime should slow down target identification. But on the other hand, at the sublexical levels, the residual activation left by prior presentation of shared segments should speed up target sublexical processing. How these effects combine still needs to be addressed by simulations. Moreover, as NAM was conceived for the processing of short words, and therefore does not take overlap location into account when computing the neighborhood, no difference clue to overlap location should be observed in the priming effects. In contrast to PARSYN (see above), the connectionist models TRACE (McClelland & Elman, 1986; see also Frauenfelder & Peeters, 1990, 1998) and Shortlist (Norris, 1994; Norris, McQueen, Cutler, & Butterfield, 1997) are technically (although not theoretically) unable to deal with priming e ffects. The major limitation of these models is that their current architectures offer no connection between lexical nodes that do not overlap in time. Primes and targets presented consecutively will therefore be processed completely independently.
The first aim of the present study was to specify the representations involved in the facilitatoty effect of final overlap between bisyllabic items, using both behavioral and electrophysiological measures. The experiments were conducted in French. In this language, the specific role of the syllable in speech processing has been widely documented using different paradigms including sequence detection [4] (Mehler, Dommergues, Frauenfelder, & Segui, 1981), phoneme monitoring [5] (Segui, Dupoux, & Mehler, 1990), attentional allocation [6] (Pallier, Sebastain-Galles, Felguera, Christophe, & Mehler, 1993), phonological migrations [7] (Kolinsky, Morais, & Cluytens, 1995), and word-spotting [8] (Dumay, Banel, Frauenfelder, & Content, 1998; see also Dumay, Content, & Frauenfelder, 1999). Three types of overlap were compared in the present study. Primes and targets with CV.CVC phonological structure shared either the last phoneme (the coda), the last two phonemes (the rime), or the last three phonemes (the syllable). The predicted pattern of results depends on what form of representation underlies final priming between bisyllabic items. If the final overlap effect depends upon the amount of shared acoustic-phonetic information, the effect should increase linearly with the number of shared phonemes. By contrast, different patterns of results would be expected if priming relies upon specific linguistic units like the rime and/or the syllable. Exclusive reliance on rime units should lead to effects of similar magnitude for rime and syllable overlap, but to no effect of coda overlap. Likewise, exclusive reliance on syllabic units should lead to facilitation from syllable overlap only. If both units are involved, the syllabic effect should be larger than the rime effect and there should be no effect of coda overlap.
The amount of priming due to target recognition was assessed using both words and pseudowords as targets--as words only are stored, only words can be recognized. One possible interpretation of final phonological priming, which is compatible with the general framework of a cohort-like model of word recognition, is that the effect occurs during the lexical selection phase (Radeau et al., 1995). The main idea is that the functional role of the representations activated by the final information in the prime is to speed up lexical selection of the target once the target cohort is activated. According to this view, final overlap priming should be restricted to word targets; no effect should be observed for pseudowords because these are not stored in the lexicon.
Experiment 1 used behavioral measures (reaction times [RTs] and error rates) to assess final phonological priming, while participants performed target shadowing. An additional aim of this experiment was to determine whether the effect was modality-specific. To this end, two conditions of presentation were compared. In one condition both primes and targets were presented auditorily, while the other condition used a crossmodal procedure, in which primes were presented visually and targets auditorily.
The rationale underlying the unimodal versus crossmodal comparison is...
Read the full article for free courtesy of your local library.
|
Bookmark this article
Print this article
Link to this article
Email this article
Digg It!
Add to del.icio.us
RSS
