13.07.2015 Views

effet du nombre des graphèmes en Anglais - Aix Marseille Université

effet du nombre des graphèmes en Anglais - Aix Marseille Université

effet du nombre des graphèmes en Anglais - Aix Marseille Université

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

228App<strong>en</strong>dice Isummed frequ<strong>en</strong>cy of fri<strong>en</strong>ds -words with the same spelling pattern and the same pronunciation- and thesummed frequ<strong>en</strong>cy of <strong>en</strong>emies -words with the same spelling pattern but a differ<strong>en</strong>t pronunciation-, Jared,McRae, & Seid<strong>en</strong>berg, 1990 ; Treiman et al., 1995).In contrast, in the LDT, feedforward inconsist<strong>en</strong>cy effects are much less clear. To the ext<strong>en</strong>t that the LDTdoes not require an overt pronunciation, it is also less likely to be s<strong>en</strong>sitive to feedforward consist<strong>en</strong>cy (Jaredet al., 1990). Two more rec<strong>en</strong>t studies (Brown, 1987 ; Jared et al., 1990) that used more carefully controlledstimuli than older studies failed to find an effect. In contrast, Stone et al. (in press) provided one of the firstexperim<strong>en</strong>tal demonstrations of a feedforward consist<strong>en</strong>cy effect in the LDT using English-speaking participants(see Pugh, Rexer, & Katz, 1994, for an earlier demonstration).Stone et al. (1997) found that mean LDT-lat<strong>en</strong>cy to feedforward inconsist<strong>en</strong>t words was 48 ms longerthan for feedforward consist<strong>en</strong>t words wh<strong>en</strong> all words were feedback consist<strong>en</strong>t. For words that were feedbackinconsist<strong>en</strong>t (i.e., whose phonological body maps into more than one spelling, such as /_ip/ in DEEP andHEAP), the effect decreased to 8 ms. This suggests that previous studies might have failed to detect the effectbecause they did not control for feedback consist<strong>en</strong>cy. Rec<strong>en</strong>tly Ziegler et al. (in press c) replicated thiseffect in Fr<strong>en</strong>ch. They obtained similar effects to Stone et al. : a 55 ms effect for feedback consist<strong>en</strong>t wordsthat decreased to 13 ms wh<strong>en</strong> feedback inconsist<strong>en</strong>t words were used. To the ext<strong>en</strong>t that this effect can besuccessfully replicated, and giv<strong>en</strong> that the LDT requires no overt pronunciation, the feedforward consist<strong>en</strong>cyeffect in the LDT provi<strong>des</strong> stronger evid<strong>en</strong>ce for bidirectional influ<strong>en</strong>ces of orthographic-phonological processesin visual word recognition than the results from the naming task reported above.Effects of sound-to-spelling (feedback) consist<strong>en</strong>cy are a rec<strong>en</strong>t discovery in psycholinguistics. For English,feedback consist<strong>en</strong>cy effects have be<strong>en</strong> reported both in the visual lexical decision task and in the lettersearch task (Hooper & Paap, in press ; Stone et al., in press, Ziegler & Jacobs, 1995 ; Ziegler et al., inpress c). For the pres<strong>en</strong>t chapter, we conc<strong>en</strong>trate on the effect reported by Stone and collaborators. In twolexical decision experim<strong>en</strong>ts, Stone et al. found a reliable feedback consist<strong>en</strong>cy effect. Words withphonological bodies that could be spelled more than one way pro<strong>du</strong>ced slower correct "yes" responses andmore errors than words with phonological bodies that could be spelled only one way. In their Experim<strong>en</strong>t 2,they used a factorial <strong>des</strong>ign that included four types of words : (1) bidirectionally consist<strong>en</strong>t words such asDUCK, in which the spelling body (_UCK) could be pronounced only one way, and the pronunciation body(/_uk/) could be spelled only one way, (2) feedforward inconsist<strong>en</strong>t words such as MOTH, in which thespelling body could be pronounced more than one way (e.g., BOTH), but the pronunciation body (/_oth/)could be spelled only one way, (3) feedback inconsist<strong>en</strong>t words such as HURL, in which the spelling bodycould be pronounced only one way, but the pronunciation body could be spelled more than one way (e.g.,GIRL), and (4) bi-directionally inconsist<strong>en</strong>t words such as WORM, in which the spelling body could bepronounced more than one way (e.g., DORM), and the pronunciation body could be spelled more than oneway (e.g., FIRM). Stone et al. found that lexical decision performance was equally affected (longer RTs andmore errors) for feedforward inconsist<strong>en</strong>t words, feedback inconsist<strong>en</strong>t words, and bi-directionally inconsist<strong>en</strong>twords. Only words that were both feedforward and feedback consist<strong>en</strong>t pro<strong>du</strong>ced better performance. Bidirectionallyinconsist<strong>en</strong>t words did not affect performance over and above of what was obtained for wordsthat were inconsist<strong>en</strong>t only one way (feedforward only or feedback only).Ziegler et al. (in press c) replicated Stone et al.'s results in English in more carefully controlled conditionsin Fr<strong>en</strong>ch. They excluded the possibility that the feedback consist<strong>en</strong>cy effect obtained in English resultedfrom a failure to match feedback consist<strong>en</strong>t and inconsist<strong>en</strong>t items on a number of orthographicneighborhood variables. This replication is of particular interest. Since statistical analyses showed that thestructure of Fr<strong>en</strong>ch and English with respect to feedback inconsist<strong>en</strong>cy is highly similar for these two languages(Ziegler et al., 1996 ; in press a), similar feedback consist<strong>en</strong>cy effects were predicted for English andFr<strong>en</strong>ch. This being the case, their results join those of Stone et al. to suggest that visual word perception isaffected by both feedforward and feedback consist<strong>en</strong>cy.An important aspect of this effect is that feedback inconsist<strong>en</strong>cy may explain small and/or unreliableconsist<strong>en</strong>cy effects in previous studies. Ziegler et al. (1996) analyzed all Fr<strong>en</strong>ch words that would traditionallyhave be<strong>en</strong> classified as "consist<strong>en</strong>t" on the basis of spelling to phonology correspond<strong>en</strong>ces (87.6% of allmonosyllabic words). In traditional experim<strong>en</strong>ts on consist<strong>en</strong>cy effects, these "consist<strong>en</strong>t" items serve ascontrol items against which the processing cost of inconsist<strong>en</strong>t items is tested. Ziegler et al. (in press a)calculated that 77.4% of these presumably "consist<strong>en</strong>t" items were, however, feedback inconsist<strong>en</strong>t. Thus,small and/or unreliable consist<strong>en</strong>cy effects in previous studies may have resulted from the possibility thatthe major part of the presumably consist<strong>en</strong>t control items was feedback inconsist<strong>en</strong>t. Another interestingaspect of feedback inconsist<strong>en</strong>cy is that it should be an important variable for crosslinguistic research onspelling. If multiple possibilities of mapping phonology into spelling affect spelling performance, it shouldbe harder in a feedback inconsist<strong>en</strong>t language (e.g., Fr<strong>en</strong>ch) than in a relatively feedback consist<strong>en</strong>t language(e.g., Spanish).Considering these argum<strong>en</strong>ts, it seems clear that psycholinguistic experim<strong>en</strong>ts should be controlled forfeedback consist<strong>en</strong>cy and further research is needed to specify its influ<strong>en</strong>ce. In this respect, the pres<strong>en</strong>t attemptto give a formal account of this effect will, if successful, provide us not only with a tool for makingpredictions, but also with a formal means for stimulus selection and control. For example, simulations byMROM-P could be used in the planning phase of an experim<strong>en</strong>t -together with statistical analyses- to makesure that the stimuli are well matched on the feedback consist<strong>en</strong>cy variable.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!