Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm UniversityA H*-specified suffix such as –ar in lekar‘games’ would hence be expected to producean ungrammatical word if combined with astem associated with a clashing L*. The wordwould have to be reprocessed, which would bereflected in a P600 effect. No such effect wouldbe expected for suffixes that usually co-occurwith Accent 1, such as –en in leken ‘the game’,since they are assumed to be unspecified forword accent.Roll et al. (2009c) found the expected dissociationwhen they compared 160 sentencescontaining words with either the H*-specifiedsuffix –ar or the unspecified suffix –en, andstems phonetically associated with a H* or aL*, obtained by cross-splicing. The effectswere compared with another 160 sentencescontaining words involving declension errors,such as lekor or leket, which have 1 st and 5 thdeclension instead of 2 nd declension suffixes,and therefore yield a clash between the lexicalspecification of the suffix and the stem. Theresults were similar for declension mismatchingwords and words with a H*-specifying suffixinaccurately assigned an Accent 1 L* (Figure5). In both cases, the mismatching suffix gaverise to a P600 effect at 450 to 900 ms that wasstronger in the case of declension-mismatchingsuffixes. The combination of the lexically unspecifiedsingular suffix –en and a H* did notyield any P600 effect, since there was no specification-mismatch,although –en usually cooccurswith Accent 1.Figure 5. ERPs for the Accent 2-specifying pluralsuffix –ar (L*PL), the word accent-unspecifieddefinite singular suffix –en (L*SG), as well the inappropriatelydeclining –or (L*DECL) and –et(L*NEU), combined with a stem associated withAccent 1 L*. The clash of the H*-specification of –ar with the L* of the stem produced a P600 similarto that of the declension errors at 450–900 ms.The study showed that a lexicalized prosodicfeature has similar effects on morphologicalprocessing as other lexicalized morphologicalfeatures such as those related to declensionmarking.Summary and conclusionsIn the present contribution, two prosodic featureswith different degrees of grammaticalizationand their influence on language processinghave been discussed. It was suggested that aSwedish left-edge boundary tone has arisenfrom the grammaticalization of the physicalconditions on speech production represented byGussenhoven’s (2002) Production Code.Probably stemming from a rise naturally associatedwith the beginning of phrases, the tonehas become associated with the syntactic structurethat is most common in spoken languageand most expected at the beginning of an utterance,namely the main clause. The tone has alsobeen assigned a specific location, the last syllableof the first prosodic word. When hearingthe tone, speakers thus increase the activationof main clause structure.However, the tone does not seem to be fullygrammaticalized, i.e. it does not seem to beable to override syntactic cues to subordinationin situations where both main and subordinateembedded clauses occur. Even when hearingthe tone, speakers seem to be nevertheless biasedtowards subordinate clause structure afterhearing the subordinate conjunction att ‘that’and a sentence adverb in preverbal position inthe embedded clause. However, embeddedmain clauses are easier to process in the contextof a left-edge boundary tone. Thus we can assumethat the H tone activates main clausestructure. Further, the boundary tone influencedacceptability judgments, but only in the absenceof word order variation in the test sentences.The combination of syntactic cues suchas the conjunction att and subordinate word order(S–SAdv–V) thus appears to constitute decisivecues to clause structure and cancel outthe potential influence the initial H could havehad in reprocessing an embedded clause as amain clause.A fully grammaticalized prosodic featurewas also discussed, Swedish Accent 2, whoseassociation with the stem is accounted for by aH* lexically specified for certain suffixes, e.g.plural –ar (Riad, in press). When the H*-specification of the suffix clashed with a L*70