Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm UniversityFigure 2. ERPs at the sentence adverb ju ‘of course’in sentences with embedded main clauses of thetype Besökaren menar alltså att familjen kännerju… ‘(lit.) The visitor thinks thus that the familyfeels of course…’ In the absence of a precedingleft-edge boundary tone (∅, grey line), there wasincreased positivity at 300–400 (P345) and 600–800ms (P600).The effect found in Roll et al. (2009a) henceindicates that in the absence of a left-edgeboundary tone, a sentence adverb showingmain clause word order (S–V–SAdv) is syntacticallyrelatively unexpected, and thus triggersreprocessing of the syntactic structure. Theword order judgments confirmed the effect.Embedded main clauses associated with a leftedgeboundary tone were rated as acceptable in68% of the cases, whereas those lacking a tonewere accepted only 52 % of the time. Generallyspeaking, the acceptability rate was surprisinglyhigh, considering that embedded main clausesare limited mostly to spoken language and areconsidered inappropriate by many speakers.In other words, the tone was used in processingto activate main clause word order and,all other things being equal, it influenced thenormative judgment related to correct word order.However, if the tone were fully grammaticalizedas a main clause marker, it would beexpected not only to activate main clause structure,but also to inhibit subordinate clausestructure. That is to say, one would expect listenersto reanalyze a subordinate clause as anembedded main clause after hearing an initial Htone.In a subsequent study (Roll et al., 2009b),the effects of the left-edge boundary tone weretested on both embedded main clauses and subordinateclauses. The embedded main clauseswere of the kind presented in Figure 1. Correspondingsentences with subordinate clauseswere recorded, e.g. Stofilerna anser alltså attafghanerna inte intog Persien… ‘(lit.) The oldfogies think thus that the Afghans not conqueredPersia…’ Conditions with embeddedmain clauses lacking left-edge boundary tonesand subordinate clauses with an initial H tonewere obtained by cross-splicing the conditionsin the occlusion phase of [t] in att ‘that’ andintog ‘conquered’ or inte ‘not.’For embedded main clauses, the ERPresultswere similar to those of Roll et al.(2009a), but the effect was even clearer: Arather strong P600 effect was found between400 and 700 ms following the onset of the sentenceadverb inte ‘not’ for embedded mainclauses lacking a left-edge boundary tone (Figure3).Figure 3. ERPs at the disambiguation point for theverb in–tog ‘conquered’ (embedded main clauses,EMC, solid lines) or the sentence adverb in–te ‘not’(subordinate clauses, SC, dotted line) in sentenceslike Stofilerna anser alltså att afghanerna in–tog/te… ‘(lit.) The old fogies thus think that the Afghanscon/no–quered/t…’ with (H, black line) orwithout (∅, grey line) a left-edge boundary tone.Embedded main clauses showed a P600 effect at400–700 ms following the sentence adverb, whichwas reduced in the presence of a left-edge tone.Thus, it was confirmed that the left-edgeboundary tone increases the activation of mainclause structure, and therefore reduces the syntacticprocessing load if a following sentenceadverb indicates main clause word order. Asmentioned above, however, if the tone werefully grammaticalized, the reverse effect wouldalso be expected for subordinate clauses: Aleft-edge boundary tone should inhibit the expectationof subordinate clause structure. However,the tone did not have any effect at all onthe processing of the sentence adverb insubordinate clauses. The left-edge boundarytone thus activates main clause structure, albeitwithout inhibiting subordinate clause structure.68