Linguistics

This essay investigates change and stability in the alignment system and its expression in morphology in Arawakan languages. The verbal alignment system and the agreement-pattern of 32 Arawakan languages were compared to each other and maps showing the geographical distribution as well as the distribution in genealogical subgroups were generated. Moreover, the nominal alignment system and case-marking which occurs in two of the investigated languages were studied as well as the correlation
of alignment systems, agreement-patterns and word order. It was focused on changing patterns and in the cases where no language-internal explanation of a change could be found, the languages were compared to non-related contact-languages.

It was found that the alignment system is quite stable whereas the changes that had happened were very varied. There were found changes due to internal factors and such caused by contact-languages. While internal changes affected the morphology before changing the alignment system, contact-induced changes could affect either the alignment system, or the morphology or both of them.

Findings from a pilot study on the distribution, function and phonetic realization of the filler EH in interviews from SweDialOOO interviews are presented. The results show that EH occurs almost exclusively after function words at the beginning of constituents. The ...

The main goal of our current research is the development of the Swedish prosody model. In our analysis of discourse and dialogue intonation we are exploiting model-based resynthesis. By comparing synthesized default and fine-tuned pitch contours for dialogues under study we are able to isolate relevant intonation patterns. This analysis of intonation is related to an independent modelling of topic structure consisting of lexical-semantic analysis and text segmentation. Some results from our model-based acoustic analysis are presented, and the implementation in text-tospeech-synthesis is discussed.

The issue of the focal accentuation of contextually 'given' travel domain concepts in task-related Swedish dialogues is examined. An explanation for this focal accentuation is proposed using concepts within Centering theory, i.e. it is assumed that it is domain-related 'Backward Centers' that are associated with the prominent focal accents. The variation in the timing of the focal H¯is further accounted for by relating it to different orderings of the Optimality theory Align XP and Wrap XP constraints with respect to Align Focus. Support for the theoretical assumptions is obtained from a perception test using constructed manmachine Initiative-Response dialogues where the 'machine' Responses vary as to the type of F0 contour on the Backward Centers (focal vs non-focal and early vs. late timing of the focal H¯).

Prosodic and lexical correlates of 'clause-like' and 'paragraph-like' boundaries associated with the Swedish discourse marker men 'but/and/so' are examined. Men-tokens in spontaneous monologues were labelled as to their boundary-status, first using text-only data. The 'strong' tokens (labelled identically by all labellers) were subsequently seen to be correlated with clear differences in the prosodic and lexical parameters examined. This tendency was not found for the corresponding 'weak' tokens which were subsequently relabelled using both text and speech nor for the database as a whole. A test using a neural network trained using strong tokens is seen to be able to correctly categorize 90% of the strong me«-tokens as to their associated boundary-type. The results show that discourse markers along with their prosodic and lexical correlates constitute a constellation of important information for understanding how segmentation of speech is produced and understood.

The Swedish cue word men ‘but’ can mark the boundary between both different topic units as well as topic-internal units in spontaneous speech. The goal of this study is to see if these two functions of men can be distinguished on the basis of their local prosodic correlates and co-occurring lexical items. Men-tokens in spontaneous narrations were labelled as to their function, first using text-only data. The ‘strong’ tokens (categorized identically by all labellers) were subsequently seen to be clearly differentiated into two classes on the basis of related prosodic parameters and co-occurring lexical items. This distinction was, however, not found for the corresponding ‘weak’ tokens which were subsequently relabelled using both text and speech nor for the data-base as a whole. A test using a neural network trained using strong tokens was seen to be able to correctly categorize 90% of the strong men-tokens as to their associated boundary-type (topic-shift vs. topic-internal). The re...

36 Computational Extraction of Lexico-Grammatical Information for Generation of Swedish Intonation Merle A. Home Marcus KD Filipsson ABSTRACT By using a minimal amount of syntactic information in combination with information on lexical word class and morphological and lexico-...

The phonological trace of perceived words starts fading away in short-term memory after a few seconds. Spoken utterances are usually 2-3 s long, possibly to allow the listener to parse the words into coherent prosodic phrases while they still have a clear representation. Results from this brain potential study suggest that even during silent reading, words are organized into 2-3 s long 'implicit' prosodic phrases. Participants read the same sentences word by word at different presentation rates. Clause-final words occurring at multiples of 2-3 s from sentence onset yielded increased positivity, irrespective of presentation rate. The effect was interpreted as a closure positive shift (CPS), reflecting insertion of implicit prosodic phrase boundaries every 2-3 s. Additionally, in participants with low working memory span, clauses over 3 s long produced a negativity, possibly indicating increased working memory load.

This electrophysiological study investigated how right-and left-edge prosodic boundary tones interact in the processing of syntactic structure. Swedish sentences of the type 'Peter hit Larry(NP2) and Jason(NP3) fell/hard…' were used. A verb ('fell') requires a clause boundary between NP2 and NP3, whereas an adverb ('hard') implies continuation of the first clause, which incorporates NP3 as a coordinated object. The effects of right-edge prosody associated with NP2 and left-edge prosody associated with NP3 were tested. Results suggest interaction between prosodic right-and left-edge boundary cues both at the earliest stages of processing the left-edge boundary tone on NP3 and at the immediately following word category distinction. Right-edge boundary tones on NP2 yielded an early positive deflection (P200) and a later positivity (CPS). Left-edge tones on NP3 showed a P200 effect only if preceded by a right-edge boundary on NP2. In the absence of a prosodic right-edge boundary, left-edge tones instead yielded an early negativity (N100), suggesting that they were unexpected. At the following word category distinction point, adverbs, showing continuation of the first clause, produced an anterior negativity when preceded by both right-and left-edge prosodic boundaries. The negativity is thought to reflect detection of a syntactically incorrect word category. Syntactically un-preferred constructions with an adverb following NP3 received generally lower acceptability ratings and gave rise to a P600 effect in all conditions. Syntactically preferred constructions with verbs following NP3 showed a similar P600 only when not preceded by either right-or left-edge boundary tones.

Results indicating that high stem tones realizing word accents activate a certain class of suffixes in online processing of Central Swedish are presented. This supports the view that high Swedish word accent tones are induced onto word stems by particular suffixes rather than being associated with words in the mental lexicon. Using event-related potentials, effects of mismatch between word accents and inflectional suffixes were compared with mismatches between stem and suffix in terms of declension class. Declensionally incorrect suffixes yielded an increase in the N400, indicating problems in lexical retrieval, as well as a P600 effect, showing reanalysis. Both declensionally correct and incorrect high toneinducing (Accent 2) suffixes combined with a mismatching low tone (Accent 1) on the stems produced P600 effects, but did not increase the N400. Suffixes usually co-occurring with Accent 1 did not yield any effects in words realized with the nonmatching Accent 2, suggesting that Accent 1 is a default accent, lacking association with any particular suffix. High tones on Accent 2 words also produced an early anterior positivity, interpreted as a P200 effect reflecting preattentive processing of the tone.

High and low tones on Swedish word stems are associated with different classes of suffixes. We tested the electrophysiological effects of high and low stem tones as well as tonally cued and uncued suffixes. Two different tasks were used involving either choosing the suffixdependent meaning of the words, or pressing a button when the word ended. To determine whether effects were in fact due to association of tones with lexical material, delexicalized stimuli were also used. High tones in lexical items produced an increase in the P2 component in both tasks, interpreted as showing passive anticipatory attention allocated to the associated upcoming suffix. This effect was absent for delexicalized forms, where instead an N1 increase was found for high tones, indicating that the high pitch was unexpected in the absence of lexical material, and did not lead to anticipatory attention. A P600 effect was found for uncued high-associated suffixes in the semantic task, which was also where the largest increase was found in reaction times. This suggests that the tonal cues were most important when participants were required to process the meaning of the words.

The Swedish word att /at/ is associated with two grammatical functions: a) (part of) a subordinate conjunction and b) an infinitive marker. This pilot study, with 5 short spontaneous discourses from 3 male speakers shows a correlation between pauses after att, and aspiration of /t/ in att, where aspiration can be interpreted as a kind of final lengthening involving the release phase of the stop consonant. We also show a tendency for att with aspiration to be associated with the grammatical function of subordinate conjunction. Further, we are able to show a tendency for the infinitive marker to be unaspirated in the normal case, while the subordinate conjunctions are characterized by final aspiration in 40 % of the cases.

Based on the results from three Event-Related Potential (ERP) studies, we show how the degree of grammaticalization of prosodic features influences their impact on syntactic and morphological processing. Thus, results indicate that only lexicalized word accents influence morphological processing. Furthermore, it is shown how an assumed semi-grammaticalized left-edge boundary tone activates main clause structure without, however, inhibiting subordinate clause structure in the presence of competing syntactic cues.

Phonetic markedness regarding linguistically relevant tonal patterns (Accent 2, boundary tones) in Central Swedish is discussed. Both tonal markedness and F0 turning points are assumed to be important cues for anticipatory attention to grammatical structure during speech processing. Empirical evidence from neurolinguistic and psycholinguistic experiments for the assumed relation between anticipatory attention and marked tonal patterns' association with Swedish word and clause structures is presented.