Graduiertenkolleg Sprachliche Repräsentationen und ihre Interpretation


Forschungsprogramm: Degree of Specification in phonetic and phonological Representations (Neurolinguistic Correlates of Underspecification)

Ziele des Programms

My research interests relate to the areas which are relevant for the programme of the Kolleg: the representation of the phonetic, phonological and prosodic aspects of linguistic knowledge and the neuroanatomical correlates of stored and rule-generated language knowledge. The leading problem related to the research is the question of the degree of specification of linguistic, in particular phonetic, phonological and prosodic representation (cf. Stand der Forschung). It is usually claimed that phonetic knowledge of a language is so highly specific that it has to be stored. Phonetic representations are argued to be fully specified in all respects. The acoustic, articulatory and auditory aspects of speech sounds are argued to constitute a single phonetic module, which moreover, is to be stored in a separate distinct part of the brain. Such a radical option for stored knowledge is untypical for claims concerning linguistic representations. This radical position has been questioned on the linguistic phonetic grounds where specific underspecified phonetic representations have been argued for. Also the neuroanatomical evidence for the single ,phonetic module' has stired a lot of controversy in the scientific community. Phonological knowledge, like other aspects of grammatical knowledge, is argued to be represented by two types of mechanism. The regular (morpho-) phonological forms are argued to be generated by rule, whereas the irregular and exceptional formations are argued to be stored by means of an associative net. Recent advances in Optimality Theory stress the role of the associative nets as a general alternative to approaches in which generativity results from the use of rules. Limitations of such approaches are at a pivot of theoretical linguistic discussion of the day and will constitute one of the central theoretical issues of the Kolleg. The neuroimaging data which has been recently provided, supports linguistic models in which rule-governed forms and exeptions are processed by separate mechanisms. This new and complex data is, however, open to quite controversial interpretations. Prosodic knowledge is probably one of the linguistic modules in which the storage aspect plays the least important role. Prosody is argued to be generated by highly regular processes. It is claimed that an independent device, the Prosody Generator is collecting information about the segmental spellout of the ''intonational meaning'' and the surface syntactic structure, and sets prosodic frames and phonetic parameters. Because prosody has to integrate such a variety of input information it is hard to imagine how the frames and parameters it provides can be fixed or stored in any specific way. A more realistic view of prosody is that of a system of violable constraints. The neuroanatomical correlates of prosody build a substantial part of current DFG projects [DFG-Aktenzeichen DO 536]. The methodological and empirical basis of these projects will be made available to the members of the Kolleg.

I would contribute to the general research programme of the Graduiertenkolleg in the following ways:

Stand der Forschung

From the very moment when distinctive features were introduced in phonological theory, underspecification theory has been subject to rather violent swings of fashion.

In the original Prague School conception of representation nondistinctive features were universally absent even if their phonetic attributes were present on the surface. Within the system of distinctive sound properties Trubetzkoy (1939) identified characteristic opposition types (privative, proportional, one-dimentional [binary]). He formulated criteria to determine which of the universally possible 'correlations', that he defined as the conjunctions of sound properties, were phonologically irrelevant in a system. Such irrelevant properties were PERMANENTLY UNDERSPECIFIED, or simply speaking, nonspecified within the system.

The view of nonspecification of certain features was taken over by early generative phonology. Unlike the mono-stratal Prague School phonology, generative phonology in the early stages was strongly derivational in nature. The underlyingly unspecified features were allowed to be filled in by phonological rules during phonological derivation, yielding completely specified surface representations. This proviso for TEMPORARY UNDERSPECIFICATION which allows for features present on the surface to be left out of underlying lexical representations, came under attack already very early in generative phonology (Stanley 1967). Stanley argued that descriptive convenience (fewer features in structural description) and simplicity of language specific derivation (more general rules) were apparently the only benefits of temporary underspecification. One should expect, Stanley claimed, that discrepancies in feature specification between underlying lexical representations and surface representations follow from general linguistic principles and not from descriptive convenience. At that time it was not possible to immediately name such general principles, and in SPE chapter 8 it was proposed that underlying representations are fully specified for all features, redundant or not. The model of FULL SPECIFICATION has been characteristic of most of the descriptive and theoretical work in generative phonology in the tradition of the SPE.

More pertinent reasons for temporary underspecification arose due to the study of tonal and harmonic systems (Ringen 1975; Goldsmith 1976). It was argued that features from the surrounding environment spread to underlyingly toneless syllables and featureless vowels. The question of which features are universally available for insertion into unspecified lexical entries has been explicitly raised within the framework of Lexical Phonology (Kiparsky 1985). This line of research, taken up most prominently by Diana Archangeli, led to the discovery of features and feature values which show a high degree of inactivity and invisibility in phonological systems, even if they are contrastive within these systems. Kiparsky (1985), and particularly Archangeli (1988) argued that this phonological inertness should be coded by (universal) nonspecification of such features in underlying representations. The amount of internal and external evidence that was collected for what became to be called RADICAL UNDERSPECIFICATION is impressive (cf. Archangeli 1988; Stemberger 1991, 1992; Paradis \& Prunet 1991). However, no general principle for determining which features can be universally underspecified and under which conditions these underspecified features (and segments) are inserted into the phonological derivations could be provided.

Radical underspecification has come under increasingly radical criticism which, in its most extreme form, suggested eliminating the concept of underspecification altogether from feature theory (Mohanan 1991; Broe 1993). Another critical reaction to radical underspecification has been to restrict it to a small core of reliable cases, particularly those cases where an underspecified feature is truly predictable (Clements 1987, Steriade 1987, McCarthy \& Taub 1992). Close scrutiny of these solid cases has shown that underspecification can be restricted to nondistinctive feature values. In other words, surface contrast excludes underspecification. This model, called CONTRASTIVE or RESTRICTED UNDERSPECIFICATION, has been used to form a theoretical basis for the type of underspecification required in widely accepted nonlinear (autosegmental) phonological and phonetic analyses. Even this restricted version of underspecification theory has recently been challenged. Steriade (1995) argues that most of the compelling evidence for restricted/contrastive underspecification disappears if one allows privative features, indirect licensing conditions, and permanent underspecification for some classes of segments. In Steriade's system, privative features are, like asymmetrical oppositions in Trubetzkoy's original system, such features for which apparently underspecified values simply do not exist. The set of candidates for privative features offered by Steriade (1995: 147-158) includes [nasal], [labial], [coronal], [dorsal], [round], [high], [pharyngeally expanded], [vibrating vocal cords], [nonperipheral], [back], [lateral], [retroflex] and (presumably) [front]. The only remaining binary features appear to be [anterior], [distributed/laminal], [sonorant] and [consonantal]. This radical change of the feature system has clear consequences for the theory of underspecification; it drastically reduces the number of phonological contrasts that might be underspecified. Indirect licensing conditions (cf. Steriade 1995: 159-165) explain the apparent cases of underspecification due to positional neutralisation. These cases arise when features or feature values are predictable on the basis of other properties of the morpheme (e.g. Yokuts vowel harmony, Guarani nasal spreading, coda devoicing in German, Polish, Catalan, etc.). Steriade does not provide specific reanalyses of all those cases; however, she does offer a research perspective (licensing conditions) under which these cases may be reanalysed.

What remains are cases of PERMANENT UNDERSPECIFICATION. These cases involve segments in which some features and feature values appear neither underlyingly nor at any subsequent derivational stage. Segments that start out and remain fully or partially underspecified include laryngeal consonants (which are placeless), pharyngeal consonants (which are nasalless), schwa-like vowels (which are targetless), plain coronals (which are [-lateral]), etc. Nonspecification can be universal, as in the case of pharyngeals (which cannot be nasal) or laryngeals, which have no place features (Stemberger 1993). Nonspecification can be also language specific . For example, Latin underlying [l] in the adjectival suffix {-alis - nav-alis 'maritime' - dissimilates to [r] if the stem contains a preceding [l] - sol-aris 'solar'. Dissimilation is blocked by the nonlateral liquid [r] - flor-alis 'floral'. Other consonants, like a plain coronal such as [t] do not block the dissimilation - milit-aris 'military'. Steriade claims that [r] blocks dissimilation due to its [-lateral] specification. Nonliquid consonants like [t] are permanently nonspecified for [lateral] in Latin and thus transparent to the Latin {-alis-dissimilation rule. Their nonlaterality is a fact of the phonology of Latin, since a number of languages contrast central vs. lateral obstruents.

The question whether permanent nonspecification (universal or language specific) is transferred from phonology to phonetics also has been an important issue in recent research. In a seminal paper, Keating (1988) examined phonological theories of underspecification in view of the persistence of nonspecification straight down to the motor planning level. Keating argued that 'when phonetic rules build trajectories between segments, an unspecified segment will contribute nothing of its own to the trajectory' (Keating 1988: 281). Hence, the theory of phonetic underspecification predicts a correlation between phonological nonspecification and phonetic transparency. In particular, this theory predicts that transitions between fully specified segments will be short and crisp, and that they will be blurred and stretched when underspecified segments are present in the string. In the meantime a lot of supporting phonetic evidence has been offered for this prediction. Work on glottal transparency by Keating and others (cf. Stemberger 1993; Vollmer 1997) and publications on other aspects of transitional coarticulation (cf. Pierrehumbert \& Beckman 1988; Cohn 1989; 1993, Vollmer 1997; Zsiga 1997) point out that underspecification may persevere into phonetics. This research shows that phonetic data can be used to make inferences about the lack of specification at higher levels of representation. The TARGET = SPECIFICATION relation between phonetic and phonology is obviously too simple. Boyce et al. (1991) show, however, that even if targets for underspecified segments exist, temporal constraints prevent them from being detectable in the acoustic or articulatory signal. Phonetic theories attempt to account for this apparent invisibility in different ways. In a TARGET-AND -INTERPOLATION approach like that of Keating (1990) it is argued that the width of the window, which is defined as the contextual variability of a feature value, is responsible for transparency and invisibility. Keating explains: For some segments this window is very narrow, reflecting little contextual variation; for others it is very wide, reflecting extreme contextual variation. Window width thus gives a metric of variability. There is no other 'target' associated with the segment; the target is no more than this entire contextual range (Keating 1990: 455).

In the GESTURAL SCORE-TASK DYNAMIC MODEL of ARTICULATORY PHONOLOGY (Browman \& Goldstein 1989; Byrd 1996) lexical distinctions are made by the presence or absence of gestures and by the phasing relations between the gestures. Byrd (1996: 148-162) argues in contrast to ARTICULATORY PHONOLOGY that not all phasing relations between gestures are specified lexically in the gestural score. Coordination between associated gestures is assumed to be variable but constrained to particular ranges specific to the types of gestures involved, for example: V-to-C, C-to-C and V-to-V. I refer to these ranges as PHASE WINDOWS. The ultimate timing relations are actualised concurrently with a dynamic model that converts the output of the linguistic gestural model (the phonology) into articulator movements. (Byrd 1996: 148)

Thus, phonetic underspecification is assumed to be an important part of the task dynamic model. Byrd (1996: 159-162) argues that for some types of segments timing has to be lexically specified. When temporal relations are crucial in making phonological contrasts they have to be fully specified. Candidates for such fully specified segments (segments with a NARROW PHASE WINDOW) are multiply articulated stops, ejectives, implosives, clicks and other 'multigesture' segments. According to current phonetic theory these segments are special in the sense that they are in a way 'overspecified'.

This overview of literature on underspecification in phonology and phonetics implies that a concept of permanent nonspecification of features, gestures and their timing relations is necessary.

Eigene Vorarbeiten (Auswahl der neueren Arbeiten)

Dogil, Grzegorz; Ackermann, Hermann; Grodd, Wolfgang; Haider, Hubert; Kamp, Hans; Mayer, Jörg; Riecker, Axel; Röhm, Dietmar; Wildgruber, Dirk; Wokurek, Wolfgang (2004) Brain dynamics induced by language production in Thomas Pechmann and Christopher Habel, editors, Multidisciplinary Approaches to Language Production pp. 397-431 Mouton de Gruyter, Berlin.

Dogil, Grzegorz; Ackermann, Hermann; Mayer, Jörg; Riecker, Axel; Wildgruber, Dirk (2004) Das Sprechnetzwerk im menschlichen Gehirn: Evidenz aus der funktionalen Kernspintomographie und aus der Klinik in Horst Müller, editor, Neurokognition der Sprache pp. 191-211 Stauffenburg, Tübingen.

Dogil, Grzegorz; Frese, Inga; Haider, Hubert; Röhm, Dietmar; Wokurek, Wolfgang (2004) Where and how does grammatically geared processing take place - and why is Broca's area often involved. A coordinated fMRI/ERBP study of language processing Brain and Language 89(2):337-345.

Dogil, Grzegorz (2003) Naturally! Festschrift für Wolfgang Ulrich Dressler in Chris Schaner-Wolles; John R. Rennison and Friedrich Neubarth, editors, Interfacing language and speech in the human brain pp. 97-103 Rosenberg & Sellier, Wien.

Dogil, Grzegorz (2003) Understanding Prosody in Gert Rickheit; Theo Herrmann and Werner Deutsch, editors, Psycholinguistics: An International Handbook pp. 544-566 Mouton de Gruyter, Berlin.

Dogil, Grzegorz (2002) How we say When it happens in Hans Kamp and Uwe Reyle, editors, Intonation of Aspectual Meaning: Remarks on NOCH in German pp. 1-17 Max Niemeyer Verlag, Tübingen.

Dogil, Grzegorz; Riecker, Axel; Ackermann, Hermann; Wildgruber, Dirk; Mayer, Jörg; Haider, Hubert; Grodd, Wolfgang; Kamp, Hans (2002) The Speaking Brain: a tutorial introduction to fMRI experiments in the production of speech, prosody and syntax Journal of Neurolinguistics (15):59-90.

Möbius, Bernd; Dogil, Grzegorz (2002) Phonemic and postural effects on the production of prosody in Bernard Bel and Isabelle Marlien, editors, Proceedings of the Speech Prosody 2002 Conference pp. 523-526 Laboratoire Parole et Langage, Aix-en-Provence.

Dogil, Grzegorz; Mayer, Jörg (2001) Sprache und Gehirn: ein audio-visuelles Tutorial.   [URL]

Dogil, Grzegorz; Möbius, Bernd (2001) Toward a perception based model of the production of prosody Journal of the Acoustical Society of America 110(5, Pt. 2):2737.

Dogil, Grzegorz; Möbius, Bernd (2001) Towards a model of target oriented production of prosody in Proceedings of the European Conference on Speech Communication and Technology (Aalborg, Denmark) volume 1 pp. 665-668 ISCA.

Dogil, Grzegorz; Riecker, Axel; Ackermann, Hermann; Wildgruber, Dirk; Grodd, Wolfgang (2000) Opposite hemispheric lateralization effects during speaking and singing at motor cortex, insula and cerebellum NeuroReport 11:200-208.

Mayer, Jörg; Haider, Hubert; Dogil, Grzegorz; Ackermann, Hermann; Erb, Michael; Riecker, Axel; Wildgruber, Dirk; Grodd, Wolfgang (2000) Cognitive substrate of syntactic operations - evidence from fMRI NeuroImage 11(5):302.

Riecker, Axel; Ackermann, Hermann; Wildgruber, Dirk; Mayer, Jörg; Dogil, Grzegorz; Haider, Hubert; Grodd, Wolfgang (2000) Articulatory/Phonetic Sequencing at the Level of the Anterior Perisylvian Cortex: A Functional Magnetic Resonance Imaging (fMRI) Study Brain and Language 75(2):259-276.

Dogil, Grzegorz (1999) Acoustic landmarks and prosodic asymmetries in Proceedings of the 14th International Congress of Phonetic Sciences (ICPhS) pp. 128-134 San Francisco.

Dogil, Grzegorz (1999) The phonetic manifestation of word stress in Harry van der Hulst, editor, Word Prosodic Systems in the Languages of Europe pp. 273-334 de Gruyter, Berlin.

Dogil, Grzegorz (1999) West Slavic in Harry van der Hulst, editor, Word Prosodic Systems in the Languages of Europe pp. 877-896 de Gruyter, Berlin.

Dogil, Grzegorz; Jilka, Matthias; Möhler, Gregor (1999) Rules for generation of TOBI-based American English intonation Speech Communication 28(2):83-108.

Mayer, Jörg; Dogil, Grzegorz; Wildgruber, Dirk; Riecker, Axel; Ackermann, Hermann; Grodd, Wolfgang (1999) Prosody in speech production: A fMRI study in Proceedings of the 14th International Congress of Phonetic Sciences (ICPhS) pp. 635-638 San Francisco.

Mayer, Jörg; Dogil, Grzegorz; Wildgruber, Dirk; Riecker, Axel; Ackermann, Hermann; Grodd, Wolfgang; Haider, Hubert (1999) Die funktionelle Neuroanatomie der Prosodiegenerierung MS. Universität Stuttgart.

 

Themen geplanter Dissertationsprojekte


Verzahnung innerhalb des Kollegs

Thematische Beziehungen innerhalb des Kollegs:


Literatur

Archangeli, D. (1988). Aspects of underspecification theory. Phonology 5. 183-207. Avery, P. \& K. Rice (1989). Segment structure and coronal underspecification. Phonology 6. 179-200.

Boyce, S. E., R. A. Krakow \& F. Bell-Berti (1991). Phonological underspecification and speech motor organisation. Phonology 8. 219-236.

Browman, C. \& L. Goldstein (1989). Articulatory gestures as phonological units. Phonology 6. 201-251.

Byrd, D. (1996). A phase window framework for articulatory timing. Phonology 13. 139-169.

Clements, G.N. (1985). The geometry of phonological features. Phonology Yearbook 2. 225-252.

Clements, G.N. (1987). Phonological feature representation and the description of intrusive stops. In A. Bosch, B. Need \& E. Schiller (eds.) CLS 23: Parasession on autosegmental and metrical phonoilogy. Chicago: CLS. 29-50.

Clements, G.N. (1993). Underspecification or nonspecification? Travaux de l'Institut de Phonétique de Paris (Nouvelle série) 1.55.-86

Clements, G.N. \& E. Hume (1995). Internal organization of speech sounds. In J. Goldsmith (ed.) The handbook of phonological theory. Cambridge: Blackwell. 245-306.

Goldsmith, J. (1976). Autosegmental phonology. PhD dissertation, MIT.

Halle, M. (1959). The sound pattern of Russian. The Hague: Mouton.

Halle, M. (1992). Phonological features. In Bright, W. (ed.) International encyclopedia of linguistics. Oxford: Oxford University Press. 207-212.

Keating, P.A. (1988). Underspecification in phonetics. Phonology 5. 275-292.

Keating, P.A. (1990). The window model of coarticulation: articulatory evidence. In Kingston \& Beckman (1990). 451-470.

Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2. 85-138.

Ringen, C. (1975). Vowel harmony: theoretical implications. PhD dissertation, Indiana University.

Stanley, R. (1967). Redundancy rules in phonology. Lg 43. 393-436.

Stemberger, J. (1991). Radical underspecification in language production. Phonology 8. 73-112.

Stemberger, J. (1992). Vocalic underspecification in language production. Lg 68. 492-524.

Stemberger, J. (1993). Glottal transparency. Phonology 10. 107-138.

Steriade, D. (1987). Redundant values. CLS 23:2. 339-362.

Steriade, D. (1995), Underspecification and markedness. In Goldsmith (ed.) The handbook of phonological theory. Cambridge: Blackwell. 114-174.

Trubetzkoy, N. Grundzüge der Phonologie. Göttingen: Vandenhoeck und Ruprecht.

Vollmer, K. (1993). Untersuchung der Koartikulation bei gest"orter Sprache: Experimentalstudie. MA thesis. University of Bielefeld.

Vollmer, K. (1997). Koartikulation und glottale Transparenz. Doctoral Dissertation, University of Stuttgart. Phonetik-AIMS 3.5 (Working Papers of the Chair of Zsiga, E. C. (1997). Features, gestures, and Igbo vowels: an approach to the phonology-phonetics interface. Lg 73. 227-274.