Department of General Linguistics
University of Helsinki | Faculty of Arts

Conference of Linguistics

Call for Papers
Hotels & accommodation
Links and local information


The dialectics of grammatical change, or why life cycles are not always cyclic

Dahl, Östen

Grammaticalization processes have variously been described as the outcome of a struggle between opposing forces, one constructive and one destructive, leading to an eternal cycle (the "cyclical theory") or, as a harmonic simultaneous movement along a set of parallel scales (the "concerted scales model"). Although these models - at least described in this way - seem rather different, what they have in common is that the initial and final points of a grammaticalization process are identical - essentially, one starts with zero and ends with zero. As an alternative, I propose to see grammaticalization and similar processes by which grammatical structures are built up as "dialectic": two antagonistic kinds of process are indeed involved, one which increases redundancy and one which decreases it, but under normal circumstances this does not lead to a cancelling-out but rather to a "synthesis", a stable state in which redundancy is exploited gainfully in the system - "smart redundancy".

[ Top of the page ]

English Interdental Substitution

Ahn, Sang Cheol

The purpose of this paper is to describe the various types of English interdental substitution within the framework of Optimality Theory. For this purpose, I will first category the substitution patterns into four major types and discuss the phonological implication of each pattern: Type 1 (Russian, Thai):[s]/[z], Type 2 (German, Japanese):[t]/[d], Type 3 (Dutch positional variation): [t, s]/[d, z], Type 4 (Korean positional variation): [sâ]/[t]. I will then propose that Ident-manner be exploded for [cont] and other manner features such as [stiff vocal folds] and [sibilant].

Specifically, as Type 1 is the case of the emergence of the unmarked, we can simply use the constraint hierarchy where markedness constraints dominate faithfulness constraints. For Type 2, however, we need to promote the faithfulness constraint, Ident-manner, so that it can dominate the markedness constraints such as *cont and *stop. As for the positional variation in Types 3 and 4, I will argue for positional markedness (Zoll 1998), rather than positional faithfulness (Beckman 1997). In Dutch, for example, the stops appear word-initially, while fricatives word-finally (Lombardi 2000). As for Korean, I will propose to employ the laryngeal feature [stiff vocal folds] for Ident constraint, i.e., Ident[stiff. v. f.]. For the extension of the current analysis to other fricative substitution, certain skeletal information is required as well for a satisfactory analysis. For this, following Ahn & Iverson (2003), I will also propose to employ the so-called geminate hypothesis for Korean tensed consonants. During the discussion, I will also discuss alternative accounts and point out their problems such as the lack of uniformity in explanation (Lombardi 2000, Lee & Cho 2002, Idsardi & Son 2003, etc.). Finally, I will show how this current analysis can be used for the substitution of other English fricatives. According to Gough (1999), for instance, the voiceless interdental fricative is substituted by [f], rather than [s], in Africaans or in certain southern dialects of American English. That is, the speakers of these languages seem to preserve the soft friction of the target interdental fricative in phoneme substitution. For this peculiar substitution, we can also invoke a constraint like Ident[sibilant] which is ranked higher than other faithfulness constraints like Ident[place] and Ident[cont]. That is, Ident[cont] has to be promoted above the markedness constraint *cont, so that the fricative [f] wins over the (less marked) stop [t].

[ Top of the page ]

Syntactic Productivity

Barðdal, Jóhanna

Productivity has been a much-debated concept within morphology and word formation for years, while syntactic productivity has hardly been the subject of any attention within the research community. A closer investigation of the adjective productive reveals that the term is used in at least 13 different ways in the linguistic literature and that at least two, if not three, different concepts of productivity are the basis for these different usages, i.e. extensibility, regularity and generality. This entails that syntactic productivity can either be regarded as regular performance, as in the Chomskyian tradition in which syntactic productivity is taken to be the speakers ability to generate sentences never encountered before, or it can be regarded as the extensibility of argument structure constructions to new verbs which enter the language, as has been done in the Cognitive Linguistics paradigm. It has also been a cause of dispute whether productivity is an all-or-nothing phenomenon or whether it is gradable. A problem that arises in the latter case has to do with which criteria to use to decide between different degrees of productivity and, hence, which criteria to use to draw the line between low-productivity and non-productivity, for instance. I will argue, following Clausner & Croft (1997) that productivity is a function of type frequency and coherence, with the term coherence referring to internal consistency. On this view syntactic productivity is taken to be predictable from a constructions type frequency and semantic coherence and the inverse correlation between the two. This means that the higher the type frequency, the lower degree of semantic coherence is needed for a construction to be extended to new verbs, and vice versa, that the lower the type frequency of a construction, the higher degree of semantic coherence is needed for the construction to be extended to new verbs. On this approach, the most frequent, regular and general construction, such as the ordinary nom-acc construction, is also going to be the construction extended to most new verbs, while on the other end of the cline, the least frequent construction will only be productive in the case of a high degree of semantic overlap. Such coinages, however, are usually taken to be the result of analogical processing, which has been excluded as belonging to the category of productive processes. On the approach presented here, these two represent the extreme poles of the productivity cline, or two sides of the same coin. From this view, both the classical definition of productivity and the more-narrow concept of analogy can be derived, as well as different degrees of productivity.

[ Top of the page ]

The Emergence of San Andresan Spanish

Bartens, Angela

In this paper, we shall examine the emergence of an autochthonous variety of Spanish on San Andrés, Colombia. San Andresan Native Islanders speak an English-based creole descended from Jamaican. For sociohistorical reasons, the younger generations have been shifting from different varieties of English to Spanish. Competence in Caribbean Standard English is mostly limited to elderly people. Among adult L2 speakers of Spanish, the prestigious norm of the interior of the country is dominant while school-aged L2 speakers adopt more and more phonetic and lexical features from the speakers of less prestigious Colombian coastal varieties now spoken by the majority of the immigrants to the island.

We have submitted the data from recordings of both L1 and L2 speakers of Spanish to VARBRUL analysis. Because of the existence of many studies of the same feature in diverse Spanish varieties which may serve as a base of comparison we have chosen postnuclear ?s as our dependent variable. We argue that an autochthonous variety of San Andresan Spanish with a clear coastal Colombian flavor is in the process of crystallization.

[ Top of the page ]

Mirror neurons and verb meaning

Dimitrova-Vulchanova, Mila

In this talk I discuss a possibility for the representation of verb meaning based in current linguistic research (e.g., problems with theta-roles, as discussed in Dowty 1991, Zaenen 1993, Pustejovsky 1995, Dimitrova-Vulchanova 1998, among others) and representations/feedback from other modalities (on the assumption that language has a rich interface with other cognitive faculties) (cf. among others, Jackendoff 1997, 2003; Koenig et al.2001, 2003; van der Zee & Nikanne 2000, cf. also Pustejovsky's (1995) qualia structure).

I will demonstrate the case at hand on cross-linguistic data from verbs of motion and claim that the classical picture given e.g., in work by Talmy (1985, 2000) is basically flawed in being too crude and based on the conflation of criteria. I will argue for a more fine-grained (and richer) representation of motion along the following lines:
- Nikanne 2002, van der Zee 2000 propose details of the angle of the path, [+/-object rotation], [+/-path rotation]), van der Zee & Watson 2003 propose spatial and contextual features in the meaning of prepositions;
- motion can be decomposed into features/specification of the path per se (i.e. the translational component) and specification of figure orientation (e.g. a rotational component; 0/00 manner); the 3 levels assumed in Nikanne 2002
- the possible types of motion (scenes) can be predicted on the basis of the values of the two basic components of motion (e.g. 4 basic types of trajectory multiplied by 2 [+/- figure rotation] .

I will claim that
- even verbs traditionally described as manner verbs (e.g. run, swim, fly, fall, roll) lexically encode a path parameter (cf. Koenig et al. 2001, 2003, for lexically encoded information as the information immediately accessed upon recognition of a word)
- there is a distinction between two types of lexically encoded information: - fixed parameters (usually to do with trajectory/path orientation; less commonly with path origin/end (fall [fixed | vertical & bounded path ]: 35% into/down/to PPs vs. an equal % of "PP-bare" occurrences in Brown); this is a distinction similar to Tortora's (1998) Further Specification Constraint, and - specifiable parameters (e.g. path origin/end/length in run). I will also bring in support found in mirror neuron studies (Gallese et al. 1996, Rizzolatti et al. 1996, Rizzolatti et al. 2002, Fogassi & Gallese 2002, Stamenov & Gallese 2002, Buccino et al. 2001).

[ Top of the page ]


Eide, Kristin M.

This paper advocates the idea that what universally separates verbs from other lexical categories is the presence of the inherent features tense and/or finiteness. This implies that every verb form includes either a tense feature or a finiteness feature, or both. Moreover, I will argue that these tense and finiteness features are interpretable features with semantic implications, thus not erased during the derivation. Encoding temporal relations, these features are easily detectable to the language acquirer. The tense element of any verb can be formalized semantically as a dyadic predicate function, which takes the preceding event as its first argument and the event variable of its own verb as the second argument. In the case of finite verbs, the first argument is (typically) the speech event S, whereas non-finite tense elements take as their first argument the event variable of the preceding verb. Thus, a finite tense element is deictic, whereas a non-finite tense element is always anaphoric. Syntactically, the tense element is construed as the extended projection of a verb; thus TP has to be replicated for every verb (including auxiliaries) in a verb sequence. This gives rise to a composisiontal tense system, where no tense element has scope over the entire proposition; only over "its own" VP. In Reichenbachian terms, this gives rise to the possibility of an infinite number of Rs in a verbal sequence. I show that modal constructions, perfect constructions, and various other constructions can be accurately described by means of this compositional tense system.

Julien (2001) suggests that every sentence in every language contains one TPFUT(URE) and one TPPAST, where TPFUT universally scopes over TPPAST. The present paper takes a slightly different view, in suggesting that TPFUT and TPPAST are both subject to parametrisation. Thus we find languages like Hua, employing a binary tense system of future and non-future tenses (cf. Haiman 1980, Comrie 1985). This is explained on the assumption that Hua employs TPFUT only. Most Germanic languages select for TPPAST only, thus the verb forms of these languages are marked as either past or as non-past. Languages like Irish and Turkish seem to employ both a TPFUT and a TPPAST, where both features are present in one and the same verb form. Languages like Burmese are claimed to lack any marking for tense proper (cf. Comrie 1985). Thus, I hypothesise that Burmese is a language which selects for no TP at all.

[ Top of the page ]

Word Order in the Swedish Midfield - an OT Approach

Elisabet Engdahl, Kersti Börjars, & Maia Andréasson

Most accounts of Swedish clause structure assume that the subject position is rather fixed. This holds both for generative approaches which adopt a binary branching structure (e.g. Platzack 1998) and for grammatical descriptions which adopt Diderichsen's field schema (e.g. Teleman et al.1999). On both approaches, it is assumed that subjects in Swedish main clauses appear either initially or immediately following the finite verb.

There is, however, considerable variation in the Swedish midfield. Subjects may be preceded by a range of adverbials as well as pronominal objects. This variation is not free but depends, as we will show, on syntactic, scopal, phonological, morphological and information structural factors. For this reason we believe that a constraint based account is most suitable. We will outline an approach in Optimality Theory (OT) using violable OT-constraints to account for the word order options and preferences. For instance, the preference for subjects to precede sentential adverbs may be offset in case the adverb takes scope over the subject, as shown in (1).

(1) Där vill aldrig någon / någon aldrig bo.
'Nobody ever wants to live there' / 'Somebody never wants to live there'

If the subject is rhematic, it may follow an adverbial, as shown in (2).

(2) Då skulle antagligen alla grodorna / alla grodorna antagligen dö.
'Then presumably all the frogs would die' / Then all the frogs would presumably die'

Conversely, if a pronominal subject follows a sentence adverbial, it is interpreted as rheme and cannot be unstressed.

(3) Då kommer tyvärr 'VI /*vi för sent.
'Then it's unfortunately us that will be late.'

Short thematic objects may precede subjects (so called Long Object Shift) as in (4a), but only if the form of the object is morphologically distinct from a subject.

(4a) Där mötte honom François Mitterand.
'There François Mitterand met him.'
(4b) Där mötte dom François Mitterand.
'There they met François Mitterand.' *'There FM met them.'

We adopt the partial constraint ranking in (5) which we will compare with Sells' (2001) analysis.

(5) TOP-L >> HEAD-L >> [ V < COMPL ] >> [OBJIND> [SUBJ> GROUND> SUBJ-L >> ADV-L >> NEG-L >> OBJ-L

One advantage of this approach is that it captures the interaction between constraints relating to different components of the grammatical system in a way that allows for cross-linguistic testing.

Platzack, C. (1998) Svenskans inre grammatik - Det minimalistiska programmet. Studentlitteratur.
Sells, P. (2001) Structure, Alignment and Optimality in Swedish. CSLI.
Teleman, U., S.Hellberg & E.Andersson (1999) Svenska Akademiens Grammatik. Norstedts.

[ Top of the page ]

Androcentricity in Finnish Personal Nouns

Engelberg, Mila

I will discuss the 'people=male' bias as formulated by Silveira (1980) in Finnish human nouns. In Silveira's hypothesis androcentricity is bidirectional: people are more likely perceived as male than female (people=male) and men are more likely seen as people than are women (male=people). Experimental evidence for the bias has been found in Turkish (e.g. Braun 1998) and English (Hamilton 1991). I will provide examples of androcentricity from the Finnish media and report findings from a series of experiments with native speakers of Finnish.

The atypicality of women as people manifests itself in several ways. A woman can be described as e.g. "ihminen ja nainen" 'a human being and a woman' or "suomalaisen vaimo" 'wife of a Finn'. There are also conventional female compounds with "ihminen" 'human being' as the main element, e.g. "naisihminen" 'female human being' (*miesihminen 'male human being'). Androcentricity probably also explains double gender marking in feminine forms, e.g. "naislaulajatar" female+woman singer and "naiskaunotar" female+beauty, belle 'a female beautiful woman' (or 'a female representative of feminine beauty' < naiskauneus 'feminine beauty').

Two studies tested the people as male bias. Ninety-five elementary school pupils, 92 high school students and 108 senior citizens participated in Study 1. Male participants at all age levels illustrated text paragraphs referring to "ihminen" as a species significantly more frequently with male than with female drawings.

In Study 2, 109 high school pupils and 111 university students perceived the referents of such gender neutral category names as e.g. "tyypillinen televisiokatselija" 'a typical TV viewer' significantly more often as male than as female.

Study 3 examined the male as people bias. Overall, 773 high school pupils producing news headlines for various news stories referred to a male referent's gender significantly less frequently than to a female referent's gender.

Braun, Friederike 1998. Prototype theory and covert gender in Turkish. In Koenig, Jean-Pierre (ed.): Discourse and cognition. Bridging the gap. Stanford, Center for the Study of Language and Communication, 113-122.
Engelberg, Mila 2001. Ihminen ja naisihminen - suomen kielen piilomaskuliinisuus. Naistutkimus-Kvinnoforskning 4, 23-36.
Hamilton, Mykol C. 1991. Masculine bias in the attribution of personhood. People=Male, Male=People. Psychology of Women Quarterly 15, 393-402.
Silveira, Jeanette 1980. Generic masculine words and thinking. Women's Studies International Quarterly, Vol. 3, 165-178.

[ Top of the page ]

[V-N] Compound Nouns Formation in Italian

Ferrari-Bridgers, Franca

The Syntactic Premise
In the spirit of Chomsky (1995-98), I will argue that Merge and Move are the grammatical operations responsible for the formation of [V-N]-compound nouns (see example 1). The premise for a syntactic analysis are constituency tests, such as (a) wh-movement and (b) head deletion in coordination (see example 2), which suggest that the two elements forming the compound are, respectively, VP and NP, and not simple lexical heads.

Proposal for a Syntactic Derivation
Because N functions as the internal argument of V, I assume that a [V-N] compound starts out as a Larsonian VP that at some point in its derivation merges with a Generic Aspect Projection (Cinque 1999) (henceforth GAP). Habitual Generic Aspect morphemes (Dahl 1985) as functional heads of the GAP trigger verb movement from which derives the final V-O constituent order. Finally, a null non-anaphoric N (henceforth N0) merges with GAP to conclude the derivation of the initial VP generating an NP.

A Four Step Analysis
1. As a first step, I will demonstrate that the end vocalic segments [-a,-i,-i] on V are habitual-generic aspect morphemes, i.e., GAP's heads, and not thematic vowels (henceforth TV) as assumed by Scalise (1992) and by Vogel (1993), nor 2nd singular person of the imperative as proposed by Prati (1931), nor 3rd singular forms of the present tense as argued by Dardano (1978). In addition to the obvious formal differences between [-a,-i,-i] and TVs [-a,-e,-i] or Tense markers [-a,-e,-e], I will show the following:
(A) Compound nouns formed with verbs that insert the inchoative affix -sc-, e.g., lui puli-sc-e ' he cleans', falsify the TV hypothesis. For these verbs, the verb stem plus TV is not a sufficient base for compounding (see example 3).
(B) The requirement of imperative verbs to be followed by DPs and not by bare objects conflicts with Prati's imperative hypothesis (see example 4).
(C) The impossibility to merge a N0 with a tensed verb undermines the present tense hypothesis (see example 5).
2. As a second step, I will show that the presence of GAP semantically justifies and explains why the object in the compound must be a bare noun. [V-N] compound nouns are generic propositions whose predicate selects for a generic NP. In Italian, bare nouns can have a generic reading (Chierchia 1996) and bare nouns are generic objects only when the predicate selecting them has a habitual rather then episodic reading (Longobardi 1999).
3. As a third step, I will demonstrate that the presence of N0 as the head of the compound noun is syntactically motivated for DP-VP agreement purposes. In fact, the agreement mismatch between D and the internal object of the compounds (see example 6) indicates that the internal object is not the head of the compound.
4. Finally, I will argue that N0 can only be merged with Aspect to generate an NP and specifically with a GAP to form a [V-N]-compound. The merger with GAP is morphologically justified by the derivational nature of GA morphology (Bergsland 1994). Not only do [V-N] compound nouns have verbs marked by GA as bases, but so do deverbal agentive/instrumental nouns formed with the derivational morpheme "tore" (see example 7), as GAP is the syntactic locus for deverbal nominal formation.

Data and Selected Reference
1a. tagli-a carte (paper cutter)
1b. accend-i sigari (cigar lighter)
1c. apr-i scatole (can opener)

2a. WH-movement
Un apri-che hai comprato?
An open-what did you buy?
What kind of opener did you buy?

2b. Head deletion in coordination
Colleziono accendisigari e sigarette americani
I collect American cigar-and cigarette lighters

3a. pulire (to clean)
Vsterm+TV+infinite marker

3b. Il pulisci scarpe (shoe polish)
pul+i+sc+i scarpe

3c. * il puliscarpe
* pul-i scarpe
*VStem+ TV scarpe

4a. Imperative forms are incompatible with bare objects
accendi i sigari, per favore
light the cigars, please!
4b. *accendi sigari, per favore

5a. *l'apre bottiglie ( indicative present)
*l'apriva bottiglie (3rd. sg. imperfect)

6a. il (sg.mas) taglia cart-e (pl, fem)
6b il (sg.mas) tira sassi (pl.mas)

7a. Il compra-tore (buyer)
7b. Il bevi-tore (drinker)
7c. Il servi-tore (servand)

[ Top of the page ]

"All Things Considered" - the Pragmatic Function of a Type of Adverbial Adjunct

Fretheim, Thorstein

When a speaker modifes an English sentence by producing a gerund phrase like considering his age or a participal phrase like given her weakness for sweets, she is adding a contextual premise, something that must be a condition intended to make the utterance relevant, maybe even a condition of the conditional antecedent type that constrains the truth conditions of the consequent. There is often a small step from recognizing that a proposition P represents a condition for the truth of Q to the inference that the truth of P causes Q to be true. Nevertheless, (2a) below is a more likely, or even more correct paraphrase of (1) than (2b).

(1) Considering his age he did well.
(2) a. He did well in spite of his age.
b. He did well because of his age.

Although the adjunct in (1) appears to have a meaning that points to the premise on which the communicator's speech act rests, we read a concessive relation into (1).

It will be demonstrated that Sperber & Wilson's Relevance Theory offers an elegant and psychologically satsifactory way of accounting for what is communicated when we use (1) and similar sentences. I have been examining tokens of the type of adverbial adjunct exemplified in (1) in a bi-directional translation corpus, The English-Norwegian Parallel Corpus (, and have found that such constructions require an interpretation at two levels of analysis simultaneously. At one level there is a causal relation between the proposition of the adjunct and the main clause proposition, at a different level the relation seems to be concessive, or 'anti-causal'.

Data of the sort examined will be shown to vindicate the theoretical distinction made by relevance theorists between 'ground-floor explicatures' and 'higher-level explicatures' of utterances, the latter communicating the speaker's propositional attitude or the kind of speech act performed.

[ Top of the page ]

Semantics and Grammar of Taxis in Nivkh

Gruzdeva, Ekaterina

Temporal relations between the situations p and q referred to in a sequence of two verb forms can be analyzed in terms of category of taxis that determines time location of situation p relative to situation q, regardless of the speech moment. In other words, taxis relations connect two verb forms, one of which, an independent form, encodes a situation q that defines a reference point for another situation p expressed by a dependent form.

The aim of the paper is to discuss the parameters along which the semantic types of taxis relations may be analyzed and to present a detailed taxis study of Nivkh (Paleosiberian, isolate, agglutinative, SOV), which is a good example of almost a 'pure' taxis language. It is characterized by regular use of numerous converbs and the sentence-final position of the finite verb form. In Nivkh, an absolute time is typically marked only on the finite (= independent) verb form which establishes a real temporal domain (non-future vs. future), whereas various taxis relations are expressed by a set of converb (= dependent) forms.

Basic taxis relations are known to embrace the following ones: Simultaneity, Anteriority, and Posteriority. I believe that the crucial parameters for further analysis of taxis are: (a) temporal boundedness / unboundedness of both situation p and situation q, which is established by identifying of initial and terminal boundaries of the situations, and (b) the correlation between temporal boundaries of situation p and those of situation q. The correlation between temporal boundaries involves (i) the existence / absence of time interval between temporal boundaries of the situations, (ii) the size of the time interval, and (iii) the completeness / interruption of situation p. Moreover, on every level of the hierarchy, taxis relations closely interact with semantic properties of verb forms that may denote situations pertaining to states, processes, or events, as well as correlate with coreference / non-coreference of subjects of these verb forms.

[ Top of the page ]

Defocalization Strategy - General Suppression Phenomena of Phonetic Forms in the Canonically Realized Position

Hosono, Mayumi

I argue that the issue of an 'apparent imperfection' of a human language (Chomsky 2000) will shift from whether a language has displacement to whether a phonetic form is suppressed in the position where it should canonically be realized. Pesetsky (2000) argues that the difference between over and covert wh-movement is phonological: the distinction is whether a wh-phrase is pronounced in situ or in the higher position (Pesetsky 2000:8). This idea should be extended to the other kinds of movement, since all kinds of movement share a property that a 'moved' element is pronounced in the higher position for some reason.

The problem is that the principled way has been lacking to determine which position is pronounced and why that position is pronounced. Assuming that the features are, as the occurrences of the same component (Chomsky 2000), distributed to several places in the derivation, the point will be why in the case of 'movement' the phonetic form of the in-situ occurrence is lost though nothing should prevent it from being realized there. A key to solve this problem will, I suppose, lie in the saturation process that Rizzi (1986) proposes. He states that a theta-role is saturated (and its phonetic form is lost) when its referential content can be recovered from the context (Rizzi 1986:508). His statement can be extended in the way that when there exists 'something' (e.g., some morphological marking or the abstract entity like the context) that enables us to confirm the relevant element, the phonetic form of the element can be omitted in the position in which the element should canonically be realized. Then, I propose to formulate the saturation process as defocalization strategy in general, which can suppress the phonetic form of an element in its canonical position when some 'manifestation' that lets us confirm the relevant element can be found.

The explanation will be given as follows: in all the cases of wh-movement, DP-movement, and verb movement, movement cross-linguistically takes place. The phonetic form of the original position of the moved element can be suppressed due to the morphological realization like [wh] of a wh-phrase, Case, and an agreement morpheme respectively. I argue that the cross-linguistic difference in the presence/absence of movement (of all kinds) will be accounted for in terms of which position, either in situ or in the higher position, is pronounced.

[ Top of the page ]

Stylistic Fronting

Hrafnbjargarson, Gunnar Hrafn

Stylistic Fronting (also known from Swedish linguistics as "kil-konstruktionen", or the wedge construction) has been claimed to be an optional leftward movement of various elements, (e.g. an adverb, a participial, a verbal particle, etc.) to IP-Spec in sentences with no overt subject, such as subject relative clauses. Stylistic fronting is found in Icelandic, (1), and Faroese, (2), but not in the Mainland Scandinavian languages, i.e. Danish, Norwegian and Swedish, (3).

(1) Is. Þetta er maður sem lesið hefur t margar bækur
This is man that read has many books
'This is a man that has read many books'

(2) Fa. Hatta er ein maður sum lisið hevur t mongar bøkur
This is a man that read has many books
'This is a man that has read many books'

(3) Da. *Dette er en mand som læst har t mange bøger
This is a man that read has many books

Stylistic Fronting is quite common in Old Danish and Old Swedish but as Falk (1993) observes, the loss of V°-to-I° movement and the loss of Stylistic Fronting took place simultaneously in Old Swedish. This also seems to hold for Middle Danish.

As pointed out by Platzack (1988), Stylistic Fronting can also be found in relative clauses with an overt subject pronoun. However, Falk (1993) shows that the examples that seem to involve Stylistic Fronting in relative clauses with overt subject pronouns, do not involve Stylistic Fronting, instead, their unexpected word order is caused by the lack of V°-to-I° movement. As I will show, data from Middle Danish support Falk's (1993) hypothesis. Stylistic Fronting in relative clauses with an overt subject pronoun can be found in Icelandic. Interestingly, there is a difference with respect to which elements can undergo Stylistic Fronting in relative clauses of this type and relative clauses with no overt subject. In the former, only heads can undergo Stylistic Fronting, whereas in the latter both heads and maximal projections can. My claim is that there must be two different landing sites for the two different types of Stylistic Fronting, which cannot be accounted for in previous analyses of Stylistic Fronting, such as Holmberg (2000).

[ Top of the page ]

Computational Testing of Five Swahili Dictionaries

Hurskainen, Arvi

An accurate and effective method of testing the performance of a dictionary is to convert it into a morphological analyser, which then can be used for comparing a text and the dictionary. In this method, the analyser is made to mirror the information given in the dictionary. It reveals missing headwords as well as deficiencies or mistakes in supplementary information regarding the use of the word.

A Swahili morphological analyser, based on two-level morphology, was used as a base for constructing dictionary-specific versions of five current Swahili dictionaries. These include two authoritative dictionaries compiled by the Institute of Kiswahili Research, University of Dar-es-Salaam, as well as one based on corpus data.

Testing was carried out with three different sub-corpora extracted from the Helsinki Corpus of Swahili. Each of these represents distinct text categories, i.e. older newspaper texts, fiction (including some scientific text in the field of literary research and health care), and texts from a number of newspapers from the year 2002. The monolingual dictionary was also tested against itself. The results show that the monolingual dictionary used about 1,000 such words in explanations as were not found as headwords. The two authoritative dictionaries contained a large number of such words, in some word classes even 50% of words, as were not found in the corpus, while several hundred fairly common words were missing. On the other hand, the dictionary compiled on the basis of a corpus of about one million words, although its size is only 75% of the two authoritative ones, had a slightly better total coverage and far fewer unused words than the larger dictionaries. Especially noteworthy is that while the larger dictionaries had omitted a number of common words, omissions in the smaller one were almost exclusively rare words.

In addition to statistical data, this testing system produces categorised lists of missing words with frequency information, as well as examples of usage from the corpus for each word. The limitation in the use of this method is, however, that one has to be well acquainted with the structure of the analysis program and have access to it.

[ Top of the page ]

Historical Productivity of Some Derivational Suffixes in Icelandic

Indridason, Thorsteinn G.

Historical productivity of some derivational suffixes in Icelandic The study of historical productivity can give us some valuable insights into the forces that expand the vocabulary of a language from one period to another. In my paper I will present a study on historical productivity of some derivational suffixes in Icelandic. With the use of data from Icelandic texts from 1550 to the latter part of the 20th century one gets a fairly clear picture of how the productivity of these derivational suffixes has changed and possible explanations for that. Some derivational suffixes that were relatively productive before seem to have lost their productivity while others have increased it. This indicates that it is necessary to study historical productivity in order to say something sensible about synchronic productivity. Furthermore, I'll show an example of an imported derivational suffix that has become productive, i.e. it has began producing new words from indigenous bases, and I'll also show the birth of a derivational suffix and how it's productivity has developed through time.

I'll also discuss some problems in using corpuses of written language in linguistic research; the difference in productivity of imported and indigenous derivational suffixes; and the nature of productivity in general.

Some references
Bauer, Laurie. 2001. Morphological Productivity. Cambridge University Press, Cambridge.
Plag, Ingo. 1999. Morphological Productivity: structural constraints on English derivation. Mouton de Gruyter, Berlin and New York.

[ Top of the page ]

A Phrase Structure Grammar for Treebank Use

Johannessen, Janne Bondi & Lars Nygaard

Phrase structure grammars (PSGs) can be rather simple as far as linguistic sophistication is concerned. Unless they are equipped with either some kind of unification mechanism for semantic information, or some kind of slash feature like in GPSG (ref.), they will usually overgenerate massively.

However, if the input that the phrase structure grammar takes is already partially analyzed, obviously the results of the PSG grammar are expected to be much better. Voutilainen (1998) discusses the idea of using POS taggers for syntactic parsing: While it is expected to be helpful for the parser that a lot of disambiguation has already been performed by the tagger, there is also a danger in that mistakes made by the tagger will be transported to the syntactic parser. Voutilainen concludes that the output of the parser gains a lot by using tagged input.

The University of Oslo has a morphological CG tagger which, like other CG taggers, has a relatively high accuracy (recall 99 %, precision 95.4 %.) In addition, it performs some syntactic analysis, assigning function tags like SUBJECT, OBJECT etc. on the head word of each "phrase" in the usual CG way (Karlsson et al 1987). It should be stressed, however, that the syntactic part of the tagger is very rudimentary and has a rather low performance (at this stage we have no measure for its performance).

Syntactic annotation is a central part of the development of a treebank. A minimum requirement for any treebank is to present syntactic chunks rather than just simple words. The dependency style CG syntactic annotation does not mark chunks or phrases. However, Voutilainen (2001) shows how a dependency CG grammar output for Swedish can be used as input for a grammar that explicitly states all the syntactic relations in the input sentences. Bick, likewise, describes a system in which a syntactic CG has been used as a basis for automatically constructing "ordinary" phrase structuretrees, to be used in treebanks for, among others, Danish.

In this paper, we describe how the output of the Norwegian University of Oslo CG tagger, with its acceptable morphological performance and not so acceptable syntactic performance, has been used as a basis for automatic phrase structure chunking. The results are interesting, in that very few and very simple PSG rules can handle most sentences. Compared with Bick's system, our grammar is small and simple; his module for Danish consists of 1350 rewriting rules, ours of around 200. There is currently no treebank for Norwegian, so we have as yet no possibility of accurately testing our grammar. However, with our grammar, the first Norwegian treebank is within reach.

Bick, E. A CG & PSG Hybrid Approach to Automatic Corpus Annotation. Ms. Southern Denmark University.
Hagen, K., J. B. Johannessen and A. Nøklestad. 2000. A Constraint- Based Tagger for Norwegian. In Lindberg, C.-E. and S. Nordahl Lund (eds.): 17th Scandinavian Conference of Linguistics, vol. I. Odense: Odense Working Papers in Language and Communication, No. 19, vol I.

[ Top of the page ]

The Role of Contrast in Pronoun Interpretation: A Look at the Referential Properties of the Estonian Pronoun 'tema'

Kaiser, Elsi

The form of referring expressions is said to be connected to the accessibility/topicality of their referents: The most reduced referring expressions refer to highly accessible referents; more marked expressions to less accessible referents (e.g.Ariel 1990). In languages with full and reduced pronouns, full forms are said to refer to less accessible referents (Bresnan 2001).

I use corpus data to test the predictions these accessibility-based claims make for the Estonian long pronoun 'tema'(he/she) in relation to the short form 'ta'(he/she). In particular, I explore how these predictions relate to Pajusalu's claim (1995,1997) that 'tema' is used for referents being compared/opposed to other referents. I conclude that an accessibility-based approach cannot capture the referential properties of 'tema'.

Combining the claim that full forms refer to less accessible referents with the finding that subjects are more accessible than objects/obliques (e.g.Brennan, Friedman, Pollard 1987), leads to the prediction that 'tema' will not be used for subjects. However, in my corpus (50 tokens of 'tema' in subject position, randomly selected from four novels), in the grammatical roles of the antecedents of 'tema,' there is a slight preference for subjects (ca.48%) over other arguments (objects/obliques combined=ca.24%). This does not match the prediction that 'tema' is used for lower-accessibility/non-subject referents.

Instead, almost all corpus uses of 'tema' involve explicit or implicit contrast. In some cases, the contrast between the antecedent of 'tema' and other referent(s) is explicit, but in many cases it is implicit and not inferable without context. At first glance (and out-of-context), these two uses of 'tema' look different. However, I show they can be unified by the notion 'contrastive topic' in the sense of Buering (2002). Buering uses 'contrastive topic' to mean a member of a salient set that contrasts with other members, as specified by the predicate. I extend this notion to Estonian, and discuss in detail what it means to claim that 'tema' is used for contrastive referents. Crucially, the data show that the antecedent of 'tema' is not interpreted contrastively in the discourse preceding the sentence with 'tema.' Thus, the use of 'tema' provides new information about its referent (that it is contrastive) and differs from 'normal' pronouns which are just 'retrieval instructions' for the intended referent.

Conclusion: The referential properties of 'tema' do not fit accessibility-based predictions, since use of 'tema' hinges on contrast, not accessibility. We should not try to map all referential forms onto a unified accessibility scale.

[ Top of the page ]

Stribevis af eksperter, drøssevis av sjansar, hinkvis med kaffe: An Emerging Plural of Abundance in the Scandinavian Languages?

Kinn, Torodd

This paper discusses a subgroup of words ending in -vis found in Danish, Norwegian and Swedish. The words in question have in common that a) -vis is suffixed to a measure noun, b) their meaning is one of abundant quantity, and c) their syntax resembles that of bare plural indefinite measure nouns. In a typical usage, they appear initially in a phrase functioning as an NP (e.g. subject, object of preposition) and are followed by a preposition (Danish: af, Norwegian and Swedish: av or med) and an NP. Simplified authentic examples: Danish: Massevis af pensionister kan ikke klare sig. Norwegian: Folket kjøper tonnevis med aviser. Swedish: Det har blivit kilometervis av stavgång under träningarn a. Another usage is common in Danish and Norwegian, but rare in Swedish. Here the word in -vis is used alone as the object of a preposition, typically i, and this PP functions as an adverbial or a postnominal adjunct. Examples: Danish: Huset havde stået tomt i årevis. Norwegian: Dokumenter i kilovis vitner om seriøse utredninger. Swedish: Hon travade omkring i timmavis. The full class of words in -vis also includes adverbs like naturlig(t)vis and distributive adverbs/adjectives like kommun(e)vis. In the literature, the abundance subclass is usually not distinguished from the distributives, and the words are variously classified as derived adverbs or compound nouns. It is sometimes observed that the construction type N-vis af/av/med NP is structurally opaque (Western 1921, Teleman et al. 1999), and it has been argued that the words are derived quantifiers (Kinn 1998, 2001).

This paper will sketch the diachrony of the constructions involved and discuss their synchronic structure. The latter issue is related to the category of the words in -vis. Corbett (2000) discusses plurals of abundance in several languages and points to cases where plurals have emerged from distributives. Based on this, an alternative to the old assumptions will be discussed: Could abundance words in -vis be better regarded as an emerging inflectional category of measure nouns, a plural of abundance?

Corbett, Greville G. (2000): Number. CUP, Cambridge.
Kinn, Torodd (1998): Drøssevis av forsømte avleiingar: ord på -vis. Nordica Bergensia 19.
Kinn, Torodd (2001): Pseudopartitives in Norwegian. Dr.diss., Bergen.
Teleman, Ulf et al. (1999): Svenska Akademiens grammatik. Svenska Akademien, Stockholm.
Western, Aug. (1921): Norsk riksmŽåls-grammatikk. Aschehoug, Kristiania.

[ Top of the page ]

"This article analyses ..." - Research and Writing in the Writings of Researchers

Kinn, Torodd

This paper presents preliminary results from the project Cultural Identity in Academic Prose (KIAP). KIAP is a doubly comparative study of scientific research articles in English, French, and Norwegian within the disciplines of social economy, linguistics, and medicine. The project asks whether there exist cultural identities in academic discourse, and - if so - whether these are more tightly connected to the language of the author or the discipline. The project is primarily linguistic, and cultural identities are sought in phenomena of language use. KIAP focuses on the voices of the writer and other researchers, and areas of special attention include pronoun usage, metadiscourse (Dahl to appear), bibliographical references (FlŽøttum 2003c), and expressions of epistemic modality.

Fløttum (2003a) has studied the use of the pronouns I, we, and one (and corresponding ones) and looked (Fløttum 2003b) at what kind of activities are connected to the authorial I by looking at the verbs in French that are used with je as the subject. Of particular interest are verbs for research (like English analyse and compare), textual organization (e.g. conclude and underline), and opinions (e.g. reject and think). The use of these tells us a good deal about the way writing researchers choose to represent their own activities. It is known that there are differences between the disciplines with respect to how visible the writers are in the text (FlŽøttum 2003a).

This paper reports a study of the use of a selection of such verbs in 60 Norwegian articles (20 from each discipline). It presents the relative frequency of the verbs, the syntactic constructions they appear in (active, passive, impersonal passive, etc.), and the subjects that they are used with. It discusses to what extent the disciplines differ, and whether the differences merely reflect the fact that the authors perform different research activities, or if they can be assumed to be evidence of differing cultural identities of writing in the three disciplines.

Dahl, Trine (to appear): "Textual metadiscourse in research articles: a marker of national culture or academic discipline?".
Fløttum, Kjersti (2003a): "Personal English, indefinite French and plural Norwegian scientific authors?" Norsk Lingvistisk Tidsskrift 21/1.
Fløttum, Kjersti (2003b): ""Je" et le verbe". Tribune 14.
Fløttum, Kjersti (2003c): "Polyphony and bibliographical references". Kjersti Fløttum and François Rastier (eds.): Academic discourse. Multidisciplinary approaches. Novus, Oslo.

[ Top of the page ]

Avoid Ambiguity: On the Encoding of Causee, Recipient and Beneficiary in Tritransitives

Kittilä, Seppo

The present paper discusses tritransitive constructions from a cross-linguistic perspective. The given constructions comprise causativized ditransitives (like 'the man made the boy give the book to me) and ditransitives with an additional beneficiary ('the man gave the book to me for the woman'). The paper focuses on cases in which two of the roles under study are present simultaneously. As opposed to other studies concerned with the notion (especially causatives, like Comrie 1975/1976 and Song 1999), we will make the claim that the morpho-syntactic nature of tritransitives arises from the need to resolve potential ambiguity overtly. The approach adopted here is thus more semantically motivated than the previous ones. What makes the scrutinized roles interesting is their identical coding (e.g. in the dative) in a number of languages. Furthermore, they all typically refer to animate participants. Thus, the presence of two of the roles in question yields ambiguity, since neither the marking nor the animacy of the arguments disambiguates the semantic role assignment. This is clearly different from canonical ditransitives, in which the differences in animacy suffice for disambiguating the semantic role assignment. Languages therefore need to resort to other mechanisms in order to distinguish between the roles. The most relevant of these are rigidity of the order in which the relevant arguments appear, changes in the verb morphology (applicativization), and changes in the case marking of one of the arguments.

The claim proposed here is supported by a number of facts. First, languages in which the marking of the roles is inherently different, usually permit tritransitives without further modifications. Second, resorting to unorthodox ways of marking is confined to cases displaying potential ambiguity, i.e. it occurs only if two of the roles are present. This also means that changes in the argument marking occur only if neither the marking nor the animacy resolves the ambiguity. Third, there are only a few languages that lack disambiguating mechanisms for these cases altogether. And fourth, the avoidance of ambiguity blocks the formation of tritransitives in some languages (e.g. Epena Pedee). Additional evidence for our claim comes from other seemingly similar cases. For example, in Retuarã, Recipient and Theme are both marked on the basis of animacy alone, but if they both have an animate referent, the word order becomes rigid (Theme-Recipient).

[ Top of the page ]

A New Cognitive Framework for Russian Aspect

Kravchenko, Alexander V.

This paper is an invitation to approach the phenomenon of aspect from a theoretically different angle (as compared to traditional aspectual studies) looking, in the first place, at purely grammatical evidence, both morphological and syntactic. First, I will question the validity of the traditional approach to the semantics of verbal aspect based on the features "boundedness" and "totality" which, within the current theoretical paradigm, constitute the semantic invariant of aspectuality. I will argue that these features are not inherent in the meaning of all perfective verbs, so they cannot be regarded as the semantic invariant.

Then, I shall look at some intriguing data from Russian syntax and morphology and analyze the derivational patterns for base non-derived verbs.The morphological paradigm for the Russian perfective and non-perfective verbs will be summed up in 3 tables showing clear relationships between the form and meaning both within each verb class and for all 3 classes as a whole. It will be shown further, in a subsection dealing with the syntax of the Russian aspect, that aspectual oppositions reflect different cognitive statuses of events expressed by paired verb lexemes: observed events are categorized in the form of aspectually marked verbs, whereas aspectually unmarked verbs categorize events without reference to observation (Kravchenko 1993). These observations will be followed by a discussion of the meaning of aspect, where I hope to show that the true nature of aspect in Russian is radically different from what it is traditionally believed to be. I will conclude by defining the grammatical meaning of aspect as "indication to the source of information about the event" which can be definite (based on observation), or indefinite (based on speaker's knowledge), thus relating it to the cross-linguistic phenomenon of evidentiality.

The suggested cognitive framework for Russian aspect parallels a similar framework for English (Kravchenko 1990; 1997; 2002), providing a unified methodology for typological aspectual studies.

Kravchenko, A. V. (1990). "K kognitivnoi teorii vremeni i vida." Filologicheskije nauki, 6.
Kravchenko, A. V. (1993). "K probleme nabludatelja kak sistemoobrazujushchego faktora v jazyke". Izvestija akademii nauk. Serija literatury i jazyka, 52:3. 45-56.
Kravchenko, A. V. (Ed.) (1997) The English verb: A new grammar for everyone. Irkutsk Linguistics University Press.
Kravchenko, A. V. (2002). "A cognitive account of tense and aspect: resurrecting "dead" metaphors." Anglophonia. French Journal of English Studies, 12. 199-212.

[ Top of the page ]

On the Role of Phonological and Morphological Complexity in Writing

Leiwo, Matti & Åsa Nordqvist

In this paper we will present some data on the role of linguistic complexity in writing processing. Traditionally, writing has been studied from the point of the product, and less is known about the process behind the product. In order to study the process, we have used the ScriptLog program (StrŽömqvist & Karlsson 2001) which makes it possible to study the writing process in real time.

The Finnish orthographic system is at the extreme end on the scale of writing systems in the sense that there is an almost one-to-one correspondence between the graphemes and the phonemes, and one can decode the written forms in phonological form by knowing the single value of the graphemes, with a few exceptions. Most words can also be divided into syllables by one simple rule. Hence, the study of the process of writing Finnish gives good possibilities to study the role of phonological and phonotactic features and syllable and word structure in this process.

Thirty-nine 9-year-old children performed a nonsense word dictation task in which the words they were asked to write on a portable computer consisted of two-syllabic nonsense words. The stimuli words were (near) minimal pairs of short and long vowels or consonants (e.g. sinti - sintti, sintti - siinttu, siinttu - siinttuu, sinti - syntti). The transition time, i.e. the time that elapses between two key strokes/letters, is seen as reflecting processing of linguistic complexity. Our hypotheses are that the decision of whether the segment is short or long is made before (rather than within) the short or long segment and that the marking of long quantity is more difficult. Thus, the transition times in front of short and long stops should be longer than the average transition time within words and this result would be even more pronounced in the cases of long quantity. We have also constructed a hypothetical linguistic complexity scale and related it to the transition times in the different words. Our hypothesis is that increased phonological and morphological complexity will increase the difficulty of writing and thereby affect the transition times. In addition, an analysis of misspellings and corrections of the words is made.

The study is part of the projects "Language Development and Dyslexia during the First School Years", supported by the Academy of Finland, and "Jyväskylä Longitudinal Study of Dyslexia".

[ Top of the page ]

Social History and Language Type: What Could the Balkans Show Us?

Lindstedt, Jouko

It is an old idea in linguistics that some grammatical features of a language could be dependent on the social environment in which it is spoken, but this assumption is difficult to prove empirically. Creole languages may present the clearest case, as the simplicity of their grammars must be due to their special social history (McWhorter 2001). Trudgill (2002) has defended linguists' old intuition that high-contact languages in general tend to simplify and regularize their grammars. The Balkans, a linguistic area that has been studied for nearly two hundred years, should be a good test case for such a hypothesis. The task of Balkan linguistics should be (1) to describe the Balkan linguistic type, (2) to study the social history of the Balkan languages, and (3) to explain what kind of sociolinguistic and interactional processes could be the causal mechanism between the social and linguistic facts.

Ad (1). As pointed out by Thomason and Kaufman (1988: 96), among others, in Sprachbund situations an isomorphism of structures is a more likely outcome than outright simplification. The Balkan languages favour explicit and analytic marking of syntactic functions (Lindstedt 2000), though their verb systems have remained notoriously complex. Grammatical Balkanisms are the biggest common denominators of the Balkan languages rather than typologically simplest structures. They are often what have been characterized as highly borrowable structures (Dahl, ms.), though it is often difficult to show the first source of the borrowings in the Balkans.

Ad (2). The social history in the Balkans is characterized by complex ethnic mixing in two multi-ethnic empires - first the Byzantine, then the Ottoman Empire. Nation-states have shorter history here than in most of Europe. Balkan "Kommunikationsgemeinschaft" (Walter 1991) must be an ancient phenomenon, though historical documentation is scanty. Much could be learnt from similar situations in other parts of the world, such as the Indian subcontinent.

Ad (3). The link between social and linguistic history could be found by looking for the greatest concentration of Balkanisms in time, place, and social structure. Areally, Balkanisms concentrate in the centre of the Balkan Peninsula; sociolinguistically, in the middle of the prestige scale of the Balkan languages; and on the temporal scale, in the Balkan Slavic languages, which were later arrivals than Albanian, Greek, and Balkan Romance, but earlier than Romani and Ladino. This triple middle-of-the-scale phenomenon shows that it was an interplay of several factors that produced the Balkan linguistic type.

Dahl, Ã-sten (manuscript). The growth and maintenance of linguistic complexity.
Lindstedt, Jouko. 2000. Linguistic Balkanization: Contact-induced change by mutual reinforcement. Pp. 231-246 in D. G. Gilbers & al. (eds.), Languages in Contact. Amsterdam & Atlanta, GA: Rodopi. (Studies in Slavic and General Linguistics, 28.)
McWhorter, John H. 2001. The world's simplest grammars are creole grammars. Linguistic Typology 5:2/3, pp. 125-166.
Thomason, Sarah Grey & Kaufman, Terence. 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press.
Trudgill, Peter. 2002. Linguistic and social typology. Pp. 707-728 in Chambers, J. K. & Trudgill, Peter & Schilling-Estes, Natalie (eds.). 2002. The handbook of language variation and change. Malden, Massachusetts & Oxford: Blackwell.
Walter, Hilmar. 1991. Sprachgemeinschaft - Kommunikationsgemeinschaft - Balkansprachbund. Balkansko ezikoznanie - Linguistique balkanique 34: 3-4, pp. 3-10.

[ Top of the page ]

Perception Verbs Are (also) Control Verbs: A Norwegian Perspective

Lodrup, Helge

In Lexical Functional Grammar (as in versions of Minimalism), raising and obligatory control are closely related phenomena. This paper discusses an especially striking case, the perception verb construction with an object and an infinitive, as in (1)-(2).

(1) Vi ser gassen gloede 'we see the gas glow'
(2) Vi ser ham stupe 'we see him dive'

I assume that the semantic subject of the infinitive is a syntactic object of the perception verb. It cannot be a thematic object in a sentence like (1); there is subject-to-object-raising. It is not as clear in (2), however. I will argue that the perception verbs can be both raising verbs and control verbs. (This idea has been proposed now and again, see Barron 1999.) This means that (2) can be a case of raising or control.

I will present Norwegian evidence for the control analysis, including the following:

- The semantic subject of the infinitival complement can be a complex reflexive (Hellan (1988:123), as in (3).

(3) Han hoerte seg selv synge 'he heard himself sing'.

- The semantic subject of the infinitival complement can be the object of a preposition, as in (4); this excludes a raising interpretation.

(4) Vi saa paa ham stupe /?? paa gassen gloede 'we looked at him dive / at the gas glow'

- The perception verb construction is spreading from perception verbs to other verbs with similar meanings (see Karker 1996 on Danish). Some examples are in (5). Some of these verbs can only be used as control verbs.

(5) forestille seg 'imagine', huske 'remember', merke 'notice', oppleve 'experience', tenke seg 'imagine', vise 'show'

It has often been observed that the perception verb construction is different from the more general raising to object construction with verbs like _anta_ 'assume', _forvente_ 'expect'. Perception verbs (and some of the verbs in 5) can also be used in this general raising to object construction, as in (6).

(6) noe vi ikke kan se aa vaere tilfelle 'something (that) we cannot see to be the case'

The syntactic and semantic properties of the verbs are then different from their use in the perception verb construction - and they do not have the competing control analysis.

[ Top of the page ]

"'Ajak' of all Trades": Problems with Categorizing Balinese "ajak" in Discourse

Luna, Edmundo

The determination of "lexical categories" has persistently been an elusive issue within linguistic analysis, as resolving the issue is far from straightforward for many languages. One distinction that is presumed to be applicable in most languages is "verbs" versus "everything else"; however, there are cases where even this distinction is not warranted. Studies such as Hopper and Thompson (1984) refine such problematic factors in determining "lexical categories" by including discourse-functional based considerations. However, there may be room for refining these factors further, such as accounting for differences across discourse genres. Using both written and spoken discourse-based corpora, this study will show that the Balinese morpheme 'ajak' (traditionally described as a "verb" meaning 'to take with one, take along, invite to do sth with one' [Barber 1979]) necessitates this additional consideration if one considers its usage-based frequency across discourse genres. In written narratives, 'ajak' occurs most often as a legitimate "verb", i.e. it frequently occurs with the verbal morphology available in Balinese, which include the nasal prefix 'ng-' and the third person suffix "-a", as shown in (1) and (2):

(1) "Apang ya nyak makurenan ngajak I Satigga"
so.that 3(L) want MA-spouse N-AJAK TITLE S.
'so that she wants to be married with I Satigga." [MAMADIK]

(2) Wayan Tamba ngosong luhne ajaka ka bale dangin.
W. T. N-lift.up woman-3:POSS AJAK-PT to pavilion in.east
'Wayan Tamba lifted up the woman (and went) with her to the east pavilion.' [MATEGUL]

However, in spoken discourse genres, the status of "ajak" is much more obscure: tokens representing 'ajak' without any overt verbal morphology are much more prevalent. An example is shown in (3):

(3) ...Apang selamat ajak onyang,
so.that auspicious AJAK all
'so that everything is well,' [Collected by the author, 2003]

In this example, 'ajak' does not have any overt verbal morphology. This does not mean that 'ajak' cannot semantically act as a verb here, since "verbs"- or predicates-in-series are possible in Balinese. Rather, this means that the distinctions between "verbs" and other lexical categories are not as clear in spoken discourse genres in Balinese.

Thus, this study will demonstrate that the determination of "lexical categories" lies not only in the various linguistic devices used within discourse in general, but also within frequency-based variations found in the available discourse genres in any particular language. (380 words)

Barber, C. C. 1979. A Balinese-English dictionary, vol. 1-2. Aberdeen University Library Occasional Publications 2. Aberdeen: Aberdeen University.
Hopper, Paul J. and Sandra A. Thompson. 1984. The discourse basis for lexical categories in universal grammar. Language 60(4): 703-752.

[ Top of the page ]

Grammatical Complexity of Languages: A Cross-linguistic Pilot Study

Miestamo, Matti

The received view among linguists is that all languages are equally complex -- only the loci of complexity may vary from language to language. This view has recently been challenged by e.g. McWhorter (2001) and Kusters (2003). There is no principled reason why all languages should be equal in their overall complexity or why complexity in one grammatical domain should be compensated by simplicity in another. My paper addresses these questions from a cross-linguistic point of view. Ranking the world's languages in terms of complexity is not an interesting question per se, but developing a means for talking about complexity is crucial for answering these highly relevant questions. For the purpose of studying whether creoles tend to have simpler grammars than non-creoles, McWhorter (2001) proposes a metric for measuring the overall structural complexity of languages. The metric pays attention to overt signaling of phonetic, morphological, syntactic and semantic distinctions beyond communicative necessity: a grammar is more complex than another to the extent that 1. its phonemic inventory has more marked members, 2. its syntax requires the processing of more rules, 3. it gives overt and grammaticalized expression to more fine-grained semantic and/or pragmatic distinctions, and 4. to the extent it uses inflectional morphology. In a study focusing on the complexity of verbal inflection in four genealogically defined language groups, Kusters (2003) proposes three central principles the violation of which entails higher complexity: 1. Transparency -- clarity of the relation between meaning and form, 2. Economy -- restriction of the number of overtly signal led categories, and 3. Isomorphy -- identity of the order of elements in different domains.

The issue of complexity has not been addressed from a general typological perspective. In this pilot study, I examine how the criteria proposed by McWhorter, Kusters and others can be used in a sample-based typological study of complexity, and test their applicability using a small genealogically and areally stratified language sample. The study aims at a deeper methodological understanding of the typological study of complexity. It will be a step towards the goal of establishing a metric of complexity that can later be used in more extensive typological research.

Kusters, Wouter. 2003. Linguistic Complexity, the Influence of Social Change on Verbal Inflection. Ph.D. Dissertation, University of Leiden. Utrecht: LOT.
McWhorter, John H. 2001.The world's simplest grammars are creole grammars. Linguistic Typology 5.125-66.

[ Top of the page ]

Pitch Accent without Pitch, Voicing without Voice: Studies to Scottish English and Norwegian Whisper

Mills, Timothy & Hannele Nicholson

Pitch accent without pitch, Voicing without voice: Studies to Scottish English and Norwegian whisper Prosodic cues such as lexical tone and pitch accent depend on the rate of vibration of the vocal folds. Voicing contrasts are signalled by the presence or absence of such vibrations. In the absence of vocal fold vibration (such as in whispered speech), how do listeners perceive these linguistic features?

Previous studies have shown that pitch is perceptible in whispered Chinese, Thai, Norwegian, and Swedish. We look at other contrasts that are based on vocal fold vibration: pitch accent in Norwegian and voicing contrasts in Scottish English.

Nicholson and Teig (in press) presented perceptual tests of Norwegian pitch accent. The current paper furthers these analyses and presents results from Scottish English voicing contrasts. Both studies show better-than-chance perceptual performance in distinguishing minimal pairs in whispered speech, even when controlling for syntactic and semantic context. Such a result indicates that the whispered acoustic signal contains sufficient secondary cues to compensate and preserve the distinction.

An initial acoustic analysis of Norwegian shows that there is a correlation between correct perception and word duration. This suggests that duration may be an important cue. Vowel duration has also been identified as a useful cue to voicing contrasts in Scottish English (Mills, Forthcoming).

Further acoustic analysis of the Norwegian data will look at syllable and vowel durations, intensity, and formant values as further possible cues to pitch accent.

Ongoing research in the Scottish voicing contrasts is being conducted to establish a more detailed picture of the perceptual performance, based on place and manner of articulation and vowel context. Acoustic analyses will also be undertaken to test further cues to voicing, such as consonant duration and vowel formants, in whispered speech.

We hope the current work will enhance our understanding of the strategies used in the production and perception of normal speech.

Mills, Timothy. Forthcoming. Cues to voicing contrasts in whispered Scottish obstruents. MSc. dissertation, University of Edinburgh.
Nicholson, Hannele and Andreas Hilmo Teig. In press. How to tell beans from farmers: cues to the perception of Norwegian pitch accent in whispered speech. Proceedings of the 19th Scandinavian Conference of Linguistics, Tromso, Norway.

[ Top of the page ]

On the Existence of Spatial Adverbs in Grammar: a Functional Account of the Problem of Intransitive Prepositions

Moreno, Ana Ibanez

The purpose of this paper is to give a brief overview of how spatial prepositions and adverbial particles have been developed in the literature, and to present a new classification of such items. Thus, after a review of descriptive grammars' treatment of prepositions until the present moment, some of the main ideas related to prepositional analysis proposed by the main contributors (Gawron 1985, O'Dowd 1998, etc.) to this theoretical field are presented. Besides, some innovative contributions to the literature are added, in the line of functional grammars (Van Valin & LaPolla 1997). Finally, in this work an attempt is made to define the clear boundaries that permit to distinguish adverbs, operators and prepositions, since this issue is the origin of some controversy in the field.

Traditionally, locational adverbs and prepositions have been considered as interrelated word classes, which is mainly due to the fact that they share many of their forms. They are distinguished in terms of their syntactic organization. Thus, prepositions are always followed by an object complement, while adverbs appear alone. Besides this, prepositional objects have to be, according to tradition, noun phrases (hereafter NPs). However, in the past recent years there has emerged a new line of discussion in this respect: some authors like Huddleston & Pollum (2002) defend the view that locational prepositions can be followed by an object complement which performs any function (e.g. prepositional phrase -PP-). Even more, these authors (2002), as well as others (Van Valin & LaPolla 1997) state that prepositions can be transitive, if they are followed by an object complement, or intransitive, if they are not. This idea, as can be realized, is controversial in the sense that the function of adverbs as locational is eliminated and substituted by the function of spatial prepositions.This new classification brings about a number of problems to the grammatical organization of languages. Therefore, it is critically examined in this paper, and a number of irregularities that it brings about are presented. After that, some proposals are given that attempt to solve some of the existent problems to classify locational expressions.

[ Top of the page ]

From silence to sound: Early spoken language development in deaf children who receive a cochlear implant

Nordqvist, Åsa

The number of congenitally deaf children who receives a cochlear implant (CI), i.e. a device that provides direct electrical stimulation to the auditory nerve and allow for the perception of sound sensation, is steadily increasing. The children are also operated upon at younger ages than previously. Today many children are implanted when they are between one and two years of age which means that the children get access to auditory input during a phase that is extremely important for the process of language acquisition. These children have in common with normally-hearing peers a cognitive ability to acquire language, however, the spoken language acquistion process might be affected by the fact that, firstly, there is a certain period of time, before and after birth, when no auditory stimulation and no spoken input are available to the child at all, and secondly, after the implant is switched on, the hearing is still limited and the child has to learn to recognise and distinguish the phonologically and grammatically relevant features of speech.

The study reports through longitudinal case studies the spoken vocabulary and grammatical development of four young Swedish congenitally deaf children during the two first years with the cochlear implant. The children were between 1 year and 8 months and 2 years and 5 months when their implants were connected. Four normally-hearing control children were also included in the analyses. The data were coded and analyzed for the children's vocabulary size and content (total number of tokens, tokens per minute, types and parts of speech) and grammatical complexity (occurrence of inflectional morphology and mean length of utterance (MLU) scores).

The results show that there are clear individual differences in the spoken language development among the children with CI. In some respects the children with CI showed a similar development as the normally-hearing controls, e.g. in the order grammatical categories and parts of speech were acquired. However, there were differences that can be related to the CI-children's hearing loss and a more limited ability to perceive the spoken language around them: Their speech intelligibility was lower, the MLU-curves flatter and they had less inflectional morphology. The latter finding can be explained by the fact that the grammatical morphemes in spoken language tend to be relatively unstressed in comparison to content morphemes, and is likely to be more difficult for these children to perceive.

[ Top of the page ]

Generic NPs in Dutch

Oosterhof, Albert

In this paper, I present the results of my corpus research into genericity. I will focus on the subject-NPs of generic sentences in Dutch. In Dutch there are four types of NP which can function as subject-NP of generic sentences : definite singulars, definite plurals, indefinite singulars and bare plurals.

1a.De ijsbeer leidt een zwervend bestaan. (Singular definite NP)
The polar bear leads a roving life.
b.De ijsberen leiden een zwervend bestaan. (Plural definite NP)
The polar bears lead a roving life.
c.Een ijsbeer leidt een zwervend bestaan. (Singular indefinite NP)
A polar bear leads a roving life.
d.Ijsberen leiden een zwervend bestaan. (Plural indefinite NP)
Polar bears lead a roving life.

First, I examine some differences between Flemish Dutch and Standard Dutch. The most striking observation is that in Flemish Dutch bare plurals are used far more frequently than in Standard Dutch. In Standard Dutch the definite singular appears more frequently.

Second, I draw attention to the application possibilities of these four types of NP. In the literature about genericity in English, there have been noted at least two semantic contrasts. In the first place, singular indefinite NPs cannot be used as subjects of so-called kind predicates. A second phenomenon is that English singular definite NPs cannot refer to a kind that is 'too general'. It is, for example, more natural to attribute a generic predicate to the polar bear than it is to attribute it to the bear or the mammal.

In my talk, I compare these intuitions about English with the results of my corpus research into Dutch. I will argue, contra Cohen (1999), that bare plural subject-NPs of generic sentences are ambiguous between a kind-referring reading and a non-kind-referring reading, while definite singulars refer to kinds unambiguously. This analysis explains that the taxonomic category the NP refers to, has an influence on the form of the NP. Moreover, this viewpoint offers an explanation of the interpretations of bare plurals in contexts where a definite singular gets a representative object interpretation (cf. Krifka et al. 1995 : 83).

Cohen, Ariel. 1999. Think Generic! The Meaning and Use of Generic Sentences. CSLI Publications, Stanford, California.
Krifka, Manfred, Francis J. Pelletier, Gregory N. Carlson, Alice ter Meulen, Godehard Link, and Gennaro Chierchia. 1995. 'Genericity : An Introduction.' In Carlson, Gregory N. and Francis Jeffry Pelletier (eds.) 1995. The Generic Book. Chicago : University of Chicago Press.

[ Top of the page ]

Where is the head? On numeral plus noun constructions and their headedness in Estonian and Izhorian

Payne, John & Katrin Hiietam

As has been recently highlighted by Koptjevskaja-Tamm & WŽälchli (2001), in the Baltic linguistic area the numeral in a noun phrase determines the singular number (and sometimes the case) of the noun and dependent adjective. This is illustrated in examples (1) and (2):

(1) need kakskümmend ilusat pilti
this.NOM.PL twenty.NOM beautiful.PART.SG photo.PART.SG
'these twenty beautiful photos'

(2) nende kahekümne ilusa pildi hind
this.GEN.PL twenty.GEN beautiful.GEN.SG photo.GEN.SG price.NOM
'the price of these twenty beautiful photos'

Example (1) is a nominative plural noun phrase, as reflected in the form of the demonstrative need 'these'. However, the nominative numeral kakskŽümmend 'twenty' appears to determine the form of the noun and dependent adjective: these must be in the partitive case and singular. By contrast, example (2), which illustrates the non-nominative paradigm, is a genitive plural noun phrase. Here only the singular number of the noun and dependent adjective is determined by the presence of the numeral: the genitive case of the noun and dependent adjective agrees with that of the phrase.

It is a natural conclusion, based purely on case and number-coding properties, that the numeral is the head of the construction (Eesti Keele Grammatika II, 1993). The wider syntactic properties of the Finnic construction are however never discussed, in striking contrast with the detailed discussion of the analogous construction in Russian by Mel'chuk (1985). In this paper, on the basis of two Baltic-Finnic languages, Estonian and Izhorian, we show that, as in Russian, the coding properties of the Finnic construction are in some respects misleading: a variety of arguments indicate that the syntactic head is the noun. For example, case resolution rules for coordinate constructions treat numerals as dependents rather than as heads:

(3) kas üks või kaks numbrit
either one.NOM or two.NOM number.PART
'either one or two numbers'

The nominative üks 'one', unlike any other numeral, requires a nominative rather than a partitive noun. Case resolution is nevertheless possible, in contrast with the ungrammaticality which arises with verb and noun heads which govern distinct cases.

The discrepancy between coding and syntactic properties supports Zwicky's (1993) proposal that morpho-syntactic and syntactic notions of headedness are distinct, and requires an analysis in which case and number assignment rules do not necessarily coincide with the axes of syntactic projection.

[ Top of the page ]

Sphere Semantics for Aspect

Pietarinen, Ahti-Veikko

To get a logical grip of aspectual distinctions of verbs has proved elusive. Point-based structures of time and the classical Priorian approaches to temporal phenomena have turned out to be too weak. The interval-based approaches have also lacked sufficient expressivity. I will propose a different model of time based the notion of spheres. I will argue that by such semantics, many aspectual distinctions may be characterised in a unifying logical manner.

Sphere semantics is derived from Tarski's (1929) geometry of solids. Given a universe of closed spheres, possible-worlds semantics is created by defining an alternativeness relation in terms of tangentiality. As soon as we have that, we may define the fundamental notions of external and internal tangents, external and internal diameters, concentricity, and many other notions derived from these.

In sphere semantics, the notions of events, states, processes or episodes are not evaluated in segments of times as in interval semantics, but in closed spheres, which are primitives of the universe. Consequently, the problem in interval semantics, namely that of taking intervals as the primitive notion dispensing with the notion of time points at its extremities, is seen to evaporate.

I will characterise the English progressive as the occurrence of continuous action on spheres that does not terminate unless a sphere is exited via external tangents. Accordingly, the latter notion of tangential exit defines the possibility of a completion. The treatment is mereological in resorting only to the 'part of' and disjointness relations. Being a 'part of' is then used to differentiate between events that hold in homogeneous and those that hold in heterogeneous spheres. Non-duratives are null- diametric spheres of time.

Further, we get a qualitative notion of possible worlds. External diametricity defines a passage from one event to another being maximally protracted, while internal diametricity captures the notion that two subevents are maximally distant. The latter means that two spheres are produced: one for an event that has a beginning (atelicity) and the other for an event that has an ending (telicity); that some subevent both begins and ends requires centric tangentiality.

Since distances between subevents do not have to be maximal, for internal tangencies non-maximality defines an event-internal earlier-than relation, useful in tackling the internal constituency of events referred to by uninflected verbs (primary aspect). By suitable restrictions, external tangencies define the sphere-theoretic analogue to a branching time logic but with a richer structure: as different 'branches' may also overlap, situations that refer to spheres with a common ancestor may share common parts. Overlaps may likewise concern internal properties of primary aspect.

A comparison of this approach with interval-based semantics establishes that the former is a richer framework for the logical treatment of natural language aspect than the latter.

[ Top of the page ]

Contact-induced Changes in Finnish Romani

Pirttisaari, Helena

This paper discusses the interaction between contact-induced and internally motivated changes in Finnish Romani, with particular attention to verbal morphosyntax. I intend to show that Finnish Romani manifests moderate to heavy structural borrowing, resulting in simultaneous simplification and complication of the system. The paper focuses on variation and change, adopting the theoretical concepts of functional approach (e.g. Dressler 1996), contact linguistics (e.g. Thomason & Kaufman 1988) as well as research on obsolescence (e.g. Dorian 1989).

A central question is the extent to which changes in dying languages can be attributed to (i) internal universals, (ii) contact influence from the dominant language(s), or (iii) multiple causation. Do the linguistic effects common to language-contact situations, as loss of structural complexity and an overall increase in semantic transparency, even hold true for Finnish Romani? It is of further interest whether the speakers have preserved some original ways of coding meanings, or adopted the structures of the dominant contact language.

The verbal system of Finnish Romani shows a complex picture of archaisms and innovations. As a residue of the late Middle-Indo-Aryan ergativity system, finitized participles are continuously used in past tense. At the same time, Finnish interference is manifested in various parts of the verbal system. First, Finnish structures as impersonal passive or agent participle occur. Second, morphological oppositions not found in Finnish are lost, as the future tense or the category gender, and occasionally even number in finitized participles.

The significance of the bilingualism of the speakers is evidential. The language has become a Para-Romani-like variant, being morphosyntactically Finnish, although with Romani morphology and some characteristic features of Romani. It has been shown that Romani often uses inherited forms in this way to "copy" the structure of an European contact language. On the one hand, the system is simplifying due to attrition and on the other hand, the contact situation triggers complication of the system, in which even overgeneralization of marked features occurs. This points - in accordance to Thomason and Kaufman (1988) and against Dressler's (1996) and other naturalists' view - to the inappropriateness of the concept of markedness and internal universals in contact situations.

Dorian, Nancy 1989 (ed.). Investigating obsolescence. Studies in language contraction and death. Cambridge University Press.
Dressler, Wolfgang Ullrich 1996. Language Death. Rajendra Singh (ed.), Towards a Critical Sociolinguistics. CILT 125, 195-210. Amsterdam: Benjamins.
Thomason, Sarah Grey & Terrence Kaufman 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press.

[ Top of the page ]

The Futhark and the Fur Trade: On the Adaptation of an Alphabet by its Users

Poussa, Patricia

In Scandiavia the futhark, or runic alphabet, went through two main early forms. Early Runic, a 24-letter form (AD c200-500), was replaced in Scandinavia in the Viking Period (c800-1050) by the Younger Futhark, a 16-letter form. Epitaph stones inscribed in this alphabet are particularly frequent in in the MŽälar region of Sweden (Jansson 1987:187). I ask: whence their reading public?

Wiik's general model of the role of prehistoric lingua francas in the contact between proto-Uralic, proto-Baltic and protoGermanic groups ( Wiik 2002:30-52) seems to provide a possible social motivation for the 16-letter futhark, as a means of writing long-distance business messages on birchbark or wood, in the area of the limes norrlandicus. The limes is a climactic border which marks the boundary between deciduous and peri-arctic woodland, running from the Oslo fjörd to the Mälar region of Sweden and along the southern edge of the Finnish peninsula. During the Iron Age it divided the northern hunter-gatherer economy from the southern agricultural economy. A new class of farmer-traders apparently arose at various locations along this border during the Merovingian Period (c550-800), who controlled the distribution of artic goods such as furs, dried fish, seal-oil, etc., to southern Europe. This class, I suggest, may provide our answer.

The apparent phonological reductions suggest pidginisation, or a trade jargon. Since an alphabet is not a language, I shall argue that the reduced letter set could have represented proto-Finnic as well as proto-Scandinavian speech. The basing of the Younger Futhark on a trade jargon used outside Scandinavia proper would explain why its genesis has proved so puzzling for runic scholars working within the family-tree model of language development, and why the language contact here proposed is not obvious from the place name evidence of the Scandinavian peninsula.

As B. F. Jansson has shown, Swedish viking epitaphs often refer to eastern trading voyages, and short-stave- and staveless runes save writing-time and space. Thus the Younger Futhark itself may also have evolved somewhere along the eastern routes, cut on wood or bark. In Eura in south-western Finland, for example, there are extensive Merovingian Period inhumation cemeteries whose graves have yielded Arabic coins and a mixture of Scandinavian and eastern luxury metal goods the right kind of location for our hypothesis, but unfortunately not wet enough to preserve wood/bark well (unlike medieval Novgorod). We must, therefore, await new evidence from northern archaeologists, who may yet surprise us.

[ Top of the page ]

Clause Type and Clause Order in the Comprehension of Temporal Structures: Experiments with Finnish Children

Pyykkönen, Pirita, Jussi Niemi & Juhani Järvikivi

The interpretation of the order of events in temporal expressions in children in different ages has been shown to depend on a variety of factors, such as, order-of-mention, i.e., linearity (Clark 1971), clause type, i.e., main vs. sub ordinate clause (Amidon & Carey 1972) and the nature of relationship between the events, i.e., logical vs. arbitrary (French and Brown 1977). Recently Pyykkönen et al. (submitted) studied the effects of sentence structure and linearity in the comprehension of Finnish temporal conjunctions (ennen kuin 'before' and sen jälkeen kun 'after') and temporal converb constructions in 8-12 year-old Finnish children. The results showed that the 8- year-old children relied mainly on the linear order of the actions depicted in the sentences as well as on pragmatic feasibility. In contrast, the 10 and 12- year-olds were found to rely more on grammatical structure (morphosyntactic marking). However, the study did not address the question whether the clause type (main clause [M] vs. sub ordinate clause/converb construction [S]) and clause order (M+S vs. S+M) would affect the interpretation of temporal expressions. The present study investigated the effect of clause type and clause order in the comprehension of temporal sentences in native Finnish children in ages 8, 10 and 12. More precisely, a series of off-line questionnaire experiments investigated the interpretation of relative clause structures (ennen kuin 'before', sen jälkeen kun 'after') with conjunctions as well as sentences with temporal converb constructions. The order of clauses (M+S vs. S+M) was counterbalanced between two groups of participants per age group, as follows:

(1) Temporal conjunctions

Markus söi aamupalan sen jälkeen kun heräsi
'Markus had breakfast after he woke up'

Sen jälkeen kun Markus heräsi, hän söi aamupalan
'After Markus woke up, he had breakfast'

(2) Temporal converb constructions

Jorman kävellessä mökille Jyrki löi hevosta
'While Jorma was walking to the cottage, Jyrki hit/was hitting the horse'

Jyrki löi hevosta Jorman kävellessä mökille
' Jyrki hit/was hitting the horse, while Jorma was walking to the cottage'

In our presentation we will discuss our results from of the three age groups of children and focus on the effects of clause type and clause order as well as the effect of age.

Amidon, A. & Carey, P. (1972) Why five-year-olds cannot understand before and after. Journal of Verbal Learning and Verbal Behavior, 11, 417-423.
Clark, E.V. (1971) On the acquisition of the meaning of before and after. Journal of Verbal Learning and Verbal Behavior, 10, 266-275.
French, L. A. & Brown A. L. (1977) Comprehension of before and after in logical and arbitrary sequences. Journal of Child Language, 4, 247-256.
Pyykkönen, P., Niemi, J., & Järvikivi, J. (submitted) Sentence structure, temporal order and linearity: Slow emergence of adult-like syntactic performance in a free word order language.

[ Top of the page ]

Distinctive [voice] Does Not Imply Regressive Assimilation

Ringen, Catherine & Pétur Helgason

In a recent paper, van Rooy and and Wissing (2001) distinguish between what they call the "broad interpretation" and the "narrow interpretation" of the feature [voice]. According to the broad interpretation (Lisker & Abramson 1964, Kingston & Diehl 1994), languages with a two way [voice] contrast may implement this contrast phonetically with any two of the following: voice onset precedes plosive release, voice onset immediately follows plosive release, voice onset substantially lags behind plosive release. According to the narrow interpretation (Jakobson 1949:389, Keating 1990; Iverson & Salmons 1995; Jessen 1989, 1998; Jessen & Ringen 2002) [voice] is employed only when actual vocal fold vibration is present. According to van Rooy and Wissing languages with distinctive [voice], on the narrow interpretation, always have regressive voice assimilation.

The purpose of this paper is to show that Swedish employs the feature [voice] on the narrow interpretation: voice onset precedes plosive release in utterance initial position, (voiced) plosives are produced with vocal fold vibration intervocalically and word finally, but there is no regressive assimilation of [voice].

Six native speakers of Central Standard Swedish were recorded in a sound-treated room at Stockholm University. The speakers read a list of words containing stops from both stop series of Swedish, referred to here as fortis vs. lenis stops. The stops occurred in word-initial position, in intervocalic position and in word-final position, as well as in word-medial and -final clusters. The vast majority of instances of word-initial lenis stops exhibited prevoicing. All intervocalic lenes were voiced and practically all word-final lenes were voiced as well.

Among the test words were two suffixes, one /-t/ and one /-d/e.

köpa 'buy' kö/p+t/ > kö[pt] 'bought (ppart)' kö/p+d/e > kö[pt]e 'bought (past)'

väga 'weigh' vä/g+t/ > vä[kt] 'weighed (ppart)' vä/g+d/e > vä[gd]e 'weighed (past)'

Our data indicate clearly that in underlying mixed voice/voiceless clusters, [-voice] is dominant and the cluster is voiceless on the surface. Hence the claim that languages with narrowly defined voiced plosives exhibit regressive assimilation of voice is incorrect. In Swedish both progressive and regressive assimilation to [-voice] is found.

We present an OT account of the Swedish data which involves both features [±voice] and [±spread glottis].

[ Top of the page ]

A PF Account of the Scandinavian DP

Roehrs, Dorian

PROPOSAL: This paper argues for a novel surface-oriented account of the Scandinavian DP, involving semantically vacuous head movement of N at PF (cf. Boeckx & Stjepanovic 2001). Following Santelmann (1993) and Julien (2002), I assume a lower position where the overt determiner merges. Unlike these authors, I follow Roehrs (2002), who extends ideas of Bruge (1996), in arguing that the determiner MOVES to DP to license D (Longobardi 1994):

(1) [ deti+D [ adjective [ deti [ noun ]]]]

PARAMETRIC ASSUMPTIONS: (i) Taraldsen (1990) argues that the head noun moves across a possessor to incorporate into the (overt) determiner. I propose that the time of noun incorporation (NI) may vary; (ii) Scandinavian determiners can license D under adjacency, extending the proposal of Agbayani (2001) for subject wh-questions in English; (iii) considering the lack of prenominal genitive DPs, I argue that Icelandic has a PF constraint that allows the deletion of the upper unstressed or nonliterary copy of the determiner (cf. Franks 1998, Boskovic 2001).

ILLUSTRATION: (A) modified DP: the overt determiner is not adjacent to D. Old Icelandic follows if NI precedes movement of the determiner to D ({} = deleted):

(2) [ [maðrk+inn]i+D [ gamli [ {maðrk+inn}i [ {maðrk} ]]]] (Old Icelandic)

The double definiteness effect in Norwegian and Swedish follows if NI precedes copy deletion, making the determiner part of a complex word and thus not available for deletion (Nunes 1995). The (different) phonetic realizations of the determiner copies are due to Distributed Morphology (Halle & Marantz 1993):

(3) [ deni+D [ gamla [ mannk+eni [ {mannk} ]]]]

In Danish and literary Icelandic, NI follows copy deletion, leading to the deletion of the lower copy (in fact, here NI only occurs in unmodified DPs, see below):

(4) [ deni+D [ gamle [ {-eni} [ mand ]]]]

Similar to Danish and literary Icelandic, NI follows copy deletion in non- literary Icelandic. Unlike those two languages, the assumed PF constraint on copy deletion deletes the upper copy:

(5) [ {inni}+D [ gamli [ maðurk+inni [ {maðurk} ]]]]

(B) unmodified DP: D can be licensed under adjacency. With no movement of the determiner to D, there is only one copy of it. Independent of the time of NI, we derive the same pattern in all Scandinavian languages:

(6) [ D [ mannk+en [ {mannk} ]]]

word count: 397

[ Top of the page ]

Lexicalization as a Means of Grammaticalization

Rostila, Jouni

Lexicalization and grammaticalization have often been considered forces diametrically opposed to each other (cf. Lehmann 1989). My purpose is to show that lexicalization can also form a device that makes grammaticalization possible. It can serve to obscure the lexical meanings of items and thus enables them to acquire new, more abstract meanings typical of grammaticalization. A basic example of the first stage of such a process can be seen in the development of complex prepositions like 'in light of', 'atop', 'aufgrund', 'på grund av', where the lexicalization of the phrase as a whole frees the noun from its lexical meaning and allows the phrase to acquire a more grammatical status (that of a preposition) along with a more grammatical meaning. As an instance of such a process, I am going to offer an alternative to Detges & Waltereit's (2002) proposal as to how nouns like 'pas', 'step' etc. are grammaticalized as standard expressions of emphatic negation. According to their proposal, the lexicalization of phrases like 'not ... a step', 'ne ... pas' does contribute to the development of items like 'pas' to standard expressions of emphatic negation, but is not a crucial factor in this development. In my view, however, it is precisely the lexicalization of the phrase that sets the stage for the grammaticalization of such items. It amounts to the conventionalization of the pragmatic inference required to understand 'not ... a step' correctly (i.e. as NOT ... AT ALL), thus constituting the reanalysis Detges & Waltereit ascribe to other factors. As a result of becoming parts of a lexicalized phrase, items like 'step' and 'pas' lose their concrete meanings in the context of the phrase and can be grammaticalized further to markers of emphatic negation applicable to any verb, not just to verbs of motion, as was originally the case. A prerequisite for the second development stage - the second reanalysis, to be more accurate - is, however, that phrases like 'not ... a step' occur frequently enough to give the impression that 'step' can express AT ALL in the context of any verb.

I am also going to demonstrate that essentially the same account applies to the development of (German) prepositional object constructions and the grammaticalization of their prepositions (cf. Rostila, in press).

Detges, Ulrich & Richard Waltereit (2002): "Grammaticalization vs. Reanalysis: A Semantic-Pragmatic Account of Functional Change in Grammar." Zeitschrift für Sprachwissenschaft 21.2, 151-195.
Lehmann, Christian (1989): "Grammatikalisierung und Lexikalisierung." ZPSK 42, 11-19.
Rostila, Jouni (in press): "Zur Grammatikalisierung bei Präpositionalobjekten." In: Leuschner, Torsten & Tanja Mortelmans (eds.), Grammatikalisierung im Deutschen. Berlin: de Gruyter.

[ Top of the page ]

Where Anteriority is Located in the German and Swedish Perfect

Rothstein, Björn

GOAL: In this paper I address the question how the anteriority in German and Swedish Perfect is distributed over the Auxiliary and the Participle. In a form like Jag har ätit (I have eaten) anteriority could be either expressed by the auxiliary (auxiliary-hypothesis: Bierwisch (1996), WUNDERLICH (1997), STECHOW (1999), the past participle (participle-hypothesis: MUSAN (1998/2003) or by both (compositional-hypothesis: EHRICH (1992), KLEIN (2000)). I show that neither the auxiliary nor the past participle independently from one another can express anteriority, but that an anteriority reading becomes possible when a past participle is combined with the Perfect auxiliaries have and be. Other auxiliaries (passive) will block the participle s anteriority effect.

THE PROBLEM: None of the hypothesises mentioned above is able to explain where anteriority resides in the Perfect. Auxiliary-deletion in Swedish reduced relatives and elliptic constructions in general provide evidence against the auxiliary hyphothesis, the werden/bli-passives that have by default no anteriority reading (Bilen blir köpt (Car-the is bought)) are evidence against the past participle hypothesis and the compositional hyphothesis cannot be tested in constructions with missing auxiliaries.

THE ARGUMENTATION: When investigating the past participle hypothesis one must consider all participle constructions. I assume two different types of participle constructions: adjectival and verbal participles (LEVIN&RAPPAPORT (1986), RAPP (1998) that can be distinguished by prefixation, distribution, argument structure, pseudo-cleft ... . Adjectival participles are formed by conversion from V to A with a stativizer (KRATZER (2000)) and have stative meaning (GIVÓN (1984)), therefore they will not be compared to the Perfect constructions. The remaining verbal participle constructions (Perfect, kommen (come)+Past Participle and werden/bli- (become), bekommen- (get) and gehŽören (should be)-Passive) will be compared to one another. Since there is no category change from V to A, I assume no stativizer. These verbal participle constructions have different temporal meaning. The passives express no anteriority, but most Perfects do: (I morse har jag ätit (I have eaten this morning)) (Existential-Perfect). Some Perfects do not express anteriority: (Jag har alltid Žälskat dig (I have always loved you) (Universal-Perfect).

The verbal participle allows for anteriority reading with those auxiliaries not expressing processes (werden, bli (become), kommen (come), bekommen (get), gehören (should be)). Have and be are statives and therefore they can locate the participle in time. Universal- and Existential-Perfect can be explained by (c)overt U- and E-reading adverbials (always vs. this morning).

[ Top of the page ]

Adnominal Modifiers in the History of Lithuanian: Syntactic Functions and Word Order

Say, Sergey

The paper focuses on the diachronic development of the word order structures in Lithuanian. In contemporary Lithuanian adjectives and genitives precede their head nouns; contrarily to the cross-linguistically usual word order patterns the relative order of these modifiers is Gen-Adj-N. In the Old Lithuanian written texts (16-17 centuries) position of the genitive and adjectival modifier is often claimed to be relatively free. It is, however, argued here that genitives expressing possessive meaning favoured postposition; these genitives were used referentially, that is, served as referential anchors (to use Koptjevskaja Tamm's terms) of their heads. Prepostion was typical of non-referential (non-anchoring) genitives that played the role of mere qualifier. Grammaticalisation of the genitival position was largely due to the interaction between various types of adnominals. Firstly, Lithuanian genitive modifiers are widely used non-referentially and can express various meanings that are expressed in other languages by denominal adjectives. Thus, in the process of grammaticalisation of word order non-referential use of the genitive may have been viewed as the core of this construction and served as the basis for grammaticalisation. Secondly, the functions of relational adjectives and genitives overlap in Lithuanian even more than in other languages. Translational practice in the Old Lithuanian period shows that Lithuanian genitives were easily identified with the adjectives of other languages. It is argued that the preposition of agreeing modifiers played an analogical role in the levelling of the position of other adnominal dependents with respect to the noun and that the uniform pre-head position of nominal modifiers in Lithuanian was achieved by harmony-by-extension development in terms of Harris and Campbell (1995:212). An approach assuming a functional convergence of non-referential genitives with adjectival modifiers in Lithuanian captures the ability of the Lithuanian (non-referential) genitives to be coordinated with adjectives, which ability is often viewed as a manifestation of belonging to the same syntactic category. It is argued that judging from their combinatory abilities, three major classes of Lithuanian adnominals must be distinguished: (1) anchoring genitives, (2) non-anchoring genitives and relational adjectives, (3) qualitative adjectives. A comparison of some more fine-grained word order facts from the two Baltic languages corroborates the above hypothesis that in the history of Lithuanian (but not Latvian!) the ordering discrepancies between anchoring and non-anchoring genitives have been analogically levelled and that the pattern typical of descriptive (non-referential) genitives extended onto other genitives.

[ Top of the page ]

Complexity of Right-branching Structures

Sinnemäki, Kaius

Mainstream syntactic research has concentrated on the simple clause and a moderately complex clause. A reason for this, I believe, is that recursion (or downward iteration) in natural language is theoretically limitless, thus making it seem unrevealing to investigate further. However, earlier corpus studies have established that there are limits to left-/right-branching and centre-embedding. Therefore, recursion is limited. The present study tries to find in what way complex right-branching clauses behave in different genres, and why they behave in that way.

First, a simple method is put forth in order to compare the complexity of different genres. The number of words and clauses in a sample are compared to the number of sentences that have right-branching clauses at or below the depth -3 (where the depth of the matrix clause is 0). Secondly, the behavior of different clause types (complement, relative, adverbial) is studied in regard to depth. The corpus consists of several million words of running text from Finnish newspapers, magazines and novels.

The three genres form a continuum of complexity running from newspapers to novels via magazines. More interestingly, the language in a literary style called stream-of-consciousness proves to be strikingly more complex than the other genres considered. In addition, complements tend not to "thrive" at deep levels, whereas other clause types do.

Two factors are suggested to explicate the question why right-branching clauses at or below depth-3 are rare and even avoided in some genres, but exploited as a technique in stream-of-consciousness. One factor is that in order to understand a sentence as a whole, the clauses need to be related to one another. However, in right-branching structures clauses are directly related only to their governing clause; their relation to the main clause is only indirect. The second is that late attachment and coordination of subordinate clauses argue for holding in memory the depths of clauses in right-branching structures. These factors together make it difficult to relate a clause at a low depth to the main clause, and push memory capacity to its limits. To account for the behaviour of different clause types, some clues are sought from their general functions in texts.

[ Top of the page ]

K-structure - an Interlingua

Søgaard, Anders

Standard LFG architecture posits at least two levels of syntactic information, which licenses transfer-based machine translation. But in order to deal with more complex translation puzzles, due to sense extension and "partial" translation equivalents, one needs some level for semantic information. For most complex puzzles, a very rich metaphysics is necessary.

One way of representing ontological information is by k-structures. Puzzles that are defined and solved by reference to k-structures include translation of prepositions with certain sense extensions in one language but not in another, and ambiguous compounds to be translated by phrasal constructions. Consider (1)-(4) for translation of prepositions, and (5)-(8) for translation of compounds:

(1) the baby is by the blackboard
(2) babyen er ved tavlen
(3) the student is at the blackboard
(4) eleven er ved tavlen

(5) costello di plastica
(6) plastic knife
(7) costello da plastica
(8) plastic knife

Such translation puzzles are solved by k-structure unification. In (1), the formal quale of blackboard is modified, whereas in (3), the telic quale is modified. Similarly, the ambiguity of (6) and (8) arise from modifying different quales of knife. Even more complex puzzles, involving eventuality and taxonomic information are discussed.

(9) bút-lông ['pen'+'hair']
(10) brush
(11) tô-quô´c ['father'+'land']
(12) ancestor country

Examples (9)-(12) are probably as complex as non-stylistic translation problems come. Less complex problems, also to be discussed, include the translation of redundant compounds, such as palm tree (in Danish, palme), compared to non-redundant compounds, such as apple tree (in Danish, æbletræ).

Evidence that k-structure is a relevant tool for linguistic theory comes from acceptability and entailment phenomena. A k-structure is a well-defined feature structure comprising information about eventuality, qualia and other relevant semantic features. Our theory of k-structures does not conflict with standard LFG architecture. A projection from c-structure into k-structure is defined, similar to that from c-structure into f-structure. But for machine translation purposes, it might make us prefer an interlingual approach to more traditional, transfer-based approaches. I.e., we might let k-structures carry the entire load and just occasionally refer to f-structures by inverse projections.

[ Top of the page ]

Old English: Language with Active Alignment

Toyota, Junichi

English of any periods is unambiguously considered a language with nominative-accusative alignment. However, such treatment of grammar causes problems in diachronic analysis, since language can change its alignment during the course of change. English is a part of Indo-European (IE) languages. IE languages all descended from a common ancestral language, Proto-Indo-European (PIE), which is known to possess active alignment. Thus, the change in alignment in IE languages is from active to nominative-accusative. However, such change is often not so clear-cut, and some earlier grammatical characteristics often remain unchanged. In this paper, we analyse one such case, i.e. the word order change, in English from IE perspectives and explain what was earlier word order and why it changed into SVO.

The word order in earlier English is described in various ways, e.g. V-final, SOV, OV, etc. This unclear distinction is, in our view, due to the characteristics of topic-prominent language (Li and Thompson 1976), i.e. the earlier word order was based on the discourse topicality/focus, and the basic order was not so clearly detectable. This was also the case of ancestral IE languages. However, it is possible to consider that the earlier basic word order in English is SOV, due to the order in the subordinate clause, which is one of the environments known to preserve the earlier word order better (Givón 1979). In our view, this word order is a residue of PIE, in which the basic order is also considered SOV. The only difference is that English, unlike ancestral languages like Hittite (cf. Luraghi 1990), does not use any particles indicating topicality relationship, in a sense of obviative marking. From roughly around late Middle English, SVO as basic order emerged. This indicates that the subject became more sensitive to the discourse-topicality, i.e. the language changed from topic-prominence to subject-prominent language. Around the time of the word order change, other residues of earlier active characteristics disappeared from English, although they still remain in some other IE languages. For example, impersonal verbs are still functional in a number of IE languages, but not in English. Thus, once the earlier active alignment in ancestral languages is considered, we can provide another insight in the historical change of English.

[ Top of the page ]

Explaining the Great Gender Shift

Trosterud, Trond

According to Steinmetz (2001), the Germanic languages have undergone a "Great Germanic Shift", where a system with default n and a hierarchy n>m>f has been replaced by m and m>f>n. This is due to loss of the m marker -R in nomsg. Icelandic still has this suffix, and hence also n as default gender.

Modern Norwegian has m as default gender, and Old Norse (ON) has n; Norwegian thus behaves as expected, given Steinmetz' theory. But although there is a correspondence between {-r/no -r} and default {n/m}, many facts related to the attested gender shifts in the history of Norwegian do not fit Steinmetz' one-factor explanation.

During the transition from Proto Norse (PN) to ON there are several shifts from n to f. In ON, most nouns are m (39% m, 32% f, 28% n), not n. Among ON loan words, the share of n is even smaller (50% m, 27% f, 23% n). The position of n was weak already before the -r loss. An analysis of the gender shift should account for this.

In PN, declension (stem) class membership was detectable from the form, and gender was detectable from stem class membership. During the syncopy period, the stem vowel disappeared, and declension class membership was detectable from oblique forms only. In the new system, gender had to be determined by other means. The semantic basis for the Indo-European three-gender system is the link between biological sex and grammatical gender. Neuter thus falls outside the semantic base of the system. As witnessed by the loan word data, applying the main m-assignment rules (nouns denoting males, persons, living beings, concrete entities, tops, long objects, ships, time, ... are m) to the set of loan words simply gave twice as many m as n nouns.

The paper argues for a distinction between two types of default gender, with default n as a weak default gender, rooted outside the semantic base of the gender system, and m as a strong default gender, rooted in the semantic base of the system. As long as n had support in an overt declansion class system, it was upheld, but the change to covert declension class systems paved the way for large-scale gender shifts.

Steinmetz, D. 2001: The Great Gender Shift and the attrition of neuter nouns in West Germanic: The example of German. Rauch & Carr (eds.) New insights in Germanic Linguistics II:201-224. New York: Peter Lang.

[ Top of the page ]

The Relation between Mind and Language: The Innateness Hypothesis and the Poverty of the Stimulus

Vallauri, Edoardo Lombardi

The existence of an innate Universal Grammar (UG) is one of the most important assumptions of what may be called "Chomskyan" linguistics, and has relevant consequences on the consideration of the very nature of thought and man himself. The Poverty-of-the-Stymulus Argument (PSA) is always cited, starting from Chomsky's earliest works, as relevant evidence in favour of the innateness hypothesis: language acquisition is regarded as impossible in the absence of some specifically linguistic knowledge, that must be built-in to the mind/brain in order to allow children to make the most of environmental stimulus. Usual criticisms to the PSA mainly come from extralinguistic fields, such as neurology, paleontology, ethology, psycholinguistics, studies on language acquisition. This contribution will try to show some weak points of the PSA on a specifically linguistic level, both from a quantitative and a qualitative point of view. From a quantitative point of view, the following (among others) will be argued:

- that children have much more time and energy at their disposal for acquiring their language than is usually recognized;
- that the grammar proposed as a possible UG by innatist scholars, namely Generative Grammar (mainly in its Principles and Parameters version) often implies interpretations of linguistic structures that are more complex than those actually required for the children to acquire the same structures;
- that the very problem for the children is not the acquisition of syntax, but the acquisition of the lexicon, which is much more cumbersome.

From a qualitative point of view, the following will be argued and exemplified:

- that the innatist tradition highly underestimates analogy as a resource for the child to build his own grammar from the incomplete stimulus he receives from the environment;
- that knowledge and consciousness of reality surrounding the speech acts are also underestimated, but actually play a major role in allowing the child to build his internal grammar;
- that the role of "negative information", conceived as the fact that some structure simply (but systematically) does not occur in the stimulus, is also underestimated. It will be also shown that the high degree of convergence of all known grammars does not need to be explained by means of one grammar in the brain, but simply results from a series of constraints that are pragmatic in nature, or directly derive from the definition of a system designed for communication.

[ Top of the page ]

A Contrastive Pragmatic Analysis of two Concessive Markers - Norwegian riktignok and Hungarian jóllehet

Vaskó, Ildikó & Thorstein Fretheim

The Hungarian concessive marker jóllehet (lit: well-maybe) introduces either an embedded or a matrix clause, like Swedish fast, exemplified by (1) and (2), respectively, but in (3) it is an adverb.

(1) Jóllehet nyelvészetet tanultam, nem beszélek sok nyelvet.
"Though I studied linguistics, I don't speak many languages."

(2) Nem beszélek sok nyelvet. Jóllehet nyelvészetet tanultam.
"I don't speak many languages. Though I studied linguistics."

(3) Jóllehet nyelvészetet tanultam, de nem beszélek sok nyelvet.
"Admittedly, I studied linguistics, but I don't speak many languages."

Jóllehet and Norwegian riktignok (lit: true-enough) are oftentimes translational correspondents, but riktignok is much less grammatically versatile, it is always an adverb. (4) equals (3).

(4) Riktignok studerte jeg lingvistikk, men jeg snakker ikke mange språk.

The appearance of riktignok in (5b) without a following men-conjunct makes that utterance sound unfinished, though in an acceptable way. With Hungarian (6) there is no analogous feeling of incompleteness.

(5) Jon flytta fra Byåsen til Lade.
a. Husleia var lavere.
b. Husleia var riktignok lavere...
"John moved from Byåsen to Lade."
a. "The rent was lower." (= at Lade)
b. "The rent was admittedly lower..." (= at Byåsen)

(6)János Egerbôl Budapestre költözött. Jóllehet kevesebb volt lakbér.
"John moved from Eger to Budapest. The rent was lower, though." (= in Eger)

While the follow-up utterance with jóllehet in (6) is consistent with the speaker's failure to understand why János moved, the follow-up utterance (5b) with riktignok implicates that the speaker knows why Jon moved, and expects the hearer to be able to work out the reason if no men-conjunct is forthcoming, in which case the hearer is instructed to construct a 2nd conjunct mentally, based on accessible pragmatic implications, e.g. the assumption in (5b) that Jon moved from Byåsen to Lade for a reason. Unlike other "but" constructions, "Riktignok P, men Q" can be shown to give P and Q the same argumentative orientation (Anscombre & Ducrot 1977) if that interpretation is supported contextually; "Jóllehet P, de Q" shows the same tendency, albeit not so strongly. Lexical definitions of riktignok and jóllehet in terms of relevance-theoretic procedural semantics will be offered.

[ Top of the page ]

Unmarked Tone Accent in Scandinavian

Wetterlin, Allison & Aditi Lahiri

Recent synchronic analyses of Norwegian and Swedish tone accents suggest that affix morphemes can be lexically specified for Accent 2, Accent 1 or are accent neutral (Withgott & Halvorsen 1984, Riad 1998, Kristoffersen 2000). However, these analyses cannot account for many specific affixes, e.g. the infinitive suffix -a in Swedish which, as Riad suggests, induces Accent 2, when considering the prefixed Accent 1 be- verbs such as "beskriva". These analyses also have difficulty accounting for compounding in every dialect other than central Swedish.

This study offers a new analysis of tonal accents in Norwegian and Swedish. By looking at prefixes, compounds and present tense in Norwegian and Swedish, we arrive at a new analysis of tone accent, proposing only one lexically specified Accent, i.e. Accent 1. The benefits of our analysis will be illustrated by clarifying many of the exceptions that could not be gracefully provided for in previous theories

Kristoffersen, Gjert. 2000. The Phonology of Norwegian. Edited by J. Durand, The Phonology of the World's Languages. Oxford: Oxford University Press.
Riad, Tomas 1998. "The origin of Scandinavian Tone Accents". Diachronica 15. 63-98
Withgott, Meg and Per-Kristian Halvorsen. 1984. Morphological Constraints on Scandinavian Tone Accent. Center for the Study of Language and Information, Stanford University. Report. No. CSLI-84-11.

[ Top of the page ]

Typology of Repetitive and Negation - Similarities and Differences

Wälchli, Bernhard

In his Lezgian grammar, Martin Haspelmath observes a strong parallelism between sentence negation and repetitive (the expression for 'again'). In Lezgian both negation and repetitive are expressed either by a prefix (for a closed class of verbs) or periphrastically by means of the negative or repetitive form of the auxiliary verb awun 'to do' (negative tawun 'not do', repetitive qhuwun 'do again'). This paper investigates to which extent the typologies of negation and repetitives are similar in a more general cross- linguistic perspective. Notably the following similarities are striking:

· In both typologies there is, among others, a verbal type and an auxiliary type.
· Both repetitive and sentence negation can be expressed analytically (by particles, verbs or auxiliaries) and morphologically (by affixes).
· In both typologies the lexical verb sometimes has a different form than in the affirmative/non-repetitive (Consider Spanish Volvió a enviar 'sent again' vs. envió 'sent').
· In both typologies one meets astonishingly many cases of obligatory or optional double marking (such as French ne...pas in negation).
· For both negation and repetitive, scope is important. Both markers can have different scope (at least if they are expressed by particles).
· Both repetitive and sentence negation are universal in the sense that every language has an explicit marker or construction for their expression. · Both repetitive and negation are source for various paths of grammaticalization. The regressive may, for instance, become a coordinating conjunction or a prohibitive. Negation markers may become markers of comparison, subordinate conjunctions or interrogative particles.

Besides of many similarities there are, of course, also many differences, among other things the following:

· Sentence negation has a higher text frequency, which makes that it often has a more grammatical character than the repetitive.
· The repetitive is a presupposition trigger, the negation is not.
· Negation and repetitive have different functions in narrative discourse. The negative, having collateral character, is often backgrounding whereas the repetitive often marks central, foregrounded passages in a text.

The most obvious difference is, however, that the typology of sentence negation has been amply studied whereas little is known about the typology of the regressive. The paper focuses therefore on the typology of the regressive which is investigated in a sample of 100 languages from all continents.

[ Top of the page ]

Mapping Chinese Morphemes Across Millenia (I): A Diachronic Study of di and its Cognates

Yap, Foong Ha, Pik-Ling Choi, & Kam-Siu Cheung

This paper traces the development of ©Â³ di 'bottom' in ancient Chinese. Previous studies offer competing accounts on the origin of this morpheme. Lu (1943) linked the origin of ©Â³ di to ªÂÌ zhe on functional grounds; Wang (1980) argued that ©Â³ di is derived from ¤Â§ zhi on phonological grounds; Jiang (1999) claims that the grammatical uses of ©Â³ di emerged from its lexical uses essentially via the locative sense; Shi and Li (2002) propose that the emergence of grammatical uses of ©Â³ di was triggered by analogy with the emergence of the classifier system in Chinese. Using computerized databases on Chinese ancient texts (e.g. CHANT) and recently excavated bamboo and silk writings from the Warring Period (5th century BC), we focus instead on tracing the polysemy network not only of ©Â³ di but also its cognate morphemes, as well as non-cognate morphemes that are homonymous with it.

Our findings reveal an older morpheme ¥Â di 'root, origin' that over time was recruited to serve multiple functions. Our diachronic analyses reveal the following developments:

(i) each individual morpheme, whether cognate or non-cognate, evolved semantic subclusters of their own, often with increasingly abstract senses over time;

(ii) overlaps between the semantic subclusters tended to be relatively infrequent and generally transient;

(iii) overlaps between the source morpheme and its cognates tended to be restricted to the more abstract meanings;

(iv) the morpheme ©Â³ di 'bottom' that eventually replaced the source morpheme likewise tended to restrict overlaps with cognates to the more abstract meanings, and it went on to acquire more abstract senses;

(v) overlaps between the source morpheme and non-cognate homonyms such as ©Âè di 'reach' paved the way for ©Â³ di 'bottom' to acquire an even wider range of abstract meanings;

(vi) this now highly versatile morpheme ©Â³ di 'bottom' began to serve grammatical functions (e.g. pronominal 'this', interrogative 'what, how') as attested in texts from the Tang period;

(vii)overlaps with the source morpheme ¥Â 'origin, root' and with its more abstract and grammatical cognate ©Â³ 'bottom' also made it possible for the non-cognate morpheme ? 'reach' to grammaticalize into an interrogative morpheme, although this development has not survived into modern times.

The above observations highlight a process of 'convergences among abstract senses', and shows how cognate and non-cognate morphemes, sometimes seeping across semantic subclusters, often retreat after a brief foray, leaving the more abstract contender(s) to take flight along the grammaticalization trajectory.

(400 words)

Jiang, Lansheng. 1999. Chusuoci de lingge yongfa yu jiegou zhuci ¡Â§di¡Â¨ de youlai. Yuyanwenzixue 6: 91-101. Lu, Shuxiang. 1943. Lun ¡Â§de¡Â¨, ¡Â§de¡Â¨ zhi bian ji ¡Â§de¡Â¨ zi de youlai. Jingling Qilu Hanxi Zhongguo Wenhua Huikan 3.
Shi Yuzhi & Li, Charles N. 2002. The establishment of the classifier system and the grammaticalization of the morphosyntactic particle de in Chinese. Language Sciences 24:1-15. Wang Li. 1980. Hanyu Shigao. Beijing: Zhonghua Shuzu.

[ Top of the page ]

Disambiguation of Syntactic Functions in Norwegian: Modelling Variation in Word Order Conditioned by Animacy and Definiteness

Øvrelid, Lilja

Norwegian has a fairly rigid word order, and marks syntactic functions in large part through structural positioning, as opposed to languages with a freer word order, where morphological case is a crucial disambiguator in this respect. In Norwegian, pretty much all constituents may topicalize, however, when the object in a simple transitive construction topicalizes, disambiguation of the syntactic functions of subject and object is at risk. In linear surface structure there are no structural indicators as to the syntactic functions of the two arguments in such a construction. The work described below has originated from a problem of automatic disambiguation of syntactic functions in Norwegian, and a need to discover facets of word order and argument realization apart from solely structural criteria.

Much recent work (Aissen to appear, Lee, 2002a) in Optimality Theoretic syntax, has focussed on the influence of hierarchies of markedness on phenomena like case marking, word order and argument realization. In particular, languages exhibit a strong tendency for subjects to be animate and definite, and objects to be inanimate and indefinite. What happens then, when arguments do not adhere to these generalizations? So-called 'freezing effects' have been observed in diverse languages (Bloom 1999, Lee 2002b, Morimoto), when arguments violate the above generalization. These are characterised by a 'freezing' to canonical word order when arguments are either maximally marked, or too alike.

In this talk we will take a closer look at the effect that animacy, in particular, has on word order in Norwegian, thus, also on disambiguation of syntactic functions. Norwegian does not have any categorical freezing effects and marks for morphological case only to a very limited extent. Might, however, similar freezing effects be present in Norwegian as a strong statistical tendency? To find an answer to this, an analysis of 1000 transitive sentences was performed. We found that Norwegian strongly adheres to a principle that the subject should not be lower in animacy than the object. Also, when the subject-object configurations were maximally marked, a clear freezing tendency could be observed. In sentences where the arguments were of equal animacy, there was a strong preference for canonical word order, as well as a clear asymmetry in definiteness in cases with non-canonical order.

Further results will be reported in the talk, as well as a demonstration of how the observed variation in word order, as conditioned by constraints pertaining to properties of animacy and definiteness, may be modelled using stochastic Optimality Theory (Boersma and Hayes, 2001).

Judith Aissen. Differential Object Marking: Iconicity vs. Economy. To appear in Natural Language & Linguistic Theory.
Douglas B. Bloom. 1999. Case Syncretism and Word Order Freezing in the Russian language. Master's Thesis, Stanford University
Paul Boersma and Bruce Hayes. 2001. Empirical Tests of the Gradual Learning Algorithm. Linguistic Inquiry 32, 45-86.
Hanjung Lee. 2002a. Prominence Mismatch and Markedness Reduction in Word Order. Natural Language and Linguistic Theory.
Hanjung Lee. 2002b. Crosslinguistic Variation in Arguments Expression and Intralinguistic Freezing Effects. In Tania Ionin and Heejeong Ko and Andrew Nevins (eds.), MIT Working Papers in Linguistics 43, 103-122.

[ Top of the page ]


Muokattu viimeksi: Tuesday, 16-Dec-2003 21:12:28 EET