Ordered alphabetically according to first author's surname.
For workshop and plenary paper abstracts, see separate pages.
Allan 1995 argues for a division of labour between the lexicon and the encyclopedia of which it forms a part: the lexicon contains formal, morphosyntactic, and semantic specifications of listemes and the encyclopedia contains other kinds of information about listemes, e.g. their etymology, and information about the denotata of listemes.
The present paper is about the conversational implicatures (hereafter referred to simply as 'implicatures') that result from the two maxims of quantity identified by Grice 1975 and subsequently discussed by (amongst others) Atlas and Levinson 1981, Horn 1984, and Levinson 1995. The question I seek to resolve is whether Q[uantity] implicatures should be entered in the lexicon, or whether they constitute encyclopedic information.
Conversational implicatures are pragmatic (Grice 1975, Gazdar 1979, Levinson 1983): they arise from the use of language in particular contexts. They differ from entailments in being defeasible. We might suppose some of them to be at the interface between the lexicon and the encyclopedia, but this paper argues that quantity implicatures have a place in the lexicon.
In other work (Allan 1999), Q1 implicatures, deriving from the first maxim of quantity, are included in lexical entries. Consider the NP three birds. It is part of the lexical knowledge of the language user that the lexical meaning of three includes both the indefeasible (logical) meaning "greater than or equal to three (at least three)" and the defeasible Q1 implicature identifying the default meaning "equal to three (exactly three)". For instance, if 1 is true then 2 is necessarily true.
(1) Five birds were shot.
(2) Three birds were shot.
I.e. three birds necessarily means "at least three birds". However, to utter 2 when 1 is the case would be misleading since three birds is standardly understood to implicate "exactly 3". The implicature is canceled, or at least suspended, in (3).
(3) Three birds were shot, if not five.
I shall argue that what Jackendoff 1983, 1985, 1990 refers to as 'preference conditions' on lexical items are implicatures deriving from the second maxim of quantity augmented with the Atlas and Levinson principle of informativeness, a combination here referred to as Q2 -- a quantity 2 implicature. Jackendoff incorporates preference conditions within his lexical entries. For instance, the lexical meaning of bird includes both an indefeasible part identifying the class of creatures, and a defeasible part "capable of flight". Surveying as many examples of quantity implicature as time permits, I find that all Q implicatures based on a single lexical item are noted in the lexicon entry. Nonlexical implicatures arise from collocations of lexical items and can perhaps be located within the encyclopedia of which the lexicon is a part.
It is known that in some languages a reduplicated adjective 'distributes' over its head noun. Consider the following examples from Georgian (Gil 1988:1043):
|'heavy suitcases' [lit. 'heavy heavy suitcases']|
According to Gil (1988:1043) in (1a) it is possible that the suitcases are collectively heavy though they may be light individually. On the other hand, in (1b) only one interpretation is possible: every suitcase is individually heavy. This phenomenon is not restricted to adjectives: reduplicated numerals also force a distributive interpretation. Consider the English sentence 'Two men carried three suitcases' (Gil 1995:324). The ‘two men’ may have acted individually (distributive) or collectively (non-distributive). The distributive interpretation can be forced by adding ‘each’ as in 'Two men carried three suitcases each'. Gil (1988) shows that in Georgian a reduplicated quantifier forces a distributive reading (Gil 1988:1044):
|'two men carried three suitcases [each]'|
[lit. 'two men carried three three suitcases']
In his analysis of this phenomenon, Gil (1988), building on the typological classification proposed in Gil (1987), argues that Georgian belongs to a class of languages ('Type B') which can be characterised by the following cluster of properties:
|(3)||a.||no obligatory marking of (in)definiteness|
|b.||existence of adnominal distributive numerals|
In this paper I present data from Amharic which questions Gil's proposed classification. First, I show that Amharic has adnominal distributive numerals while requiring definite NPs to be marked obligatorily. Thus, clearly (3a) and (3b) do not cluster in Amharic. Furthermore, I show that there is no dependency between the number feature of the head noun and the distributive reading of the reduplicated adjective. Gil (1988) notes that in constructions such as (1b), the distributive interpretation is blocked when the head noun is singular. In the case of Amharic, the distributive reading is available irrespective of the number marking of the head noun, as can be seen in (4):
|'heavy suitcase(s)' [lit. 'heavy heavy suitcase(s)']|
This is not surprising given the independent fact that the plural morpheme is optional if plurality is marked on the specifier. I argue that the distributive meaning induced by reduplicated adjectives and quantifiers in Amharic must be examined in the context of the lexical semantics of the predicate. This shows that reduplicated adjectives and quantifiers are not semantically homogeneous.
There is a growing body of cross-linguistic studies which shows that languages vary tremendously in the way they organise and package information about spatial topological relations (see Levinson 1996 and references therein). Most of the literature concentrates, however, on the expression of such information in adpositions neglecting the role of verbs. In this paper, I compare the grammatical constructions used in describing the localisation of entities in two Kwa (Niger-Congo) languages in contact -- Ewe and Likpe (Ghana), paying particular attention to the verbs used in such structures. I will demonstrate that in both languages there is a division of labour in packaging spatial information between the elements in a locative construction. Both languages are also similar in using spatial region terms, designated postpositions, to code the search domain information. In spite of these similarities and the influence of Ewe on Likpe, I will argue that there are striking differences between the two languages with respect to the use of verbs in locative constructions. Spatial relational information is packaged in a single locative verb in Ewe. In Likpe, on the other hand, the coding of spatial relational information is shared between a set of a dozen or so contrasting spatial verbs and a general locative preposition. The consequences of this difference between the two languages will be explored. The extensions of the verbs in both languages for the expression of other meanings such as existence, possession and caused location will be discussed in the context of grammaticalization, the localist hypothesis and the typology of systems of spatial description. The study underscores the fact that even two languages with similar geo-cultural background and superficially similar formal strategies for spatial description may show differences in the semantic distribution of spatial information.
A number of points in the analysis of Algonquian languages (particularly Potawatomi) have played important roles in the development of the theory of "A-Morphous Morphology." These issues include (a) the nature of disjunctive or `blocking' relations in morphological systems; (b) the importance of many-to-many relations between formal elements and the content they signal: (c) the nature and internal structure of morphosyntactic representations; (d) the possibility of restructuring or `inversion' operations internal to morphology; and (e) the relation between affixes and clitics.
Earlier treatments of these and other points growing out of the discussion of Algonquian have been the subject of comment and criticism from a variety of perspectives, including those of Articulated Morphology (Steele, 1995), Distributed Morphology (Halle & Marantz, 1993; McGinnis, 1999), and more directly syntactic accounts such as that of Dechaine, 1999. The present paper will evaluate and respond to these alternatives, and propose clarifications and modifications where warranted in the descriptive and theoretical accounts offered of Algonquian within the overall perspective of A-Morphous Morphology.
In the descriptive and typological linguistic literature, Australian Aboriginal languages, especially those of the Pama-Nyungan family, have become notorious for their complex systems of case marking, showing such characteristics as split-ergative morphological coding (Silverstein 1976, Dixon 1979, 1994, Blake 1977 etc.), interactions of marking with animacy (Aristar 1997), multiple levels of case marking ("case stacking") (Dench and Evans 1988, Simpson 1991, Planck 1996, Andrews 1996, Nordlinger 1998), and case affixation on verbs (Dench and Evans 1988, Blake 1996).
The focus of this paper is argument encoding in linked, typically non-finite, clauses in a range of languages to explore the ways in which predicate-argument grammatical functions are encoded in such constructions cross-linguistically within Australia. I will draw on data from a range of languages, primarily from the Pama-Nyungan family, but also including some non-Pama-Nyungan materials. The aim is to develop a descriptively adequate typology of argument encoding in linked clauses which can serve as a basis for further theoretical and historical research (but see Nordlinger 1998, Dench 1994, Evans 1995).
Clause combining in Australian languages is typically achieved by hypotactic linking of dependent clauses, which commonly occur structurally on the margins of the matrix clause and code in their (non-finite) verb morphology both clause type and cross-clausal reference relations (with many systems encoding switch-reference, ie. sameness or difference of subjects between the linked clauses — see Austin 1981). In many languages the types of non-finite construction of interest here are relatively rare textually, and when they do occur they generally consist of just a non-finite dependent verb without any arguments. Co-occurrence of such verbs with one argument, let alone two, is even rarer although it is clear that the structural principles outlined here are well motivated in the languages concerned, and can be confirmed through elicitation.
We will show that there are seven strategies for argument encoding that Australian languages adopt in linked clause constructions. In a number of languages strategies co-exist and it is possible to array clauses in a formal hierarchy of clause linkage according to such parameters as the argument coding strategy, encoding of cross-clausal reference, case agreement on the dependent verb, and expression of tense/aspect/mood categories. We will explore such formal hierarchies and their relation to proposed functional hierarchies of clause linkage (see Silverstein 1976, 1980, 1993, Van Valin 1993, O’Dowd 1992), concluding that much empirical research needs to be undertaken before it is possible to show that the semantics and pragmatics of clause linkage in any way predicts the distribution of the formal characteristics detailed here.
This paper examines the development of infinitival complementation in the variety of Pennsylvania German spoken in Waterloo County, Canada. In particular, it focuses on the rise of the so-called fer-zu 'for-to' construction. This construction is totally absent in Standard German, although a 'for-to' purposive clause does occur in certain regional dialects (e.g. Pfaelzisch). What is striking about the Pennsylvania German construction, however, is the rapid disappearance of the marker zu and the expansion of single fer. For example, we now find fer clauses as the complement of adjectives, nouns, verbs and as postposed subjects. The construction can also occur with or without a following overt subject.
|He is too lazy to work|
|She has a pattern to make a dress|
|I promised to help him|
|It is unusual for John to read hard books|
Fer is clearly taking over the functions of zu, but has expanded to include even the complements of verbs previously taking a bare infinitive.
|Would you like to have a cup of coffee?|
In this paper we address a number of questions concerning the rise of fer and its current status. These questions relate especially to the rapid expansion of the construction, the accompanying loss of zu and the consequences this has had for the role of fer within the clause structure. In particular, we argue that fer is moving away from its position as complementiser and is becoming more closely associated with the verbal complex.
There has been a significant loss of indigenous languages through 200 years of European settlement in Australia. Linguistic research has dramatically increased over the past 30 years and has led to the publication of dictionaries in several languages. Dictionaries are a tool that can assist in the maintenance and preservation of threatened languages.
The paper will report on consultation discussions with the Kunwinjku people and plans for a Kunwinjku dictionary. The author is one of several researchers working on a dictionary of the Kunwinjku (Gunwinggu) language of western Arnhem Land in the Northern Territory of Australia.
This paper will discuss the presentation format of several published dictionaries in Australian languages. Most dictionaries have a citation form in the vernacular with a meaning in English, suggesting that the intended readers are non-native speakers. The Kunwinjku people are looking for a different approach. They wish meanings of Kunwinjku words to be given in both Kunwinjku and English. The paper will outline plans to develop a 'bilingualised' dictionary for the Kunwinjku language. Following a Kunwinjku citation form, it is proposed meanings will be given in both Kunwinjku and English.
This paper will seek to demonstrate two things. First that the categories of classical rhetoric (Greek then Roman) (Sonnino 1968), can be used to demonstrate artistic strategies in the Warlpiri narrative Jajirdikirli (in Cataldi and Rockman 1993). In doing this the differences beween Jajirdikirli and other Warlpiri narratives will, where needed, be addressed. Many examples are given.
Second if the artisitic strategies in Jajirdikirli can legitimately be described using these categories (that is, if what constitutes for example the grand style in the Graeco-Roman tradition can be usefully translated cross-culturally and cross-linguistically), then the narrative Jajirdikirli deserves consideration as a type of oral epic, a Warlpiri composition whose style we might be able to describe as grand or high. This section of the paper discusses the ways in which, if there is Warlpiri high style, it can be recognised.
The paper will conclude that, if these are so, the status European culture affords to its classical epics (the Odyssey is the chosen example) can also be given to some Warlpiri (and other indigenous Australian) narratives.
A further suggestion of the paper is that the categories of classical rhetoric can provide a useful way of describing and evaluating other oral narratives in the indigenous Australian corpus.
This paper forms part of a broader project aiming at investigating the intonational phonology of Lebanese Arabic. The framework adopted in this research is a line of work on intonational phonology triggered by Pierrehumbert (1980). The theory recognizes four major components for describing intonational patterns: tune, relative prominence, phrasing and pitch range. It also describes processes of tune-text association. The present paper concentrates on the tonal component of intonation description in Lebanese Arabic, including tune-text association. Two major forms of data are used: 1. Laboratory experiments, controlling for segmental effects and varying lexical stress and nuclear accent location. 2. A map-task corpus complementing laboratory speech with more naturally-occurring conversational data.
A ToBI-style transcription of the data is carried out according to which seven pitch accent types and two levels of intonationally marked constituents are identified (the intermediate and intonational phrase). The present tonal inventory makes a few alterations to that obtained by a previous study conducted by the author (Chahal, to appear). It also motivates the intermediate phrase level both phonologically and phonetically. The basic tunes of Lebanese Arabic are here described according to the permissible combination of various pitch accents and edge tones. A downstepping tune is recognised in addition to the declarative, question, continuation and plateau contours observed in Chahal (to appear). Tune-text association in Lebanese Arabic is reached through the association of of edge tones to their respective phrases and the association of pitch accents to lexically stressed syllables. In unmarked or default cases, nuclear pitch accents associate to the rightmost content word of the phrase. Default phrasal stress is therefore right-headed in the language. Finally, tune-text association in Lebanese Arabic is given a metrical representation involving the levels of prosodic structure relevant to its intonational structure.
It has long been known that there is a regular correlation, at all levels of the grammar, between expressive meaning and 'marked' structure (in a Praguian sense, cf Mukarovsky 1964 ). The following examples illustrate:
i) Expressive syntax. In English, all predicates must contain a verb phrase - except in certain 'exclamatory' constructions:
|c.||A semantic account, my eye!|
ii) Expressive morphology: In English, expletives occur in a range of structures not available to other classes. For example, they alone can 'irrupt' into words and lexicalized expressions, irrespective of the category of the base (McMillan 1980):
iii) Expressive phonology: Expressive lexis is often phonologically aberrant. Exclamatory particles, for example, often contain segments found nowhere else in the language's phonology, as in these examples from English:
|(3)||Shh(t)! Psst!||syllabic obstruents|
|Phew!||bilabial fricative, voiceless vowel|
Phenomena such as these present a challenge for Optimality Theory (Prince & Smolensky 1993), which views the grammar as a device for generating the least-marked of a set of candidate output structures.
Based on the cross-linguistic recurrence of data like the above, this paper proposes that two (prelinguistic) principles are involved in the generation of ALL outputs (cf Fowler 1986):
|(4)||a.||Perceptual Salience: 'More-marked' candidate structures are perceptually more salient than 'less marked' ones available in the same context|
|b.||Potential for Interpretation: The occurrence of a 'more-marked' structure, where a 'less marked' one is available in the same context, invites pragmatic interpretation.|
It is further proposed that the following linguistic principle interacts with 4a&b in the selection of all outputs:
|(5)||Functional Evaluation: Given a set of candidate output structures, select the candidate whose Perceptual Salience (4a) and Potential for Interpretation (4b) best complement the semantic/functional context.|
Standard OT would no doubt explain phenomena those in 1 -3 above by semantically-driven reranking of constraints, so that an otherwise more-marked candidate structure becomes least-marked, thus preserving "selection of the unmarked". A more 'radical' - or just more traditional (Praguian) - view is that both less-marked and more-marked structures are available to the grammar, with semantic criteria determining which is selected. The merits of both approaches will be compared. In either case, semantics, rather than purely formal concerns, drives output selection.
This paper presents a syntactic and semantic account of degree adjuncts in English using data derived from several standard corpora (which facilitated the exploration of intervarietal comparisons, and of collocational patterns with particular adjuncts, e.g. “deeply hate/admire/value/ *like/*prefer”). Several elicitation tests were also used to supplement the corpus data (for instance in testing collocational restrictions, and the intensity levels of adjuncts as reflected in their patterns of modification).
While degree covers a range of cases such as 'accomplishment' (e.g. “to completely conquer”) and 'extent' (e.g. “to smoke a lot”) - often overlapping with such other semantic classes as manner and frequency - in prototypical cases degree adjuncts express gradable properties that can be plotted along a scale of intensity. The binary division of this scale posited in Quirk et al.'s (1985) “Comprehensive Grammar” into 'amplifiers', which are said to 'scale upwards' from an 'assumed norm', and 'downtoners', which 'scale downwards', was found difficult to apply in practice, with a number of degree adjuncts occupying points around the middle. A five-category classification is suggested ('extreme high', 'high', 'mid', 'low', 'extreme low'). The high classes were found to be larger and more lexically varied, the low classes smaller and more grammaticalised.
In addition to the five scalar categories, two categories that generate special additional implicatures, 'approximating' (e.g. “almost”, “practically”) and 'relative' (e.g. “enough”, “more”), are recognised.
The majority of studies examining copula constructions concentrate primarily on the range of semantic possibilities which can be covered by 'copula' constructions (e.g. Declerck 1988, Hengeveld 1992). This paper presents some results of a recent cross-linguistic survey of the syntax of copula constructions, looking only at those constructions used in languages to encode the core copula concepts of identity ("that man is my brother") and group membership ("that man is a teacher"). It will be shown that there are four main strategies which languages use to a verbal copula, a non-verbal copula, an inflection on the predicate noun or a "zero copula". Any language may use one or more of these strategies, with alternations dependent on a variety of factors, the most common being tense/aspect/mood, person, polarity, and the concept encoded.
Given that there are few (if any) copula constructions which involve a transitive verb - that is, which have the two arguments marked in the same way as and having the same behaviour as the Agent and Patient arguments of a Primary Transitive Verb such as 'kill' - one issue which arises is the grammatical function or relation (A, S, O, Subject, Object, etc.) which is played by each of the two NPs in these constructions (cf. Andrews 1985). It is sometimes claimed that copula constructions have an intransitive subject (S) and a Copula Complement or Predicate Noun (e.g., Dixon 1994, Stassen 1992); this will be examined and found to be true in some cases, at least when there is a verbal copula. However in other cases it is not clear that there is any justification for considering either NP in a copula construction to be an S, with neither NP exhibiting language-internal S or subject properties. In the majority of languages without a verbal copula, one or both NPs in a copula construction appear to share many properties with S, for example they commonly have the same case marking, and they are therefore often analyzed as being S. However these apparent shared properties between S and the NPs in copula constructions are almost always the sharing of the most unmarked feature. In languages where S has marked features (for example, where there is a marked Nominative case), the NPs in copula constructions are often distinct from S (but not always; cf. Comrie 1997), and remain in the unmarked case while S is in the marked Nominative case.
This might suggest that in some languages at least, it is necessary to introduce Copula Subject and Copula Complement as two new and distinct core grammatical functions, beside A, S and O. However there are problems with this as well; as noted above, there are many languages with two or more distinct constructions used to encode concepts of identity and group membership, and the syntactic and morphological properties of each of the two NPs in a copula construction can differ greatly in different construction types, suggesting that in many languages neither Copula Subject nor Copula Complement can be considered to be unitary grammatical functions.