Language Evolution: The Windows Approach addresses the question: "How can we unravel the evolution of language, given that there is no direct evidence about it?"
The LINGUIST List is dedicated to providing information on language and language analysis, and to providing the discipline of linguistics with the infrastructure necessary to function in the digital world. LINGUIST is a free resource, run by linguistics students and faculty, and supported primarily by your donations. Please support LINGUIST List during the 2016 Fund Drive.
Date: Sun, 12 Mar 2006 19:17:23 +0300 From: Alexandre Arkhipov Subject: Co-Compounds and Natural Coordination
AUTHOR: Wälchli, Bernhard TITLE: Co-Compounds and Natural Coordination SERIES: Oxford Studies in Typology and Linguistic Theory YEAR: 2005 PUBLISHER: Oxford University Press
Alexandre Arkhipov, Moscow State University
Co-compounds (< coordinating compounds) are also known as dvandva compounds, pair words (Russian 'parnye slova'), and copulative compounds (German 'Kopulativkomposita'). Mordvin t'et'at- avat 'parents', literally 'father-mother', is an example of co-compound. These are the expressions of natural coordination which appear as word-like, rather than phrase-like, formally tight coordination.
Natural (vs. accidental) coordination conjoins inherently semantically close concepts belonging to the same taxonomic level, in a way that the meaning of the whole construction is a conceptual unit which is closely related to the meaning of the parts and belongs to a superordinate taxonomic level in respect to the parts. Examples of natural coordination are 'brother and sister', 'fork and spoon', 'eat and drink' as opposed to 'my uncle and his book' or 'read and swim'. Natural coordination, like some other semantic types of coordination, often tends to be expressed by formally tight (vs. loose) constructions. Generally, the tightness of a construction correlates with the number of morphemes and boundaries intervening between its parts (see Chapter 3). For example, a coordination of two nouns is looser if they are connected by an overt coordination marker and if each one bears its own articles, inflection markers and adpositions; it is tighter with the articles omitted, no overt coordinator and a single inflection marker attached to the whole. Tight coordination patterns, then, may expose different features, being more word-like (cf. Georgian ''dá- dzma'' 'siblings', literally 'sister-brother', without any morpheme intervening and bearing a single stress and a single final inflection) or more phrase-like (cf. English ''brother and sister'', two phonological words separated by a coordinator).
Co-compounds (=CCs) are treated as a cross-linguistic class type, i.e. as a type of language-specific classes which share a number of recurrent features. CCs usually take a more word-like pattern, but rarely can all CCs in a given language be undoubtedly classified as words. Also, the formal properties of CCs vary across languages along with the properties of coordination patterns, hence the problem of definition (see Chapter 4).
The book investigates the formal, semantic and use-specific properties of CCs both in particular languages and cross-linguistically.
Chapter 1, 'Introduction', is an overview of the problems to be discussed and of methods to be used in the rest of the book. In Chapters 2 and 3 co-compounds are treated along with their phrase- like counterpart. Chapters 4 to 7 consider co-compounds on their own. The book is accomplished with Conclusions (Chapter 8), Appendices showing genetic affiliation and location of the languages cited, Bibliography and Indices.
Chapter 2, 'The marking patterns of natural coordination', makes a tour of the marking strategies used to code natural coordination. Two main options are relational and non-relational marking. Marking is relational if it is the relation between the coordinated elements (coordinands) which is coded by an overt marker (e.g. English ''and''). Natural coordination rarely uses overt markers of its own, but can also be distinguished by the absence or reduction of a 'normal' coordination marker. The absence of marker has, though, limited distinctive power, as in many languages coordination may be normally zero-marked.
Coordination has a non-relational marking if the markers used in the construction encode some properties of the coordinands which do not concern the relation of coordination itself; cf. the marking of definiteness, number, possession etc. on coordinand(s). This non- relational marking may be distinctive if the patterns are different for different types of coordination. E.g. possession may be marked on both coordinands in ordinary coordination, but only once in natural coordination, as in Eastern Armenian; this is what Bernhard Wälchli (hereafter BW) calls 'distinctive non-relational single marking'. There are two more options, viz. double and zero marking. Non-relational double marking may involve double dual, plural, comitative, proprietive, additive focus or possessive marking. Systematic distinctive zero marking appears to be confined to the definiteness category, occasionally involving also number and adpositions. All three subtypes receive explanation in terms of concurrence between two iconic motivations: that of minimal distance and that of symmetry. The issue of syntactic (a)symmetry of single marking constructions is also discussed; it is argued that group inflection phenomena support the claim of non-isomorphism between phonology and syntax rather than the syntactic asymmetry of the construction.
Chapter 3, 'Tight coordination', explores the formal properties leading to the tight-loose contrast, and also the semantic nature of coordination that can be realized as tight coordination. The first formal parameter is the length of coordinands: the shorter the coordinands, the tighter is the construction. The second parameter of tightness pertains to the patterns of coordination: the less material (words, affixes, coordinators, intonation breaks etc.) is separating the two slots, the tighter is the pattern. One important observation is that typically a language does not have two strictly distinct patterns, a loose one and a tight one, but rather a whole range of patterns which occupy different places on the loose-to-tight continuum; Turkish nominal coordination is taken as an example of such a range of patterns. However, it is not always possible to decide unequivocally which one of two patterns is tighter or looser. Moreover, the degree of tightness of a construction is intertwined with the degree of grammaticalization of particular affixes and adpositions; if, say, a preposition is repeated with each coordinand in one case but is used only once in another, it is not always obvious whether the difference lies in the tightness of two constructions or in the degree of grammaticalization of the preposition.
The larger part of the chapter is devoted to various semantic types of coordination which can be expressed by tight constructions. Most of these types (group coordination, intersective coordination, enumeration, and pseudo-coordination) are shown to be independent of the natural coordination. E.g. group coordination may conjoin a natural pair, but this is not obligatory; at the same time, elements that naturally belong together may participate in a situation separately, as in distributive contexts (''Brother and sister went to different schools''). Some other kinds of coordination that turn out to be relevant for natural coordination are overlapping, non-exhaustive listing, and disjunctive coordination. The low degree of contrast between the coordinands, characteristic of natural coordination, also correlates with tighter constructions.
In Chapter 4, 'Co-compounds as a lexical class type', it is argued that co-compounds should be defined as a more basic concept than compounds; they are considered as a functional-formal class type. Discussing various problems connected to the definitions of word and compound (prosodic vs. grammatical words; non-compositionality or 'unitary' semantics of compounds; special compounding forms of stems; fixed order, etc.), the author decides not to dwell upon the notion of word or compound. The approach chosen here is to prove CCs to be a valid lexical class (vs. grammatical classes such as plural or future); examples of other lexical classes include middle as described by Kemmer (1993), diminutives, light verb with specifier constructions ('instruction + give > teach', 'work + do > to work'), or noun incorporation as described by Mithun (1984). Lexical classes typically have 'a clearly discernible semantic core'; they are highly idiosyncratic as to participation of individual elements, with lexeme- internal vacillation; they also do interact with other lexical classes.
An important feature of a well-established lexical class is its productivity in two aspects: it should be productive in forming occasional highly context-oriented units, which, on the other hand, should be able to lexicalise and penetrate the permanent part of the lexicon. It is also essential that co-compounds as a lexical class ''have a strong tendency to be formally non-distinct from other linguistic phenomena'', such as sub-compounds, serial verbs, and phrase coordination. Even if the class of co-compounds in a given language is usually formally distinct from coordination, almost always one can find examples which cannot be unambiguously attributed to one of those.
In Chapter 5, 'A semantic classification of co-compounds', some ten semantic types of CCs are discussed. The classification is based on the semantic relation between the parts of CC and the whole compound. One and the same CC may be attributed to different types, depending on its actual meaning in a given context. Not all semantic types can be strictly distinguished. There are five basic types: (i) Additive CCs denote the sum of its parts (Georgian ''xel-p'exi'' 'hand- foot'). (ii) Generalizing CCs are a sort of universal quantifiers, they denote general notions such as 'all', 'always' and 'everywhere', depending on the type of object referred to by their parts (Tagalog ''araw-gabi'' 'day and night'). The parts often have opposite meanings (day and night, big and small, here and there). (iii) Collective CCs represent collective types of which their parts denote typical representatives, but do not exhaustively list the whole (Chuvash ''erex-săra'' 'vodka/wine-beer > alcoholic beverages'). (iv) In alternative CCs the whole is the disjunction of the parts ('two- three > two or three'). They occur even in weakly co-compounding, e.g. West European, languages. They are semantically close to the (v) approximate CCs ('two-three > several').
Several other types (synonymic, ornamental, imitative, figurative, and scalar CCs) are considered non-basic. The basic types conform strictly to the prototype of natural coordination, while each of non- basic types deviates from the features of natural coordination (within part-part relations, part-whole relations, or the meaning of the whole). (vi) In synonymic CCs, the meaning of the whole is almost identical to the meanings of both parts (Khalkha Mongolian ''üzel bodol'' 'view- thought > opinion'), thus is not taxonomically superordinate. (vii) In ornamental CCs, one of the parts does not contribute to the meaning of the whole, thus the two parts are not on the same taxonomic level (Mordvin ''vir'-ukštor'' 'forest-maple > forest'). (viii) Imitative CCs have a meaningless part phonologically similar to the meaningful part (Turkish ''çoluk çocuk'' '[IMITATIVE] child > wife and family'), so that there is no coordination strictly speaking. (ix) The meaning of figurative CCs belongs to another semantic domain than the meanings of the parts (Mandarin Chinese ''mao2 dun4'' 'spear shield > contradictory, inconsistent'). (x) In scalar CCs there is maximal contrast between the parts, the whole denoting an abstract scale between these two extremes (Tibetan ''srab-mthug'' 'thin-thick > density'). The synchronic relations between different semantic types of CCs do not necessarily hold in diachrony: e.g. ornamental CCs do not derive from semantically close synonymic CCs. Several other types of compounds share similar properties with CCs, like appositional compounds (French ''wagon-restaurant''), intermediate compounds (English ''southwest''), echo words (Turkish ''doktor moktor'' 'doctor and/or the like') etc.; the closest to the CCs are ideophone compounds like English ''ding-dong''.
The last section of the chapter discusses the various semantic factors that favour the use of semantically redundant CCs, e.g. synonymic CCs: emphasis, generalization, contrast, non-referentiality under the scope of negation, and pictorality. These contextual factors may neutralize the differences between types of CCs and also between CCs and related types of compounds.
Chapter 6, 'Areal distribution in the languages of Eurasia', is devoted to the comparative frequency measurements of co-compounds in texts in various languages, principally those of Eurasia. The main observation of the chapter is that, with few exceptions, the ''co- compounding'' decreases steadily from continental South-East Asia (Mandarin, Vietnamese, White Hmong) to the west (West European languages) as well as to the north. The main exceptions to this areal vector are Basque in the West which has a moderate level of co- compounding (higher than its neighbours), and Modern Tamil and Modern Uyghur which have too few CCs.
The calculations have been made using two kinds of parallel texts: Universal Declaration of Human Rights or excerpts therefrom (=UDHR), and Gospel according to Mark (=Mark). Seven levels of co- compounding are distinguished. It is shown that geographic location is sometimes more significant here than the genetic affiliation; e.g., genetically closely related Turkic languages do not behave so uniformly as the several distinct linguistic families in the Caucasus region.
Translations of Mark are also used to count the relative frequency of different semantic types of co-compounds. It is argued that the proportions of CC types in a given text depend more on the overall frequency of CCs than on other factors. There are some correlations; notably, the proportion of non-basic (synonymic + ornamental + imitative) CCs increases along with the overall frequency of CCs, i.e. in highly compounding languages, while the proportion of generalizing CCs decreases at the same time (their absolute number remaining roughly the same). These tendencies prove then to work in language- internal variation (case study of Erzya Mordvin). Here two fiction texts and a traditional epos are compared, revealing great variation between the three in frequencies of CCs, with the proportions of their semantic types sticking to the predictions.
Finally, non-Eurasian languages are discussed in an appendix to Chapter 6. Beyond Eurasia, CCs are found in New Guinea and more scarcely, in the Americas (Quechua, Chinantec and Mixe, Tzotzil). They seem to be very rare in Africa and Australia, as well as in Creole languages.
In Chapter 7, 'Some considerations about the diachronic evolution of co-compounds', the two main perspectives on the evolution of CCs are discussed: their evolution as formal patterns vs. their evolution as lexical classes. The role of textual markedness in the destiny of CCs and their affinity to the folk poetry are also addressed. It is argued that normally formal patterns for CCs emerge from tight coordination patterns, although spontaneous evolution is not excluded (viz. for mere juxtaposition patterns). The class of CCs may also arise as an extension of tight coordination into word-domains, without introducing a new formal pattern.
The formation of a class of CCs is different from the evolution of specific formal patterns which do not necessarily form classes. It is argued that lexical classes of CCs evolve gradually, so that every highly co-compounding language should have passed the stages of low and moderate level of co-compounding. What prevents CCs from being used in non-co-compounding languages is their high degree of textual markedness. The more CCs exist already in a language (or in a particular register or style), the easier it is for a new CC to lexicalise; so the development of a class of CCs is self-accelerating process. The same idea accounts for the fact that non-basic CCs begin their expansion only after a certain level of co-compounding is reached, as they are still more textually marked than the basic CCs.
On the whole, the book reviewed here is well-structured and presents a fresh view upon a phenomenon important to many languages of the world, containing a considerable amount of language data. Co- compounds turn out to be a cluster of formal and functional features, and consequently, the book should be studied carefully by linguists interested in a number of subjects such as compounding, coordination, comitative, associative plural, and lexical semantics within domains specific to co-compounding. A number of basic linguistic notions are reconsidered, such as word and compound, markedness and lexicalisation. One of the main points of the book is to show the importance of continuous variables in typology, which allow one to account for language-internal as well as cross-linguistic variation.
A good deal of the study is centred around Erzya Mordvin in which the author is proficient, but in whole the data of various languages are sufficiently balanced for this type of study as far as it does not pretend to use a representative sample of the world's language families.
The semantic classification of CCs is probably the most subtle issue in the book, and the author himself acknowledges repeatedly that no sharp border can be settled between some classes, as well as between CCs and some related phenomena. Some questions still remain controversial. Thus we find echo-words like Turkic m- and Yiddish shm-formations (cf. Turkish ''doktor moktor'' 'a doctor or something'; Russian ''tancy-shmancy'' 'dances and similar amusements') distinguished from imitative CCs. BW does not consider echo-words to be CCs because they have only one lexical slot, the second (echo) part being generated by a common phonological rule. Meanwhile, the imitative part of an imitative CC is by definition meaningless and does not occur independently, so it has hardly more rights than an 'echo' to occupy a lexical slot. The only synchronic difference here seems to lie in the regularity of the phonological correspondences between the meaningful and the meaningless parts, which is not a clear-cut distinction; hence a continuous scale between the two types. It appears, then, that both types can equally be considered or not considered (non-basic) co-compounds.
In some cases, the terminology used might be reconsidered. For instance, in Chapter 5, BW uses Cruse's (1986) terms to name the compounds like ''husband.PL-wife.PL'' 'husband and wife' converses because ''A is the reference point for B and B for A''; he then states that compounds for bipartite tools such as 'bow and arrows' ''are also a kind of converse, where A and B are each other's reference points''. This extension of the term ''converse'' does not seem very appropriate. While a husband cannot be a husband of no one, but always and only of his wife or wives, a bow does not require an arrow to be a bow. Here we find another type of relations otherwise known as complementer-complemented relations (Russian ''dopolnitel'- dopolniaemoe'', cf. Voroncova & Rakhilina 1998); in these relations, the two parts do not form a whole but they are necessary for each other to function properly.
A remark is to be made on the book's presentation. Due to the significance of hyphens for the subject of the study and to their frequency in the cited examples, BW keeps hyphens as an orthographic means to separate parts of co-compounds, and uses dots instead as a morpheme separator, e. g. (Erzya Mordvin ex. 5, p. 144; palatalization is marked with apostrophe):
This usage seems somewhat confusing and not very readable, as it deviates from the widely accepted use of the hyphen as a morpheme delimiter, at the same time running into conflict with the use of dot to denote cumulation in the glossing line; see also the ''Leipzig glossing rules'' in http://www.eva.mpg.de/lingua/files/morpheme.html. (BW sometimes also uses colon in glosses, presumably to denote cumulation, but never mentions its value explicitly.) I would instead suggest replacing the ''compounding'' hyphens with a special symbol, such as underscore.
The book is not free from a small number of misprints and slips which, in rare cases, complicate the interpretation of the text; cf. in 5.4.8 'Conclusions' on page 170, ''Echo-words and affirmative- negative co-compounds occur with the same or often similar functions as co-compounds'' instead of ''...affirmative-negative compounds...''.
To summarize, the research presented in the book is an original work of high interest to many specialists. One could also mention that BW's Ph.D. dissertation, which was the basis for this published study, was awarded the 2005 ALT Junior Prize (see http://www.lancs.ac.uk/fss/organisations/alt/news/05-36.HTM#sec1), although he declined it.
I am grateful to Natalia Shibasova (Institute of Linguistics, Russian Academy of Sciences) for her remarks on a draft of this review.
Cruse, D. A. (1986) Lexical Semantics. Cambridge: CUP.
Kemmer, Suzanne. (1993) The middle voice. Amsterdam, Philadelphia: John Benjamins.
Mithun, Marianne. (1984) The evolution of noun incorporation. Language 60, pp. 847-895.
Voroncova M. I.; Rakhilina E. V. (1994) Physical object nouns and prepositional constructions. Znak: A collection of papers in linguistics, semiotics and poetics in commemoration of A. N. Zhurinskij. [Predmetnye imena i predlozhnye konstrukcii. Znak: Sbornik statei po lingvistike, semiotike i poetike pamyati A. N. Zhurinskogo.] Moscow: Russkij uchebnyj centr MS, 1994, pp. 181-190.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Alexandre Arkhipov is a researcher at Lomonosov State University of Moscow. His Ph.D. thesis (2005) was devoted to the typology of comitative constructions. His research interests also include Basque and Daghestanian linguistics, fieldwork and electronic resources for field linguists, and semantics of prosody.