Locative subject constructions such as the English existential there-construction and the Chichewa locative construction described by Bresnan & Kanerva 1989 pose difficulties for theories of linking arguments to grammatical roles, as well as being interesting in their own right. Comparison of the Chichewa locative subject construction to its counterparts in other Bantu languages, and the English there-construction to related constructions in other Germanic languages, reveals the existence of a grammaticalization path from locative subject constructions to impersonals. Locative phrases in locative inversion constructions take on at least some subject properties, then are extended into functions such as existential and weather expressions which can be construed as having locative subjects, but then are reanalyzed as impersonals, ending up being used for impersonal situation types that have nothing to do with locative subjects. I argue that locative subjects and related locative constructions can be handled by the causal model of linking developed in Croft (1991, 1996) better than by localist and prototype theories of linking, and present an analysis of the semantic changes that underlie the locative-to-impersonal grammaticalization process.
One of the basic principles in establishing a genetic grouping of languages is that typological traits must be excluded from consideration at the initial stage of establishment; only substantive traits (form-meaning pairings) are relevant. Yet typological evidence is still frequently appealed to in recent discussions of the methodology for establishing genetic classifications and in recent proposals for genetic affiliations. It is suggested that this is due to a conflation of typological and substantive traits in theoretical discussions of the role of 'grammar' and 'structure' in establishing genetic classifications of languages.
One of the most challenging phenomena in syntax and semantics is how events and their participants are encoded by clauses in languages. In this chapter, we focus on the relation between verbs and argument structure constructions (ASCs). The central dimension of the semantics of argument structure constructions is force dynamic interactions between participants. Following Croft (2012), the interaction is postulated to be a causal chain. Noncausal relations are often construed as directed. Our analysis goes beyond Croft's in defining types of change that each participant undergoes in an event, and elaborating types of force-dynamic interactions. This allows us to analyze event classes such as 'application', 'transfer of possession' and 'caused motion'. Verbal semantics is also force-dynamic, but allows force-dynamic networks that involve cyclicity and possibly branching. Types of participant change and many types of force-dynamic interactions are common to both ASC causal chains and verbal networks. Some mismatches represent construals of events, such as construing the Force verb kick as a motion verb in He kicked the football across the stadium. Others represent metaphorical mappings from ASC causal chains to verbal networks, such as the use of the physical domain argument structure construction [Sbj Verb Obj to Obl], as in Linda taped the picture to the wall, in the social domain of transfer of possession, as in Jerry loaned his skateboard to his brother.
We describe two mathematical representations for what have come to be called "semantic maps", that is, representations of typological universals of linguistic co-expression with the aim of inferring similarity relations between concepts from those universals. The two mathematical representations are a graph structure and Euclidean space, the latter as inferred through multidimensional scaling. Graph structure representations come in two types. In both types, meanings are represented as vertices (nodes) and relations between meanings as edges (links). One representation is a pairwise co-expression graph, which represents all pairwise co-expression relations as edges in the graph; an example is CLICS. The other is a minimally connected co-expression graph — the "classic semantic map". This represents only the edges necessary to maintain connectivity, that is, the principle that all the meanings expressed by a single form make up a connected subgraph of the whole graph. The Euclidean space represents meanings as points, and relations as Euclidean distance between points, in a specified number of spatial dimensions. We focus on the proper interpretation of both types of representations, algorithms for constructing the representations, measuring the goodness of fit of the representations to the data, and balancing goodness of fit with informativeness of the representation.
Stories are typically represented as a set of events and temporal or causal relations among events. In the metro map model of storylines, participants are represented as histories and events as interactions between participant histories. The metro map model calls for a decomposition of events into what each participant does (or what happens to each participant), as well as the interactions among participants. Such a decompositional model of events has been developed in linguistic semantics. Here, we describe this decompositional model of events and how it can be combined with a metro map model of storylines.
Computational resources such as semantically annotated corpora can play an important role in enabling speakers of indigenous minority languages to participate in government, education, and other domains of public life in their own language. However, many languages — mainly those with small native speaker populations and without written traditions — have little to no digital support. One hurdle in creating such resources is that for many languages, few speakers would be capable of annotating texts — a task which requires literacy and some linguistic training — and that these experts' time is typically in high demand for language planning work. This paper assesses whether typologically trained non-speakers of an indigenous language can feasibly perform semantic annotation using Uniform Meaning Representations, thus allowing for the creation of computational materials without putting further strain on community resources.
Different types of languages evolve in situations of human social contact, depending on the nature of the contact and the attitudes on the part of the speakers towards the societies in contact. Three socially-defined language types are useful for classifying contact languages. One may distinguish between an esoteric language, used for communication within a speech community, and an exoteric language used for communication between different speech communities. The third language type, a neogenic language, results when speech communities merge (e.g., in colonisation). Each language type involves a continuum depending on the degree and asymmetry of the social contact between speech communities. This chapter describes these continua, the "contact" or "mixed" languages that they account for, and their structural linguistic correlates.
In this paper we present Uniform Meaning Representation (UMR), a meaning representation designed to annotate the semantic content of a text. UMR is primarily based on Abstract Meaning Representation (AMR), an annotation framework initially designed for English, but also draws from other meaning representations. UMR extends AMR to other languages, particularly morphologically complex, low-resource language. UMR also adds features to AMR that are critical to semantic interpretation and enhances AMR by proposing a companion document-level representation that captures linguistic phenomena such as coreference as well as temporal and modal dependencies that potentially go beyond sentence boundaries.
This paper presents a "road map" for the annotation of semantic categories in typologically diverse languages, with potentially few linguistic resources, and often no existing computational resources. Past semantic annotation efforts have focused largely on high-resource languages, or relatively low-resource languages with a large number of native speakers. However, there are certain typological traits, namely the synthesis of multiple concepts into a single word, that are more common in languages with a smaller speech community. For example, what is expressed as a sentence in a more analytic language like English, may be expressed as a single word in a more synthetic language like Arapaho. This paper proposes solutions for annotating analytic and synthetic languages in a comparable way based on existing typological research, and introduces a road map for the annotation of languages with a dearth of resources.
This paper introduces a representation and annotation scheme for argument structure constructions that are used metaphorically with verbs in different semantic domains. We aim to contribute to the study of constructional metaphors which has received little attention in theoretical and computational linguistics. The proposed representation consists of a systematic mapping between the constructional and verbal event structures in two domains. It reveals the semantic motivations that lead to constructions being metaphorically extended. We demonstrate this representation on argument structure constructions with Transfer of Possession verbs and test the viability of this scheme with an annotation exercise.
This paper extends Croft's theory of argument realization to include nominals that denote events, instead of participants. Events are represented as force dynamic interactions between participants and their subevents. That is, each participant is associated with its own subevent; participants are related to each other through force dynamic interactions as part of a causal chain (Croft 2012). There is extensive cross-linguistic evidence that a participant’s place in the causal chain determines its argument realization. In this paper, we tested three hypotheses about the argument realization of event nominals against English sentences from VerbNet. First, we define an event nominal as any nominal that refers to an event, regardless of whether it is morphosyntactically derived from a verb. Our three hypotheses are (i) event nominals correspond to participant subevents, (ii) event nominals follow the same argument realization rules as their associated participant, and (iii) when both a participant and its subevent are realized as arguments, the subevent (i.e., the event nominal) is construed as subsequent to its participant in the causal chain. We find support for these hypotheses in the VerbNet data. In addition, we discuss more complex cases of argument realization with event nominals.
This paper presents an annotation scheme for modality that employs a dependency structure. Events and sources (here, conceivers) are represented as nodes and epistemic strength relations characterize the edges. The epistemic strength values are largely based on Sauri and Pustejovsky's (2009) FactBank, while the dependency structure mirrors Zhang and Xue's (2018b) approach to temporal relations. Six documents containing 377 events have been annotated by two expert annotators with high levels of agreement.
This paper proposes a novel representation of event structure by separating verbal semantics and the meaning of argument structure constructions that verbs occur in. Our model demonstrates how the two meaning representations interact. Our model thus effectively deals with various verb construals in different argument structure constructions, unlike purely verb-based approaches. However, unlike many constructionally-based approaches, we also provide a richer representation of the event structure evoked by the verb meaning.
Developers of cross-lingual semantic annotation schemes face a number of issues not encountered in monolingual annotation. This paper discusses four such issues, related to the establishment of annotation labels, and the treatment of languages with more fine-grained, more coarse-grained, and cross-cutting categories. We propose that a lattice-like architecture of the annotation categories can adequately handle all four issues, and at the same time remain both intuitive for annotators and faithful to typological insights. This position is supported by a brief annotation experiment.
Many English tense and aspect semantic contrasts are not currently captured within Abstract Meaning Representation (AMR) annotations. The proposed framework augments the representation of finite predications in AMR to include a four-way temporal distinction (event time before, up to, at, or after speech time) and several aspectual distinctions (including static vs. dynamic, habitual vs. episodic, and telic vs. atelic). We validate this approach with a small annotation study of sentences from The Little Prince and report details of ongoing discussion to refine the framework. This will en- able AMR to be used for NLP tasks and applications that require sophisticated reasoning about time and event structure.
Although English grammar encodes a number of semantic contrasts with tense and aspect marking, these semantics are currently ignored by Abstract Meaning Representation (AMR) annotations. This paper extends sentence-level AMR to include a coarse-grained treatment of tense and aspect semantics. The proposed framework augments the representation of finite predications to include a four-way temporal distinction (event time before, up to, at, or after speech time) and several aspectual distinctions (including static vs. dynamic, habitual vs. episodic, and telic vs. atelic). This will enable AMR to be used for NLP tasks and applications that require sophisticated reasoning about time and event structure.
We present a rich annotation scheme for the structure of mental events. Mental events are those in which the verb describes a mental state or process, usually oriented towards an external situation. While physical events have been described in detail and there are numerous studies of their semantic analysis and annotation, mental events are less thoroughly studied. The annotation scheme proposed here is based on decompositional analyses in the semantic and typological linguistic literature. The scheme was applied to the news corpus from the 2016 Events workshop, and error analysis of the test annotation provides suggestions for refinement and clarification of the annotation scheme.
The increasing availability of large digital corpora of cross-linguistic data is revolutionizing many branches of linguistics. Overall, it has triggered a shift of attention from detailed questions about individual features to more global patterns amenable to rigorous, but statistical, analyses. This engenders an approach based on successive approximations where models with simplified assumptions result in frameworks that can then be systematically refined, always keeping explicit the methodological commitments and the assumed prior knowledge. Therefore, they can resolve disputes between competing frameworks quantitatively by separating the support provided by the data from the underlying assumptions. These methods, though, often appear as a ‘black box’ to traditional practitioners. In fact, the switch to a statistical view complicates comparison of the results from these newer methods with traditional understanding, sometimes leading to misinterpretation and overly broad claims. We describe here this evolving methodological shift, attributed to the advent of big, but often incomplete and poorly curated data, emphasizing the underlying similarity of the newer quantitative to the traditional comparative methods and discussing when and to what extent the former have advantages over the latter. In this review, we cover briefly both randomization tests for detecting patterns in a largely model-independent fashion and phylolinguistic methods for a more model-based analysis of these patterns. We foresee a fruitful division of labor between the ability to computationally process large volumes of data and the trained linguistic insight identifying worthy prior commitments and interesting hypotheses in need of comparison.
Storyline research links together events in stories and specifies shared participants in those stories. In these analyses, an atomic event is assumed to be a single clause headed by a single verb. However, many analyses of verbal semantics assume a decompositional analysis of events expressed in single clauses. We present a formalization of a decompositional analysis of events in which each participant in a clausal event has their own temporally extended subevent, and the subevents are related through causal and other interactions. This decomposition allows us to represent storylines as an evolving set of interactions between participants over time.
Charles Fillmore developed both the theory of construction grammar and the theory of frame semantics. Fillmore's original frame semantic model included broad frames such as Commercial Transaction and Risk. FrameNet II (the current incarnation) has a complex lattice of frame-to-frame relations, in which the original frames are often abstract frames, and/or have been decomposed into distinct concrete frames. The frame-to-frame relations are of different types beyond simple taxonomic relations. The concrete frames are often more restricted in their argument structure construction possibilities than the original, more general frames. We argue that some frame-to-frame relations effectively represent the internal structure of the core events in the frame; the event structures are associated with different argument structure constructions. Introducing event structure into the frame semantic representation offers the potential to simplify the FrameNet lattice. Brief analyses of the Commercial Transaction and Risk frames are presented.
Current work on universal dependency schemes in NLP does not make reference to the extensive typological research on language universals, but could benefit since many principles are shared between the two enterprises. We propose a revision of the syntactic dependencies in the Universal Dependencies scheme (Nivre 2015, Nivre et al. 2016) based on four principles derived from contemporary typological theory: dependencies should be based primarily on universal construction types over language-specific strategies; syntactic dependency labels should match lexical feature names for the same function; dependencies should be based on the information packaging function of constructions, not lexical semantic types; and dependencies should keep distinct the "ranks" of the functional dependency tree.
The fundamental importance of lexical categories is uncontroversial within both formal and functional approaches to grammatical analysis. But despite the familiarity of this topic and its foundational nature for grammatical description and analysis, it is paradoxically not among the best-studied or -understood topics from either the functionalist or formalist perspective. Both schools of linguistic theory have inherited their basic assumptions and instincts about lexical categories from the structuralist practice of distributional analysis. We briefly survey approaches to the various lexical categories. We then comment on a few issues of strategic value that arise from these approaches, including the importance of clearly distinguishing roots, stems, words, and syntactic units when it comes to issues of lexical categories; the importance of recognizing when distributional tests are similar across languages in principled ways; and the need for the choice of distributional tests to be informed by theoretical hypotheses.
The relationship between typology and cognitive linguistics was first posed in the 1980s, in terms of the relationship between Greenbergian universals and the knowledge of the individual speaker. An answer to this question emerges from understanding the role of linguistic variation in language, from occasions of language use to typological diversity. This in turn requires the contribution of discourse analysis, sociolinguistics, and evolutionary historical linguistics as well as typology and cognitive linguistics. While cognitive linguistics is part of this enterprise, a theory of language that integrates all of these approaches is necessary.
What are comparative concepts and how are they related to language-specific categories used in language description? Three general categories of comparative concepts are defined here: purely functional comparative concepts, and two types of hybrid formal-functional concepts, constructions and strategies. The two hybrid types provide more explicit and precise definitions of common typological practice. However, a terminological issue is that Western grammatical terms are frequently used to describe strategies, which are not universal, rather than constructions, which are. Language-specific categories appear to be radically different from comparative concepts, because the former are defined distributionally whereas the latter are defined in universal functional and formal terms. But language-specific constructions have functions, that is they are instances of constructions in the comparative sense; and their form is an instantiation of a strategy. Typology forms generalizations across language-specific constructions, in both their form and their function. Finally, a major issue is the confusion of terminological choices for language-specific categories. Four rules of thumb for useful labeling of language-specific categories, largely following best descriptive practice, are offered.
We use a mathematical model to examine three phenomena involving language change across the lifespan: the apparent time construct, the adolescent peak, and two different patterns of individual change. The apparent time construct is attributed to a decline in flexibility toward language change over one's lifetime; this explanation is borne out in our model. The adolescent peak has been explained by social networks: children interact more with caregivers a generation older until later childhood and adolescence. We find that the peak also occurs with many other network structures, so the peak is not specifically due to caregiver interaction. The two patterns of individual change are one in which most individuals change gradually, following the mean of community change, and another in which most individuals have more categorical behavior and change rapidly if they change at all. Our model suggests that they represent different balances between the differential weighting of competing variants and degree of accommodation to other speakers.
Causal and temporal relations among events are typically analyzed in terms of interclausal relations. Yet participants in a monoclausal event interact causally as well, and monoclausal events unfold in temporal phases. We propose an annotation scheme for the resulting analysis of event structure types. The annotation scheme is based on a fine-grained analysis of aspectual structure combined with a novel analysis of physical event types based on proposals in the theoretical linguistics literature. By decomposing complex events in a clause, we will ultimately model the overall dynamic causal network of entities interacting over time described in a text.
How universal is human conceptual structure? The way concepts are organized in the human brain may reflect distinct features of cultural, historical, and environmental background in addition to properties universal to human cognition. Semantics, or meaning expressed through language, provides indirect access to the underlying conceptual structure, but meaning is notoriously difficult to measure, let alone parameterize. Here, we provide an empirical measure of semantic proximity between concepts using cross-linguistic dictionaries to translate words to and from languages carefully selected to be representative of worldwide diversity. These translations reveal cases where a particular language uses a single "polysemous" word to express multiple concepts that another language represents using distinct words. We use the frequency of such polysemies linking two concepts as a measure of their semantic proximity and represent the pattern of these linkages by a weighted network. This network is highly structured: Certain concepts are far more prone to polysemy than others, and naturally interpretable clusters of closely related concepts emerge. Statistical analysis of the polysemies observed in a subset of the basic vocabulary shows that these structural properties are consistent across different language groups, and largely independent of geography, environment, and the presence or absence of a literary tradition. The methods developed here can be applied to any semantic domain to reveal the extent to which its conceptual structure is, similarly, a universal attribute of human cognition and language use.
The term 'functional' or 'functionalist' has been applied to any approach to grammar that puts the primary emphasis on explanation in terms of factors outside the formal structure of language. The four most significant functional approaches can be characterized in terms of the types of functional explanations that are put forward and the position taken vis-à-vis formalist explanations of grammar. More recently, a dynamic, usage-based approach to the functionalist analysis of grammar has been developed.
The close relationship between biological evolution and language was noted by Darwin himself in an oft-quoted passage from The Descent of Man: "the formation of different languages and of different species, and the proofs that both have been developed through a gradual process, are curiously parallel”" (Darwin, 1882). In fact, the development of evolutionary theory in biology was inspired in part by the advances in historical linguistics in the early nineteenth century. In the twentieth century, evolutionary theory did not have much influence on linguistics, in part due to the advent of structuralism and the focus on synchronic linguistic analysis. In the past two decades, there has been a considerable increase in interest in the relationship between language and evolution. The interest has been manifested in three areas of recent research. The first is the evolutionary origin of the human language capacity, a topic that was actively avoided in linguistics for a century. The second is the employment of techniques of phylogeny reconstruction from biology in the analysis of genetic families of languages. Finally, there is the application of theories of evolutionary processes to language change, based on the hypothesis that, as Darwin wrote, the two are strikingly parallel.
In the past thirty years, a variety of approaches to linguistic analysis have emerged under the name of "functionalism," including construction grammar, cognitive linguistics, typology and universals research, the usage-based model, and evolutionary approaches to language change. All of these approaches have shed new light of the relationship between grammatical form and meaning in language. In the twenty-first century, these seemingly diverse approaches to language can be integrated into the study of language as a complex adaptive system.
This chapter argues that an evolutionary cultural approach to language not only has already proven fruitful, but it probably holds the key to understand many puzzling aspects of language, its change and origins. The chapter begins by highlighting several still common misconceptions about language that might seem to call into question a cultural evolutionary approach. It explores the antiquity of language and sketches a general evolutionary approach discussing the aspects of function, fitness, replication, and selection, as well the relevant units of linguistic evolution. In this context, the chapter looks at some fundamental aspects of linguistic diversity such as the nature of the design space, the mechanisms generating it, and the shape and fabric of language. Given that biology is another evolutionary system, its complex coevolution with language needs to be understood in order to have a proper theory of language. Throughout the chapter, various challenges are identified and discussed, sketching promising directions for future research. The chapter ends by listing the necessary data, methods, and theoretical developments required for a grounded evolutionary approach to language.
Language change can be understood as an evolutionary process. Language change occurs at two different timescales, corresponding to the two steps of the evolutionary process. The first timescale is very short, namely, the production of an utterance: this is where linguistic structures are replicated and language variation is generated. The second timescale is (or can be) very long, namely, the propagation of linguistic variants in the speech community: this is where certain variants are selected over others. At both timescales, the evolutionary process is driven by social interaction and the role language plays in it. An understanding of social interaction at the micro-level—face-to-face interactions—and at the macro-level—the structure of speech communities—gives us the basis for understanding the generation and propagation of language structures, and understanding the nature of language itself.
A variety of mechanisms have been proposed in sociolinguistics for the propagation of an innovation through the speech community. The complexity of social systems makes it difficult to evaluate the different mechanisms empirically. We use the four-way typology of mechanisms proposed by Baxter et al. (2009), and define them mathematically in such a way that the effects of different mechanisms in the trajectory of a change can be modeled. The model suggests that the widely observed empirical pattern of an S-curve temporal trajectory of change can be captured only if the mechanisms for propagation include replicator selection, that is, differential weighting of the competing variants in a change, except under highly specialized circumstances that probably do not hold in speech communities in general
Language is not the result state of some type of learning process, but is a process itself, so that there is complete continuity in kind between what an infant is doing and what an adult is doing with language. This chapter describes three fundamental language processes undertaken by children and adults. Language is part of joint action: language is a conventional coordination device for communication, which in turn is an effective coordination device for joint action. Language itself is behavior: speakers replicate linguistic structures in utterances for social purposes. The nature of the replication process guarantees that replication generates variation and ultimately language change (cultural evolution). Finally, language is also verbalization: the construal of human experience into lexicogrammatical form.
Language universals are usually thought of as properties holding of all languages. But very few, if any, such universals exist, due to the extreme structural diversity of languages. Instead, there are many typological universals, which allow for variation but constrain it or at least limit its distribution. This is true even of linguistic categories. Formal (grammatical or lexical categories) are not universal, but are constrained by the structure of conceptual space, as demonstrated by a multidimensional scaling analysis of adpositional semantic data from Levinson et al. (2003). Even so, broad conceptual categories are not universal either. Instead, what is universal is the holistic conceptualization of highly particular situation types, and the conceptual relationships that hold among them. This conclusion is confirmed by the analysis of data on within-language variation in verbalization from Croft (2010).
Using our interdisciplinary research collaboration as a case study, we discuss the question of whether formal modeling and empirical approaches can be successfully integrated into a single line of research. We argue that to avoid an undesirable disconnect between the two, one needs considerable time and patience for a science–humanities collaboration to bear fruit. In our collaboration and, we believe, in science=humanities collaborations in general, certain shared goals are required for success, including: starting with simple models before moving to more complex models; the importance of continually comparing models with empirical data where possible; and a focus on explaining statistical patterns rather than accounting for single data points individually.
Smith offers a critique of the theory of parts of speech in Croft (1991, 2001) inter alia. Smith tries to make a functionally-based universal-typological theory of parts of speech provide an answer to the problem of defining word classes and giving those classes the same names across languages ("noun"; "adjective"); this is not possible and not what I intended. Smith conflates semantic properties with pragmatic properties, and he conflates different pragmatic properties that cannot be conflated. There are challenging issues in defining pragmatic functions and their linguistic reflexes, but Smith's critique only briefly touches on them.
The origins of language change, particularly grammatical change, appear to be unobservable. But the first step in language change, innovation, can be observed in the production of synchronic variation in sound change. The same can be done for morphosyntactic change such as grammaticalization by comparing alternative verbalizations of the same experience in a controlled situation. Examples of innovation in lexical semantic change and grammaticalization are examined using the twenty parallel English narratives of the Pear Stories. Morphosyntactic variation is pervasive in the Pear Stories narratives and the alternative verbalizations show that morphosyntactic change is drawn from a pool of synchronic variation. These results disconfirm the traditional theory of morphosyntactic change, in which innovation is rare and special mechanisms are require to produce it. Instead, grammaticalization, and language change in general, originates in the variation inherent in the verbalization of experience.
Language has a fundamentally social function. Processes of human interaction along with domain-general cognitive processes shape the structure and knowledge of language. Recent research in the cognitive sciences has demonstrated that patterns of use strongly affect how language is acquired, is used, and changes. These processes are not independent of one another but are facets of the same complex adaptive system (CAS). Language as a CAS involves the following key features: The system consists of multiple agents (the speakers in the speech community) interacting with one another. The system is adaptive; that is, speakers' behavior is based on their past interactions, and current and past interactions together feed forward into future behavior. A speaker's behavior is the consequence of competing factors ranging from perceptual constraints to social motivations. The structures of language emerge from interrelated patterns of experience, social interaction, and cognitive mechanisms. The CAS approach reveals commonalities in many areas of language research, including first and second language acquisition, historical linguistics, psycholinguistics, language evolution, and computational modeling.
Language is a complex adaptive system: Speakers are agents who interact with each other, and their past and current interactions feed into speakers' future behavior in complex ways. In this article, we describe the social cognitive linguistic basis for this analysis of language and a mathematical model developed in collaboration between researchers in linguistics and statistical physics. The model has led us to posit two mechanisms of selection — neutral interactor selection and weighted interactor selection — in addition to neutral evolution and replicator selection (fitness). We describe current results in modeling language change in terms of neutral interactor selection and weighted interactor selection.
Studies of language change have begun to contribute to answering several pressing questions in cognitive sciences, including the origins of human language capacity, the social construction of cognition and the mechanisms underlying culture change in general. Here, we describe recent advances within a new emerging framework for the study of language change, one that models such change as an evolutionary process among competing linguistic variants. We argue that a crucial and unifying element of this framework is the use of probabilistic, data-driven models both to infer change and to compare competing claims about social and cognitive influences on language change.
Constructional analysis of corpus data can contribute to the analysis of a semantic frame, as demonstrated by a small corpus study of English eat and feed. The EAT/FEED frame forms part of a taxonomy of frames including the superordinate frame of CONSUME and subordinate frames of human vs. animal eating; construction and metaphor data in the corpus shows that English covertly distinguishes human and animal eating. The EAT frame includes three phases (intake, process, and ingest), differentiated lexicogrammatically. The EAT frame also includes three domains in its domain matrix: physical, biological (nutritional) and social, all clearly differentiated by distinct constructions in the corpus. An examination of metaphors with eat and feed in the corpus demonstrate that the target domain contributes image-schematic structure to the metaphorical mapping, contrary to the Invariance Hypothesis.
Trudgill (2004) proposes that the emergence of New Zealand English, and of isolated new dialects generally, is purely deterministic: it can be explained solely in terms of the frequency of occurrence of particular variants and the frequency of interactions between different speakers in the society. Trudgill's theory is closely related to usage-based models of language, in which frequency plays a role in the representation of linguistic knowledge and in language change. Trudgill's theory also corresponds to a neutral evolution model of language change. We use a mathematical model based on Croft's usage-based evolutionary framework for language change (Baxter et al. 2006), and investigate whether Trudgill's theory is a plausible model of the emergence of new dialects. The results of our modeling indicate that determinism cannot be a sufficient mechanism for the emergence of a new dialect. Our approach illustrates the utility of mathematical modeling of theories and of empirical data for the study of language change.
Cognitive linguistics starts from a set of fundamental hypotheses about the nature of grammar: grammatical structures and processes in the mind are instances of general cognitive abilities; grammar is symbolic, and thus meaning is an essential part of grammar; meaning is encyclopedic; and meaning involves conceptualization (construal). These hypotheses contrast with those found in formal linguistics, and align cognitive linguistics with functionalist theories. But language is not just a cognitive ability - if it were just that, we would not need to speak, and might not even need language at all. Language is a central feature of human social interaction, and cannot be fully understood outside of that fact. Cognitive linguistics must reach out and embed itself in a more general social theory of language.
The fundamental hypotheses of cognitive linguistics must be reformulated to reflect this fact. Linguistic cognitive abilities represent an individual's knowledge of linguistic convention, which is a coordination device for communication, which in turn is a coordination device for joint action. Our encyclopedic knowledge (that is, linguistic meaning) is situated in common ground among speakers; its basis in shared expertise and shared experience emerges from the communities of practice in which speakers partake. The symbolic unit of grammar is actually a semiotic triangle of form, meaning and the community in which the form-meaning pairing is conventional. The conceptualization of experience in grammar formulates or "packages" that experience for the purpose of communication. This social reformulation of the fundamental hypotheses of cognitive linguistics in turn has consequences for the cognitive aspect of language. In particular, I will argue that linguistic meaning is indeterminate and partly socially constructed, which leads to a variationist, utterance based, evolutionary model of grammar.
There are two broad theoretical traditions in research on syntactic universals, the typological and the generative. While the theoretical differences between the two approaches are well known, less known but equally important is the fact that the methods used to find and justify syntactic universals are very different between the two approaches. The generative approach, illustrated with Baker's (2003) analysis of verbal vs. nonverbal predication, proceeds "one language at a time", and draws on whatever constructions are deemed relevant to support the universal underlying structure. However, the "one language at a time" method often produces universals that fail a wider crosslinguistic comparison; but the choice of constructions used to counter anomalous evidence is an instance of methodological opportunism (Croft 2001). A rigorous method of syntactic argumentation is proposed, which is close to best practice in typological research.
Both qualitative concepts and quantitative methods from evolutionary biology have been applied to linguistics. Many linguists have noted the similarity between biological evolution and language change, but usually have only employed selective analogies or metaphors. The development of generalized theories of evolutionary change (Dawkins and Hull) has spawned models of language change based on such generalized theories. These models have led to the positing of new mechanisms of language change and new types of selection that may not have biological parallels. Quantitative methods have been applied to questions of language phylogeny in the past decade. The focus has been on widely accepted families with cognates already established by the comparative method (Indo-European, Bantu, Austronesian). Increasingly sophisticated phylogeny reconstruction models have been applied to these families, to resolve questions of subgrouping, contact and migration. Little progress has been made so far in analyzing sound correspondences in the cognates themselves
A fundamental fact about grammatical structure is that it is highly variable both across languages and within languages. Typological analysis has drawn language universals from grammatical variation, in particular by using the semantic map model. But the semantic map model, while theoretically well-motivated in typology, is not mathematically well-defined or computationally tractable, making it impossible to use with large and highly variable crosslinguistic datasets. Multidimensional scaling (MDS), in particular the Optimal Classification nonparametric unfolding algorithm, offers a powerful, formalized tool that allows linguists to infer language universals from highly complex and large-scale datasets. We compare our approach to Haspelmath's semantic map analysis of indefinite pronouns, and reanalyze Dahl's (1985) large tense-aspect dataset. MDS works best with large datasets, demonstrating the centrality of grammatical variation in inferring language universals and the importance of examining as wide a range of grammatical behavior as possible both within and across languages.
Why is there grammar? I argue that the verbalization of experience can provide an explanatory model of the function of grammatical categories and constructions. Chafe's model of verbalization provides a model of how the unanalyzed, unique whole of experience is broken up into parts that are categorized into types, that is, the lexical roots of an utterance. We must add further processes to Chafe's model, because the speaker also must convert the types represented by the lexical roots back to the particulars of the experience, and put the parts back together into the original whole. These additional processes express the aspects of experience that become grammaticalized. A functional classification of grammatical categories and structures in terms of their role in verbalization is outlined.
The distribution of grammatical units (GUs) across intonation units (IUs) is analyzed in a corpus of 2072 intonation units of Wardaman monologic oral narrative (Merlan 1994), and compared to a previously published study of English (Croft 1995) and several other languages. Since English and Wardaman are structurally very different languages, any common patterns in the mapping of grammatical units to intonation units would be of considerable interest as potential grammar-discourse universals. The Full GU Condition - IUs are almost always full GUs - holds in Wardaman as well as English and other languages. Both English and Wardaman employ a substantial number of grammatically independent noun phrase intonation units. Three factors constrain the occurrence of GUs in a single IU in English, in descending order of strength: parallelism, complexity and distance. All three factors also hold in Wardaman in the same order of strength. The behavior of the IU-GU mapping in Wardaman supports critiques of the analyses of arguments as adjuncts and of modifiers as appositive phrases. On the other hand, spoken English displays more grammatical characteristics similar to Wardaman than prescriptive written English.
Aarts (2004) argues that the best way to model grammatical categories is a compromise preserving Aristotelian form classes with sharp boundaries on the one hand, and allowing gradience in terms of the number of syntactic properties that a category member possesses on the other. But the assumption of form classes causes serious theoretical and empirical problems. Constructions differ in their distributional patterns, but no a priori principles exist to decide which constructions should be used to define form classes. Grammatical categories must be defined relative to specific constructions; this is the position advocated in Radical Construction Grammar (Croft 2001). Constructionally defined categories may have sharp boundaries, but they do not divide words into form classes. Nevertheless, the most important traditional intuitions for parts of speech (Aarts' chief examples) are reinterpretable in terms of crosslinguistic universals that constrain distributional variation but do not impose Aristotelian form classes, gradable or not, on the grammars of particular languages.
'Radical' templatic phonology is a template-based approach to segmental phonological representation. The central hypothesis is that the segmental phonological structure of words is represented as language-specific phonotactic templates, in the sense used in the developmental literature. Template-based organization of the early lexicon has been identified in children acquiring several different languages. It is the result of a usage-based abstracting or 'induction' process, based on both babbling practice (phonetic production) and input experience with specific adult phonological patterns. The resulting templates thus constitute patterns that reconcile (or 'adapt') the model provided by target words with the child's own phonetic repertoire of syllables or word shapes - typically extending or building on the forms initially 'selected' for first word production, in which adult and child forms show a close match. In adult phonology segment categories-natural classes, or features-are best defined in terms of their occurrence in positions in the templates in individual languages, not as independent universal categories. After reviewing the status of segment categories and their phonetic basis in contemporary phonological theory we present crosslinguistic evidence of pervasive variation in both phonetic realization and phonological distribution patterns, evidence that supports the template construct.
In Explaining Language Change (2000), I develop an evolutionary model of language change. This model is based on the philosopher of science David Hull's generalized analysis of selection, which is abstracted from biological specifics. In this model, I identify the replicators - the central units that evolve - as linguistic structures in utterances; they are replicated every time we open our mouths to speak. The interactors - whose behavior causes replicators to be replicated differentially - as the speakers who choose alternative forms to express meanings in particular social contexts. Variation is generated by phonetic factors in replication (for sound change) and functional factors (for grammatical change). Selection, that is, differential replication, is driven by social factors. In this paper, this model is outlined and updated, and used to address a number of outstanding issues in historical linguistics, including the role of children in language change, the invisible hand, the use of phylogeny reconstruction algorithms in genetic historical linguistics, the significance of convention and intentional behavior in language change, types of grammatical change, the integration of internal and external factors, the integration of sociohistorical linguistics and 'traditional' historical linguistics, and finally the integration of synchronic and diachronic linguistics.
This chapter describes an evolutionary framework for analyzing language change that integrates functional-typological and variationist sociolinguistic approaches to historical linguistics. The model is evolutionary in that it is a theory of how Hull's generalized analysis of selection is to be applied to the processes of language change in contemporary human languages, such as the history of English. As such, the evolutionary framework is not a theory of how the (ultimately biological) ability for human language evolved, though it has important implications for such a theory. Nor is it a theory of the relationship between human linguistic prehistory and human biological prehistory, though it implies certain ways to apply biological methods to linguistic problems. Instead, it is a theory of language change that takes as its starting point variation across individual speech events, across speech communities in a society, and across languages. I focus on functional-typological approaches to language and language change and how they are integrated in an evolutionary framework.
We present a mathematical formulation of a theory of language change. The theory is evolutionary in nature and has close analogies with theories of population genetics. The mathematical structure we construct similarly has correspondences with the Fischer-Wright model of population genetics, but there are significant differences. The continuous time formulation of the model is expressed in terms of a Fokker-Planck equation. This equation is exactly soluble in the case of a single speaker and can be investigated analytically in the case of multiple speakers who communicate equally with all other speakers and give their utterances equal weight. Whilst the stationary properties of this system have much in common with the single-speaker case, time-dependent properties are richer. In the particular case where linguistic forms can become extinct, we find that the presence of many speakers causes a two-stage relaxation, the first being a common marginal distribution that persists for a long time as a consequence of ultimate extinction being due to rare fluctuations.
Fifty years ago, Joseph Greenberg put forward the now widely-accepted classification of African languages. This book charts the progress of his work on language classification in Oceania, the Americas, and Eurasia, in which he proposed the language families Indo-Pacific, Amerind and Eurasiatic. It shows how he established and deployed three fundamental principles: that the most reliable evidence for genetic classification is the pairing of sound and meaning; that nonlinguistic evidence such as skin colour or cultural traits, should be excluded from the analysis; and that the vocabulary and inflections of a very large number of languages should be simultaneously compared. The volume includes Joseph Greenberg's substantive contributions to the debate his work provoked and concludes with his writings on the links between genetic linguistics and human history.
William Croft's introduction focuses on the substance and the development of Professor Greenberg's thought and research within the context of the discussion they stimulated. He also includes a bibliography of scholarly reactins to and developments of Joseph Greenberg's work and a comprehensive bibliography of his publications in books and journals.
Cognitive linguistics argues that language is governed by general cognitive principles, rather than by a special-purpose language module. This introductory textbook surveys the field of cognitive linguistics as a distinct area of study, presenting its theoretical foundations and the arguments supporting it. Clearly organized and accessibly written, it provides a useful introduction to the relationship between language and cognitive processing in the human brain. It covers the main topics likely to be encountered in a course or seminar, and provides a synthesis of study and research in this fast-growing field of linguistics. The authors begin by explaining the conceptual structures and cognitive processes governing linguistic representation and behavior, as well as syntactic representation and analysis, focusing on the closely related frameworks of cognitive grammar and construction grammar. This much-needed introduction will be welcomed by students in linguistics and cognitive science.
Radical Construction Grammar is a model of construction grammar. I will take as a starting point the hypotheses of "vanilla [basic] construction grammar" and briefly outline the differences between it and RCG. Vanilla construction grammar makes the following hypotheses about the nature of syntactic representation: (i) a uniform representation of grammatical structures - constructions - as symbolic units of varying degrees of schematicity and complexity; (ii) organization of constructions into taxonomic networks (a third hypothesis shared by some but not all models is a usage-based approach to grammatical organization). RCG proposes three further hypotheses about grammatical representation: (a) constructions are the basic units of representation; grammatical categories are defined in terms of the constructions in which they occur and in terms of semantic maps on conceptual space; (b) syntactic relations between elements of a construction are unnecessary given the presence of symbolic relations between the elements and the components of the construction's semantic structure; (c) syntactic constructions do not form discrete language-independent structural types, but vary across languages along dimensions of syntactic space. For each hypothesis of RCG, logical arguments and (a tiny sample of) typological arguments in favor will be presented.
Comparison of the grammars of human languages reveals systematic patterns of variation. Typology and universals research uncovers those patterns, to formulate universal constraints on language and seek their explanation. A comprehensive introduction to the method and theory used in typology-universals research is presented. The theoretical issues raised range from the most fundamental - on what basis can the grammars of diverse languages be compared? - to the most abstract - what is the role of functional and historical explanations of language universals? Language universals in phonology, morphology, syntax and semantics are extensively illustrated.
The second edition has been thoroughly revised and rewritten, to bring in typological research in the past decade, including new methodologies such as the semantic map model and questions of syntactic argumentation; discussion of current debates over deeper explanations for specific classes of universals, and comparison of the typological and generative approaches to language.
Verbs appear to change their meaning when put into particular constructions. For example, bake means 'bake x and give x to someone' in the ditransitive construction. Two alternative analyses have been proposed: to derive the meaning of bake in the ditransitive construction by a lexical rule, or to have the meaning of bake in the ditransitive construction be predictable from the semantics of the construction. An analysis of the behavior of different verb classes in the ditransitive construction indicates that both analyses are partly right. The ability to enter the ditransitive construction is verb-class-specific, or even verb-specific, but the meaning of the verb (class) found in the ditransitive is peculiar to the ditransitive construction and clearly involves a family resemblance to the meaning of other verb classes found in the construction. I argue that the simplest way to capture these facts is to represent verb-class-specific constructions and verb-specific constructions in the grammar of English.
Linguistics and evolutionary biology have substantially diverged until recently. The chief reason for this divergence was the dominance of essentialist thinking in linguistics during the twentieth century. Croft (2000) describes a thoroughgoing application of Hull's (1988) generalized theory of selection to language change. In this model, tokens of linguistic structure in utterances ('linguemes') are replicators and speakers are interactors. Current debates in the philosophy of evolutionary biology (e.g. Sterelny & Griffiths 1999) are then applied to language change. Hull's generalized theory is post-synthesis: it recognizes a distinction between replicator and interactor and is independent of levels of biological organization. Biological issues such as mechanisms of inheritance (e.g. Lamarckism) and of selection (e.g. intentional behavior) are simply irrelevant to the generalized theory of selection outside biology. However, there are many striking parallels between biological evolution and language change that are likely to be consequences of the generalized theory of selection, including flexibility of adaptation to the environment, emergent structure, evolutionary conservatism, vestigial traits, exaptation, and the absence of "progress". The evolutionary theory of language change is not evolutionary psychology, but it is mimetics; this approach is defended against Sterelny & Griffith's criticisms.
Radical Construction Grammar presents a profound critique of syntactic theory and argumentation, and offers a genuinely new approach to syntax based on the fact of grammatical diversity. Recent syntactic theories are essentially formal models for the representation of grammatical knowledge, and posit complex syntactic structures in the analysis of sentences. The result has been a endless cycle of new and revised theories of syntactic representation.
Radical Construction Grammar argues that this approach to syntax is incompatible with the grammatical variation found within and across languages. This book defends three fundamental theses: (i) constructions are the primitive units of syntactic representation, and grammatical categories are defined by constructions, not the other way around; (ii) the only syntactic structures are the part-whole relations between a construction and the syntactic elements that make it up; (iii) not only are grammatical categories construction-specific, but constructions are language-specific. In other words, syntactic structure is almost entirely language-specific; attempts to find a universal formal model are doomed to failure.
Radical Construction Grammar integrates concepts from typological theory and construction grammar to uncover the genuine universals of grammar. Constructions are represented as complex symbolic units pairing form and meaning. The semantic map model of typological theory is used to map category distributions on a largely universal conceptual space. Universals of grammar are found in the mapping of meaning onto form. Systematic patterns of grammatical variation provide evidence for the topography of conceptual space, which in turn reflects the geography of the human mind.
We present an analysis of the argment linking of the commercial transaction frame in English, Russian and Japanese. The commercial transaction frame is semantically quite complex, in that there are two transfers in opposite directions (money goes from buyer to seller and goods go from seller to buyer). Also, the three languages examined have verbs linking either buyer or seller to subject position, and either money or goods to object position; the remaining two participant roles are linked to obliques. We use Croft's analysis of causal structure in argument linking to model the variation and constraints on linking. The model uses an underlying uniform representation of the commericail transaction frame, built in turn on the transfer of possession frame, and allows for different construals of the event in different languages. The choice of oblique cases or adpositions for the oblique participant roles are motivated by the use of the same case/adposiiton in other semantic frames. English and Russian construe the commerical transaction frame in essentially the same way. Japanese construes one aspect of the frame differently, leading to different argument linking patterns in that language.
The term functional or functionalist has been applied to any approach to grammar that puts the primary emphasis on explanation in terms other than the formal structure of a language. Three types of functional explanations are generally offered. Processing explanations attribute aspects of grammatical structure to limitations on the human ability to produce or comprehend the structure.Cognitive explanations attribute aspects of the analysis of the grammatical phenomenon its conventional meaning. Discourse explanations make reference to the structuring of information for the purpose of communication. There are four widely practiced functional approaches. The autonomist functionalist approach accepts a formal analysis of grammatical structure, but proposes that many constraits on the operation of grammatical rules are explained by semantic factors and/or information structure in discourse. Mixed formal-functional analyses offer functional accounts of some aspect of grammatical structure, and formal accounts for other aspects. Typological functionalists argue that cross-linguistically valid universal constraints on grammatical structure have their source in cognitive and/or discourse factors, and purely structural generalizations are language-particular. Cognitive linguistics argues that the basic structural elements of grammar are not purely formal categories but represent a conceptualization of experience encoded in language.
Ever since the origins of both linguistics and evolutionary biology in the 19th century, scholars have noted the similarity between biological evolution and language change. Yet until recently neither linguists nor biologists have developed a model of evolution general enough to apply across the two fields. Even in linguistics, the field is split between the historical linguists who study change in language structure, and the sociolinguists who study social variation in the speech community.
Explaining language change represents the first thoroughly worked out framework for language evolution, building on the pioneering ideas of Richard Dawkins and David Hull in biology and philosophy of science. Its central thesis is that the locus of language change is the utterance in social intercourse. Linguistic innovations emerge from the remarkable complexity of communication in social interaction. Once innovations occur, they are propagated through the equally complex social structures of the speech communities we participate in. Explaining language change provides a framework for assessing current theories of language change, and advances new ideas about grammatical reanalysis, conventional and nonconventional use of language, the structure of speech communitites, language mixing, and the notion of "progress" in language change. Explaining language change reintegrates sociolinguistics and historical linguistics, weaving together research on grammatical change, pragmatics, social variation, language contact and genetic linguistics.
The major parts of speech (noun, verb, adjective) are not categories of particular languages, but are language universals. Linguists have used distribution of words in constructions to justify parts of speech membership. But no sound theoretical basis has been provided to justify choice of tests for membership, leading to disagreement and confusion. In fact, the variation in the occurrence of constructions and in the distribution patterns of words across languages and within languages demonstrates that lexical classes are language-specific and construction-specific. A radical construction grammar model is proposed to represent this state of affairs. The universals of parts of speech are manifested in conceptual space, with principles such as typological markedness defining prototypes in the formal expression of conceptual categories found in conceptual space.
This essay deals with one of the fundamental aspects of the semantics of cognitive linguistics, namely the the nature of construal operations, and this entailis reference to the conceptualization of the described situation in the mind of the language user. In expounding the conceptualist approach to semantics, the essay highlights its differences with respect to formal semantics and analyzes particular forms of imagery (scanning, scalar adjustment, profile-base, attention). In doing so, it also elaborates a number of general points concerning the two different types of appraoch to semantics. This essay also conducts a critique of the use by artificial intelligence of certain phenomenological concepts.
Despite differing theoretical views within cognitive semantics there appears to be a consensus on certain fundamental theoretical constructs: (i) The basic semantic unit is a mental concept; (ii) concepts cannot be understood independent of the domain in which they are embedded; (iii) conceptual structures represent a construal of experience, that is, an active mental operation; and (iv) concept categories involve prototypes and are organized by (at least) taxonomic relations. Although 'concepts', 'domains', 'construal' and 'category structure' go by different names, the basic constructs are essentially the same across researchers in cognitive linguistics. We examine a fifth theoretical construct, 'image schemas' (recurring basic conceptual structures), and argue that image schemas are a subtype of domain. We begin with the theory of domains proposed by Langacker, which is similar to Fillmore's theory of frame semantics. Langacker distinguishes two types of domains, locational and configurational; we argue that it is concepts in domains that are locational or configurational, not the domains themselves. We then analyze image schemas and show how they function like domains, in which are found both locational and configurational concepts.
Ever since the pioneering work of Berlin & Kay 1969 and Talmy 1972, there has been a fruitful interaction between cross-linguistic research and cognitively oriented linguistic research. Still, there are some significant methodological differences between the two. Also, as in any healthy scientific paradigm, there are some important disputes within cognitive linguistics. In this talk, I will suggest that the typological approach to grammar can address some of the issues being debated in cognitive linguistics, in particular fundamental issues of categorization and the relationship between form and meaning.
Generative grammatical theories (including their nontransformational offshoots and some formal-functional hybrids) have assumed that there are universal categories such as 'noun', verb', 'subject' and 'object'. At least some cognitive linguists have concurred that these universal categories exist, but instead have offered conceptual semantic schemas to characterized them. Cross-linguistic research has uncovered a greater degree of variation than expected with respect to parts of speech and the encoding of participant roles (and much else besides) Typologists have formulated cross-linguistic universals governing this variation but have not pursued the consequences of this for the mental representation of the knowledge of individual languages by speakers.
The conclusion that must be drawn from typological research is that there are no universal grammatical primitives, in the sense of an inventory of universal categories, some or all of which are found in the grammars of particular languages. Instead, the universals that we use the names 'noun', 'verb', 'subject' and 'object' for are patterns in the topography of conceptual space (an approach that represents a significant contribution from cognitive linguistics to typology, incidentally). These patterns typically occur as radial categories with some sort of prototype structure. This conclusion suggests that cognitive linguists should not expect to find universal schematic meanings for categories such as 'noun', 'subject', etc., but rather a prototype-extension pattern for these categories. This does not mean that there is no semantic conceptualization or construal associated with universals of grammatical categories (though see below). It does imply that the construal is the result of the prototype-extension structure rather than following from a universal category schema (category boundary).
The role of construal in the form-function relation gives rise also to a related apparent conflict between functional-typological linguistics and cognitive linguistics. A central principle of functional-typological linguistics is iconicity: that the structure of language reflects the structure of experience. (It should be added that iconicity interacts with other functional principles such as economy.) Typologists see iconicity as an organizing principle for grammatical structure, which means that the semantic structure partially determines grammatical form.
In cognitive linguistics, a central principle is that allegedly meaningless grammatical categories and constructions do have meaning: they represent a construal or conceptualization of experience (which can range from conventional imagery to a more active process), which changes with the use of different constructions, including different constructions across languages. In other words, construction form specifies a construal of experience by the speaker - the opposite assumption of causal directionality from iconicity.
There is truth to both points of view. Grammatical constructions can impose a conceptualization on experience; but examples of grammaticalization demonstrate that constructions themselves change their meaning, which should be interpreted as certain prototypical conceptual structures asserting themselves over the conceptualization imposed by the source construction.
Not all formalists are alike; nor are formalist theories so deeply interconnected that functionalists must reject everything done by all formalists, or be required to accept all of formalist theory by accepting some part of it. Features of the formalist research tradition are identified that could be adopted (or could converge with) the functionalist research tradition. Certain criticisms by formalists of conclusions drawn from functional analyses should be accepted by functionalists, and functionalists must address representational issues more closely. Finally, both formalists and functionalists must recognize the limits to synchronic analysis and the need to examine cross-linguistic evidence before offering explanations.
Cognitive linguists (and other linguists, notably generative linguists) argue for certain mental representations of grammatical and lexical knoewledge based on introspective linguistic data. It is argued here that introspective linguistic data can only restrict the range of possible mental representations a priori. That is, one cannot determinately posit a particular model of mental representation without evidence from actual usage or psycholinguistic experiments. Four points on a probable continuum of mental representations are described, from the most specific representation to the most general: independent entries (or homonymy), polysemy, (grammatical) derivational rules and pragmatic rules (or monosemy). The introspecitve evidence that can exclude one or more of these models is described, and other intermediate models are discussed. Essentially, grammatical and semantic idiosyncrasies are evidence for excluding the more general models; but grammatical and semantic generality is not a priori evidence for exluding the more speciifc models. The reason for the probable continuum of models of mental representation is that dynamically, metnal representations of linguistic knowledge change over time, for a single speaker and for a specch community.
Children's overextensions (e.g. referring to a pomegranate as apple) raise intriguing questions regarding early word meanings. Specifically, how do object shape, taxonomic relatedness, and prior lexical knowledge influence children's overextensions? The present study sheds new light on this issue by presenting items that disentangle the three factors of shape, taxonomic category, and prior lexical knowledge, and by using a novel comprehension task (the screened-alternative task) in which children can indicate negative exemplars (e.g. which items are not apples). 49 subjects in three age groups participated (Ms = 2;0, 2;6 and 4;5). Findings indicate: (1) Error patterns differed by task. In production, errors were overwhelmingly due to selecting items that matched the target word in both shape and taxonomic relatedness. In comprehension, more errors were based on either shape alone or taxonomic relatedness alone, and the nature and frequency of the overextensions interacted with prior lexical knkowledge. (2) Error patterns also differed markedly based on the word being tested (apple vs. dog), in both comprehension and production. (3) As predicted, errors were more frequent in production than comprehension, though only for children in the two younger age groups. Altogether, the study indicates that overextensions are not simply production erros, and that both taxonomic relatedness and object shape play a powerful role in early naming errors.
A considerable number of theories of argument linking from a wide range of theoretical perspectives have been proposed in recent years Although there are significant differences between these approaches, it would not be unfair to say that all of them have all or almost all of the following salient theoretical constructs, arranged here in "layers":
Event structure. Some sort of decompositional representation of the causal/aspectual structure of events is taken as the semantic foundation for argument linking.
Thematic roles and a thematic role hierarchy. There is a consensus that thematic roles are not theoretical primitives but are derived from event structure. However, once thematic roles are defined in terms of event structure, they are then usually arranged in a thematic role hierarchy which provides a ranking of participant roles for the purposes of argument linking.
Syntactic arguments or "super-roles". It appears to be a near-consensus that the thematic role hierarchy is insufficient for determining argument linking, and another layer of representation is necessary. Here the theories fall in two camps. One camp proposes a syntactic level of argument structure; another argues that in fact this layer is semantic.
Linking rules. Finally, there are rules specifying how participants in participant roles are linked to the appropriate casemarked NPs using the concepts in the preceding three layers.
In this paper, I outline an alternative view, earlier versions of which can be found in Croft 1991 and various papers, and which will be described in detail in a forthcoming book (Verbs: aspect and argument structure). The positing of thematic role hierarchies, syntactic arguments, and/or super-roles still requires certain linkings to be idiosyncratically stipulated. The fundamental problem with the theories referred to above is that once thematic roles are derived from the semantic structure of events, the rest of event structure semantics is so to speak thrown away. Yet two other aspects of event structure are crucial for argument linking: force-dynamic relations and the verbal profile. In fact we can do without the thematic role hierarchy and syntactic arguments/super-roles if we employ these two elements of event structure, and end up with a linking theory that is simpler than the alternatives - making reference to event semantics only and using a small number of universal linking rules - and comparable in its range of predictions.
The resulting model, including the relevant features of event structure, can be compared to the aforementioned theories in the following way:
Event structure. A rich event semantic structure is assumed, from which thematic roles (to the extent that they are useful in linking, which is in fact not much) can be defined.
Force-dynamic relations among participants. Instead of a thematic role hierarchy, a ranking of participants in terms of their force-dynamic relations to each other is argued to be critical for linking, in fact more important than type of thematic role in the usual sense. One participant outranks another if it is antecedent to the other in the causal chain (in terms of transmission of force). Establishing the causal chain involves a conceptualization (or construal) of the event on the part of the speaker, particularly for noncausal relations between participants; these appear to follow universal cognitive constraints. (Hence force-dynamic structure is a complex derivation from "raw" event structure.) Alternative force-dynamic conceptualizations of an event are also possible, giving rise to alternative linking patterns for certain semantic classes of events.
Verbal profile. Different verbs may be defined from the same force-dynamic conceptual semantic structure of an event, depending on what segment(s) of the causal (force-dynamic) chain is actually denoted by the verb. I use the term profile to describe the denoted segment of the causal chain, following Langacker's Cognitive Grammar. The specification of the verb profile largely does the work of syntactic relations/super-roles in the other theories.
Linking rules. If we use the force-dynamic structure of events, and recognize the verb profile as part of the semantic representation of a verb in a particular language, then the following four universal linking rules can be established (X > Y = 'X antecedes Y in the force-dynamic chain'):
1. The verbal profile is delimited by Subject and Object (if any)
2. Subject > Object
3. Antecedent Oblique > Object > Subsequent Oblique
4. Subject > Incorporated Noun > Object (if any)
Linking Rule 1 makes clear the essential role of the verb profile in argument linking. Linking Rules 2-4 exploit the force-dynamic relationship among participants in an event. Linking Rule 3 implies that oblique cases can be divided into two types for the purposes of linking (see Croft 1991). Linking Rule 4 is necessary to account for patterns of noun incorporation as the latter is generally defined in the cross-linguistic literature (see Croft 1991).
The theory of metaphor proposed by Lakoff & Johnson (1980) and Lakoff (1993) involves a mapping of conceptual structure from one semantic domain to another. We investigate properties of these conceptual domain mappings by comparing them to morphological derivational relations. Schematicity and productivity are properties that Bybee (1985) and Langacker (1987) propose for characterizing morphological derivational relations, which we apply to our analysis of metaphor.
Metaphors are argued to vary in their degree of semantic schematicity: domain relations function as generalizations over specific metaphorical expressions. We also demonstrate three points on a continuum of productivity: conventional metaphors are highly productive, metaphorically motivated transparent idioms are semi-productive, and opaque idioms are unproductive. Each point is compared with an example of morphological productivity having a corresponding conceptual organization. The results demonstrate that semantic productivity can be characterized in the same way as morphological productivity, suggesting that form and meaning are organized by the same principles.
Hull (1988) uses recent developments in the theory of biological evolution, in particular rigorous application of the population theory of species, a consistently phylogenetic approach to evolutionary taxonomy and a proposed resolution of the dispute over which levels natural selection operates, to propose a general analysis of selection processes which he then applies to conceptual change in science. Hull's model of selection is applied to language change. It is argued that the the utterance plays the central role in linguistic selection, and causal mechanisms by which linguistic selection - language change - occurs are proposed. The final sections consider the possibility that selection occurs also at higher levels of linguistic organization, and suggest how language contact may be accounted for in terms of phylogenetic reticulation.
Zwicky (1985) and Hudson (1987) present a wide range of morphosyntactic and semantic criteria for defining the notion of a 'head' in grammatical theory, which represent the standard criteria for headhood, and apply them to a set of six English constructions. Zwicky argues that the criteria yield different results, but Hudson argues that most of them converge on a particular constituent. I widen the empirical base of the analysis to additional constructions and languages, and present a cognitive semantic analysis of headhood.
Most of the syntactic criteria discussed by Zwicky and Hudson pertain to functions other than headhood. Concord and agreement phenomena always involve agreement with a noun phrase (C. Lehmann 1982). Distributional equality and obligatoriness depend critically on the preferences of the language with respect to anaphora (particularly zero anaphora), ellipsis and "headless" constructions, which also seem to be independent of headhood. I analyze the phenomena of subcategorization and government in a construction grammar framework (Fillmore 1988, Langacker 1987). In this framework, subcategorization frames and their "governed" morphosyntactic form are constructions which are listed in the lexicon in the same way as single lexical items. Constructions have their own semantic function and are found in a many-to-many relation with their lexical "fillers". Headhood is at best indirectly related to these phenomena.
This leaves the fundamental semantic intuition voiced by Zwicky and reinforced by Hudson, that "in a combination X+Y, X is the 'semantic head' if, speaking very crudely, X+Y describes a kind of the thing described by X." I will make this notion less crude by employing definitions of the major syntactic categories based on the propositional acts of reference, predication and modification (Croft 1991) to define the phrases they head. However I will argue, contra Hudson, that there is a case for headhood for both the "content word" (e.g. N in NP, V in a clause) and the relevant "functional category" (e.g. Det in NP, AUX/INFL in a clause). This is because reference and predication are not conceptual primitives, and can be pragmatically decomposed, following suggestions by Langacker (1991). This accounts in a general way for the variation in the morphosyntactic locus for grammatical categories across languages.
However, it does not provide a means to decide between the functional category and the content word as the head of the expression. If we shift to a diachronic perspective, evidence from grammaticalization and syntactic reanalysis reveals an asymmetry in the behavior of the two head-like elements: the functional category gets grammaticalized while the content word does not, and the functional category can even be affixed to the content word. I argue that the content word, that is, the 'primary information bearing unit' (PIBU) is thus best considered as the head of a syntactic constituent.
The concept of markedness has played a significant role in certain linguistic theories, most importantly the Prague School and the Greenbergian school of typology. Greenberg explicitly acknowledges a debt to Trubetzkoy and Jakobson of the Prague School for the formulation of the theory of markedness. However, the concept of markedness underwent a dramatic change in Greenberg's hands that has no obvious antecedent in Prague School theory. Specifically, the Prague School used markedness as a asymmetry in the structural organization of the grammars of particular languages, but Greenberg redefined markedness as a universal asymmetry in the properties of functionally-defined grammatical categories.
The Prague school notion of markedness is a binary opposition between two values of a grammatical parameter, such as masculine-feminine. Trubetzkoy calls it a 'privative' opposition, and clearly defines it as a system-internal and hence language-specific phenomenon. Moreover, in examples from Principles of Phonology, Trubetzkoy suggests that one phonetic value could be marked in one context and unmarked in another. This of course fits with structuralist principles. Jakobson follows Trubetzkoy exactly in this respect; in fact, he argues that the same value can be simultaneously marked and unmarked in a single grammatical system. Unlike American structuralists, however, both Trubetzkoy and Jakobson discussed phonological typology and universals. But neither linguist tried to connect markedness of particular phonetic values with universals, because the former are system-internal.
Greenberg, in his monograph on language universals (1966), discusses the criteria used by Trubetzkoy and Jakobson for markedness, but then proceeds to link them to universals, without commenting on the inconsistency with structuralist principles. As a consequence, changes to the concept of markedness in typology change it almost beyond recognition as an outgrowth of Prague School markedness. Markedness becomes the property of a pattern of cross-linguistic variation, not a system-internal trait that applies to forms in particular languages. Greenberg dispenses almost immediately with binariness, allowing for grammatical hierarchies, as well as other more complex interactions. Finally, the criteria used by the Prague School and the criteria central to typological markedness demonstrates that the concept undergoes a major empirical change also. For the Prague School, neutralization is the basic criterion, though "zero sign" is a primary indicator. For Greenberg, textual frequency is the causal origin of (morphosyntactic) markedness, and the cluster of other properties that dovetail with textual frequency does not in general include neutralization (though it does include zero sign). In sum, the conceptual change in the concept of markedness is radical, reflecting the radical differences between structuralist and typological theory.
Functional analyses of grammatical phenomena, and the functionalist approaches that advocate them, are appealing to those who believe that an integrated view of language structure and language function is desirable. But functional analyses have been held to founder on basic grammatical facts that are taken to support the autonomy of grammar. The concept of autonomy is a complex one, and at least two different notions are found in current linguistic theory: arbitrariness and self-containedness. What these notions of autonomy is claimed to apply to varies as well: either to the syntactic component of the grammar, or (more recently) to the grammar itself, with respect to change, use, and acquisition. The arbitrariness of syntax must be accepted; and many functional analyses are compatible with self-containedness. However, mixed formal/functional analyses provide an argument against the self-containedness of syntax, and in fact even many formal theories of syntax accept non-self-containedness. The arbitrariness of grammatical knowledge must also be accepted; and many functional analyses of the dynamic process affecting grammar are compatible with self-containedness. An argument against the self-containedness of grammar comes not from these functional analyses but from sociolinguistics.
We examine a corpus of English oral narratives in order to explore the relation between intonation units (IUs) and grammatical units (GUs). IUs are almost always grammatical units (GUs) - the Full GU Condition - but the mapping is not one-to-one. We examine what causes a GU to be broken across IUs, and identify three major constraints: parallelism, syntactic complexity in general, and distance. The converse of these principles, closeness and simplicity, determine the assignment of two GUs to a single IU. We examine the consequences of our analysis for grammaticalization theory, which includes the evolution of syntactic closeness, and for construction grammar and related processing models, which relates syntactic simplicity inversely to token frequency. We propose the IU Storage Hypothesis to link together IUs, grammatical simplicity and high token frequency.
Kiyomi (1992) argues that neither animacy nor shape can be established as the primary semantic distinction in noun classification systems. However, if one distinguishes different types of classification systems according to the grammatical and ultimately semantic-pragmatic function - noun classes, numeral classifiers, possessive classifiers, and classifiers of spatial predication - then subtle but regular cross-linguistic generalizations can be made. There is a different hierarchy of semantic distinctions associated with each type of classifier. A number of apparently anomalous cases can be explained in terms of diachronic processes of grammaticalization, by which semantic distinctions found with one type of classifier are occasionally transferred to a construction characterized by another type of classifier. The distinctions found with each classifier can be accounted for by the function of the construction in which they are found.
This paper will discuss issues in the representation of grammatical voice based on a cognitive semantic analysis of verb meaning. A central part of this analysis is the hypothesis that an event encoded by a single verb form (as opposed to several verb forms chained together in some fashion; cf. Haiman and Thompson 1989) is conceived of as a self-contained event by the speaker (encoder). I will argue that other semantic properties that are associated with verbal voice and voice-like forms, in particular transitivity (Hopper and Thompson 1980; Givón 1981), and control (agentivity) and affectedness (Barber 1975; DeLancey 1984; Klaiman 1988) contribute to the conceptualization of an event as self-contained.
Various classifications of speech acts have been proposed in the philosophical literature, but no consensus has been arrived at. It is proposed here that cross-linguistic structural properties of utterances be used as evidence for an empirically justified classification of speech acts. I argue that structural properties of the sentence type as a whole should be used to define linguistic speech act categories, such as word order and/or relative position of the major constituents (e.g. word order inversion of English questions, and second-position modal/evidential markers), presence or absence of major constituents (e.g. the missing subject of the English imperative), and alterations of the semantic head of the sentence (i.e. the main verb and its inflections).
If we follow these structural criteria, then we find a pattern of three continua of sentence types, expressing knowledge, action and emotion, all rooted in the declarative type. The knowledge continuum extends from declarative to interrogative via epistemic/evidential modality and biased questions. The action continuum extends from declarative to imperative via deontic modals, optatives and hortatives. The emotion continuum extends from declarative to exclamative via emphatic and evaluative sentences. The continua represent degree of expected response from the addressee. Two further primary conceptual dimensions in the linguistic classification of speech acts are polarity (positive/negative) and politeness.
Metaphor and metonymy do not occur in isolation; they are triggered in utterances in particular linguistic (and extralinguistic) contexts. They pose an interesting problem from the point of view of semantic composition in that the metaphorical or metonymic interpretation of the parts (the individual words) appears to be determined by the interpretation of the whole construction in which they are found. Much of this is determined by the domain in which the words are to be interpreted. Domains play a central role in the definition of a metaphor as a mapping of conceptual structure from one domain to another. Domains also play a significant (though not defining) role in most metaphors and some related lexical ambiguities, as the highlighting of particular domains in a domain matrix. The processes of domain mapping and domain highlighting are governed by the requirement that a dependent predication (in the sense of Langacker 1987) and all of the autonomous predications it is dependent on must be interpreted in a single domain; this is 'the conceptual unity of domain.' This is only one of several 'conceptual unities' imposed by a whole construction on its component parts.
The functional-typological approach, and affiliated approaches, are sometimes portrayed as "radical" departures from recent approaches to linguistic analysis (usually, generative grammar), or even as something very new, usually in contrast to structuralism. While there is undoubtedly much that is new to the functional-typological approach to language, like any other intellectual movement it did not arise out of a vacuum and it owes a good deal to its predecessors. Likewise, its future is dependent on the forces that shape intellectual history. This paper describes some basic features of the functional-typological approach (as currently practiced largely in the United States, and as I perceive them) and situate them in the context of the history of American linguistics and broader intellectual currents as well.
The class of mental (or "psych") verbs, verbs of perception, cognition and emotion, pose difficulties for theories of argument linking because of variation in the linking of experiencer and stimulus. In some languages, mental predicates link the experiencer to the subject position, while in other languages, or even in different predicates in the same language, the experiencer is linked to a nonsubject position (object or oblique). This paper argues for a resolution of the problem in terms of causal or force-dynamic structure, such that the subject "acts on" the object in dynamic predicates.
Mental verbs can be divided into three causal-aspectual types: causatives such as remind and surprise; activities such as think, watch and mourn; and states such as see, know and like. Causative verbs consistently link experiencer to nonsubject position and stimulus to subject across languages, because they encode how a stimulus brings about a change in mental state in the experiencer. Activity verbs consistently link experiencer to subject position across languages, because they encode the mental activity and attention towards a stimulus. State verbs vary in their linking cross-linguistically because they are force-dynamically neutral.
Mental situation types display anomalous behavior because they are in fact force-dynamically bidirectional: the experiencer directs her/his attention to the stimulus but the stimulus causes a change in mental state of the experiencer. Activity verbs highlight the former direction of causation and causative verbs highlight the latter direction. Stative verbs highlight neither, or both at once. If so, then we predict that in some languages, both experiencer and stimulus will be encoded like subjects, or both like objects, or by either linking, with the alternative linkings construing the state as more experiencer- or stimulus-oriented. All of these predictions are borne out.
Seeking to develop the first satisfying unviersal definitions of basic grammatical units of the major syntactic categories such as "noun", "verb", and "adjective" and of the principal grammatical relations, "subject", "object" and "oblique", William Croft employs the methods of cross-linguistic or typological analysis and cognitive liguistics.
The typological method has demonstrated that certain grammatical criteria provide universal definitions of "marked" and "unmarked" grammatical categories. Drawing on evidence from a vide variety of languages, Croft shows that the traditional semantic definitions for the major syntactic categories (e.g. nouns denote persons and things) represent the unmarked correlation or "prototype" for the pragmatic functions of reference, predication and modification. Turning to recent work in cognitive linguistics and discourse analysis, Croft argues that the pragmatic functions represent universal ways of conceptualizing experience for the purpose of communicating information in language.
Croft then applies this hypothesis to grammatical relations between the verb and its dependent subject, object and oblique noun phrases, arguing that one cannot account for cross-linguistic patterns of marking using semantic roles such as "agent" or "instrument". He proposes instead an "idealized (prototypical) cognitive model" that human beings use in conceptualizing events based on causal relations among participants in the action, which largely determines the choice of subject, object and oblique forms. This model accounts for cross-linguistic patterns in the use of prepositions, postpositions, case affixes, verbal voice and other derived verb forms.
Evidence from the synchronic typology of markers of negation in declarative sentences is presented to demonstrate that negative existential predicates are a significant historical source of negative declarative markers. The argument is an example of the method of the dynamicization of a (synchronic) typology, which allows one to construct hypotheses of universal diachronic processes based on synchronic data from historically unrelated language types, a large but relatively untapped source of evidence for the diachronic or dynamic theory of language. This study also illustrates a number of issues in diachronic typological method and the evolution of related grammatical structures.
There are a great number of nouns, verbs and adjectives that do not fit the "normal" or prototypical characterization of those syntactic categories as objects, actions and properties respectively. Such words have more in common semantically with minor syntactic categories (so-called function words) and inflectional categories. In this paper, I present a conceptual framework for both prototypical and nonprototypical words and morphemes, based on the analysis in Syntactic Categories and Grammatical Relations.The prototypes for noun, verb and adjective are semantic classes combined with the propositional acts (Searle 1969) of reference, predication and modification. These three propositional acts can be characterized as the major propositional acts; but there are also three minor propositional acts: categorizing situating, and selecting. The nonprototypical words, function words and inflectional morphemes encode the minor propositional acts for objects, actions and possibly also properties (modification is clearly not a primary major propositional act).
Categorizing objects is achieved by animacy/gender classes and categorizing actions by valence (transitivity) and Aktionsart. Entities may be situated in a physical dimension - space for objects, via deixis, place expressions and spatial adpositions, and time for actions, via tense, temporal deixis and temporal expressions. Entities may also be situated in mental space (Fauconnier 1985) - accessibility or identifiability for referring expressions, via definiteness and referentiality (encoded by articles), and in mental space via epistemic, deontic and alethic modality, attitude marking, and speech act types. Selecting the particular objects referred to from the object type is achieved by countability/individuation inflections and constructions. Selecting the particular actions predicated from the action type is achieved by aspectual inflections and constructions. Properties are situated on a metric scale by gradability operators (intensifiers, downtowners) and by numerals. Properties are also selected by measure terms.
What is a possible verb? Unlike objects, events do not come easily individuated in human perception and cognition. I argue that there is an idealized cognitive model (ICM; Lakoff 1987) that human beings bring to the conceptualization and individuation of events. The idealized cognitive model for events is: (a) events are segments of the causal network; (b) events involve individuals acting on other individuals (transmission of force); (c) transmission of force is asymmetric (d) simple events are nonbranching causal chains. More specifically, possible verbs are defined in terms of a three-segment CAUSE-CHANGE-STATE chain, such that causatives profile all three segments of the chain, inchoatives the last two segments, and statives the final segment. A comparative study of zero vs. overt coding of causative, inchoative and stative expressions of the same event in 33 semantic classes in English, French, Japanese and Korean, reveals patterns of typologically marked and unmarked expression of events, for which explanations can be offered in terms of commonsense knowledge of the events in the world.
[Jerry Hobbs, William Croft, Todd Davies, Douglas Edwards, and Kenneth Laws]
In the TACITUS project for using commonsense knowledge in the understanding of texts about mechanical devices and their failures, we have been developing various commonsense theories that are needed to mediate between the way we talk about the behavior of such devices and causal models of their operation. Of central importance in this effort is the axiomatization of what might be called "commonsense metaphysics". This includes a number of areas that figure in virtually every domain of discourse, such as granularity, scales, time, space, material, physical objects, shape, causality, functionality, and force. Our effort has been to construct core theories of each of these areas, and then to define, or at least characterize, a large number of lexical items in temrs provided by the core theories. In this paper we discuss our methodological principles and describe the key ideas in the various domains we are investigating.
The representations of adjectives and their adverbial counterparts in logical form raises a number of issues in the relation of (morpho)syntax to semantics, as well as more specific problems of lexical and grammatical analysis. This paper addresses those issues which have bearing on the relation of properties to events. It is argued that attributes and context play only an indirect role in the relation between properties and events. The body of the paper addresses the criteria for relating surface forms to logical form representations, and offers a unified analysis of adjectives and their adverbial counterparts in logical form while maintaining a clear distinction between operators and predicates; this requires the postulation of a factive sentential operator and the relaxation of the one-to-one syntax-semantics correspondence hypothesis. Criteria for determining the number of arguments for a predicate are established and are used for the analyses of phenomena such as passive-sensitivity, lexical derivational patterns, and gradability.