Drawing insights from cognitive psychology, neurology, and conversation analysis, Alexander Haselow offers in this book an emergentist approach to grammar. The two dominant approaches to syntactic theory, Saussurean and American structuralism, and Chomskyan generative grammar, whose grammar conception is composed of canonical patterns, have long been focusing on internal structures of pure but often isolated sentences. Linguistic structures are treated as finished products in fixed-coded rules and decontextualized configurations. Consequently grammatical analysis is to integrate different elements into unified abstract superstructures, focusing on morphosyntactic and semantic dependency relations with other constituents. Without reference to cognitive constraints and interactive features, the relatively static and monolithic knowledge system, which cannot be reconciled with cognitive, psycholinguistic and neurolinguistic findings on syntactic processing, is largely inappropriate for describing structures produced in spontaneous speech.
Given the human parallel ability to process linear sequences without hierarchization like [ABCD] and hierarchically structured linear sequences like [A[B[CD]]], Haselow argues that language structure is not fixed but continually reshaped by adapting established patterns in conversational interactions. Grammar under this view is a dynamic structuring system encompassing two domains, in which linguistic structures are incrementally created in the time flow of spontaneous speech. Language therefore is a complex adaptive system arisen from speakers’ mental organization, and a speaker’s grammatical knowledge is not reduced to well-formed syntactic units, but is the ability to continually adapt an emergent structure to the local cognition contingencies (such as perception, attention, and working memory) and social interaction. Taking authentic spoken English from a corpus as observation objects, Haselow advocates that grammar, especially syntax, rests upon two interactive cognitive principles of linearization – microgrammar and macrogrammar – hence the Grammatical Dualism Assumption (GDA), a dualistic approach characterizing properties such as temporality, dialogicity, and situation-dependency of linguistic structures.
In line with the basic principles of structuration theory (Giddens Reference Giddens1984) and integrationalism (Harris Reference Harris2012), both positing that language structure is the result of interaction between social action and cognition, Haselow proposes an interfield approach as metatheoretical frame to integrate grammatical analysis, conversation analysis, and (neuro)cognition of language. Specifically, grammatical analysis probes into internal organization of linguistic structure, whereas conversation analysis deals with social action in concrete context. The (neuro)cognition of language contributes to the tools needed to explain cognitive constraints, mental organization, and neural representations of language production and processing in real time. The three parts all contribute some particular meanings or functions to spontaneous speech, and grammar consequently cannot be forced into one uniform analytic apparatus or a static linguistic object. This naturally suggests a shift in syntactic theory from a static to a dynamic perspective, ‘from the language system as an entity isolated from the speaker to the interaction between the system and the speaker’s creativity’ (48).
Since grammar is both a knowledge system and an emerging activity, grammatical description consequently should not be merely production-based, but knowledge-based and activity-based, which motivates Haselow to advocate GDA. Accordingly, grammar as a dualistic organization and creation of linguistic structure in spontaneous speech is guided by two principles. To be specific, microgrammar is based on linearization tied to hierarchization. It deals with the internal organization of emergent syntactic units, exhibiting asymmetric and binary morphosyntactic properties. Without morphosyntactic and semantic embedding, macrogrammar is based on linearity without hierarchization. It involves the organization of cognition (speech planning discourse), text organization (textual coherence and sequential embeddedness), and interaction (management of speaker–addressee relationships).Typical macrogrammar expressions can be distinguished from items with or without a lexical etymology (e.g. anyway, you know or oh, wow), simple and complex internal structure (e.g. well or if you want), autonomous or bound (e.g. right or then). However, such dualism assumption does not mean that macrogrammar is superimposed upon microgrammar, but assumes that both operate in parallel in producing structures. Microgrammar encompasses principles for hierarchical creations with propositional structures, while macrogrammar establishes relations between microgrammatical segments, and relations between emergent structures and cognitive, discourse-organizational and interactional aspects.
Haselow then provides an in-depth analysis of two magnifications of macrogrammar – field and linearization. Fields are gradually emerging places for fixed structures, whose formats, however, are not defined prior to the act of speaking. Three kinds of fields, all syntactically not integrated into microgrammatical or hierarchical relationships, are identified and specified in detail. The initial field, the production of a talk unit such as initial adverbs, discourse markers, or interjection, is pragmatically incomplete and morphosyntactically autonomous. The function is to serve as addressee-oriented processing or attention management, but cognitively it projects an upcoming structure. The medial field, in which dysfluencies such as hesitations, repetition and repair phenomena occur frequently, selects a suitable structural format from possibilities, reflecting cognitive loads on planning processes at the conceptual or syntactic level. The expressions in this field ‘serve different functions, such as smoothing out dysfluencies and indicating different kinds of ongoing speech planning activity’ (156). In the final field, the structural units such as final particles, tag questions or general extenders allow speakers to deal with a variety of communicative tasks before potential talk transition or continuation with a new idea. The function is to balance speaking and planning, integrate segments into a coherent talk unit, and maintain contact with the addressee, but without illocutionary force or receiving semantic properties within the unit.
A second manifestation of macrogrammar is the linearization of structural segments, or the syntactic design and relation of microgrammatical segments. Spontaneous speech, due to lack of planning time, is structured for maximally efficient processing in unidirectional flow. Drawing on the processing efficiency principle (Hawkins Reference Hawkins2004), Haselow identifies three processing principles, motivating three structural formats of talk units in spontaneous speech. Firstly, minimal structures or situational ellipses may reduce cognitive efforts and mental activation costs by minimizing forms and domains, but may result in ambiguous interpretations if context and co-text do not provide sufficiently clear cues. Secondly, segment-chaining structure may minimize the number of hierarchical structure relations through information distributions over smaller syntactic segments. This allows for rapid production and interpretation, but cannot anticipatethe ultimate structural, semantic, and pragmatic representation. Finally, far-reaching projections may maximize real-time processing by providing and anticipating the ultimate representations of a unit, but may take a risk of working memory overload from temporarily storing the projection and domain size. The three formats are based on ‘mere linearization’ (103), suggesting that microgrammar and macrogrammar are interactive in shaping syntactic configurations in spontaneous speech.
Haselow finally turns to some neurological evidence for GDA by observing correspondences of microgrammar and macrogrammar. Firstly, novel speech is involved in left hemisphere (LH) activity, and formulaic speech is mainly associated with the right hemisphere (RH). Secondly, LH-damaged patients tend to produce speech automatisms and formulaic speech, showing impairments of fully generating predicative hierarchization and semantic compositionality, while the domain communication deficit for RH impairments is discourse structure, pragmatic aspects, and organizations of speaker–addressee interaction. Both brain domains and neural activities in linguistic processing, Haselow concludes, strongly suggest that language is not processed in a uniform way, but microgrammar and macrogrammar are processed in different areas, which provide direct evidence for GDA.
Summarily, GDA pulls syntacticians back to focus on observed spontaneous language rather than merely on grammaticality judgments. It further broadens perspectives of linguistic analysis by considering not only the final products, but also the linguistic structures involving cognition and communicative tasks in dynamic conversational contexts. An integrated approach to grammatical analysis, given the pervasive scientific compartmentalism in academic research, is necessary. This book provides a framework to incorporate spoken varieties into the integral parts of language structure, different from the ‘purely’ sentence-based descriptions of language structures, in which cognitively and interactionally grounded structural features are often ‘neglected, deleted, marginalized, or even thrown into the pragmatics waste-box’ (294). For example, the exclusion of various adverbs, which, according to Haselow, are macrogrammar, would lead to inadequate description since macrogrammar is a part of grammatical knowledge. To fully grasp a sentence or interpersonal communication therefore necessarily requires the inclusion of macrogrammar.
Though the scope of the book, as the title clearly defines, is confined to spoken English, its general conclusions can be safely extended to other languages. Without any doubt macrogrammar exists pervasively in spontaneous speech in human communication, and the book from this perspective is of great theoretical importance for further research in both grammatical analyses and typological generalizations. More importantly, such extension is necessary for constructing linguistic theories, if some more powerful and explanatory principles, guided by Occam’s razor, are to be constructed and developed. If an extension is made, some problems and challenges may come out immediately, however.
One challenge is the permeable boundaries between microgrammar and macrogrammar in terms of hierarchical relations and illocutionary forces, which are empirically inadequate operational criteria even in spoken English. Actually Haselow states explicitly that the two ‘are not strictly separated in concrete speech activities’ (260), but should ‘be conceived of as a continuum’ or ‘as gradient rather than discrete’ (35, 117–118). When extended to other languages, the prospects may appear to be quite different. For example, different final particles in Chinese, which according to Haselow’s definition are macrogrammar, may change sentence meanings dramatically. Such particles are organized as a hierarchy in the sense that their syntactic positions cannot be alternated freely when two or more particles are clustered (Paul Reference Paul2015). Similarly, Cinque (Reference Cinque1999: 106) proposes a hierarchical relation and projection of various adverbs, modeling the multidimensional nature of meaning representations, which is now cross-linguistically verified. A third piece of evidence is from discourse agreement, such as politeness markers in Japanese. More obviously, in Souletin, an eastern dialect of Basque, a sentence with the same propositional meaning when uttered to different addresses – a male or a female friend, a social superior or more than one addressee – is marked with differential locutive agreements (Miyagawa Reference Miyagawa2017). Such grammatical agreements, which according to Haselow’s definition are macrogrammar, play an important role in driving syntactic operations, indicating that they may be a different part of microgrammar.
Taking another perspective, it is brodadly acknowledged that language processing, according to extensive body of neurolinguistic studies, correlates with both LH and RH activities. However, the neurolinguistic evidence, even leaving aside some uncontrolled factors such as interpersonal performance variations and assessment focus, is not robust enough to support GDA, as Haselow admits that ‘the generalizations are to be taken with care’ (262). The corresponding areas of the human brain may seem to ‘have evolved to subserve the human capacity to process syntax, which is at the core of the human language faculty’ (Friederici Reference Friederici2017: 219). This means that LH and RH may share the combinatorial principle but differ in the building blocks that they employ in the process of constructing structures.
Syntax is dynamic per se when it is regarded as a process from lexical items to structured representations. Adopting the current syntactic operation of Merge, the single and basic linguistic design which is argued to systematically combine two syntactic elements into hierarchical structures (Berwick & Chomsky Reference Berwick and Chomsky2016), we can envision that all syntactic constructions, whether microgrammar or macrogrammar, are the externalizations of Merge. For example, sedimentation, the product of ‘long-term process that derives from the routinization of particular action patterns’ (19), essentially is the product of Merge at least in the beginning, but some meanings are bleached during language use and change. Thus, the book provides a provocative and innovative prospect to probe the relations between macrogrammar and microgrammar for further research.
