Susan Hunston’s Pattern, Construction, System: A Unified Approach to Grammar and Lexis provides an ambitious and in-depth account of verb complementation patterns from Pattern Grammar (PG), organizes them into constructions following Construction Grammar (CG), and then, drawing from concepts in Systemic Functional Grammar (SFG), arranges them into semantic networks based on their process meaning type, offering a coherent, usage-based framework for understanding how patterns, constructions and systems interact in language.

Chapter 1, ‘Introduction’ (pp. 1–10), sets out the theoretical and methodological foundations for this integration, situating each framework in its historical and functional context. While each tradition has developed independently, all three are usage-based and treat grammar and lexis as a continuum, rather than as discrete syntactic slots filled by lexical items. Accordingly, Hunston explains, the ‘job of the grammarian’ differs across frameworks: in PG, it is to observe large corpora to identify recurrent grammatical environments for verbs, nouns and adjectives (Francis et al. Reference Francis, Hunston and Manning1996, Reference Francis, Hunston and Manning1998); in CG, it is to model the network of form–meaning pairings that capture speakers’ mental representations of language, from highly idiosyncratic expressions to more regular sentence patterns (Goldberg Reference Goldberg2006); and in SFG, it is to model language as a system of choices that communities of speakers draw on to make meaning in social interaction (Halliday & Matthiessen Reference Halliday and Matthiessen2014). Hunston brings together language as both a mental and a social phenomenon, showing that PG, a corpus-driven approach for discovering recurrent patterns linked to words, can dialogue with SFG and CG without privileging one framework over another in order to illuminate how language works.

In chapter 2, ‘Background to Pattern Grammar and Construction Grammar’ (pp. 11–24), Hunston outlines the historical background to the Collins COBUILD English Dictionary project. Drawing on corpus-based insights from John Sinclair and further developed by Gill Francis, the project showed that differences in word senses are associated with differences in the lexical and grammatical patterns in which words occur. Focusing on verb patterns, Hunston notes that early work on the COBUILD project identified 152 verb complementation patterns (Francis et al. Reference Francis, Hunston and Manning1996), although many were rare or could be subsumed under broader patterns. As a result, only 87 were treated as core patterns and described in dedicated sections of the reference grammar. These fall into five broad types, including those where the verb is followed by a single element, realized as a phrase or clause (e.g. verb + noun phrase ‘V n’, verb + that-clause ‘V that’, verb + -ing clause ‘V -ing’, as in eat an apple, say that he would come, go swimming) or those that are followed by a prepositional phrase (e.g. ‘V at n’, ‘V into n’). Verbs sharing a complementation pattern were then grouped together, revealing that these tend to also share aspects of meaning. For instance, the ‘V on n’ pattern yielded over thirty meaning groups, including a ‘comment on’ group (comment, lecture, preach, write) and a ‘depend on’ group (depend, count, rely). Hunston acknowledges criticisms of PG, particularly its limited engagement with semantic theory, and notes that the meaning groups represent ‘something of a compromise between comprehensive coverage and manageable proportion’ (p. 14). While balancing scope against granularity, the resource remains comprehensive, coding every sense of each word for its grammatical patterns.

Hunston then explains CG, focusing on the accounts by Goldberg (Reference Goldberg2006), Hilpert (Reference Hilpert2014) and Haspelmath (Reference Haspelmath2023), leading her to characterize a construction as a conventional schema which includes at least one open slot fillable by any number of expressions from the same form-class. The inclusion of the open slot rules out both constructions which have none, such as single words and fixed phrases, and those which are based on broad generalizations, such as mood or tense. Constructions are not overtly learned – rather, they are abstractions that can productively generate different instances (i.e. they are not themselves instances) – nor does their meaning depend on the meaning of the verbs used. For example, the ‘caused motion’ construction (‘verb something somewhere’), as in He pushed the box into the corner, can be creatively extended to include Goldberg’s (Reference Goldberg1995) oft-cited example He sneezed the napkin off the table. At the same time, Hunston cites Boas (Reference Boas2008) in noting limitations of constructions, using his example of He talked himself blue in the face, where it is not possible to substitute spoke for talked. This leads to the argument of drawing on corpus studies to determine the range of words that are actually used to fill in slots, as well as their relative frequency, thus providing evidence for constructions based on usage. This evidence also helps clarify the prototypical meaning of a given construction. CG researchers further inventory constructions in a given language together with a representation of the relationships between them, creating what are termed constructicons. Hunston highlights Perek & Patten’s (Reference Perek and Patten2019) project, which combines Ruppenhofer et al.’s (Reference Ruppenhofer, Ellsworth, Miriam, Johnson, Baker and Scheffczyk2016) FrameNet (a linguistic resource that uses corpus evidence to elucidate word meanings as structured ‘frames’ with their participants and relationships) and PG, through which semantic information from FrameNet is added to the patterns. She concludes the chapter by suggesting that PG can be ‘placed at the service of Construction Grammar’ (p. 24) to provide a comprehensive method for identifying constructions and representing how they are interconnected.

In chapter 3, ‘From patterns to constructions’ (pp. 25–56), Hunston reinterprets 54 verb complementation patterns as constructions, and subsequently represents those as taxonomies or networks. A single grammatical pattern (such as ‘V into n’) can realize multiple different constructions, which thus cannot be identified solely at the level of form. Hunston proposes a hierarchical approach, in which more specific constructions (e.g. bang, smash, crash on the one hand, and blossom, change, turn on the other) are grouped under more general superordinate constructions (e.g. the ‘crash’ group and the ‘turn’ group, respectively). At the same time, Hunston shows that determining the appropriate level of generality remains problematic, requiring a compromise between capturing semantic distinctions and maintaining a workable classification. Hunston then extends the grouping by representing constructions as networks, in which constructions derived from a single pattern are related along multiple dimensions of similarity and difference. She illustrates the process of moving from pattern to construction to network using the ‘V n adj’ pattern, which yields 17 proposed constructions. Hunston demonstrates how these constructions can be represented through a left-to-right system network (as in SFG), where choices are organized sequentially: ‘V n adj’ first divides into Construal or Causation, and then, for example within Construal, into Cognition or Communication, and within Communication further into more specific constructions. Thus, networks can be extended to greater levels of delicacy, allowing increasingly detailed distinctions to be drawn between closely related constructions. At the same time, she notes a key difficulty of working with networks: fuzzy boundaries must often be overridden, as a construction cannot be represented as ‘partly this and partly that’ (p. 45), even though overlaps in meaning can and do occur. This also highlights the partly interpretive nature of constructional networks, since different configurations may emerge depending on which semantic distinctions the analyst treats as most salient.

In chapter 4, ‘Semantic role annotation’ (pp. 57–84), Hunston compares four approaches that assign semantic roles to the formal elements of constructions: Corpus Pattern Analysis (CPA, Hanks Reference Hanks2013), SFG (Halliday & Matthiessen Reference Halliday and Matthiessen2014), Local Grammar (LG, Barnbrook Reference Barnbrook2002) and FrameNet (Fillmore et al. Reference Fillmore, Johnson and Miriam2003). These approaches differ in whether semantic labels function as input to identifying constructions or emerge as output from that process, as well as in their level of specificity, from more fine-grained, such as LG’s rhetorically oriented labels, to more generalized ones. Hunston relates these differences to their underlying aims; for example, CPA focuses on distinguishing word senses, whereas SFG models broader meaning-making resources. These differences in purpose also makes it difficult to align terminology across approaches. She explains and illustrates each framework, noting the challenges of applying them to her constructions given their differing purposes and points of departure. She highlights how these considerations, such as FrameNet’s relating semantic labels to semantic fields and SFG’s process types, led to the semantic fields she identified as well as to the participant roles associated with each of the field types. Hunston further comments on examples of double coding in the case of constructions that can belong to more than one field.

Given the role of SFG in inspiring the annotation scheme for labelling the constructions and for representing them through networks, Hunston dedicates chapter 5 to ‘Background to Systemic Functional Grammar’ (pp. 85–104). She elaborates more fully on the system of transitivity, elucidating its six process types (Material, Mental, Relational, Verbal, Behavioural and Existential) and their identification through grammatical criteria and types of participant(s). She highlights the paradigmatic focus of SFG in its modelling of language as a set of meaning resources through system networks. She notes that while system networks impose discrete either/or distinctions, many linguistic categories are in fact continuous and overlapping. Thus, they could be more accurately represented topologically as gradients rather than fixed types. However, she points out that she chose the typological path for her purposes. Hunston concludes chapter 5 by situating lexis within SFG as part of a continuum from grammar to lexis, while emphasizing the difficulty of modelling this relationship in practice. She surveys a range of approaches to this problem and introduces a distinction between Meaning Networks and Systemic Networks, arguing that both are needed to account for the relationship between lexis and grammar.

The constructions established in chapters 3 and 4 form the basis for chapters 6, 7 and 8, where Hunston organizes them into networks to explore semantic fields. In chapter 6, ‘Construction in Meaning Networks: Causation’ (pp. 105–30), she demonstrates how constructions are grouped into Meaning Networks and then related to a complementary Systemic Network. Drawing on 105 Verb Argument Constructions derived from 26 verb complementation patterns, Hunston groups these into broad semantic categories (e.g. causing a state, a thought or emotion, or an action or event), and then progressively differentiates them into more specific meaning classes. For example, the ‘Cause thought/emotion Meaning Network’ is subdivided into ‘cause change in thought’ and ‘cause change in emotion’, which can be further distinguished according to features such as the presence or absence of a Cogniser and its grammatical realization. In this way, the network develops as a taxonomy of constructions, organized by increasingly fine-grained semantic distinctions, down to delicate patterns such as V n (Cog) (‘convince someone’). She contrasts the detailed, taxonomic Meaning Networks with the Systemic Network, which represents causation as combinations of features across independent dimensions, providing a compact summary of ‘what matters’ when causation is expressed in English.

Chapters 7 and 8 apply the methodology established in chapter 6 to additional semantic fields. Chapter 7, ‘Semantic fields: Cognition and communication’ (pp. 131–77), examines constructions relating to situations mediated through a thinker, perceiver or communicator. Hunston situates her approach in relation to prior work, including Halliday & Matthiessen’s (Reference Halliday and Matthiessen2014) Verbal and Mental processes, Semino & Short’s (Reference Semino and Short2004) categorization of speech, writing and thought, and Martin & White’s (Reference Martin and Peter2005) Engagement system. She then organizes the relevant Verb Argument Constructions into Meaning Networks and complementary Systemic Networks for each subfield, illustrating how constructions encode distinctions such as thought, perception, emotion and communication. Hunston also notes complications introduced by abstract and non-human subjects, which can be reinterpreted through SFG’s notion of grammatical metaphor.

Chapter 8, ‘Semantic fields: Material and relational’ (pp. 178–219), follows the same procedure for Material processes (of which there are four types: Change, Creation, Location Change, Possession Transfer) and Relational processes (with two types: Equivalence, Logical Relations). For each field, Hunston identifies and organizes the relevant constructions into Meaning and Systemic Networks, highlighting correspondences and misalignments with established SFG process types. She emphasizes that while the methodology is systematic, overlap between fields occurs, reflecting both the continuum of meanings in language and the interpretive nature of network construction.

In chapter 9, ‘Discussion and conclusion’ (pp. 220–37), Hunston provides a succinct summary of the previous chapters and considers what her study draws from and contributes to CG and SFG. With respect to CG, Hunston points out that PG’s basis in corpus lexicography allows for a more comprehensive study of constructions in terms of both number and type, including the compositional, as the literature on CG tends to focus on the non-compositional. Hunston also notes, that, while the limitation of her study is to Verb Argument Constructions, it could be extended to Adjective and Noun Argument Constructions. This leads to the suggestion that constructions can be usefully grouped along a spectrum from highly abstract, grammar-like patterns to highly specific, idiomatic expressions, with Verb, Noun and Adjective Argument Constructions occupying an intermediate position. Hunston emphasizes that this classification is analytical rather than hierarchical, i.e. no type is more central or privileged than another. At the same time, recognizing these intermediate types helps to organize the network of constructions in a language and makes it easier to identify recurring patterns, supporting a more precise and usage-informed approach to CG. This issue of categorization and/or a cline of constructions leads her to contemplate the modelling of a constructicon. Her study shows that both ‘form to meaning’ and ‘meaning to form’ networks are needed to represent the relationships between constructions more fully. Future directions include adding quantitative analyses, exploring diachronic and cross-linguistic applications, and reinterpreting PG concepts as sequences of constructions to refine and expand the constructicon.

Hunston then turns to SFG, highlighting its contributions to her study. For example, she applies the concept of network in three different ways: taxonomic networks show relationships between constructions that follow the same verb complementation pattern, semantic-field networks show relationships among constructions contributing to the same meaning domain, and Systemic Networks represent hierarchically the features that make up constructions within each field. She also remarks on the adaptation of SFG’s process types and grammatical metaphor to highlight distinctions among constructions. While borrowing concepts from SFG, Hunston also signals her departures from it. Her descriptive, corpus-based approach, for example, treats prepositional phrases as integral to verbs and annotated as Participants rather than as Circumstances. It also allows constructions to be coded under multiple semantic fields, rather than assigning a single function or analytic interpretation. As to contributions to SFG, she contemplates placing the construction rather than the word at the lexis end of the grammar–lexis continuum. Whereas traditional SFG analyses assign semantic features to verbs, Hunston shows that meaning often emerges from the combination of verb, complement and pattern, as in constructions like ‘cook food’ or ‘make a creative artefact’. This approach allows general verbs such as make or create to take on specific meanings depending on the construction, accommodates distinctions between physical and abstract entities and participant structures, and supports a usage-informed, corpus-based, semantic-field-oriented perspective. By focusing on constructions as meaningful units, Hunston bridges SFG and CG, offering a more flexible and comprehensive view of lexicogrammar.

This volume is highly welcome precisely because it so effectively brings together different linguistic traditions. As such, it serves as inspiration for linguists to go beyond their disciplinary comfort zone and to consider how different approaches can dialogue with each other to deepen our understanding of language as both a meaning-making resource and a system shaped through learning and use. Hunston’s engaging and clear style helps make readily achievable how such a dialogue can be structured and carried out. Furthermore, Hunston writes with her readers firmly in mind; the volume includes a wealth of examples and citations to the literature and is thoughtfully organized. For example, whereas chapter 6 derives the Systemic Network from previously established Meaning Networks, chapter 7 reverses this perspective, introducing the Systemic Network for each subfield before demonstrating its realization in constructions, subsequently organized into Meaning Networks. This careful reorganization underscores the soundness of Hunston’s overall design, reflecting a transition from methodological description to analytical application, foregrounding the Systemic Network as the primary organizing framework. Each chapter ends with a useful summary and a logical transition to the next.

A couple of ideas emerged from reading the text that caused some pause for further consideration. Hunston does well to comment on fuzzy boundaries when networks call for discrete ones. This recognition that linguistic categories are often gradient rather than discrete helps to explain why multiple network configurations are possible; however, it also underscores the extent to which such representations rely on analyst-imposed boundaries, potentially weakening their perceived objectivity. The reader might wonder at times whether the categories and networks would look at all similar if another analyst had carried out the study. At the same time, this flexibility can also be seen as a strength: by foregrounding the interpretive dimension of analysis while grounding her proposals in corpus-derived patterns, Hunston offers a model that is both empirically anchored and adaptable, reflecting the gradient relationship between meaning and form in language.

From an SFG perspective, I was particularly interested in Hunston’s treatment of the grammar–lexis continuum. In chapter 5, Hunston reviews some key approaches on the lexis end of the continuum. She includes Hasan’s (Reference Hasan, Cloran, Butt and Williams2015) ‘grammarian’s dream’, which demonstrates that lexis can be modelled systemically through networks of semantic features, although its labour-intensive nature limits its feasibility. Matthiessen (Reference Matthiessen2014) develops this further by linking Levin’s (Reference Levin1993) verb classes to SFG process types, thereby focusing on the mid-range of delicacy between grammar and lexis. Notably, as the book illustrates, drawing on Matthiessen (Reference Matthiessen2014), Hunston introduces a distinction between one-dimensional Meaning Networks, which capture semantic distinctions as realized in constructions, and multi-dimensional Systemic Networks, which represent the interrelated choices within the system underlying the constructions. Hunston does not explicitly link these networks to Fontaine’s (Reference Fontaine2017) lexeme–lexical item distinction, but one might tentatively interpret the relationship as follows: Meaning Networks could be seen as reflecting the lexical item, the realized instance of a word in context, capturing constructions as they occur in use, whereas Systemic Networks might correspond to the lexeme, representing the abstract paradigmatic potential underlying those constructions. Making this connection more explicit would have been a valuable step, as it would further illuminate how constructions instantiated in Meaning Networks could provide a corpus-informed bridge between the system-like grammar end and the instance-rich lexical end of the lexicogrammar, addressing Fontaine’s concerns about modelling lexis within SFG.

These reflective pauses highlight the richness of Hunston’s volume, showing how patterns, constructions, and networks open new avenues for research in corpus- and usage-based approaches that link language form with its function in human cognition and interaction. Its accessibility as an open-source resource makes this an especially valuable contribution for both teaching and research.