Linear Unit Grammar: Integrating Speech and Writing

September 2007 — Volume 11, Number 2

 

Linear Unit Grammar: Integrating Speech and Writing
Author:	John McH. Sinclair & Anna Mauranen (2006)
Publisher:	Amsterdam: John Benjamins
Pages	ISBN	Price
Pp. xxi + 185	ISBN 978 90 272 2299 2 (paper)	€33.00 euros; $39.95 USD

Linear Unit Grammar (LUG), as described in this book of the same title, is an attempt to overcome the limitations of hierarchical structures when describing utterances, especially spoken ones. Instead of the all so familiar tree diagrams, utterances are viewed as a linear sequence of chunks, phraseological units that perform certain functions. LUG follows the work by Brazil (1995), which introduced a linear finite state model for the analysis of spoken English, and also by Hunston and Francis (2000) on pattern grammar, which postulates that words are typically embedded in contexts (described as grammatical patterns).

One important distinction between LUG and its predecessors is that LUG completely abandons the traditional word classes or even phrase structure labels. There are no noun phrases to be found in the analysis. Instead, LUG assigns functional categories to the chunks which make up an utterance. While Sinclair and Mauranen initially use LUG to describe spoken language, it works equally well with written texts, as demonstrated by later examples.

Chunks are central to the analysis, but the authors decline to provide an explicit definition. Instead they state that human readers will identify them intuitively; and different people will choose more or less the same chunks when presented with a text. This vagueness might sound like heresy to people working in syntax, but does in fact nicely fit into a language model that is based on individual speakers rather than a shared common language system.

Sinclair and Mauranen present two basic types of chunks, those that organise the discourse and those that transmit the message or content. More specific sub-types also exist: The organisation chunks, for example, are subdivided into “OI” (interactive) and “OT” (text-oriented). In a dialogue, an utterance such as “oh dear," or “yes” would be an OI, whereas “because” and “first of all” would be examples of OT chunks. The message chunks are more varied: some of the subtypes mentioned are “MF” (message fragments), “M-” (incomplete), “+M” (completion of an “M-” chunk), and “MS” (message supplements, roughly adjuncts in traditional terminology). Each chunk gets assigned one of these labels.

The book is divided into three main parts: preliminaries, analysis, and theory. Each of these comprises about 50 pages, though the second is the core part.

The analysis itself consists of 5 steps, each of which is covered in its own chapter. First, the text is chunked, that is, split into the units of analysis. These are then assigned their basic type (“O” or “M”), and after that each of the respective types is further refined by the right sub-type (in separate steps for O and M chunks). The final step is the synthesis, where a set of rules is applied to ‘clean up’ the utterance. This is mainly achieved by discarding the interactive OI chunks from the input, and combining incomplete message fragments, thus getting larger units which are more self-standing.

Since this is a rather abstract description, the following example will clarify the procedure (which is quite straightforward). This extract from a conversation is analysed by Sinclair and Mauranen (p. 151), and is shown here with the chunks/unit boundaries marked:

I wondered / what happens / when you go / from one island / to the other / no / the train goes / on the ferry / oh / I see / yes

Chunk types are assigned as follows:

I wondered (M-) / what happens (+M-) / when you go (+M-) / _from one island (+M-) / to the other (+M) / no (OI) / the train goes_ (M-) /
on the ferry (+M) / oh (OI) / I see (OI) / yes (OI)

The OI elements are discarded, and the M- chunks are matched up with the +M
ones (including the incomplete ‘middle’ chunks marked as “+M-“). The final
result consists of two larger units:

I wondered what happens when you go from one island to the other / the train goes on the ferry

This last step is rather disappointing. After LUG has been compared to other grammatical formalisms in the first part of the book, one starts reading the analysis part with high hopes, expecting some radical new approach to the description of sentence structure. Instead the outcome is simply a cleaned-up version of the input that “should be able to be handled in a normal grammar” (p. 103). So the whole purpose of LUG (as developed so far) simply seems to be a pre-processing step for a conventional syntactic analysis. All that is gained by the analysis is some insight about the organisation of the utterance, but that is merely a by-product, relegated to “notes and comments on aspects of the interaction which contribute to the incremental meaning” (p. 91). I would have expected a more detailed description, one that goes beyond being a mere filter that turns messy spoken dialogue into ‘grammatically well-formed’ sentences.

However, departing from the traditional hierarchical-grammar paradigm is a bold step, and LUG is a very recent development which still needs to be developed further. In light of this, one cannot expect that all aspects of linear grammar will be covered, especially in such a small volume. The main focus of the book then is not so much introducing a fully-fledged system for describing the syntactic structure of utterances in great detail, but rather to argue that the vast majority of authentic utterances are unsuitable for being described using traditional grammars, to further support the idea of linearity as a principle of organisation, and to show how LUG works by way of analysing a number of texts from various sources.

This last purpose, in fact, is one of the book’s strong points: it is full of examples explaining how and why the different chunks are assigned their particular labels. Through this ‘gentle’ approach to an otherwise difficult topic the book almost becomes feasible as an introductory textbook. I can imagine using it in a module on linguistic theory on undergraduate level, or even perhaps as a replacement for the traditional Subject/Predicate/Object/Complement/Adjunct (SPOCA) analysis taught in introductory modules. Students will have fewer difficulties understanding the purpose of the analysis, as the labels are based on discourse function rather than on the often inadequate word class system. Also, LUG should work better on authentic texts, as it is less rigid than traditional grammars.

As it is now, LUG is an intriguing alternative to more traditional approaches to grammar, but there is still a lot of further work that needs to be done to establish it as a major formalism. From a methodological point of view, the underspecified nature of the chunks (which are, after all, the central units of analysis in LUG) seems initially unsatisfactory, but from a theoretical point of view they can be justified by looking at language as a property of individuals. Linearity as an organising principle makes a lot of sense, since language is after all a sequential stream of either sounds or letters. It is the final step, the synthesis, which does not quite seem to reach the ultimate goal of capturing the structure of the utterance and the relationship between its constituent elements.

Despite this short-fall in my expectations I recommend this book as an introduction to a different paradigm: linearity instead of hierarchy. Sinclair and Mauranen are detailed enough in their analytical account to allow the reader to follow the procedure with ease, and it should not be difficult to apply linear analysis to other texts. Furthermore, they establish the notion of larger phraseological units of meaning which are used in grammatical analysis, especially units which in themselves are not grammatically well-formed in the traditional sense. Even if the book does not convince the reader to fully abandon traditional grammar, it still makes one think more about discourse elements and their organisation. As such it fulfills the main objective of its authors, “to bridge the gap between text and grammar by introducing a new point of departure for grammatical analysis” (p. 103).

References

Brazil, D. (1995). A grammar of speech. Oxford: OUP.

Hunston, S. and Francis, G. (2000). Pattern grammar. Amsterdam:
Benjamins.

Oliver Mason
University of Birmingham, UK
<O.Masonbham.ac.uk>

© Copyright rests with authors. Please cite TESL-EJ appropriately. Editor’s Note: The HTML version contains no page numbers. Please use the PDF version of this article for citations.