A Rank Scale That Is Not A Rank Scale

 Fawcett (2010: 189-90):

One way in which a clause may be said to be a 'higher' unit than the others — as the 'rank scale' concept predicts it to be — is the fact that the unit that is shown as the 'highest' on the page in any tree diagram representation of a text-sentence is always the clause.¹ In a generative SF grammar, provision for this fact is built into the probabilities on the features in an early system in the overall network, so ensuring that a clause is always generated first. However, we may note that the informal summary of clause syntax for text analysis purposes shown in Appendix B shows that a clause is the only unit that can fill the "element" of "sentence", and this fact is given to the computer parser of texts as a relevant part of its knowledge of syntax. However, this is a fact about clauses in relation to all other units, and on its own it does not constitute evidence that there is a 'rank scale' of units in the "Categories" sense of the term.

Blogger Comments:

[1] To be clear, the rank scale does not "predict" that the clause is a higher unit than the others. The rank scale is a way of modelling constituency, and what makes the clause the highest unit, in this model, is the fact it alone is not a constituent of some other unit. A tree diagram merely represents this fact. Halliday (2002 [1961]: 43):

The units of grammar form a hierarchy that is a taxonomy. To talk about any hierarchy, we need a conversational scale; the most appropriate here might seem that of size, going from “largest” to “smallest”; on the other hand size is difficult to represent in tables and diagrams, and may also trap one into thinking in substantial terms, and a vertical scale, from “highest” to “lowest”, has advantages here. For the moment we may use both, eventually preferring the latter. The relation among the units, then, is that, going from top (largest) to bottom (smallest), each consists of one, or of more than one, of the unit next below (next smaller). The scale on which the units are in fact ranged in the theory needs a name, and may be called rank.

[2] To be clear, Fawcett's set of procedures for clause generation is inconsistent with the theory he is claiming to be developing: Halliday's superseded theory of Scale & Category Grammar. Halliday (2002 [1961]: 42):

In this view of linguistics description is, as already emphasised, a body of method derived from theory, and not a set of procedures. This has one important consequence. If description is procedural, the only way of evaluating a given description is by reference to the procedures themselves: a good description is one that has carried out the right procedures in the right order, but for any more delicate evaluation external criteria have to be invoked. Moreover, every language has to be treated as if it was unknown, otherwise procedural rules will be violated; so the linguist has to throw away half his evidence and a good few of his tools.

A theory on the other hand provides a means for evaluating descriptions without reference to the order in which the facts are accounted for. The linguist makes use of all he knows and there is no priority of dependence among the various parts of the description. The best description is then that which, comprehensiveness presupposed, is maximally grammatical: that is, makes maximum use of the theory to account for a maximum amount of the data.

[3] As this clause demonstrates, Fawcett's model is concerned with text generation by computer, not with language as spoken or written by humans.

[4] This is misleading. As Appendix 2 (pp306-7) demonstrates, Fawcett's model includes the following units, all of which are, in effect, constituents of the clause:

• nominal group,
• prepositional group,
• quality group,
• quantity group, and
• genitive cluster.

That is, Fawcett's model has a 2-level rank scale, in the "Categories" sense (see [1]), in all but name.