Thursday, 7 June 2018

David Rose Misunderstanding Generalisation And Abstraction

Why don’t we start on common ground, with the empirical technique Michael Halliday bequeathed us, the system network? Since it encapsulates the roles of generalisation and abstraction in semiosis.

A system relates features by generalisation... two or more features are grouped in a system by similarity, and distinguished by difference. But these relations are recognised by similarities and differences in the structures that realise them. This axial relation between feature and structure is abstraction, feature/value: structure/token.

The structures in our realisation statements are also generalisations. They generalise recurrent instances in text corpora. So the axial feature/structure relation is abstraction, while the instantial structure/instance relation is generalisation, and so is the paradigmatic feature/feature relation. This complex of relations is illustrated in the attached diagram.
These generalisation/abstraction relations are iterated fractally as we expand our descriptions outwards from a single system. Paradigmatic feature/feature generalisations are iterated as we expand systems in delicacy. The token/value abstraction is iterated as we expand from stratum to stratum. ...

Blogger Comments:

[1] To be clear, 'empirical' means based on, concerned with, or verifiable by observation or experience rather than theory or pure logic.  A system network is a construal of language (data) as theoretical categories organised by logical semantic relations:
  • elaboration (delicacy),
  • extension: addition (conjunction),
  • extension: alternation (disjunction), and
  • enhancement: condition (entry condition).

[2] To be clear, a system network orders features in terms of delicacy.  Halliday & Matthiessen (1999: 14):
the paradigmatic network is ordered in delicacy (subsumption, classification, specialisation), from the least delicate (most general) to the most delicate (most specific types).
[3] To be clear,  the features of a system are co-hyponyms of a superordinate type, such as [mental] vs [material] PROCESS TYPE.

[4] To be clear, the "similarity and difference relations" in a system are paradigmatic (choice) relations, whereas the "similarity and difference relations" in a structure are syntagmatic (chain) relations — the two types being of a different order.  So, for example, in the clause:
David couldn't understand
the systemic difference in choosing [mental] instead of, say, [material] PROCESS TYPE is not "recognised" in the structural "difference" between Senser and Process.

[5] To be clear, the axial relation of realisation (symbolic abstraction) obtains between syntagmatic organisation (token) and paradigmatic organisation (value).  Halliday & Matthiessen (1999: 13):
… on the one hand, syntagmatic organisation realises paradigmatic organisation; on the other hand, types in a network of paradigmatic organisation correspond to fragments of syntagmatic specification …
[6] This confuses the general–to–specific dimension: delicacy, with the token–to–type dimension: instantiation; see Halliday & Matthiessen (1999: 144-5).

[7] Contrary to the claim, the diagram does not encompass the dimension of instantiation.

[8] In addition to the problems already highlighted, two of the inconsistencies that invalidate this diagram are:
  • it misrepresents syntagmatic structure as a paradigmatic system; and
  • it construes the logical relation of 'similarity', not as a relation between features, as it does with 'difference', but as features themselves.
That is, reading the diagram, for both system and structure, the entry condition 'similarity' affords a choice between 'similarity' and 'similarity', where each 'similarity' is different from the other 'similarity'.

[9] These are claims about multiple confusions and misunderstandings, as demonstrated above.