Misconstruing Semantics As Context

Martin (1992: 289-90):

For the remainder of this chapter it will be assumed that lexical cohesion will be analysed with respect to field oppositions, and that lexical relations can [be] approached from four different perspectives within a systemic functional model.  These perspectives are:

i. from the point of view of mutual expectancy in lexicogrammar (collocation) 

ii. from the point of view of more delicate options in lexicogrammatical networks (the grammarian's dream)

iii. from the point of view of semantic relations between lexical items in text (lexical cohesion) 

iv. from the point of view of register specific oppositions (field taxonomies)

An overview of these perspectives is provided in Table 5.2. …

Table 5.2. Lexical relations across planes and strata

Context	Language
	DISCOURSE SEMANTICS	LEXICOGRAMMAR
field specific taxonomies	lexical relations (cohesion)	field neutral taxonomies (as delicate grammar)
		collocation patterns (lexical sets)

Blogger Comments:

[1] In SFL theory, lexical cohesion is a grammatical resource of the textual metafunction and field is the ideational dimension of context.  Field oppositions and taxonomies are context oppositions and taxonomies, not language oppositions and taxonomies.

[2] In SFL theory, collocation is a type of lexical cohesion and both are located in the lexicogrammar.

[3] Not just in SFL theory, register is a functional variety of language.  Field, on the other hand, is the ideational dimension of context, which is not language, but the culture as a semiotic system that has language as its expression plane.  Field is more symbolically abstract than language.

Perhaps this confusion of register with context arose from the terms 'spoken registers' and 'written registers' in which registers are classified in terms of the context (mode features) they realise — as when Martin later writes: 'though these are less spoken in register' (op cit: 312).

[4]  In SFL theory, "field specific taxonomies" — in Martin's sense — are located at the level of semantics, not context, because they are a way of organising linguistic meaning, not the culture as semiotic.  Such a taxonomy constitutes a lexically oriented domain model, a way of modelling a semantic domain.  Halliday & Matthiessen (1999: 323):

…the semantic correlate of a contextual field is a domain.  When we model the ideational semantics of a particular field, we create a domain model. … Domain models are variants of the general model.  A particular domain model specifies which of the semantic systems in the overall model are activated in a particular contextual field …

[5] As explained in the previous post, field "neutrality" is irrelevant for elaborating grammatical networks into the lexical zone.

[6] In SFL theory, 'lexical set' is opposed to 'collocation'.  That is, it describes the types of lexical cohesion in which the nature of the relation is paradigmatic (repetition, synonymy, hyponymy and meronymy).  Here the term is misapplied to the opposite type of lexical cohesion, collocation, in which the nature of the relation is syntagmatic.

Relocating Lexis Outside Language

Martin (1992: 289):

This raises the question as to what extent lexis can be treated as most delicate grammar while maintaining a lexicogrammar that is essentially neutral with respect to field. … One way of keeping the grammar neutral would be to locate specialised taxonomies in the register variable field, allowing the grammar to focus on field neutral oppositions.  This would mean for example that the similarity between morphology in linguistics and geology could be brought out in lexicogrammar and the differences between them in field taxonomies; similarly for cover in yachting and sleeping.  And analysts could make a principled decision as to whether they were analysing lexical cohesion with respect to lexicogrammatical taxonomies or field ones, depending on the purposes for which the analysis was designed.

Blogger Comments:

[1] 'Neutrality with respect to field' is irrelevant for elaborating lexis as most delicate grammar.  The lexicogrammar is the overall system of wording potential.  This misunderstanding seems to arise from giving priority to the view from below (e.g. the forms morphology, cover).

[2] This seriously misunderstands the meaning of the stratified model of SFL theory.  Field is not modelled as language; it is a higher level of symbolic abstraction than language: it is the ideational dimension of the culture construed as a semiotic system.  Locating lexical taxonomies outside language is a major category error.

[3] Construing field as a register variable confuses a higher level of symbolic abstraction than language, context, with a functional variety of language, a register.

Construing field variation as a register variable confuses instantial variation (e.g. situation type probabilities) with hyponymy (feature options in systems).

[4] The difference between morphology in the fields of linguistics and geology is characterised in the lexicogrammar — not field — by the different lexical features that specify them as distinct outputs.

[5] This confuses lexical cohesion with lexicogrammatical delicacy, and language with context.

In SFL theory, lexical cohesion is the use of repetition, synonymy, hyponymy, meronymy and collocation to establish non-structural textual relations.  A lexicogrammatical taxonomy, however, is a means of modelling the lexicogrammatical system.  A field taxonomy, on the other hand, is means of modelling the ideational dimension of context — not lexicogrammar (nor even language).

Confusing Textual Cohesion With Experiential Delicacy

Martin (1992: 288-9):

One critical consideration which has been set aside in the discussion of lexical relations to this point is that of field specificity.  In general cohesion analysts have tended not to be very specific about this point.  Categories tend to be illustrated with examples from core vocabulary, which are kept as field neutral as possible: verbs, adverbs and adjectives are favoured over nouns, and "commonsense" nouns are used rather than technical ones.  It is no accident that Hasan chooses action verbs in exploring the grammarian's dream; fields tend to differentiate themselves nominally, through technical lexis, rather than through verbs, adjectives and adverbs.  But even with verbs, the question of field slips in.

Blogger Comments:

[1] This discussion confuses two distinct domains of SFL theory:

• the grammatical system of lexical cohesion, a nonstructural resource of the textual metafunction that relates lexical items in terms of repetition, synonymy, hyponymy, meronymy and collocation; 


• the systemic specification of lexical items by extending the delicacy of grammatical networks (experiential metafunction).

[2] In SFL theory, lexical cohesion is a grammatical system concerned with textual relations between lexical items, whereas field is the ideational dimension of context.

[3] These remarks are about extending the delicacy of grammatical networks, while purporting to be about lexical cohesion.

[4] In SFL theory, field variation is modelled as a point on the cline of instantiation at the stratum of context.  For example, field variation is the ideational dimension of situation type variation, where the variation is of the probabilities of field features being selected.

Lexical Cohesion Update

Martin (1992: 288):

Finally, mention should be made of Halliday (1985: 310-2), who maintains repetition and collocation as distinct categories, and groups together synonymy, antonymy, meronymy and hyponymy under a general heading of synonymy.

Blogger Comment:

Using the dimensions of the SFL model, Halliday & Matthiessen (2004: 572) systematically distinguish the lexical relations that play a rôle in lexical cohesion; most generally, according to axis:

• the nature of relation: syntagmatic (collocation) vs paradigmatic (the rest)

and the paradigmatic according to logical relations:

• the type of expansion: extending (meronymy) vs elaborating (the rest)

and the elaborating according to experiential relations:

• attribution (hyponymy) vs identity (synonymy, repetition).

Misconstruing The Agency Of Identifying Clauses

Martin (1992: 285):

Alternatively, be could have been specified as an optional lexicalisation of the feature [single agency], which if taken up would have to be as foreclosing choices in more delicate systems.  The network could then have been simplified as in Fig. 5.8, with the choice between [exemplifying] and [exhausting] dependent on [single agency] not being realised through be.

Blogger Comment:

The proposed network is undermined by the incorrect assumption that all unassigned identifying clauses have single agency, and all assigned clauses have double agency.  In fact:

• both unassigned encoding clauses and assigned decoding clauses have single agency,
• only operative* assigned encoding clauses have double agency, and
• unassigned decoding clause have no agency at all.

		assignment
		unassigned	assigned
direction of coding	decoding	0 Agent	1 Agent (Assigner)
	encoding	1 Agent (Token/Identifier)	*2 Agents (Token/Identifier, Assigner)

The verb be serves as identifying Process in unassigned clauses only, both encoding and decoding, which means it serves in clauses with both single agency and no agency at all.

Moreover, there are identifying clauses with single agency in which the verb be does not serve as Process: assigned decoding clauses.

Falsely Claiming To Have Uniquely Classified Two Lexical Items

Martin (1992: 283):

This brings us to the point where two lexical items have been uniquely classified (define and read) and a step or two in delicacy would isolate several others (e.g. call, name, mean, denote, connote, spell, write, transcribe).

Blogger Comment:

[1] In Fig. 5.5. (ibid.),  the unique classification of define is given as:

• realise: stratum specific: semantics/grammar : wording: encoding

As demonstrated in the previous post, when the verb define serves as an identifying process, the direction of coding, like all identifying processes, can be encoding or decoding, depending on whether the Token or Value is the Identifier.  The feature [encoding], therefore, does not classify the word define.

[2] Likewise, the unique classification of read is given as:

• realise: stratum specific: grammar/phon-graphology: encoding

As demonstrated in the previous post, the verb read did not serve as an identifying process in the clause provided.  Rather, it served as a verbal process in a projection nexus.  Accordingly, not one of the features specifying read — [realise: stratum specific: grammar/phon-graphology: encoding] — actually classifies the lexical item.

Misrepresenting The Directions Of Coding And Mistaking A Verbal Projection Nexus For An Identifying Clause

Martin (1992: 282):

The discussion will now be limited to the semiotic options subclassifying the feature [realise].  These can be divided into a general set which do not specify which strata or ranks are being related and a set which is specific to either the semantics/grammar interface or the grammar/phonology (including grammar/graphology) one…

Both interfaces make a distinction between encoding and decoding, depending on whether the Token is treated as closer or further away from the expression substance than the Value.

GRAMMAR/PHONOLOGY:ENCODING

[5:17]  N-G spells ng. 

GRAMMAR/PHONOLOGY:DECODING

[5:18]  But it actually reads the big dog must be able to see the dog at all times. 

SEMANTICS/GRAMMAR…ENCODING

[5:19]  Spell means 'to represent graphically as' 

SEMANTICS/GRAMMAR…DECODING

[5:20]  'To represent graphically as' defines spell.

Blogger Comments:

[1] This seriously misrepresents decoding and encoding.  The directions of coding are not semiotic options "subclassifying the feature [realise]", and have nothing whatsoever to do with "whether the Token is treated as closer or further away from the expression substance than the Value".

The direction of coding pertains to all identifying processes and simply depends on whether the Value is used to identify the Token (decoding) or the Token is used to identify the Value (encoding), as Halliday & Matthiessen (2004: 230) make clear.

Of the three genuine identifying clauses provided, each can be construed as either decoding or encoding, as demonstrated below:

decoding: what word do the letters ‘N-G’ spell?

N-G	spells	ng
Token/Identified	Process: identifying	Value/Identifier
Given		New

encoding: what letters spell the word ‘ng’?

N-G	spells	ng
Token/Identifier	Process: identifying	Value/Identified
New		Given

decoding: what does the word ‘spell’ mean?

spell	means	to represent graphically as
Token/Identified	Process: identifying	Value/Identifier
Given		New

encoding: what word means ‘to represent graphically as’?

spell	means	to represent graphically as
Token/Identifier	Process: identifying	Value/Identified
New		Given

decoding: what meaning defines the word ‘spell’?

to represent graphically as	defines	spell
Value/Identifier	Process: identifying	Token/Identified
New		Given

encoding: what word is defined by the meaning ‘to represent graphically as’?

to represent graphically as	defines	spell
Value/Identified	Process: identifying	Token/Identifier
Given		New

This last  clause is extremely marked, not least because, in the vast majority of instances of identifying processes, it is the Token, not the Value that conflates with Subject in clauses of operative voice.

[2] This is not an identifying relational clause; it is a nexus of verbal clause projecting a quoted locution clause (despite the omitted quotation marks).

but	it	actually	reads (‘says’)	the big dog	must be able to see	the dog	at all times
1				“ 2
	Sayer		Process: verbal	Senser	Process: mental	Phenomenon	Location: time

This instance, therefore, does not construe the 'grammar/graphology interface'.  The logical relation it construes is projection, not elaboration, and the domains that are related are the material order of experience (a sayer saying) and the semiotic order of experience (a wording; i.e. lexicogrammar).

Misrepresenting 'Semiotic'

Martin (1992: 281-2):

The largest class [of exhausting processes] has to do with processes of signification.  These can be grouped into three sets.  One deals with the realisation relationships within semiotic systems; a second handles relationships between non-semiotic manifestations and their meanings; and the third has to do with meanings that are suggested rather than denoted:

REALISE
[5:12] C-A-T spells cat. 

MANIFEST
[5:13] The red stands for the blood that's been shed. 

CONNOTE
[5:14] His behaviour smacked of jealousy.

Blogger Comment:

By definition, the expression of a meaning, as when red betokens bloodshed, is semiotic.  Semiotic means 'relating to signs' and signs are necessarily both content (signifié) and expression (signifiant).

Misconstruing Unassigned vs Assigned As Single vs Double Agency

Martin (1992: 280, 281):

Identifying clauses can first be divided into those with an additional agent and those without..  This additional agent may have the function of assigning Token to Value…

Blogger Comment:

To be clear, there is only an additional Agent in an identifying clause if it is both assigned and encoding, and this is only possible in operative voice (Halliday 1994: 165).  (In encoding clauses, Token conflates with Agent.)  Unassigned decoding clauses are middle: they have no Agent.

decoding

‘yacht’	means	a light, fast sailing ship
Medium	Process:	Range
Token/Identified	identifying	Value/Identifier

encoding

‘yacht’	means	a light, fast sailing ship
Agent	Process:	Medium
Token/Identifier	identifying	Value/Identified

The accompanying system network (Fig. 5.4.) misconstrues unassigned vs assigned as single agency vs double agency.

Misrepresenting The Model Of Transitivity

Martin (1992: 278):

Where Halliday (1985) justifies his process types largely on the basis of configurations of Agent, Process and Medium, Hasan additionally draws the Beneficiary role into the picture.

Blogger Comment:

The system of process types isn't justified on configurations of Agent, Process and Medium.  Each of these — process types, on one hand, configurations of Agent, Process and Medium, on the other — represents a distinct complementary perspective on transitivity: the transitive and ergative models, respectively.

The transitive model is concerned with the ways in which processes are different, with each type of process 'characterised by process-participant configurations where the functions are particular to that type' (Halliday & Matthiessen 2004: 280).

The ergative model, on the other hand, is concerned with how the process types are all alike; how 'they all have the same grammar' (Halliday & Matthiessen 2004: 281).

Not Recognising The Stratification Of Content: Thesaurus

Martin (1992: 273, 275):

The relationship between words and ideas in a thesaurus is not a stratal one either.  Rather, words are classified notionally…

Roget's system develops the idea of general motion up to the category sailing-ship and then lists some fifty lexical items under this heading; one of these is the lexical item yacht.  This strategy reinforces the notion/wording duality invoked by Roget in introducing his work; but all that is really going on is that his classification is being arrested at a certain point in delicacy. …

Because it is organised according to meaning, a thesaurus provides a more appropriate model of lexical description than a dictionary does …

Blogger Comments:

[1] By definition, in SFL theory, the relationship between words and ideas (mentally projected meanings) is a stratal one.  It is the relation between lexicogrammar and semantics.

[2] Classifying words notionally is classifying words according to the meanings they realise.  The relation between words and the meanings they realise is a stratal one: the relation between lexicogrammar and semantics.

[3] The relation between the idea of a sailing ship and the lexical items that realise it is a stratal one: the relation between semantics and lexicogrammar.

[4] What is actually going on is: a the semantic system is elaborated to 'a certain point in delicacy', and the words that realise semantic categories are listed.  The relation between a semantic category and the words that realise them is a stratal one: the relation between semantics and lexicogrammar.

[5] In being organised according to meaning, a thesaurus provides a semantic taxonomy in which the lexical items that realise each category are listed.  In being organised according to meaning, therefore, it does not provide a model of lexis at the level of wording.  In SFL theory, lexis is modelled as most delicate grammar.  What is appropriate, in SFL terms, is the use of hyponymy as an organising principle.

Not Recognising The Stratification Of Content: Dictionary Definitions

Martin (1992: 273):

The folk-conception of a dictionary entry as a "form" and the definition as its "meaning" is of course a misleading one.  In fact the relationship between words and definitions is not stratal (i.e. words are not being associated with the semantic primes they realise).  All that is going on is that one wording is being reformulated as another.

Blogger Comment:

Halliday & Matthiessen (2004: 237) see this rather differently, because they apply SFL theory to definitions:

Naming and defining are linguistic exercises, in which the word is Token and its meaning is its Value. … 

a ‘gazebo’	is	a pavilion on an eminence
Token		Value

(= ‘the word gazebo means [names, is defined as] a pavilion on an eminence')