MAKING PLANS FOR NIGEL: Defining interfaces between computational representations of linguistic structure and output systems: Adding intonation, punctuation and typography systems to the PENMAN system. 

Chapter 3 : Graphology

previous up next 
PDF version

A hypothetical science of graphology would describe the organization of space into usable tokens.  These tokens would include writing symbols such as the alphabet and the number system, punctuation and designs. (Cummings & Simmons 1983 p.74)

3.0  Outline

This chapter and the next focus on NIGEL’s written output.  The chapter is about the addition of both a graphological stratum, and extensions to the lexico-grammar ‘above’ the interface and the new typographical output program ‘below’ the interface.  The structure is as follows.  The first part of the discussion (3.1) is concerned with the typographic system to be used.  The second section (3.2) is about the way that the systems are to be synthesised from other work in the field.  The third section (3.3) is about the mechanisms by which meanings are made through graphology.  Section (3.4) presents the graphological system itself. Finally, there is a conclusion (3.5).

While the last chapter took Halliday’s phonology as a starting point, showing how it specifies what information should be supplied by the phonology to the speech synthesiser, there is no graphology available that can be simply adopted.  The graphological system described here is a synthesis of Halliday’s (1985.a,b) and Waller’s (1980) work on written language.  In contrast to the phonology chapter, the focus here is on the graphological systems themselves, rather than the grammatical systems that they will realise; there is no system analogous to Key in this treatment.  This graphology is still under development: the first step was to find out what needs to be accounted for; the next step was to begin asking what it means.  Only when it is known what graphological structures exist, and what they mean can we make extensions to the lexico-grammar to handle ‘graphological meaning’.  The work is not yet so far advanced.  It is, however, possible to modify NIGEL so that the old rudimentary graphology, which is a series of rules for punctuation, is made to work in a “clean”, theoretically motivated way.

At the moment, the work on graphological structure is in advance of the work on meaning, it is easier to describe the ‘tactics’ of graphology than it is to to say what it means.  The meaning of graphology is painted with a broad brush in this chapter, but the mechanisms by which the meaning will be made can be depicted in more detail.

The chapter could be characterised in terms of problem and solution.  The problem is that NIGEL punctuates in an inadequate way, and has no control over typography.  The solution is to design a new system which is theoretically motivated, and descriptively general.  Two aspects of NIGEL’s inadequacy are addressed; (1) computational architecture; and (2) generative power.

1. Architectural “problems”

Figure 1.4 in the introduction showed NIGEL’s ‘dirty’ interface between lexico-grammar and output system.  This architectural problem will be solved by adding a graphological stratum, and making the interface between it and a typographic program conform to the three principles for a clean interface.

2. Generative “problems”

The graphological system described here will account for both punctuation and typographic organisation, while the old NIGEL graphology is only concerned with punctuation.  The problems are listed below.

  1. i.  The current system does not allow for an incongruent mapping between the grammatical clause-complex and the graphological sentence.  In the Jungle Groove contains four sentences (example 3.1), which would normally be punctuated as only one (cf. the grammatical analysis in section 2.1, example 2.1).  What is the meaning of the punctuation?  The current system does not answer this question. The new one will.  The mechanism which allows this involves a stratification between lexico-grammar and graphology: sentence is a graphological unit, which typically, but not always, realises the lexico-grammatical unit of clause-complex.

  2. 3.1

What you hold here is a funk bomb, primed to vaporize lethargy. A compound of full-length full-strength masterfunk. An hour or so of GET UP and go. The jungle groove.

  1. ii. The current system has no way of varying the typographic realisation of its graphology in a principled way.  Nor can it use context-specific systems for punctuating.  Two typographically different versions of the same wording, below, differ in two respects: they represent Alice’s phonological emphasis in different ways; and the first has punctuation which could loosely be described as ‘freer’ or ‘less strict’ than that the second.  How should we account for the differences?  The current system does not provide any answer.  The new one will.


“What can all that green stuff be?” said Alice, “ and where have my shoulders got to? And oh! my poor hands! how is it I ca’n’t see you?”

‘What can all that green stuff be?” said Alice. ‘And where have my shoulders got to? And oh, my poor hands, how is it I ca’n’t see you?’

  1. iii. The current system cannot handle the realisation of discourse structure features. In the Jungle Groove is divided into sections, each of which starts with a large capital letter.  These sections can be shown to correspond to lexical cohesion and reference chaining in the way that they divide up the text (Sefton, unpublished essay 1988). The new system will allow reflection of high-level textual structure in typography, through graphology.  This will be implemented as a process of realisation, the textual units that NIGEL works with will be realised in graphological units, which in turn will be realised typographically.

The solutions to the architectural and generative problems are embedded in the rest of the chapter; they will be collected together in the concluding section (3.5).


This discussion is concerned with linear text.  Not all text is linear, many texts have marginalia or other forms of parallel textual relations.  It is conceivable that the NIGEL system would be required to produce non-linear texts, but this work does not discuss the possibility.  A device for representing the typographical relations between parallel texts could be added to the formalisms discussed here.

In the interests of generality the broadest possible range of typographic phenomena is to be accounted for systemically.  That is, the boundary between the systemic-linguistic and para-linguistic is to be ‘pushed back’.  This is very important; if generation grammars are to be as ‘natural’ as possible then there should be some way of letting them control those aspects of language which have typically been described as paralinguistic.  Their inclusion in generation systems is facilitated by descriptions of the local systematicity of paralinguistic phenomena.  The major addition to systemic grammar that allows this ‘pushing back’ is a mechanism by which a generator may modify and create its own systems (see section 3.3 below).

3.1  Below the interface

It is possible to describe typographic resources systemically, indeed word processors and page-layout programs often make use of menu interfaces which have a systemic flavour.  The typographic system presented here is similar to to the one that is available in most text-based programs for Apple Macintosh computers, with which many academics will be familiar.  Each character on a screen or a printer page may be specified as to what type-face (font), what style (bold, italic, underlined etc.), what size, and what case (upper or lower) it is in.  As well, the spatial principles by which text is displayed can be specified, eg. how paragraphs are marked off by space and justified.  The output program being developed, called GEKKO, is described in the next chapter four.


Figure 3.1
A typographic graphetic system.

While it is possible to specify a font for each character individually, the choice is typically made over larger text spans.  The word processor with which this thesis was written (Microsoft WORD) allows fonts to be specified for whole paragraphs.  Paragraphs with different functions can be given a specific description, of font, style and spatial organisation.  In this text, for example, examples are marked off with a blank line before and after, and they appear in the fixed-width font Courier.

The function of a paragraph, or of any other unit is obviously a matter that concerns meaning, and so the organisation of the units over which typography may be specified will be a matter of lexico-grammar and graphology, not of typography, hence the simple typographic system here which does not include functional units like paragraphs.

The systems for BOLD, ITALIC, CAPS and UNDERLINE all have the output features ON, OFF and TOGGLE.  The first two do what one would expect; they ‘turn on’ or ‘turn off’ the relevant style for a piece of text.  TOGGLE switches between on and off states.  This is often used to represent phonological emphasis - in plain-text passages emphasis is in italics; while in italic passages it is in plain text.

3.2  Graphology

It is necessary to start almost from scratch in describing graphology because there has not been much work done on punctuation in general linguistics.  Waller, in his seminal article on the meaning of punctuation and typography notes that “it is hard to find one modern general linguistics textbook that even mentions punctuation”.  Similarly, it is hard to find descriptions of the meanings that are made through typography.

Some systemic work on punctuation has been done by Halliday (1985.a,b), but since his work was not done on the same terms as phonology, there is no explicit discussion of graphological systems as such.

In developing a graphological system it is necessary to clarify the relationship between graphology and typography.  Halliday (1985.a p.1) makes a distinction between the graphological - letters, words, and sentences - and the typographic - spaces:

Sentences follow sentences, words follow words and letters follow letters in  a simple sequence; they do not overlap, nor does anything else occur in between. The spaces that separate them - narrow spaces between letters (at least in print), wider spaces between words, and still wider spaces with accompanying full stop, between sentences - serve to mark the units off one from another. The spaces and stops are not part of the substance of writing; they are signals showing how it is organised.

It is not important at the graphological level that three different amounts of space will be used to separate letters, words and sentences; in the terms of this thesis that is the business of typography - to be handled by an output program.

The notion of graphological prosody was introduced in the introduction: any supra-segmental aspects of a written text are prosodic.  The minimal segmental units are letters, numerals and symbols, any organisation of these into words, lines, stanzas, paragraphs, documents, sentences, comma groups etc. is supra-segmental organisation, and hence is prosodic.

The actual notation will look like this:


Typographic version: Get the message?

Graphological version: [sentence:question get the message]

At a graphological level there is an abstract representation of the sentence, and a further, more delicate specification that it is a question.  Prosodies ‘belonging to’ units will be delineated with square brackets, spaces mark off words.  NIGEL will produce graphological text, and the output program will convert it to typographic text.

Converting graphology into typography is not simply a matter of producing a string of letters and symbols - it is necessary to know a number of other things about them, for example their font.  Typically, a large chunk of text is in one font, so it is not necessary to list the font for every character.  In this essay, each paragraph type has a default font; so it is possible to tell something about the role of a paragraph in the text from the type-face in which it is printed.  Figure 3.2 shows how systems for specifying font choice for a paragraph might be implemented; it shows a graphological level system which would decide on the relevant font by interrogating the semantic level: three of the inquiries have to do with the textual meta-function - the role in the discourse that the paragraph plays - while the fourth inquiry has to do with a mode-variable (cf Halliday and Hasan 1985): if the output device is a printer then standard paragraphs are in Times, but it is easier to read New York on video displays - so text is displayed in New York on the screen.


Figure 3.2
A stratified systemic representation of the way that paragraph types may be distinguished by type-face choice.

Independence between strata

Figure 3.2 illustrates the way that a stratal model of language generation would work, but it does not show the relationship between  lexico-grammar and graphology.  Further, the figure shows an ‘active’ graphology which has choosers that interrogate the environment; it has already been explained that NIGEL will have ‘virtual’, inactive graphology and phonology - not ‘active’.  Exactly what in the lexico-grammar will be realised as a graphological paragraph is beyond the present scope.  However, it is possible to discuss in principle the mechanisms that would be needed to model the relationship between lexico-grammar and graphology.

It is important to remember that sentence is not a grammatical category in SFG.  NIGEL generates clauses, related logically into clause complexes, not sentences.  There is an unmarked realisation of the clause complex as a sentence, but this is not always the case.  If NIGEL had generated example (3.4) it would consist of two clauses - in a hypotactic relationship of elaboration, and an elaborating nominal group complex (see section 2.1).


What you hold here is (1 a funk bomb

=    << primed to vaporize lethargy, >>

=2 a compound of full-length full-strength masterfunk, =3 an hour or so of GET UP and go, =4 the jungle groove).

In the old NIGEL system clause complexes are always automatically realised as sentences but the new system will be equipped for generation of more ‘playful’ graphology, like the above example.

Despite the flexible relationship between lexico-grammar and graphology there is a point of natural mapping from lexico-grammar onto graphology: the lexico-grammatical unit of word is always realised graphologically and graphetically as a word.  This correspondence is analogous to the mapping  of information units on to tone groups.  Figure 3.3 shows the relationship between the grammatical rank-scale and graphology.


Figure 3.3
The realisation of lexico-grammar in graphology: the natural mapping is from grammatical word to graphological word.

So, in regard to the word “masterfunk” from the example above: it is the responsibility of the grammar to ‘decide’ that it is one word.  Similarly the hyphenated “full-length” will be treated as having the lexico-grammatical status of a word, hence it is automatically realised as a word.

The graphological rank-scale

Having shown that there is a certain amount of independence between lexico-grammar and graphology the graphological system can be further explored.  The first step in this exploration is to work out what graphological units there are.

What are the ranks in the English graphological system?  What are the constraints on their ‘nesting’, embedding? There is no definitive answer to this.  Halliday (1985.a p.3) describes some of the alternatives, for the unit(s) between the word and the sentence:

There are likely to be some uncertainties at this point.  We might want to recognize not just one but two layers of sub-sentence, one for units separated by a colon or semi colon and one for units separated by a comma.  We might want to insist that all sentences consist of sub-sentences even if in any given text the majority may contain only one; or we might allow that some sentences consist of sub-sentences, while others consist directly of words.

The strategy adopted here is to assume graphological units which are general and unconstrained, as opposed to being highly constrained.  By using a very general descriptive process it is hoped that the greatest possible range of material can be described.  This strategy suggests working solutions both to the uncertainty about the makeup of a sentence and about the units larger than the sentence.

Up to the sentence

Letter and word are not problematical: the NIGEL system already has adequate mechanisms for generating them.

The strategy of ‘generality first, constraint second’ suggests that we adopt, for a start, a graphology with just one sort of sub-sentence, rather than two.  This sub-sentence has been named orthographic group (henceforth OGP), which can be either a colon group or a comma group.  The sentence may consist “directly of words”; there is no constraint on sentences that they consist of at least one OGP.  On the other hand, it is not assumed that a sentence can not consist of just one OGP.  The OGP is a recursive unit; that is, an OGP may be made up of other OGPs.  E.g. a colon group can contain other OGPs, such as comma groups.

The design of generative systems can be expected to modify this first, most general, graphology.  For example, question marks and exclamation marks are not found operating over pieces smaller than sentences in some registers or genres: this represents a constraint on the most general case.  There may be other constraints which are ‘universal’ and not simply register specific.  It may turn out to be better to have a larger set of sub-sentence types.

Above the sentence

There are three ranking units above the sentence.  The hierarchy has been modelled on that of WORD; again, the idea is that we should have a most-general system.  The units are, moving upwards from the sentence, paragraph, section and document.

The paragraph obviously has a very common, almost universal typographic correlate.  The graphological unit paragraph is not the same as grammatical conceptions of the paragraph such as Longacre’s (1979).  The range of typographic realisations of paragraphs is infinitely large but it generally involves the use of space in some way.  Paragraph is a recursive unit.  Thus, examples in this essay could be thought of as paragraphs, which are often embedded in other paragraphs.  In the interface code (simplified) the notation for an example paragraph embedded in another paragraph looks like example 3.5.  The lower case text represents the actual text of the document, while the uppercase text and brackets are part of the interface code.


[PARAGRAPH content text

    [ PARAGRAPH:EXAMPLE example text]

content text]

The unit of section is also recursive.  The inclusion of this unit is provisional; it is intended to provide a framework on which a very general model of graphological text can be based.

The highest order unit, the document is simply a way of bracketting the output of NIGEL in chunks that can be directed to different output devices.  The document is simply a whole text.  Document is not recursive.

A problem with this ‘most general’ approach is that, from an analytical point of view, there will be more than one way of describing the graphological structure of a text.  One answer to this problem will come from the modelling of generation: it will become clear which graphological structures are natural from a generative point of view - thus eliminating some of the analytical possibilities.  From a purely theoretical point of view, further refining of the rank-scale can not be effected until much more descriptive work has been done.

So far the graphological rank-scale has been identified.  The next section describes the prosodies that operate over the units in this rank scale from word up to sentence.  The section discusses these units as graphological entities, it is not concerned with their realisation, which typically involves punctuation marks, rather it is concerned with the way that they can be used to make meanings.  To this end, Halliday’s (1985.b) typology of punctuation marks is adapted.

Punctuation: Adapting Halliday’s treatment

Halliday, in Spoken and Written Language, (1985.b) lists three types of punctuation mark: (i) boundary markers, (ii) status markers and (iii) relational markers. The following discussion follows Halliday’s (1985.b p 32-39).  Halliday’s table, displaying this typology, is presented in figure 3.5.

















weaker clause











full stop










question mark


other functions


offer, suggestion, exclamation, call, greeting

exclamation mark



quotation, citation

first order ; or meaning

single quote

‘ ‘

second order ; or wording

double quote

“  “


any unit







( )

(compound ) word








(i) Boundary markers

The set of boundary markers comprises the full-stop, the colon, the semicolon and the comma.  According to Halliday these punctuation marks can be used to mark either grammatical or phonological boundaries.  The grammatical units that can be represented are the units of the grammatical rank-scale.  The phonological unit that can be marked is the information unit. In Halliday’s table (3.5) sentence is listed as a grammatical unit.  However, in the NIGEL grammar in particular, and in Systemic Functional Grammar in general, sentence is not a grammatical category.  It would be better to put the grammatical unit clause complex into this table as the unit that is realised by the sentence.

Furthermore, the table is not general enough in this area; it has already been shown that there is no direct correspondence between the grammatical clause complex and the graphological sentence; example 3.1 shows that a graphological sentence may realise graphological units of different size.

The difference between marking of phonological boundaries and grammatical ones can be seen in Halliday’s example (3.6) (1985.b p.37):



Freda leapt down from the gate, and as Sebastian came forward her look of recognition unmixed with any surprise, contrived to suggest that for her, the sudden appearance of someone who had been away for half her lifetime, was the most commonplace event imaginable.


Freda leapt down from the gate and, as Sebastian came forward, her look of recognition unmixed with any surprise contrived to suggest that, for her, the sudden appearance of someone who had been away for half her lifetime was the most commonplace event imaginable.

In version (b), the punctuation has followed the grammatical structure; but in version (a) it represents an interpretation of the text in phonological terms - each stretch between commas corresponding to a tone group.  Note that in neither case does the comma imply a pause, although in loud reading it is often understood that way. (Halliday 1985.b p.37)

(ii) Status markers

Status markers realise interpersonal meanings.  There are two types.  One type of status marker is usually conflated with a boundary marker to delimit sentences; the question mark, exclamation mark and the full-stop constitute this set.

The other type of status marker is the quote, either single or double: using a pair of these a writer can mark projection, typically what someone has said or thought (cf Halliday 1985.a 7.5).  Quotes can enclose text spans of any length. These text spans must be grammatical units, or groups of them, from words up to whole texts. There is a sub-type of quotation sometimes called ‘scare quotes’.  Scare quotes enclose the last two words of the last sentence.  Their function has been described as ‘meta-comment’ (Sigurd 1987).

(iii) Relational markers

The relational punctuation marks (dashes, hyphens, brackets, apostrophes) relate units of text to each other: hyphens build compounds, dashes put things in apposition, apostrophes substitute for omissions and brackets allow digression - ‘subroutining’ of text. In the system under development apostrophes and hyphens will be taken as given. The lexico-grammar will produce them as part of the wording.

Another graphological device, not considered by Halliday, realises textual functions: the highlighting of phonological prominence in reported speech:


‘What can all that green stuff be?” said Alice. ‘And where have my shoulders got to? And oh, my poor hands, how is it I ca’n’t see you?’

In this example prominence is represented by italicising of single words, thus showing the tonicity of Alice’s document.  Tonicity realises a textual function.

Meta-functional diversity in punctuation

At one level all punctuation, and most typography, must mark boundaries.  Part of the function of a prosody is to delimit the unit over which it is prosodic.  The interpersonal function of quotation marks, for example, is in addition to to a ‘boundary marking’ effect, which identifies the relevant text-piece.  Thus quotation marks perform a dual role, they delineate the quote as well as marking its status.  Similarly marking tonality with commas shows the boundaries of IUs.  Boundary markers, however, lack the extra function of status markers, they only mark off units of the grammatical rank-scale.

The interpersonal meta-function is realised by the status markers, the quotation marks and by meta-comments.  The textual meta-function is realised by the boundary marking of tonality, in punctuating phonetically and by the marking of tonicity, through typographic prosody.  Logical relations between text-pieces are realised by the relational markers.

Much of the role of punctuation is in showing the grammatical structure of text.  In punctuating grammatically (cf Halliday 1985.b) instead of ‘as you speak’ the structural organisation of language is highlighted, in a way that it is not in speech.  In spoken language the phonological rank scale is organised to package information and there is little explicit signalling of grammatical organisation.  This emphasis on grammatical structure is consistent with the high value placed on writing over speaking.  Thus there are fewer resources for commenting prosodically, and more for making highlighting the text itself, in writing (Halliday 1985.b; Waller 1980).

The units above sentence in the graphological rank-scale can be looked at on the same terms as Halliday’s treatment of punctuation.  The next section explains this view.

Typography: adapting Waller’s system

In an article entitled Graphic aspects of complex texts: Typography as macro-punctuation, Robert Waller (1980) presents a typology of punctuation made in a spirit close to Halliday’s, although they acknowledge no sources in common.  Waller’s work is essentially descriptive of a graphology which is stratified with respect to lexico-grammar and semantics, but it is not presented in those terms.  Waller’s article is important for two reasons: firstly, his description is more delicate than Halliday’s; secondly, he presents a model in which punctuation and typographic organisation are seen in the same terms.  Both of these aspects are discussed in an intertwined fashion below.

Waller's table is similar to Halliday’s, but it gives us a view of graphology along a different axis - instead of looking at the grammatical units that are realised in graphological units it shows the typographical resources that are used to delineate graphological units. The concept of graphological unit is implicit in Waller’s table; for example, the graphological paragraph would be marked of by typographic ‘space’, while a sentence is marked off with an ‘initial capital’ and a ‘full point’.




initial capital

full point





title pages








boxed inserts






oblique strokes



headings, numerals

tabular format

regular spacing/styling


quotation marks

exclamation marks

question marks

size variation

style variation

layout variation

Here is a quotation from Waller explaining this table.  The highlighting is mine:

Delineation refers to methods of indicating the beginning and end of text segments, ranging from clauses (which can be bounded by commas), through sentences (bounded by initial capitals and full points), paragraphs, chapters, parts, and ultimately to books (bounded by covers)

Interpolation refers to the insertion or juxtaposition of a short segment into a longer one in such a way that the continuity of the sentence, paragraph, page, chapter, or book is not destroyed.

Serialization refers to the organization of segments into clear structures, sets, or series.  For example this paragraph is one of a set of four that is signalled simply by the linguistic rhythm of the first two words; the set could have been numbered, bulleted or typographically distinguished from the rest of the text.

Stylization refers to the indication of a mode of discourse differing in voice or genre from the main body of text.

Waller’s typology is similar to Halliday’s: delineation directly corresponds to boundary markers; stylization to status markers; interpolation to relation markers.

While there is no Hallidayian correlate for serialization, it can be seen as a special sort of relation in Halliday’s terms.  Waller’s example (highlighted in the quote above) of serialization actually illustrates a process of serialization through lexical cohesion, rather than through any typographic resource.  A more explicit example of serialisation can be generated from example 3.1.  In NIGEL the logical meta-function is ‘responsible’ for clause and group complexes.  This logical structure is normally shown through the use of boundary markers, but it could be shown with typographic bullets:



    What you hold here is:

*    A funk bomb, primed to vaporize lethargy.

*    A compound of full-length full-strength masterfunk.

*    An hour or so of GET UP and go.

*    The jungle groove.



Waller’s most important insight is the idea that typography can be treated as macro-punctuation.  His table shows how typographic resources are used to realise the same functions over different sizes of text piece (cf Halliday 1985.a for a similar treatment of grammatical resources). For example, short quotations in this text are marked off by the use of quotation marks, while longer quotations are set off using a typographic device (they are indented over their whole length and single-spaced).

Punctuation is the single aspect of written language, for which grammatical rules exist, that does not represent words themselves but the spaces between them. It is, then, an organizational system at the micro-text level functioning in much the same way as typographic signals and the use of space at the macro-text level. (Waller 1980)

Waller sees the same functions operating at different ranks: but these ranks are not explained.  For this work the rank scale described above will be used.

It was shown above that the essential delineating role of all punctuation, both micro and macro, is prosodic.  Waller implicitly refers to this prosodic role in the above quote in talking about representing the spaces between words.

By using a rank-scale that extends from document to letter, the graphology presented here will allow NIGEL to specify prosodies from the interpersonal, logical and textual meta-functions.  However, there are a number of different ways that these meanings will be made, which have to do with the way that the strata interact.

3.3  Graphological meaning

In the course of the discussion so far, the functions of punctuation have been identified.  To explain all of the ways that graphology actually makes meaning it is necessary to return to the notion of stratification.  Meaning is made by virtue of the stratificational relationship between lexico-grammar and graphology, and between graphology and typography.  It is in the interaction between strata that meanings are made.

It is possible to identify three main modes of stratificational interaction through which graphology makes meaning.  These have been glossed, for the purposes of this thesis as, (i) meta-redundancy, (ii) choice and (ii) negotiation. This split into three mechanisms is not completely natural, as they are highly dependent on each other. Nevertheless, it is informative.

(i) Meta-redundancy

One way that graphology makes makes meaning is quite straightforward: it is embodied in the way a particular phonetic tone contour realises a particular Key selection. The correspondence between intonation and Key is, in a sense, automatic; the term for this is meta-redundancy (cf Halliday 1990, Lemke 1984). Graphological examples include: quotes, in which there is a meta-redundant relationship between the illocutionary force of a text-piece and its graphological status; and the realisation of lexico-grammatical words as graphological words which in turn are realised as typographic words.

(ii) Choice

There are also meanings made in the choice of which meta-redundancies should be marked.  These meanings fall into two classes.  (1) Some choice is conditioned by high level semiotic variables to do with register or genre with a particular register or genre having a certain set of meta-redundancies.  (2) Some choices are available across different registers. Both of these classes are exemplified below.

(1) Inter-register choice

An example of register conditioned choice is the choice between punctuating ‘as you speak’ or ‘as you write’, as discussed by Halliday (1985.b). An example of this choice was given above, (example 3.6) in which either logical-grammatical units or textual-grammatical units were marked.

Crystal and Davy (1969) and Cummings and Simmons (1983) make some attempts to describe the graphology of texts in terms of register and genre.  Typically, this involves looking at the way that grammar is meta-redundant with both typographic organisation and punctuation in different ways for different genres.

Halliday notes that this choice can apply over a whole text, or in a principled way within a text, or may be more or less random.  Presumably the level of correlation of register variables with this graphological choice would form one index of the success of a text.  If a text uses both sorts of punctuation then there is intra-register choice.

(2) intra-register choice

Choice which is made within, instead of by registers can be exemplified by the use of punctuation to suggest rhythm.  The common feature of both the examples below lies in the convention that in reading aloud (and hence in silent reading to some extent) punctuation marks such as commas, colons and stops indicate places to pause. Halliday provides one example - the (by now familiar) “what you hold here is a funk bomb” clause-complex  is another.

Halliday’s (1985.b p.38) example is a  procedural text from a maths textbook:


Take a piece of wood about 25cm long with one edge a straight edge and fix the semi-circle to the wood so that the diameter is along the straight edge. A piece of string about 10cm long has one end attached to the centre of the diameter and a small weight to the other so that it may hang freely.1

The effect of the lack of sentence-internal punctuation in the example is to “hurry the reader along” (Halliday 1985: 38).  Since there are no commas the reader won’t pause.  This constructs part of the meaning of the text.  It would be expected that this meaning might be the responsibility of those higher level systems which position the text interpersonally.

A similar sort of meaning is constructed in example 3.1 in which there is more, rather than less punctuation.  The effect here is the opposite of “hurrying along”, instead, the punctuation cuts the text up, making more pauses.  The rhythm thus created in the writing is reminiscent of the rhythm of James Brown’s lyric delivery.  The meaning here is textual.  In particular, it is intertextual: the form of this text is alluding to the form of another.

(iii) Negotiation

The third way of meaning is made when a text either implicitly or explicitly defines its own system - defined as negotiation.  The essence of negotiated systems is that they can be ‘locally’ systemic, but not conform to any global system. Negotiation occurs between different pairs of strata.  (1) At the lowest level, between the graphological stratum and the typographic stratum: different texts display different meta-redundancies, that is a paragraph may be realised typographically in different ways in different text-types.  (2) Further ‘up’ the strata: texts can negotiate special relationships between the lexico-grammar and graphology in two ways: (a) the graphological system may be enhanced - increased in delicacy, or (b) it may be decreased in delicacy relative to some ‘norm’.  Increases in delicacy must be accompanied by new typographical realisations.  The point is that it is possible to vary meta-redundancies across strata without having to tell the reader, explicitly, for example, “paragraphs in this text are not indented, nor is there any free space, rather the first two words will be rendered in all-caps”.

Graphological to typographical (1) and two sorts of lexico-grammatical to graphological negotiation (2.a,b) are discussed below.

(1) Graphology-typography negotiation

The same graphological function may be realised in a number of ways.  The example here shows this difference in realisation between two versions of the same text2. The first quote is from Lewis Carroll’s manuscript; Alice’s Adventures Underground. Phonological emphasis is represented by underlining - a prosody over a word.  In the second quote, from the book version: Alice’s Adventures in Wonderland, the phonological emphasis is represented by an italic prosody over the relevant word.  This difference is explained by the fact that handwriting is more stylistically limited than the type-face families accessible to a printer. Thus it is a register variable, mode, which conditions the negotiation (cf Halliday & Hasan 1985).


“What can all that green stuff be?” said Alice, “and where have my shoulders got to? And oh! my poor hands! how is it I ca’n’t see you?”

‘What can all that green stuff be?” said Alice. ‘And where have my shoulders got to? And oh, my poor hands, how is it I ca’n’t see you?’

(2.a) more delicacy

The In The Jungle Groove text provides some examples of negotiation between lexico-grammar and graphology. The text has two special classes of proper name.  Song names are in italics, and album names are in all-caps. Thus a more delicate graphological system is introduced without any explicit statement, rather by ‘instantiation’, which is the process by which systems are negotiated.  Typical negotiated systems are usually fairly obvious after only a few instances.

(2.b) less delicacy

Charles Bukowski’s KID STARDUST ON THE PORTERHOUSE, begins like this:


my luck was down again and I was too nervous at this time from excessive wine-drinking; wild eyed, weak; too depressed to find my usual stop-gap, rest-up job as a shipping clerk or stock boy, so I went down to the meat packing plant and walked into the office.

haven’t I seen you before? the man asked

no, I lied.

In this story there are no quotation marks, and sentences do not have initial capitals.  Thus there is negotiation at the lexico-grammatical to graphological level with the lack of quotes; and at the graphological to typographic level with the lack of capitals.  It is possible, in this opening paragraph to distinguish the quotes from the main text, although the typical meta-redundancy between quotation and quotation marks is missing.  This shows another dimension to the term meta-redundant.  In this case marking the quotes off would be redundant.

What is the purpose of this decrease in delicacy? On one hand, it helps to identify the text with respect to other Bukoswki texts.  On the other hand it has some effects on the tension between different readings of the same text.  There is both an increase and a decrease in ambiguity.  There is a decrease in ambiguity in a system without initial capitals on sentences (cf Booth 1987) in that it is possible to distinguish proper names that occur at the beginning of sentences.

More importantly for the story in question, there is increased ambiguity:


the hook which hung from the roof was dull as a mans thumb without a fingernail. you let the bottom of the steer slide back and went for the top, you poked the top part against the hook again and again but the hook would not go through. MOTHER ASS!!! it was all gristle and fat, tough tough tough.

Assuming that the exclamation point can be used on text-pieces smaller than a sentence (see section below) there is tension between at least four readings:


[SENTENCE mother ass] [SENTENCE it was all gristle and fat, tough tough tough]

[SENTENCE mother ass it was all gristle and fat, tough tough tough.]

[SENTENCE QUOTE mother ass] [SENTENCE it was all gristle and fat, tough tough tough]

[SENTENCE [QUOTE mother ass] it was all gristle and fat, tough tough tough.]

So, as well as the intertextual meaning which identifies this text as being part of a genre, the impoverished punctuation serves to increase its richness.  This can be stated anecdotally: there is tension between various readings, imbuing the text with some poetic power.

Summary of meaning making

The above has identified three ways in which meanings may be made.  Unfortunately there is no space here to identify exactly what those meanings are.  It is very important to know how the meanings are made in the interaction of strata, since this will help to evaluate the capacity of NIGEL to implement this model of graphology.  A brief evaluation can be found in the concluding chapter.  The ways that meaning is made are summarised below.

Meta-redundancy is the familiar mechanism of realisation.  It is simply the association between features in one stratum and those in another.  Choice is another familiar concept in systemics, it is the basic mechanism by which the resource of language can be seen in the action of inquiries and choosers. Thus meta-redundancy and choice require no additions to NIGEL.  Negotiation is effected by changing meta-redundancies. This will be implemented in NIGEL with a mechanism for changing and building system networks.  This mechanism has not yet been designed for the grammar, but it has been implemented in the output program (see chapter 4).

3.4  A graphological system

The graphological system itself is presented in figure 3.6.  The rank-scale (DOCUMENT, SECTION, PARAGRAPH, SENTENCE, OGP) is extended to represent a simultaneously accessible way of dividing up text as verse (STANZA, LINE).  This will allow a text to have a simultaneous organisation into sentences and orthographic groups as well as organisation into lines as in poetry.

In the case of the three-line quote that begins In The Jungle Groove these two modes of organisation (‘grammatical’ and poetic) are essentially congruent:


"Put your hand up on the box and feel this,

Lay your hand up there and feel it,

If you got any kind of soul you got to feel it."

An example of a non-congruent division into sentences and lines can be found in example 4.4 in chapter four.

The NIGEL system will make use of this network in the following way.  The system will be entered many times, starting with the feature DOCUMENT, then each time NIGEL generates a piece of text that can be realised graphologically the relevant feature is chosen from the network, right down to the specification of proper-names and words.  The same principle is already used to generate NIGEL’s grammatical structure: the grammatical rank-scale systems are entered repeatedly, building up more of the structure each time a system is entered.  The actual form of the interface code is explained in the next chapter (5).

Constraints on graphology

The Alice examples, used to illustrate negotiation in typographic realisation, have another important dimension of difference: the way that the same grammatical structure is realised in graphology differently.ways:


“What can all that green stuff be?” said Alice, “ and where have my shoulders got to? And oh! my poor hands! how is it I ca’n’t see you?”

‘What can all that green stuff be?” said Alice. ‘And where have my shoulders got to? And oh, my poor hands, how is it I ca’n’t see you?’

The differences between these two versions of example 3.2 could be accounted for by differences in the ‘wiring’, (the organisation) of the network.  In the system network shown here there is nothing to stop an OGP from being realised as an exclamation or a question.  The organisation of the network could be changed to disallow this possibility.  This could be accomplished, in a very general text generation system, by a semantic-level system being realised by a system that makes changes in the graphology network - the same mechanism by which negotiation is to work.


Figure 3.6
A graphological system.

3.5  Conclusion

This body of this chapter has proposed solutions to the three main generative problems identified in its outline:

  1. The stratal model accounts for examples like 3.1: there is a split between lexico-grammar and graphology.  It is not simply a matter of a grammatical sentence being automatically realised as a graphetic, typographic sentence.

  2. The way that phonological emphasis varies between the two versions of Alice, in its manuscript and type-set form can be accounted for by a process of negotiated realisation.  For instance, the graphological system is ‘negotiated’ by modifying its own systems.  The system is established by repeated instantiation of graphological prominence as underling in the manuscript version.

  3. Larger typographic units than the sentence, such as paragraphs, sections and documents have been included in the graphological rank-scale.  Thus Cummings and Simmons’ (1983) hypothetical graphology (quoted above), dealing with punctuation and typography is being implemented.  Waller’s (1980) macro-punctuation provides an framework by which typographic organisation can be integrated with Halliday’s (1985.a,b) work on punctuation.

The graphological system presented here is ‘natural’ with respect to structure.  That is, it reflects the syntagmatic organisation of graphological units.  In contrast to this, the Key system is organised along paradigmatic lines; the structure of the Key network reflects its organisation as a resource, not its organisation in phonetics.  This syntagmatic bias of the graphological system may change as the graphology is extended and integrated into the generation system.  One principle of organisation that might be implemented in the future is that of organisation along meta-functional lines; the network could be separated out into parts that belong to different meta-functions, as transitivity and mood are in NIGEL.

1 A. A. McMullen & J.L. Williams, on Course Mathematics 2, new edition, p.361
2 The texts are not the same throughout, but the wording in this passage is the same in both of them.