Hjemmeeksamen ENG4115 – Texture and Linguistic Structure, høst 2013

Hjemmeeksamen i ENG4115 – Texture and Linguistic Structure, høst 2013. En pdf av oppgaveteksten finnes her. 

Question 1a: Cohesive chains (similarity and identity)

For a series of linguistic signs to be perceived as text (“language that is functional” (Halliday 1989: 10)), it must have unity (Hasan 1989a: 52). According to Hasan, one “source of textual unity” is texture (1989b: 70), which is constituted by “semantic relations between ... individual messages” of a text (ibid.: 71). Semantic relations are realized formally in a text by cohesive devices (Hasan 1989b: e.g. 75, 79–80); provided that cohesive devices are inter­preted as cohesive by the reader/listener (Fries 2004: 22) – i.e., they are understood – they will form cohesive ties between items in a text (Hasan 1989b: e.g. 73, 75). Whenever “a set of items” are semantically related – or tied – throughout the whole or a part of a text, they form a cohesive chain (Hasan 1989b: 84). A text will normally consist of several cohesive chains; these will normally interact. According to Fries (paraphrasing Hasan), “two chains of equi­valent terms interact if the same experiential relation is reiterated between their members” (Fries 2004: 27). As I understand Hasan, the relative strength and comprehensiveness of a text’s chain interaction determines its degree of cohesive harmony (1989b: 93–4); this, finally, is a condition (but, according to Fries (2004: 24), not always necessary) for our experience of a text as being coherent (Hasan 1989b: 94).

Cohesive devices could be either grammatical or lexical; and semantic ties could be either co-referential, co-classificatory, or co-extensive (Hasan 1989b). In sentence 3 in the preceding paragraph (cf. t-unit-numbered list in appendix A), cohesive devices is first introduced. In 4, the same linguistic item is first repeated, and then pronominalized twice with they. All items co-refer to the exact same entity; both lexical and grammatical devices are used. This would thus constitute an identity chain (Hasan 1989b: 84). Two other, purely lexicalized, identity chains are the repeated references to Hasan (1, 2, 3, 4, 5, 8, 9, and 10) and Fries (4, 6, and 10). The former is paragraph-exhaustive (cf. Hasan 1989b: 84), implicitly (parenthetically) and explicitly (according to) projecting what is said. Cohesive chains in 6 and these in 7 are co-referential; but cohesive chain in 5 and two chains of equivalent terms and their in 8 refer to specific (albeit hypo­thetical) instances of the general set, that is, the tie between them is one of co-classification. The former might be regarded as an identity chain and the latter a similarity chain, but this would be counter-intuitive, as all four instances are objects for the same explication, and thus should belong to the same similarity chain. More abstractly, a similar argument could be made for the concept “cohesion”: The actual word is not mentioned, but there are agnate forms and grammaticalizations throughout (3 through 10), creating another near-paragraph-exhaustive similarity chain. Finally, the paragraph abounds with co-extensive terms, items that have synonymic, antonymic, hyponymic, or meronymic sense relations (Hasan 1989b: 80–1): unity (1, 2) harmony (9) and coherent (10) are near-synonyms, as are relations/related (e.g. 2, 3, and 5), tie(s) and chain(s). Text (e.g. 1) and language (1) are superordinates to series of linguistic signs (1), individual messages (2), items (e.g. 5) and terms (e.g. 8); and semantic (e.g. 2, 3, etc.) is in a meronymic (part-whole) relationship with language. These are all examples of similarity chains.

Is there chain interaction in the paragraph? Assuming we have a “unity”-chain (1, 2, 9, 10), this clearly interacts with a “relation”-chain, in a cause-effect relation (e.g. 2 (“unity constituted by relations”), and 9 (“relation determines unity”)). More generally, the Hasan identity chain interacts throughout with the “cohesion” similarity chain.

Question 1b: Theme vs. Given information

Both Theme (e.g. Berry 1995: 64; Thompson & Thompson 2009: 46) and Given (e.g. Prince 1981: 225; 232–3) are contested concepts; definitions seem to depend on theoretical stance and analytical convenience alike (cf. Berry 1995: 64: “For the purpose of this article ...”); the consequence is a profusion of terminology (cf. Chafe 1987: 35–6). To organize this answer, I will take Halliday’s definitions as my point of departure. Two broad (and inter-related) questions are relevant here: what function of language do these terms describe; and what part of language do they delimit.

The clause is “the central processing unit in the lexicogrammar” (Halliday 2004: 10); thematization is the structural organization of the clause (ibid.: 64). All clauses have a Theme^Rheme progression, where the Theme is what “serves as the point of departure of the message; it is that which locates and orients the clause within its context” (ibid.). Halliday defines the Theme as everything up until and including the first experiential element (process, participant, or circumstance; i.e., those elements that together “construe a quantum of human experience”) (2004: 79). So, returning to my example text (cf. appendix A), in sentence 9, As I understand Hasan is the first (circumstantial) element, and thus its Theme; what follows is the Rheme. Statistically, Subjects realized by nominal groups are typical themes in declarative sentences (Prince 1992); this is thus a marked Theme. Further, it is a topical theme; had I written But, surprisingly, as I understand Hasan, the elements preceding the circumstance would also be included in the Theme, the first textual (text-organizing, in this case para­tactically), the second interpersonal (in this case a modal adjunct expressing judgment).

It has been argued (cf. Thompson & Thompson 2009: 54) that the Subject should be included as an unmarked Theme after a marked Theme such as in the above example; the reasoning is that marked and unmarked Themes perform different textual functions. In this case, everything up until and including chain interaction would be part of the Theme. In this particular case, it makes sense: sentence 2, 8, and 9 all thematize a source of the following message; whereas the Subject is in each case what the clause says something about, linking them with the other clauses.

Theme is the “point of departure” of a clause; Given is what the addressee already knows about at any given time in the unfolding of a text (Halliday 2004: 93). Theme and Given belong to two different systems: Theme is a property of the thematic structure of language; Given a property of the information structure (ibid.). The thematic structure of a text is organized by its speaker; the information structure of a text is interpreted by its listener. How­ever, in the unmarked case, the “Theme falls within the Given” (ibid.), which is evident from sentence 3 in my example, where semantic relations is picked up from the Rheme in the previous sentence, and thus a Given for a (sentient) reader; cohesive devices is the salient New. Had I written cohesive devices are what formally realize …, I would have thematized a New, and relegated the Given to the Rheme. Martin and Rose (2004: e.g. 189 ff.) subsume Given under Theme (and Rheme under New); the fact that they include the Subject in their Themes seems to be a function of this choice, as they have discarded a terminology that could distinguish between thematic and information structure, and thus have to make Theme do both jobs. As Subjects typically are Given, this forces them to include Subjects under Theme, regardless of what comes before.

The identity of a Given could be recovered from within (endophorically) or without (exophorically) the text (e.g. Martin & Rose 2007: 169–70): within the text it could have already been explicitly presented, or it could be inferred (Prince 1981: e.g. 236) from the preceding co-text, as in 9 (a text’s chain interaction is an already salient collage from material in 7 and 8, making relative strength and comprehensiveness inferrable (“a chain interaction has general properties, e.g. strength”)); outside the text, it could depend on general, contextual knowledge (the I in 9 may or may not be an example).  

Question 2: Given and New information: Variations across genres

Language is used to communicate information. For information to be considered interesting by its addressee, some of it, presumably, should be new; however, if not to some extent grounded in previous knowledge, new information will be perceived as meaningless:

(1)           A: – Did you hear? They released a new album!
               B: – Who?

(2)           A: – Did you hear? The Rolling Stones released a new album! 
               B: – Cool!

In both examples, speaker A wants to inform B of a new album release; only in (2) is the information successfully conveyed. In (1) (in these admittedly constructed but hopefully realistic examples), B is only concerned with the still elusive identity of the referent for they, and not what they have done; in (2), B – building from previous knowledge that A assumes that B will have – is able to process and then respond adequately to the news.

In this example, the identity of the Given – The Rolling Stones – in (2) is easily retrieved for our appropriately cultured listener. Moreover, there is only one item that could be labeled New, a new album. However, the distribution of Given and New information in a text is rarely this neat. An initial assumption for this essay is that both factors will vary across genres: Given information will more or less easily – and for different reasons – be available to the addressee of a text; and texts will contain more or less New information per linguistic item. Moreover, the terminology in itself, and the theory it builds on, might be too vague to be useful for analyses. For example: On what grounds do we argue that something is Given in a text? To investigate information structure variance across genres, I will analyze and compare two of the texts attached to this exam, Text 1, the opinion piece “My Fellow Republicans, Put Down the Kool-Aid,” and Text 2, the short story “The Sniper.”

Prince (1981) has already established, in a comparison between an informal, oral narrative and a written theoretical treatise on linguistics, that Given information in Subject position is more explicit in the former, and that it depends more on metalinguistic reader inferences in the latter (1981: e.g. 250); and further, that the “size of the entities” differs greatly, being much more complex in the written text (ibid.: 251). Prince, however, has limited her analysis to what she terms discourse entities, objects that are established by a discourse, and realized through noun phrases (1981: 235); I will extend the analysis to cover circumstantial elements and attributes in relational clauses. In example (2), repeated here (including the elided material recovered from the preceding utterance) as (3), B’s response will be considered as New:

(3)            B: [– That they did that is] cool!

Some tentative assumptions, based solely on my intuitive understanding of the genre differences between a short story and an opinion piece, will guide my analysis. Short stories will normally narrate one or a few events, experienced by a limited number of characters – in the case of this particular story, one scene only, narrated chrono­logically. An opinion piece will normally present a perspective on a current topic, perhaps explicitly unpack aspects of it, and perhaps present premises for a suggested alternative course of action. Thus, I will assume that Given items in the short story will remain relatively stable, limited and “clearly” Given throughout the narration, whereas an opinion piece to a larger extent will develop the topic under question, and therefore consist of elements that will be less easy to retrieve.

So, the following assumptions will be tested in this essay:

• A narrative will revolve around a smaller set of Givens with a higher frequency than in an opinion piece;
•  these Given items will be more “easily” retrievable than in an opinion piece; and
• an opinion piece will introduce more complex New material.

It should be added that I have no reason to assume that my two texts are representative for the genres of which they belong; and further, that it would be unreasonable to draw general conclusions about genre differences on the basis of this very limited sample. That said, interesting findings might inspire further research.

In order to be able to conduct a relatively more detailed analysis than one would be able to do with only the categories of Given and New, I will extract a terminological framework from Prince’s very detailed “scale of assumed familiarity” (1981: e.g. 237) for Given and New information. I will, however, start with the broader view on the theoretical justification for categorizing information as Given or New, and then gradually focus in on Prince’s contribution to the discussion and my own evaluation of and extractions from this. As mentioned under question 1b, the distinction between Given and New information pertains to a text’s information structure (Halliday 2004: 88): It is a question of semantics, of the meaning of the discourse as it is construed by its addressee. The system of information organizes text into information units, which (based on my understanding of the discussion in Halliday (2004: 89) and Chafe (1987: e.g. 37)) I will consider to be the minimal unit wherein an interplay between Given and New information will occur. According to Halliday (and contra the claim on his behalf in Prince 1981: 226), the information unit “is a structure made up of two functions, … an obligatory New element plus an optionally Given” – optional because it could be elided and thus inferred as it often refers to something in the immediate co- or context (2004: 89). A special case will be at the commencement of a text, which naturally often will have completely New items (ibid.: 89). In written texts, in the unmarked case an information unit will overlap with the clause; in spoken discourse, as I understand Chafe (1987: 37), it will typically overlap with the intonation unit. The Given will typically, but not necessarily, precede the New (ibid.).

A problem with this general overview is how we go about deciding on a specific definition for the relative information status of an item, and how we can describe how and why an item receives a certain status. Disappointed with the theory on the information unit and the motivation for its terminology, Prince (1981) proposed a more refined taxonomy for items’ information status. Under the heading of assumed familiarity – the degree to which a speaker/writer can assume that his/her addressee is familiar with a concept used in a text – she divided linguistic items into an elaborate taxonomy (1981: 237). Fully aware that this has later been refined, I will still use her terminology and definitions as a foundation for the analysis below.

Prince distinguishes between items that are New, Inferrable, and Evoked. The category of Evoked (E) is further divided into Textually and Situationally Evoked. These are the items we would most clearly recognize as Givens in a text:

(4)            The Rolling Stones just released a new album. They shouldn’t have.

       (5)            Mother addressing child, indicating a shattered vase: – Was that you?

In (4), They are textually evoked, the referent retrievable from the preceding clause; similarly, have indicates (evokes) the entire predicate of the preceding clause, thus becoming an elided Given in the second sentence (and the proposition that what they have done should not have happened a New). In (5), the retrieval of the referents of that and you depends entirely on the situational context. For the addressee, both will be Evoked.

The “medial” category, Inferrable (I), refers to items that must be inferred from the co-text:

(6)            The Rolling Stones just released a new album. The first track is great.

Tracks is in a meronymic relationship to album (“an album consists of tracks”), and its meaning is therefore inferrable (and other meanings, e.g. railroad-related, are de-selected by the preceding clause). In Chafe’s terminology, this would be an example of a concept (tracks) that is semi-active after the preceding discourse (1987: e.g. 28), and therefore can be used as a “starting point” (ibid.: 37) in the following sentence (in which the attribute great is a New).

The category of New is divided into Brand-new and Unused. A Brand-new (BN) item in a text must be constructed by the hearer; an unused item is “already in the hearer’s model” (1981: 236), but new to the discourse. In my example (2) above, The Rolling Stones, according to Prince, would be categorized as New: Unused (U), whereas a new album would be Brand-new. However, on this point I will depart from Prince’s taxonomy; rather, I will collapse her Unused items under the category of Evoked. This is in part motivated by e.g. Martin and Rose’s (2007: 170–1) description of homophora, i.e. “communal reference” to culturally – rather than, but parallel to, Prince’s situationally – given entities. Moreover, I believe this choice is supported by Prince’s own familiarity scale (1981: 245), repeated in a simplified version here as example (7):

         (7)            E > U > I > BN   

       

The scale presupposes a speaker assuming what concepts the listener is already familiar with. If possible (decided by degree of listener familiarity), an entity to the left on the scale will be used before an entity to the right on the scale. So the sentence

(8)            – The band I used to dig when I was young just released a new album.

where the NP the band … is Brand-new, would only be permissible if uttered to a person unfamiliar with the band in question. If the addressee is already familiar with the referent for the NP, an item higher on the familiarity scale would be used, in this case probably the name of the band. And in that case, and even though it would be Unused in this context, it would still be treated, and understood, as Given information.

Finally, I will introduce a category of my own, which is more or less complex Brand-new or Inferrable material. This distinction is also tentative and based on a common-sense understanding of a difference between for example written and spoken discourse, or more or less complex discourse. Compare:

(9)            I shall try to bring out the plurality, priority, and problematic (empirical) status of functions in speech

       (10)          and she saw a lovely little teddybear

Example (9) is from the data in Prince (1981: 248), the treatise on linguistics; example (10) from the data in Hasan (1989b: 72), a narrative about a girl and a teddy bear. The extracts are not chosen in any way that could conceivably be called scientific, only in order to exemplify the assumption I will test on my material. The Given items are Evoked in both cases, situationally in the first and textually in the second. However, the point here is the complexity of the New information: (10) contains one Brand-new, teddybear, with two attributes; (9) contains Inferrables in two stages: functions in speech, from the preceding co-text; and the qualities the writer believes are attached to that (plurality, priority and problematic (empirical) status), which may or may not be naturally inferrable to the reader. I will assign a cut-off point for this category, and say that a Brand-new or Inferrable consisting of three or more separate lexical words will be categorized as complex (general labeling items like thing or group will not be counted).

Before embarking on the analysis, I can now rephrase my initial assumptions in the established terminology. These are as follows:

• A narrative will revolve around a smaller set of Evoked concepts with a relatively higher frequency;
• an opinion piece will gradually develop its topic and thus have a larger number of Inferrables; and
• an opinion piece will introduce more Complex Brand-new and Inferrable material.

I have analyzed the first twenty sentences from each text, in both cases what I perceive to be the orientation and the beginning of the main narrative/argumentation (refer to appendix B for the encoding of the excerpt from “The Sniper,” and C for the encoding of “My Fellow Republicans”). I have used t-units as my first unit of analysis; hypotactically embedded clauses were included in the unit, whereas paratactically joined clauses counted as separate units. The second (and main) unit for analysis was experiential concepts. I have extracted all participants and circumstances on t-unit level (but, for clarity, retained the units’ natural syntagmatic structures in the tables in the appendices), and categorized each unit according to its assumed familiarity. The clauses with extraposition in “My Fellow Republicans” were analyzed both on matrix and on subclause level (cf. Herriman 2000: 207). I have assumed the familiarity of a model reader in each case: Kool-Aid and GOP will be household names in USA, and (most importantly) are treated as such by the author.

Admittedly, the encoding was extremely difficult; the results must be taken with a grain of salt of similar in magnitude to the one suggested in Prince (1981: 248), especially regarding the opinion piece analysis. The result from the analysis is represented in Table 1.

	total items	Evoked		Inferred		Brand-new		Complex
		no	%	no	%	no	%	no	%
“The Sniper”	51	17	33.3	19	37.2	15	29.4	9	17.6
“My Fellow Republicans”	54	15	27.8	28	51.9	11	20.0	8	14.8

Table 1: Items categorized according to relative familiarity in “The Sniper” and “My Fellow Republicans”; real numbers and percentages.

The most obvious difference between the two texts is the relative share of Inferrables. Slightly more than half of the items in the opinion piece depended on mainly textual inferences made by the reader. It could perhaps also be suggested that the short story contains a larger amount of Brand-new items; several of the items that were categorized as Brand-new in the opinion piece would likely be considered either Evoked or Inferrable by some or many of its readers, depending on the extent to which one is familiar with current political trends (the opinion polls mentioned (sentence 8) being the most obvious case in point).

As is evident from the table, my suggested difference between more or less Complex Inferrables and Brand-news yielded did not differentiate between the genres; it may have been an assumption that in this case does not correspond to an underlying reality, or else a function of the encoding, or the suggested cut-off point.

I also coded the Evokeds for variation in referent. The numbers show that the short story has ten different Evokeds; of these, one item, the hero of the story, is the referent for eight of the items, whereas the other nine refer to one item each. In the opinion piece, there were also ten different Evokeds; however, in this case the distribution was somewhat different: one Evoked had four references (“the divide”), and one had three (“Fellow Republicans”); the rest had one each.

Before I return to my assumptions for a final discussion of my results, I will again stress how challenging I have found this analysis, both with respect to the actual process of coding the data, and to the categories as such, their overall functionality, and my grasp of them. I have tried to suggest alternative categorizations for some of the items in the tables in the appendices. To take just a couple of examples: 

          (11)          Around the beleaguered Four Courts the heavy guns roared

This was coded as E^BN. I reasoned that Four Courts is culturally Evoked; but this is not the case for many of us, and certainly not for this reader. Thus, for me it would rather be Inferred, as I would reason along the lines of “we are in Dublin, this is a city, it must contain boroughs and the like, Four Courts is written in capital letters, is introduced without any explicit explanations, and must thus be a an area or quarter within the city of Dublin.” How do I know (i) if this was the intention of the author, and (ii) whether this (then, today) is the received reading of this item? Moreover, my analysis was unable to account for the modifier, beleaguered, which is probably the more important here, the first indication (excluding the title) of the fact that there is a Civil War. If accounted for, the heavy guns would then rather be counted as an Inferrable (beleaguering entails guns); but even so, there would not be room in the taxonomy (as far as I can tell) for the non-neutral description of the sound of the guns. A similar problem is encountered in sentence 5 (…spasmodically, like dogs barking on lone farms) where the simile both have objects of its own (shall they be categorized?) and is definitely a non-neutral and highly elaborate description of an occurrence, thus adding perhaps several layers of concepts – new information – not only to what it locally describes, but to the overall narrative. My taxonomy is blind to this.

A further example from “My fellow Republicans”: the title:

        (12)          My Fellow Republicans, Put Down the Kool-Aid

Conceptually, both (or perhaps all three) real-world referents will be familiar to the regular reader of this column; However, the meaning of the proposition expressed in this title will probably be new, or at least surprising, to all of us – or perhaps not: Many will read through this and nod after every sentence: “He puts into words what I have said these last many years.” So is then the proposition (for some, or perhaps for most (model readers), perhaps at least to a certain degree) Inferrable from the cultural context? And how do we categorize this proposition in the first place?

With this extended caveat, I will return to my three assumptions. I will be so bold as to claim that my analysis has lent some support to the first, namely that a narrative will revolve around a very limited number of Evokeds, which will have a relatively high frequency. In addition to what I found, there are several references to the same participant that were not coded in the analysis, as these functioned as determiners for other objects. Moreover, one could speculate that this tendency would be even more pronounced if a larger excerpt had been analyzed: There is clearly a break in the story, beginning with sentence 7, between an orientation stage and a “narration proper” stage: In the first sentences, the setting (Dublin in darkness) is presented in some detail; from sentence 7 and onwards, the focus is on the sniper.

Regarding my other assumption, that opinion pieces will develop a topic and thus revolve around less stable referents that to a larger extent need to be inferred, it would seem at first blush that the data support this notion. However, I would be very hesitant to make any conclusions on this basis, for two reasons: first, that the material was so difficult to code, and that I am so reluctant to trust my own categorizations; and second, that I suspect that there could be a number of factors explaining (perhaps better) these results. One factor that springs to mind is that opinionated writing to a larger extent seeks to activate the reader, to draw on reader experiences, likes and dislikes, opinions and affinities, and that this might better explain that a reader is expected to engage more closely with this type of text.

Finally, my assumption that opinion pieces would introduce more complex material than a short story failed to be supported by the data. In this case, I will not discard the common sense assumption that complexity varies (greatly) across genres; this was also a finding in Prince’s study mentioned above. A plausible reason for this finding could also be that opinion pieces and short stories will have a comparable amount of my category of Complex, but for different and genre-specific reasons: In narratives, for extensive descriptions of scenes (cf. sentence 3, appendix B); in opinion pieces, for elaborate arguments (cf. sentence 15, appendix C). The challenge here is to develop suitable categories to describe and explain what we all seem to know is the case.

Based on the experiences from my brief excursion into what I at this stage perceive to be a quite complex and complicated field of inquiry, I will conclude by suggesting that an information structure analysis of written texts – and especially more complex, written texts – can not be complete without categorizing all experiential concepts – using a more multi-layered approach than I was able to do here. More specifically – and from the little I have read, I perceive this to be a neglected field – I believe it would inform any analysis to also include the relative familiarity of processes, that is, to what extent processes add new information into discourse, or create clues for the understanding of concepts introduced later.

Bibliography

Berry, M. (1995). Thematic Options and Success in Writing. In M. Ghadessy (Ed.), Thematic Development in English Texts (pp. 55–84). London: Pinter Publishers.

Chafe, W. (1987). Cognitive Constraints on Information Flow. In R. S. Tomlin (Ed.), Coherence and Grounding in Discourse: Outcome of a symposium, Eugene, Oregon, June 1984 (pp. 21–51). Amsterdam: John Benjamin.

Fries, P. H. (2004). What Makes a Text Coherent. In D. Banks (Ed.), Text and Texture: Systemic Functional Viewpoints on the Nature and Structure of Text (pp. 9–50). Paris: L'Harmattan.

Halliday, M. A. K. (1989). Context of situation. In M. A. K. Halliday, R. Hasan, & F. Christie (Ed.), Language, context and text: aspects of language in a social-semiotic perspective (2nd ed ed., pp. 3–14). Oxford: Oxford University Press.

Halliday, M. A. K. (2004). An Introduction to Functional Grammar (3rd Rev. ed.). (C. M. Mathiessen, Ed.) London: Arnold.

Hasan, R. (1989a). The structure of a text. In M. A. K. Halliday, R. Hasan, & F. Christie (Ed.), Language, Context, and Text: Aspects of language in a Social-semiotic Perspective (2nd ed., pp. 52–69). Oxford: Oxford University Press.

Hasan, R. (1989b). The texture of a text. In M. A. K. Halliday, R. Hasan, & F. Christie (Ed.), Language, Context, and Text: Aspects of language in a Social-semiotic Perspective (2nd ed., pp. 70–96). Oxford: Oxford University Press.

Herriman, J. (2000). The functions of extraposition in English texts. Functions of Language, 7(2), pp. 203–30.

Martin, J. R., & Rose, D. (2007). Working with Discourse: Meaning Beyond the Clause (2nd ed.). London: Bloomsbury Academic.

Prince, E. F. (1981). Toward a Taxonomy of Given–New Information. In P. Cole (Ed.), Radical Pragmatics (pp. 223–55). New York: Academic Press.

Prince, E. F. (1992). The ZPG Letter: Subjects, Definiteness, and Information-status. In W. C. Mann, & S. A. Thompson (Eds.), Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Text (pp. 295–325). Amsterdam: John Benjamins.

Thompson, G., & Thompson, S. (2009). Theme, Subject and the unfolding of text. In G. Forey, & G. Thompson (Eds.), Text Type and Texture: In Honour of Flo Davies (pp. 45–69). London: Equinox.

Appendix A: First paragraph question 1a, as numbered t-units

1.	For a series of linguistic signs to be perceived as text (“language that is functional” (Halliday 1989: 10)), it must have unity (Hasan 1989a: 52).
2.	According to Hasan, one “source of textual unity” is texture (1989b: 70), which is constituted by “semantic relations between ... individual messages” of a text (ibid.: 71).
3.	Semantic relations are realized formally in a text by cohesive devices (Hasan 1989b: e.g. 75, 79–80);
4.	provided that cohesive devices are interpreted as cohesive by the reader/listener (Fries 2004: 22) – i.e., they are understood – they will form cohesive ties between items in a text (Hasan 1989b: e.g. 73, 75).
5.	Whenever “a set of items” are semantically related – or tied – throughout the whole or a part of a text, they form a cohesive chain (Hasan 1989b: 84).
6.	A text will normally consist of several cohesive chains;
7.	these will normally interact.
8.	According to Fries (paraphrasing Hasan), “two chains of equivalent terms interact if the same experiential relation is reiterated between their members” (Fries 2004: 27).
9.	As I understand Hasan, the relative strength and comprehensiveness of a text’s chain interaction determines its level of cohesive harmony (1989b: 93–4);
10.	this, finally, is a condition (but, according to Fries (2004: 24), not always necessary) for our experience of a text as being coherent (Hasan 1989b: 94).

Appendix B: Excerpt from “The Sniper.” Concepts coded according to assumed familiarity.

Legend:

E:         Evoked

I:          Inferrable

BN:     Brand New

C:        Complex

A subscript number attached to E (E₁, E₂, etc.) indicate participant number.

T-unit no	Concepts	Assumed Familiarity
	The Sniper	BN
	The long June twilight faded into night	BN (C) BN (or I: night is preceded by twilight)
	Dublin lay enveloped in darkness but for the dim light of the moon that shone through fleecy clouds casting a pale light as of approaching dawn over the streets and the dark waters of the Liffey	E1 I I (C) I (C) I (C)
	Around the beleaguered Four Courts the heavy guns roared	E2 BN (or I: shares superordinate with sniper)
	Here and there through the city machine guns and rifles broke the silence of the night, spasmodically, like dogs barking on lone farms	I I I I (C)
	Republicans and Free Staters were waging civil war	E3 E4 I
	On a rooftop near O'Connell Bridge, a Republican sniper lay watching	I E5 E6
	Beside him lay his rifle	I I
	and over his shoulders was slung a pair of field glasses	I BN (or ?I: snipers have field glasses)
	His face was the face of a student, thin and ascetic,	I BN BN
	but his eyes had the cold gleam of the fanatic	I BN (C) (or I: guerrillas are fanatical)
	They were deep and thoughtful, the eyes of a man who is used to looking at death	E7 BN I (C)
	He was eating a sandwich hungrily	E6 BN
	He had eaten nothing since morning	E6 BN
	He had been too excited to eat	E6 BN (or I: snipers on post are excited)
	He finished the sandwich	E6 I
	and, taking a flask of whiskey from his pocket, he took a short drought	BN (C) E6 I
	Then he returned the flask to his pocket	E6 E8 E9
	He paused for a moment, considering whether he should risk a smoke	E6 BN BN (C)
	It was dangerous	E10 I

Appendix C: Excerpt from “My Fellow Republicans, Put Down the Kool-Aid.” Concepts coded according to assumed familiarity.

T-unit no	Concepts	Assumed Familiarity
	My Fellow Republicans, Put Down the Kool-Aid	E1 E2 (but perhaps BN if the entire proposition is considered)
	It's hard to understand how one person or a small group of people can convince others to do something that will end badly for them.	BN BN BN BN BN
	It is still hard to fathom how the cult leader Jim Jones could have persuaded so many intelligent people to drink poisoned Kool-Aid.	I E3 I I
	It should always have been obvious to everyone who followed him into the jungle that it would turn out badly.	I BN (or E, if one is familiar with this event) I (or E, if one is familiar with the fallout of the event)
	As a Republican watching a handful of conservative extremists push to either defund Obamacare or shut down the government, it was obvious from the start it wouldn't end well.	E4 (but here my taxonomy fails to account for the complexity) I E5 I
	To make such a prominent and painful mistake makes Republicans like me worry for the survival of our party as a national force.	I E1or 4 or both I (or BN, if this is inferrable neither from the preceding text nor the current state of affairs) E6
	The problem is, the only people who don't recognize the shutdown  as a disaster for the GOP are the people who came up with the strategy.	I BN (C) I (or E, if one accepts that these are the same as the previous people)
	Despite polls showing the American people overwhelmingly disapprove of closing the government, as well as surveys showing that forcing the shutdown may well have put the party's House majority at risk in next year's mid-terms, there are still some who defend the shutdown strategy as a great idea.	BN (or E, if one is familiar with these polls) BN (or E, if one is familiar with these surveys) I (C)
	The attempt to link the repeal of Obamacare with the threat of a shutdown has unearthed a fault line that separates Republicans from each other.	E7 I (C)
	The divide has nothing to do with ideology.	I BN (or I for the perceptive or already agreeing reader)
	It 's not about moderates vs. conservatives.	E8 I
	It has nothing to do with policy:	E8 I
	All Republicans agree Obamacare is bad policy.	E1 (probably) E9 (or BN, or the entire proposition is simply untrue)
	And it isn't about the Tea Party vs. the Establishment.	E8 I
	The real divide is between those who believe Republicans must grow the party to survive and win and those who believe we don't need to expand the party, we just need to excite the base.	E8 I (C) (or BN, if this is not inferrable from the preceding, or E, if this is common knowledge)
	No amount of polling will convince those who are content with pandering to the base that what they are doing is damaging the party.	I I (C)
	As long as the conservative media, the conservative Twitter-sphere and the conservative chattering class is happy with the "strategy," it must be working.	I (C) E10 I
	It was the "true believers" who could not see that the attempt to defund Obamacare was doomed from the start.	I I (C) I
	Even though polls clearly show the American people disagree with our approach, their views are ignored.	I I I
	What really matters to the base feeders is whether hashtags like "Make DC Listen" are trending on Twitter.	I BN (C) (or E, if one is familiar with these trends)