Talking Topically


clac 3/2000



Angela Downing


Universidad Complutense de Madrid



1.              Introduction


                  Talking topically in spontaneous conversation is both a cognitive and an interactive process. It is not something predetermined and already known to the participants in the conversation, but an achievement worked out by their mutual negotiation of the topic. Topics are constructed, not pre-existent. This means that some things we wanted to say may not get said because the topic flow has taken a different direction from that of our previous intervention. In Harvey Sacks’ words ‘talking topically doesn’t consist in blocks of talk about ‘a topic’. When you present a topic you can be assured that others will try to talk topically with what you’ve talked about, but you can’t be sure that the topic you intended was the topic they will talk to.’ (1992:762, lecture April 17, 1968)[1].


                  The analysis of topic in conversation is a great challenge for discourse analysts, for computer scientists, for those interested in the social construction of coherence, and in topic as a conceptual and cognitive domain. In this chapter I will outline and comment on a few of the issues related to topicality that have engaged linguists attention in conversation and dialogue. This will involve exploring such areas as topic introduction and consolidation, how speakers keep topics alive, the questions of discourse coherence, interestingness and point of view, topic relevance, topic drift and finally how gender may condition topic choices in conversation and the way they are handled. I am not concerned here with how topics are developed or understood in written texts.


                  Approaches to topic have proliferated in recent years. Functionalist and cognitive grammarians view topic as closely associated with the (prototypical) subject of a clause or sentence, while recognising it to be a discourse-pragmatic category (Givón 1993, Langacker 1991, Dik [1989] 1997). Among discourse analysts, by contrast, interest in topicality has followed the shift in recent decades from a focus on the sentence to a focus on discourse. The concentration on a possible ‘sentence topic’ has given way among discourse analysts first to that of ‘utterance topic’, or ‘speaker’s topic’ and then to the mutual construction of a ‘discourse topic’ in conversation by two or more participants (Linell & Korolija 1997; Chafe 1997, among others). A parallel shift in research on topic, both in written text and in conversation, has been the move from what Goutsos (1997) calls the what perspective to the how perspective: from the debate on topic as ‘aboutness’ to how it is manifested and signalled. Interest here centres on the segmentation and signalling of topic sequences: how topics are introducted, developed and closed over a stretch of discourse (Schegloff and Sacks, 1973; Button and Casey 1984; Downing et al 1998, among others). In an dynamic cognitive view of topic, Chafe 1994 and Chafe 1997 sees the discourse topic as aligned to thought and consciousness. Whether speaking or silently thinking, our focus of consciousness is constantly shifting from one thought to another, each focused idea becoming verbalised in an intonation unit. A large focus cluster cannot be activated all at once. It remains in the semi-active consciousness and for Chafe constitutes a discourse topic, through which interlocutors navigate until they judge that its content has been adequately covered (1997:42). In this respect, the construction of topic accounts for the directionality of conversation.


                  For discourse and conversation analysts, working with transcriptions, a discourse topic may be identified for purposes of analysis as either ‘global’ or ‘local’, in what is now a recognised as a useful hierarchisation. Global discourse topics cover the whole discourse or a segmented part of it, referred to as an ‘episode’ (Linell & Korolija 1997), while local discourse topics are subsumed via a coherence relationship under the umbrella of the global topic (van Oosten 1985, van Dijk 1977, van Dijk and Kintsch 1983). This conceptualisation of discourse topic is not restricted to conversation. The negotiation of topicality within the collaborative structuring of conversation, and its relation to coherence are taken up by Downing (forthcoming). In this respect coherence resides not simply in what participants say, but how they say it and when they say it (Schegloff 1990), since coherence is seen as a collaborative achievement (Bublitz 1999). Sequentiality and placing are therefore all-important.



2.              Starting up a topic


                  Topical talk involves turn-taking and, as insofar as its interactional mechanisms are concerned, can be conveniently examined within the turn-taking systematics as described by Sacks, Schegloff and Jefferson (1974). Basically, these involve taking a turn, holding a turn and yielding the turn. The topic framework or message content is usually built up after the preliminary greetings and polite remarks have been made. This is particularly noticeable in telephone conversations, which have clear openings and closings and may consist of one single topic identified as the reason for calling. There are therefore expectations on the part of the called person for the ‘first topic slot’ to be filled once the greetings are over; to such an extent that the first topic to be introduced by the caller will be interpreted by the hearer as the reason for calling (Schegloff and Sacks 1973). In face-to-face spontaneous conversation, a number of different topics may be introduced successively, especially when speakers fall silent, indicating that the current ongoing topic has been exhausted. Alternatively, topics may not be segmented from each other in this way but instead merge one into the other by shifts and drifts. These will be discussed in later sections of this chapter.


                  Speakers make use of two main strategies for the purpose of introducing a first topic or indeed a subsequent topic: they either inform, by means of an informative statement, or they enquire, by asking a question. Basically, the difference is an interpersonal one: by informing, the speaker assigns him/herself the role of topic supplier, or controller of topic, whereas by questioning or more exactly by eliciting a topic this role is offered to an interlocutor in the discourse.


2.1. Informing.

                  In using the informing strategy, one speaker provides an informative statement of something considered to be ‘newsworthy’ (Button and Casey 1984; cf Sacks’ ‘tellability’, Schank’s and Chafe’s ‘interestingness’). This is typically followed by a response, minimally an attention marker such as Mm, Did you? or  Really? provided by the other speaker; following this the topic is consolidated in the third stage, as in the following extract from part of a conversation which hinged on the subject of model trains:


                  (1)           A.            I got a good bargain this morning at the model train shop.

                                    B.             Did you?

                                    A.            Yes. I bought an engine. Really beautiful, and it was almost half the

                                                      price of the one in the shop near home......(topic continues)

                                                                                                                                                                  (author’s data)


                  The function of the minimal response is to collaboratively ratify the proposed topic by a ‘take-up’ signal which demonstrates the hearer’s attention and willingness to hear how the proposed topic is put forward. Geluykens (1992) holds that for English, the ‘take-up’ in the form of at least a minimal response is essential if the topic is to proceed. It is probably true that, in English contexts, where acknowledging the other speaker’s autonomy is a cultural script (Dirven and Verspoor 1998:152), if the response is not forthcoming, speakers will tend to repeat the introductory statement on the assumption that their partner hasn’t heard it the first time. In other language communities the requirement of topic ratification may be a less strong cultural constraint, giving rise either to an extended topic development by speaker A or to the introduction of a rival topic by speaker B such as ‘ Oh I got a bargain at the sales last week’.


2.2. Eliciting.

In eliciting a topic one speaker nominates a possible topic, again something tellable or newsworhty, by means of a question:


                  (2)           A.           How did you get on at the interview?


Such an elicitation counts as an invitation to B. to develop the topic of his/her experience of the interview. It is obvious that a question like this depends heavily on shared knowledge.


2.3. Stereotyped expressions

                  Less specific variants of the informing and eliciting strategies involve the use of stereotyped introductory expressions which alert the hearer to an upcoming topic. For instance, general informing statements starting with the injunction ‘Guess what!’ or ‘You’ll never guess what happened this morning on my way to work’ fulfil the double function of prefacing a new topic and of indicating by means of presupposition that something interesting happened and is therefore tellable.


                  Similarly, the speaker’s invitation to the hearer to provide a topic can be packaged as an inquiry of the most general kind about some newsworthy event, upto then unspecified. This is done by means of conventional formulators such as:


                  What’s new? What’s new with you?

                  What have you been up to lately?

                  What’s doing? What y’re doing?

                  Anything to relate? Anything to report?

                  What d’y’ know (AmE)


                  These are Button & Casey’s ‘topic initial elicitors’ and they are particularly sensitive to the initial stage of a conversation. They perform various functions: to segment the talk into topical chunks and to indicate the speaker’s availability for topical talk, while at the same time having no commitment to start up a particular topic of one’s own. A further feature is that they are geared to current events and provide an open though bounded domain from which events may be chosen by the hearer as possible topics (Button and Casey 1984:176).


2.4.         Getting restarted on a new topic

                  Later on in the conversation, after topic boundary turns, for instance after a silence, or the winding up of a topic, similar formulating phrases, as well as the more specific nominations, can be used to introduce or elicit further topics. Because they occur now not at the absolute beginning of a conversation but after previous topics have been talked about, the word ‘else’, ‘(an)other’ or ‘further’ typically occurs, in the informative statement or the elicitation:


                  There’s something else I have to tell you. (announcing upcoming topic)

                  Another thing, Sam. (announcing upcoming topic)

                  Anything else/further to report? (eliciting topic)

                  What else have you been doing these days? (eliciting topic)


                  Finally, topic introduction can be effected in this way even when the closure of the conversation has been initiated. This is most clear in telephone calls, where closures tend to be quite elaborate, as in (3):


                  (3)           N.            Anyway...                                                (pre-closure)

                                    H.            Anyway...           

                                    N.            So...

                                    N.            You’ll come about eight, right? (check future arrangements

                                    H.            Yeah                                                           (confirm)

                                    N.            Okay                                                           (moving into closure)


                                    -N.          Anything else to report?               (elicitation of new topic)


                                    H.            Uh...m..

                                    -H.          Getting my hair cut tomorrow

                                    N.            Oh really­?                              (conventionalised response which shows interest

                                                                                                            and invites the previous speaker to continue on topic)

                                    (continues on topic)       (adapted from Button and Casey 1984:168)


2.5.         Positive or negative responses to general topic elicitors

                  The responses seen so far to general inquiries about some interesting event have been positive, insofar as the interlocutor has something to say. But a positive response is not guaranteed:


                  (4)           N.            What’s doing?  (caller)

                                    H.            Ah, nothing        (called)                  (adapted from Button and Casey 1984)


Superficially, it might appear that the responder has two options after one of these general enquiries to produce a possibe topic: to provide a ‘newsworthy report’ or to produce a ‘no-news’ report. But since a ‘no-news’ report is likely to lead to a breakdown in communication, or at the very least throws the ball back into the first speaker’s court, a no-news report is, in CA terms, the ‘dispreferred’ option. Awareness of this fact is a matter of socialisation (children easily reply with ‘Nothing’), but adult conversationalists will make an effort to produce a potential topic even after a lengthy pause, as illustrated above, where  H. eventually adds the following piece of information:


                  (5)           H.           Getting my hair cut tomorrow.


                  Overall, then, the devices outlined here are used recursively throughout conversation, and serve to implement the negotiation of topic between interlocutors at key points in the conversation: after greetings, after a silence and after a previous topic has been wound up.


3.              Keeping a topic alive


                  Chafe (1994) sees elicitation as one of the two main ways of keeping a topic alive, the other being narration. In elicitation, the topic is driven forward (more exactly, Chafe considers that movement through the topic is driven forward) by the interaction between the elicitor and the responder. By contrast, a self-sustaining topic, which requires little or no interaction, typically takes the form of a narrative. We will turn to narrative in section 3.3. But first, a glance at the question of coherent responses as topic development.


                  Topic development of great interest to Artifical Intelligence specialists: how to teach a computer, who doesn’t know anything about it, to react appropriately, that is, to produce acceptable responses, on-topic.


Taking part in a conversation is not an obviously  hard task to the participants. On the surface it seems no more difficult than reading a story. In fact it would seem to be easier. Children conduct conversations before they can read, whereas reading comprehension is something to be studied and improved upon. Nonethless,...conversation is one of our most challenging problems. (Schank 1977)


                  Since Schank, an AI specialist, wrote these words, enormous advances have been made in our understanding of the linear structure of conversation, its sequentiality and systematicity, especially as regards the turn-taking system, the negotiated openings and closings of conversations and the relationship of conversational moves to different kinds of speech acts.


                  Likewise, topicality in conversation has made some progress insofar as it has been possible to correlate the initiation and closure of global discourse topics with the turn-taking system, as outlined in my previous section. Even the functionality of non-lexical markers such as um, er, ah, well etc in the staging of topicality has been recognised (cf Downing et al 1998).


                  Yet when we turn to the central conceptual question of the coherent development of a topic, things are not so clear. Topicality in a spoken genre such as conversation involves seeing (and for the analysts, modelling) how speakers stay on topic in conversation, gradually shifting local topics while saying something new and perceivably constructing a global domain. 


3.1. Topic and thought: Early explorations by Schank

                  Schank himself gives the clue to the problem when he expresses his firm conviction that ‘there exists a strong relationship between the rules of conversation and the rules of association of thought. When a speaker receives an input he begins to think about it. If he chooses to make a response to the input his choice represents one of the many possible thoughts he could have had in response to the input’(p.429).In other words, what you can say is a part of what you can think; that is: ‘The rules for conversation are a subset of the rules of thought’(ibid.). The problem then in modelling topicality in conversation is one involving cognition, together with sequentiality, since one speaker follows another in developing the topic. We need some kind of cognitive categories in order to model what kinds of things speakers can say while still staying on topic.  


                  In addressing this problem, Schank starts out with the important if obvious point that it is not appropriate to respond to sentences (i.e. utterances) with just anything that comes to mind. While there is no one ‘correct’ response to keep a potential topic going, there certainly do exist inappropriate responses. Consider the responses (B1-6) to A’s informative statement:



                  (6)           A.            I just bought a new hat.                                   B.1. Fred eats hamburgers.

                                                                                                                              B.2. I just bought a new car.

                                                                                                                              B.3. There is supposed to be a recession.

                                                                                                                              B.4. My hat is in good shape.

                                                                                                                              B.5. What colour?

                                                                                                                              B.6. Why are you telling me this?


                  In these hypothetical responses, Schank does not distinguish between cognitive relevance and social appropriateness. My own evaluation of the responses is as follows:


B1: cognitively irrelevant, and therefore incoherent.

B5: cognitively relevant, coherent and socially appropriate, as an inquiry relevant to an object named in A.

B2, 3 and 4: Cognitively disruptive and socially inappropriate: B2 and B4 are ‘put-downs’ which replace A’s proposed topic by B’s rival proposal. B3 is a reproach while B6 is a challenge, both uncollaborative.


                  Schank goes on to suggest a provisional definition of topic as follows: A topic is any object, person, location, action, state, or time that is mentioned in the sentence to be responded to. (This is an oft-repeated definition, which Schank will prove to be wrong, thus stated.) He proceeds to test the definition by means of a statement (7A), followed by a series of responses (7B) that maintain one or other of the topic types specified in the definition:


                  (7)           A.           John bought a red car in Baltimore yesterday.


Picking up each semantic constituent in turn, he offers:


                                    B.             i.               Cars contain motors.

                                                      ii.              John is a fisherman.

                                                      iii.            Baltimore has a good football team.

                                                      iv.            Buying things is fun.

                                                      v.              Red is quite close to pink.

                                                      vi.            Yesterday was cloudy.                 


None of these makes an appropriate response topic-wise to A, whereas the following one does:


                                                      vii.           You mean he’s not going to buy my car?


                  Why is this? Here Schank doesn’t answer this question directly. My own view is that Schank’s provisional topic rule is insufficient. No single referent named in A’s statement is itself the topic.  A’s topic is, I suggest, the whole event itself, expressable as an entity by means of a nominalisation such as ‘John’s purchase of a car in Baltimore yesterday’. Consequently, responses about Baltimore, yesterday and cars in general fail to guarantee coherence. Schank in fact identifies the clausal repesentation of the event in (7A) as an action statement, and an action statement can be followed by a statement about the effect of that action on the hearer.  This is what happens in the statement-question sequence, repeated here for convenience as (8a and ((b):


                  (8a)                          John bought a red car in Baltimore yesterday.

                  (8b)                          You mean he’s not going to buy my car?


                  From these two utterances together we can derive potential topics for future responses. That is, a sentence said in response to an input must at the same time stick to the proposed topic and introduce potential new topics, thus sticking and shifting at the same time (1977:424). This new rule replaces the provisional one which had proved inadequate.


A Potential Topic has two parts. The first part he calls the Reduced Old Topic, which derives from the input conceptualisation. The second part is a new conceptualisation that comprises the new topic shift. From the potential topic a new topic for a response is derived.


The set of concepts in A’s statement that can be used for the topic is:

                  (John buying, car (red), yesterday, Baltimore)


In the following figure which represents vii, (You mean he’s not going to buy my car?) the elements that have been paid attention to are on the left, and the new elements are on the right:


                  Fig.1.    (John buy, car)______________(someone, buy, car (mine))


This topic is then an addition, not a substitution. We keep something old and add something new. (Sentences i to vi by contrast were merely substitutions.) The conversation can branch at this point in either of the two directions, attending to either the first half (reduced Old Topic) or the second half (New topic), illustrated respectively as follows:


                  (9)           A.            Well, John needed a car in a hurry.

                  (10)        A.            No, he didn’t like your car.


                  Sentences, by this analysis, don’t have topics in isolation. Only conversations have topics, and even they only have what amount to rules for topic shifts. Sentences have potential topics that delimit the range of structures spoken in response to them (op.cit.425).


At this point it is necessary to introduce more global notions of topic: Schank introduces the concepts of SUPERTOPIC and METATOPIC, a view entirely in consonance with the cognitive and psycholinguistic models of van Oosten and Tracy among others. A supertopic is present when both the input sentence and the response contain concepts that share the same superset. For instance when an input sentence and its response share two events that are alike in kind such as BUYING/SELLING CARS, as in Fillmore’s frame concept of the commercial event, then the general event is appropriate and may be called up as a general statement. It may never surface at all,  but the important thing is its availability as a topic for discussion, as I suggest as follows:


                  (11)        Selling used cars is always a problem.

                  (12)        Buying used cars is always a risk.


                  It is worth mentioning here that supertopics are frequently generic statements, as in my two examples above. But so was the response in Bi: Cars contain motors, and this proved to be unacceptable. Why is this? The explanation again resides in the concept of ‘frame’. Examples (11) and (12) belong to the [BUY] frame, whereas Cars contain motors does not.


                  Schank’s other new concept METATOPIC. This is a comment that is inferred from the interaction of the two conceptualisations, Old and New. In the case of (vii) the metatopic might be one of the beliefs brought up by vii, such as ‘John’s betrayal’ (John has let me down’) or ‘John’s dishonesty’. It is clear that such beliefs are derived from our inference that John had promised to buy his friend’s car; this information is not stated as such but we can infer it. A metatopic can then be added to the topic development, and can be verbalised as in:


                  (13)        Yes, John does that kind of thing all the time. Do you remember when he

                                    promised to buy Al’s house?


This could then produce  a response which addresses the Supertopic BUSINESS DEALS and the new Metatopic FEELINGS as in:


                  (14)        Well, business dealings among friends often cause hard feelings.


                  It is clear that categories such as Supertopic and Metatopic are necessary. Speakers do not simply make unrelated topic-comment statements about persons, things and events, but also generalise, evaluate and exemplify, drawing upon vast amounts of contextual knowledge in their dynamic construction of topic.


                  It is impossible in the space available to summarise the wealth of Schank’s ideas. Let one further insight suffice: what I interpret as the relevance of shared cultural environments and even idiosyncratic features of individuals to the requirement of ‘interestingness’ in a proposed topic and ‘what can follow naturally’. Compare for instance the tables of responses for I JUST BOUGHT A CAMEL and I JUST BOUGHT A CAMERA in respect to the ‘associational categores’ listed on the left (Schank op.cit.:435):


                                                                        A.            I JUST BOUGHT A CAMEL


USER                                                         (I)            I didn’t know you were a camel rider

FUNCTION                                          (2)           What a you going to do with a camel?

OWNER                                                  (3)           I know someone else who has a camel.

AVAILABILITY                               (4)           Where did you find a camel seller?


                                                                        B              I JUST BOUGHT A CAMERA


USER                                                         (1)           I didn’t know you were a photographer.

FUNCTION                                          (2)           What are you going to do with a camera?

OWNER                                                  (3)           I know someone else who has a camera.

AVAILABILITY                               (4)           Where did you find a camera seller?


                  These invented responses illustrate how on the one hand, appropriateness, and on the other, the quality of ‘novelty’ or ‘interestingness’ as opposed to ‘dullness’ vary according to the shared cultural background. In our culture everyone knows someone who has a camera, where to buy one and what one might do with a camera. Not so with respect to a camel. Consequently, the responses in A about camels are appropriate in this culture, whereas those in B about cameras are absurd.


                  Arguably, the main thrust of Schank’s paper is to demonstrate that staying on topic is not a matter of making just any free association, as was illustrated in (6). It is not even a matter of merely ‘picking up’ the reference to any person, object, place etc. mentioned in the previous utterance. I argue that without relevance the topic progression will not be successful. It is perhaps strange then that Schank did not include as possible responses to John bought a car in Baltimore yesterday simple continuations such as Is he pleased with it? or I didn’t think he could afford a new car. Such responses are natural and appropriate. They also happen to confirm the cognitive grammarians’ view that the most salient topical entity is assigned subject function, while - for Givón at least -  the second most salient is assigned to object. But this we can say with the benefit of hindsight. Schank is right, however, in that too much free association must be avoided or we risk losing our listeners. ‘If we travers links, we lose the trend of the conversation for our listener.’


3.2.         Topic activation, interestingness and point of view.


                  For Chafe (1994), discourse topic is also lodged in cognition, and is ‘an aggregate of coherently related events, states and referents that are held together in some form in the speaker’s semi-active consciousness’. The focus of active consciousness scans the potential topic in a way similar to our visual scanning of a picture, picking out - activating - first one aspect and then another. For this reason, subsequent scannings of a picture, like subsequent readings of a text, can active different perceptions, details, interpretations. Not everything we scan becomes verbalised in conversation. Here Chafe’s view resonates with that of Schank: what we express linguistically is only a part of what our minds evoke.


                  And it is here that the factor of interestingness comes in again. “People will often verbalise a topic when they judge that it will be interesting to their interlocutors” (author’s emphasis) (1994:121). Chafe sees ‘interestingness’, both in life and in conversation, as lying in the unexpected. Our minds are equipped to cope with the unexpected, by confronting unexpected events with the mental representations of experiences we have stored in memory. Both in life and in conversation, when nothing unexpected happens, when there is no excitement, boredom sets in. A story that conforms to all our expectations is no story (cf Hidi and Baird 1986). This is presumably what happens when the response to ‘What’s doing?’ is ‘Nothing’; to verbalise the totally expected would have no point (cf Schank 1982). The topics worth verbalising are those that have a point, which in Chafe’s terms means a conflict with expectations; for Schank the ‘point’ is something like the relevance of an utterance to the interests of the interlocutor.


                  A conversational topic is usually verbalised from the point of view of some referent, a natural consequence of the fact that consciousness is oriented from the point of view of an experiencing self (Chafe, 1994:133; Chafe 1997). Other terms used to describe this phenomenon are ‘perspective’ (MacWhinney 1977; Givon 1999), ‘empathy’ (Kuno and Kaburaki 1977; Kuno 1987), ‘viewpoint’ (DeLancey 1981) and ‘viewing arrangements’ (Langacker 1991). Perspective is related to choices in syntax which had proved recalcitrant to explanation on formal grounds. It is the speaker’s own point of view that is most typical, but the point of view of another referent may be adopted. These possibilities have considerable relevance for written fiction.


                  Four types of evidence for point of view are provided by Chafe: i) the sheer frequency with which a referent is activated; ii) the point of view referent is favoured as a starting-point and is verbalised as subject; iii) mention of emotions or evaluations, which are directly available only to the consciousness from whose point of view something is being told; iv) the referent in question serves as the deictic centre (1994:135).


3.3.         Topical relevance in narrative


                  Grice’s maxim ‘Be relevant’ was taken up by Karen Tracy  and applied to the now generally accepted notion of a hierarchisation of topics, namely a macro or global topic and a micro or local topic. Tracy speculates on what appear to be conflicting criteria for relevance. Taking a local approach, a relevant remark would be one that chains to something in the last sentence or two of a speaker’s message, generationg the rule: “Make your remark relate to the last thing your partner talked about”. Following a global approach, a relevant remark would be one that responds to the main idea in a speaker’s message, in which case the rule would be “Make your remarks relate to the main point of your partner’s talk”. For instance, commenting topically on a narrative or story, one can either address the issue (an abstract statement of feeling or principle) or one can latch on to the event (a particular instance, example or episode that occurred in the past) in Reichman’s (1978) terms, quoted by Tracy.

                  Working in experimental conditions, Tracy used the following text:


(15)        I ran into Sara the other day; she seems like she always does - strange. We went camping with a couple

                  of other people one weekend. We’d been canoeing all day and everybody was exhausted. We got off the

                  river when it began to rain and put up our tents. It was terrible. Everybody was snapping at each other.

                  Then Sara decided to go off in the pouring rain and explore. The rest of us wanted to get in our sleeping

                  bags and get warm but Sara wanted to explore. Elli ended up going with her because we didn’t want her

                  to go alone and we couldn’t persuade her to wait till morning.


                  Applying Tracy’s terms to this narrative episode, the issue would be something like “Sara is a strange person” and the event “A camping holiday with Sara”. Subsequently, Tracy provides a set of four extensions, which informants were asked to rate for goodness as continuations of the text:


                  (a)           Yeah, she does stuff like that. She was at my apartment for dinner once and

                                    in the middle of the meal she decided that she needed to be alone so she got

                                    up and left.

                  (b)           I don’t think she’s as weird as she used to be. Ever since she and Frank split

                                    up she has seemed more normal to me.

                  (c)           Ellie and I went camping with Sara over Labor Day and had a great time.

                  (d)           There’s nothing worse than being wet when you’re camping, I went on a trip

                                    with a friend and came back early because the weather was so miserable.


                  According to the two differing criteria, the global predicts that extensions of the issue as in (a) and (b) will be seen as more relevant, while the local criterion predicts that extensions of the event, as in (c) and (d) will be more relevant.


                  We can now ask ourselves the following questions:

Q.1.        Which do you judge to be the more relevant extension(s)?

Q.2.        Do you think either type could occur naturally? (The text and extensions seem to be


Q.3.        Would you consider that what Sacks calls ‘good conversational practice’ consists of

                  mainly attending to the issue, or mainly latching on to a part of the event?


                  Tracy argued that the global rule is more likely to characterise conversational practice because of its greater consistency with discourse processing principles, presumably the top-down approach as set forth by van Dijk and Kintsch (1983). Informants supported this claim, but not very strongly. A second experiment was devised in which two contextual factors were posited as potentially important: first, perceived understandability of the main point of the speaker’s message; second, the perceived importance of the message to the speaker, that is the speaker’s emotive involvement.


                  Regarding the first condition, Tracy predicted that if the main issue was not clear, speakers would feel free to latch on to the more peripheral event. Regarding the second, it was predicted that if the issue was perceived as less important to the speaker, that is with low speaker involvement, extensions of the event would be likely.


                  These predictions were not borne out in the experiment. The results yielded strong support for the global rule, even with low comprehensibility and low involvement, in other words,  even when the organisation of the text, or some missing background information, made comprehension of the global issue more difficult, or when speaker’s involvement was low.


                  It would be interesting to check not only how informants judge extensions of both types as, but what they actually do when they produce an utterance as a topical follow-up of that of another speaker. A contrastive study of Spanish and English conversation might yield useful results.


                  I now turn briefly to two related notions: stepwise topic development and topic drift.


4.              Stepwise topic development


                  Harvey Sacks was not in favour of topic breaks:


The best way to move from topic to topic is not by a topic close followed by a topic beginning, but what we call a stepwise move. Such a move involves connecting what we’ve just been talking about to what we’re now talking about, though they are                 different. And as far as anybody knows we’ve never had to start a new topic, though we’re far from wherever we began and we haven’t talked on just a single topic’(Lecture 5, spring 1972; 1992:566.)


                  Such a procedure for Sacks would count as good conversational practice, as opposed to what he calls a ‘lousy conversation’, which is characterised by the occurrence of a large number of specific new topic starts  (ibid.). Even worse, in his opinion, are the non-conversations which never get past the exchange-of-greetings stage: ‘How have you been? Fine. How have you been? Fine.’  Sacks observes that ‘the stepwise kind of movement is the most routine thing, and ordinarily involves nothing particularly moticeable...Any next utterance is built in such a way as to be on topic with the last. That then can become a thing that is used to make jumps.’ February 17 1971, 1992:300-301).


                   Now if I inderstand him correctly, Sacks is describing two slightly different topical operations here. On the one hand, he appears to be referring to the development of a discourse topic by means of the introduction of related objects, what he calls ‘subtopical talk’, for instance, talk about a house to rent leads to talk about the yard of the house 1992;761. In effect, related objects such as house and yard could crop up naturally within the wider discourse topic of renting a house. 


                  Rather different is the deliberate provision of topic jumps by using the stepwise mechanism to ‘touch off’ a piece of information or the report of an experience, which a speaker may have been holding in store, waiting for a suitable moment to ‘fit’ it into the conversation. This is what happens in the following passage, based on Sacks’ transcription for Feb.19, 1971; 1992: 298:


                  The conversation involves a young couple (Bill and Lori) who have been to Mexico and are now being visited by their parents, Ethel and Ben). A good deal of the conversation hinges on food:


(16)        Bill:        They had some nice things. They even had like chicken liver. Y’kown, chopped chicken liver... I tasted it, it wz really horrible ‘n Lori said I better not eat it becuz...

                                    Lori:      I didn’t trust them  (     )

                                    Bill:        They probably (     )

                                    Ethel:    Lori and Bill I have something tuh tell you you probably heard about it already but just in case

you haven’t, you must not use any’ve the pottery you picked up fer any-tuh cook in. Or fer

anything edible, - havent’chu heard the news?

                                    Bill:        - from Mexico

                                    Ethel:    Becuz they have discovered, thet the coating, the glaze is so thin...


                  Ethel goes on to describe the possibility of lead poisoning. (298)


                  According to Sacks, this passage shows how speakers can rely on the natural course of conversation to touch off in their memory things they wanted to say. So instead of calling up Bill and Lori immediately when she heard about the potential risk of lead poisoning, Ethel was able to keep in mind the fact that they shouldn’t be using the pottery in order to place it when the conversation provided an opportunity.


                  In fact speakers are adept at steering the conversation so as to provide just such an opportunity. In the same conversation, Ben, sticking to the topic of Mexican food, says “I was gonna make some chile sauce..... but I didn’t have time.... I worked last night....I was up all night fighting fires so I went to sleep when I got home”. Bill says ‘You had a fire last night?’ Ben says ‘yeah’ and they talk about it (1992:299). Here Ben, who is a fireman, has led up to what he wanted to talk about - the fire the previous night and his experience of fighting it - via the stepwise device of introducing first the topic of making chile sauce, and the reason he didn’t do so. The diagram below illustrates the stepwise move.


                                    Mexican food


                                                      Mexican pottery


                                                                        using Mexican pottery for cooking could cause lead poisoning

                                                                                          intention to make chile sauce


                                                                                                            reason for not doing so - fighting a fire

                                                                                                                              talk about the fire


5. Topic Drift


                  A more recent attempt to uncover how a conversation ends up far from where it started is made by Hobbs (1990). He proposes an analogy of this phenomenon with a word game by which one is supposed to turn one word into another one letter at a time, with each intermediate step being itself a word. For instance, in seven steps one can change WIMP into JOCK:











                  Similarly, it may happen in conversation that each segment coheres with the preceding one and the following one, but there is soon little or no relation between early and later topics. This Hobbs calls topic drift. In contrast with the deliberately strategic nature of the stepwise movement as illustrated by Sacks, topic drift is aimless and less controlled. Furthermore, while stepwise development exhibits a certain duration, topic drift tends to move rapidly from one locally related topic to another.


                  In (17) from Hobbs’ data, Sam and Jeff dispatch in quick succession the topics of parents making them study, to disciplining them, to Sam’s father not liking him:


                  (17)        Jeff: On Friday, Saturday and Sunday I can go out to play. That’s all.

                                    Sam: Do you ever get a spanking?

                                    Jeff: Uh, my mother spanks me real hard.

                                                      My father, he’d do anything to keep me from...

                                                      He doesn’t care about me either.


                   Hobbs’ aim was to explore to what extent topic drift may be a matter of cognition or of convention. That is, by implication, whether control of topicality increases with cognitive maturity or is a result of social training and influences. His method was to use non-surreptitious videotaped recorded data of young speakers of different genders and age groups, from 8 to 16. All pairs were best friends. They were all told to talk about something serious and/or intimate.


                  Hobbs’ conclusions regarding the influence of cognition vs. convention are not conclusive. To some extent, the children and adolescents all handled the three coherence relations identified by Hobbs as semantic parallelism, chained explanations and metatalk. If topical coherence does increase with age, Hobbs asks, does it happen because older children are capable of longer stretches of coherent thought, or is it because  adults come to demand more highly structured talk from them? Not surpisingly perhaps, Hobbs feels that the probably answer is ‘Both’. The child is gradually made aware of conventions, and will be criticised or challenged even by peers for not being relevant. At the same time the child’s cognitive ability increases and enables him/her to comply with those conventions.



                  Nor does he address the gender question in depth. It is clear, however, that the boys, both younger and older, brought up a lot of topics of short duration,(bikes, video games, their grades at school, disciplining i.e. getting spanked by their father/mother, problems with their girl friends, with drinking) while the girls talked about fewer topics, mainly about being friends, at greater length. This difference in itself may be the reason why the girls’ talk gives the impression of greater coherence than does the boys’ which either jumps or drifts from one topic to another.


6. Gender differences in talking topically


                  The gender issue is taken up by Deborah Tannen in a fascinating article in the same volume. She addresses two elements that create integrated coherence in conversation: physical alignment and topical cohesion. Physical alignment involves the positions people adopt, such as the way they sit in relation to each other when talking, and includes eye gaze. Topical cohesion involves how people introduce and develop topics in relation to their own and others’ talk. Her analysis of videotapes reveals an extraordinay parallel between physical alignment and topical cpherence and an equally striking contrast between the genders.


                  Tannen used videotapes of four pairs of same-gender friends of different ages, some of them the same pairs of children and adolescents as those analysed by Hobbs. Right across age groups, she reports the contrast between the girls and boys to be ‘startling’ and ‘staggering’ as regards both body alignment and the handling of topical cohesion. These differences were maintained in the behavious of the pairs of young adults, even taking into account their increased cognitive competence. 


                  The girls and young women sit still, facing each other directly and close, and gaze at each other while talking, occasionally glancing away, and occasionally touching each other. They seem to be collected into the space they inhabit. By contrast, the young boys are extremely restless, both in fidgety movements and in eye gaze. The older boys and men are physically more still, but like the younger ones, they spread out in their chairs rather than gather themselves into the space they occupy. They do not sit facing each other but place their chairs at angles, and they do not fix their gaze on the other’s face; nor do they touch except in playful aggression.


                  As regards topical cohesion, Tannen’s analysis confirms that of Hobbs. The boys and men had more difficulty in adapting to the experimental conditions of sitting down and talking about something serious or intimate. Some pairs looked aimlessly around and took objects in the room, including the camera, as brief topics. The younger boys either switched abruptly or drifted from one topic to another, with considerable silences in between their numerous topics. The young men did talk on a ‘serious’ subject - marriage - from a relatively impersonal point of view. When they did discuss personal concerns or problems, each boy and young man tended to focus on his own problem, downplaying the importance of the friend’s problem.


                  The girls and young women talked on fewer topics, which were all personal, and explored them in greater depth. They did not use objects in the room as resources for talk. Their main concerns were their own friendships between themselves and other girls and how friendships are broken by disagreement. They also speculated whether they might be adopted and whether their parents might divorce.


                  Evaluating these and other differences, Tannen does not adopt a judgmental stance. She avoids claiming that the girls and women are more involved and ‘engaged’ than the boys and men when they talk. Rather, she claims, both genders are involved but they display different cultural patterns for displaying their involvement. This is the most likely explanation of the most frequent complaint by women about men- that men don’t listen to them and always talk on their own topics. And yet that is not the whole story. Head-on eye-gaze sustained between men is said to be a sign of aggression. But not between the genders. Perhaps the anwswer then is in cross-gender talk to catch the eye and hold it. Then the topics will flow.


7. Conclusion


                  In this essay I have traced in brief outline some of the key issues in talking topically. Of these the question of starting up a topic is the most straightforward. More problematic, from the point of view of analysis and modelling, is the question of HOW speakers stay on topic while contributing something new, HOW they produce acceptable responses which stick and shift at the same time, while producing something tellable (Schank 1977). Further questions were what do speakers respond to - the issue they infer to be the global discourse topic, or something immediate in the co-text, such as the last thing said. Or do they pick up illustration of the issue: the event ? They may do either; and they may do something else, which is to respond to the perceived feeling or underlying motivation of the previous speaker (Tracy 1984:460), something similar to Schank’s ‘metatopic’. They may also challenge the last speaker’s utterance. If a response is perceived as uncooperative or socially inappropriate, the remark will be seen as less relevant.  This leads us to the interesting question whether topic relevance and the ability to develop and sustain topics is determined purely on a cognitive basis or is interwoven with social judgements and gender differences, a theme which is likely to be taken up again in further research.


8. References


Atkinson, J.M. and J. Heritage, eds. (1984). Structures of Social Action. Cambridge: Cambridge University Press:

Bublitz, W. (1999). Introduction. Views of Coherence. In Bublitz, Lenk and Ventola, eds. Coherence in Spoken and Written Discourse. Amsterdam/Philadelphia: John Benjamins, 1-7.    

Button, G. and Casey, N. (1984). Generating topic: the use of topic initial elicitors. In Atkinson, J.M. and J. Heritage, eds.:167-190.

Chafe, W. (1994). Discourse, Consciousness and Time. Chigaco: University of Chicago Press.

Chafe, W. (1997). Polyphonic topic development. In T. Givón, ed. 1997:41-54.

Delancey, S. (1981) An interpretation of split ergativity and related patterns. Language 57: 626-57.

Dik, S. ([1989],1997)edited by Kees Hengeveld. The Theory of Functional Grammar, Part I, The Structure of the Clause; Part 2, Complex ad Derived Constructions. Berlin/New York: Mouton de Gruyter.

Dirven, R. and Verspoor, M. 1998. Cognitive Exploration of Language and Linguistics. Amsterdam/Philadelphia: John Benjamins.

Dorval, B. ed. (1990). Conversational Organization and its Development. Advances in Discourse Processes 38. Norwood, N.J.: Ablex.

Downing, A. (forthcoming) Negotiating topic coherence through talk-in-action.

Downing, A., Neff. J.A., Carretero, M., Martínez-Caro, E., Pérez de Ayala, S., Marín, J.  y Simón, J. (1998) Structuring and signalling topic management. In S. Embleton, ed.  Lacus Forum XXIV. Chapel Hill, North Carolina: The Linguistic Society of Canada  and the United States: 267-278.

Fillmore, C. (1975). An alternative to checklist theories of meaning. In C. Cogen, H. Thompson, G. Thurgood and K. Whistler eds. Proceedings of the Berkeley Linguistic Society. Berkeley: Berkeley Linguistic Society, 123-31.

Geluykens, R. (1992). From Discourse Process to Grammatical Construction. On left-dislocation in English. Amsterdam/Philadelphia: John Benjamins.

Givón, T. (1993) English Grammar. A Function-Based Introduction. 2 vols. Amsterdam  /Philadelphia: John Benjamins.

Givón, T. (1999) The usual suspects: The grammar of perspective in narrative fiction. Estudios Ingleses de la Universidad Complutense 7, 1999:11-48.

Givón, T. ed. (1997). Conversation. Cognitive, Communicative and Social Perspectives. Amsterdam/Philadelphia: John Benjamins.

Hobbs, J. R. (1990). Topic drift. In B. Dorval, ed. 1990: 3-22.

Hidi  and Baird  (1986). Interestingness - A neglected variable in discourse processing. Cognitive Science 10:179-194.

Kuno, S. and Kaburaki, E. 1977. Empathy and syntax. Linguistic Inquiry 8: 627-72.

Kuno, S. (1987) Functional Syntax: Anaphora, Discourse and Empathy. Chicago and London: The University of Chicago Press.

Langacker, R. (1991) Foundations of Cognitive Grammar. Volume II. Descriptive Application. Stanford, California: The Univeristy of Stanford Press.

Linell, P. and Korolija, N. (1997) Coherence in multi-party conversation. In Givón, T. ed. 1997: 167-205.

MacWhinney, B. (1977). Starting points. Language 53:152-68.

Sacks, H. (edited by G. Jefferson). (1995). Lectures on Conversation. Oxford: Blackwell’s.

Sacks, H., Schegloff, E. and Jefferson, G. (1974). A simplest systematics for the organization of turn-taking in conversation. Language 50, 4, 696-735.

Schank, R.C. (1977). Rules and Topics in Conversation. Cognitive Science, 1:421-441.

Schank, R.C. (1982) What’s the point? Cognitive Science 6, 255-275.

Schegloff, E. (1990). On the organisation of sequences as a source of “coherence” in talk-in -interaction. In Dorval ed. 1990: 51-77.

Schegloff, E. and Sacks, H. (1973). Opening up closings. Semiotica 7.4. 289-327.

Tannen, D. (1990). Gender differences in conversational coherence: Physical alignment and topical coherence. In B. Dorval, ed. 1990: 167-206.

Tracy, K. (1984). Staying on topic: An explication of conversational relevance. Discourse Processes, 7: 447-464.

van Dijk, T. and Kintsch, W. (1983) Strategies of Discourse Comprehension. San Diego: Academic Press.

van Oosten, J. (1985). The Nature of Subjects, Topics and Agents: A Cognitive Explanation. Bloomington, Indiana: Indiana University of Linguistics Club.



© Angela Downing. Círculo de Lingüística Aplicada a la Comunicación/ Circle of Linguistics Applied to Communication 3, September 2000. ISSN 1576-4737. Published in Downing, Angela, Jesús Moya, and José Ignacio Albentosa, eds., Text and talk. Cuenca 2000: Universidad de Castilla-La Mancha, 31-50.


Back to contents 3.


    [1]I am grateful to the Universidad Complutense de Madrid for the grant received under the Del Amo Programme, which enabled me to carry out the final version of this study at the University of California at Santa Barbara.