Ted Briscoe interviews Gerald Gazdar
Education Teaching Pragmatics Generalized Phrase Structure Grammar
Natural Language Processing Lexical Knowledge Representation Linguistics Conclusion
Left Up Right


EJB Was there much work on issues like presupposition in NLP in the early 1970s? I tend to think of somebody like Bonnie Webber as being the first real attempt to tackle that stuff in a way that was informed by what was going on in formal semantics and linguistics. But I guess that the issues and the phenomena were known to the Schank ian/Wilks community?

GG  Only the very basic phenomenon - that is the Strawson observation that there are these things that don't seem to behave quite like normal entailments. That was certainly known in that community. The compositionality complexities that made presupposition an interesting topic for me and other linguists were really discovered by Langendoen & Savin (1971). And the person who did most to drag them into the light of day was Lauri Karttunen in a series of papers starting in 1970 - so that stuff hadn't percolated through to the AI community.

EJB When did you first meet Karttunen then? Did you actually meet him prior to writing your thesis?

GG  No. I first met him in 1977 or 1978.

EJB It is quite clear, looking at the paper (Gazdar 1979b) which is a condensed version of the material on presupposition in the thesis, that this is very very formal work. Indeed, I remember as a PhD student looking at the book (Gazdar 1979a) and it reminded me of looking at Montague's stuff. It is nothing like as daunting as Montague but much of the feel and notation is still there and it is formal in a sense that your current work is not. It is extremely ambitious - there are promissory notes about ``here are some of the rules that one would need to do this'' but to actually define exhaustively the rules about projection and so on would surely be a massive task and there is no sort of intention to do that. So what was the value of the formalization in that work?

GG  Well, the promissory note, the one you have just mentioned, I don't think that was a massive task. It is just that that task was tedious, those things were not difficult. That bit of it, that was just identifying presupposition sources. And Karttunen (no date) had produced a comprehensive list of three dozen or so. I gave three examples of the kind of rules that were needed but if one had wanted to do the other 33 one could just have sat down and written them out.

EJB And used the Karttunen list as the source?

GG  Yes, but it would have been boring. Actually, there was a deep problem lurking in there to do with scope. But I wasn't even aware of it at the time and it didn't figure in any of Karttunen's earlier papers on presupposition projection. The other aspect of it was that in order to do it you had to adopt a view of syntax and I really wanted to keep the dissertation as syntax neutral as I could. So I didn't really want to start saying ``I will assume a such and such syntax'' so I think when I did the three examples I did, I did them on surface strings or something - I can't quite remember how I did them.

EJB That's right - the rules have a sort of flavour of ``if you see the word that'' or something without any kind of discussion of the fact that that has to be a complementizer. It was extremely bleached.

GG  That was deliberate because it wasn't interesting and because in order to do it properly one had to make commitments which I didn't want to make.

EJB But the other people who were working on this, like Karttunen - were they equally careful about not committing themselves to a particular approach? My sense was that people who were doing that kind of pragmatics assumed the vast theoretical apparatus, TG and so on as well. Was that not the case?

GG  No, not really. Karttunen's plugs, holes and filters approach was really not dependent on anything in TG. He barely mentioned it, as I recall. And, of course, he and Peters subsequently presented it in Montague Grammar terms (1975, 1979). I was working at a time when TG was the lingua franca - in a way uncontroversionally - whereas now it may still be the lingua franca, but it is controversial. I didn't really believe in it. By that stage I had become rather sceptical but, given what I was doing, the scepticism was irrelevant. It wasn't worth bringing it up as it would have been a distraction.

EJB Right. So what was the real value of the formalization? What was the - I'm not quite sure what the right word is - practical isn't quite the right word - but what were the consequences of formalizing the story about presupposition and why was that a major goal of the work?

GG  The proposal was quite a complex one - to have expressed it precisely in English would have been difficult. I think I could do it now but I think at the time I would have found it very difficult to have stated what the analysis was in unambiguous English. It was a lot of hard work to learn what I needed to learn in order to formalize it, but once I had done that work then it was straightforward to state the analysis using that default union operator.

EJB That's the earliest appearance of the notion of default in the broadest sort of sense in your work? Is that right?

GG  Yes. There wasn't a lot of work on defaults before then.

EJB I was going to say - even in terms of work on logic - there wasn't a lot of work on nonmonotonic logic or belief update or anything like that in the early 1970s. When I look at your formalization now, it looks somewhat familiar but I'm sure that when I looked at it as a PhD student it was one of the things which completely and utterly left me incapable of appreciating what you were doing. Although I had done a partial undergraduate philosophy degree and I think was quite well trained in logic, certainly by comparison with other people doing linguistics, that stuff was beyond the ken, certainly for me. That was stuff that you pretty much invented, yes?

GG  Yes, there wasn't anything. I wouldn't have invented it if there had been something around. You know, if something like Reiter 's (1980) default logic had been around, I would have used that. I wasn't in the business of making life hard for myself. But there wasn't anything.

EJB Did anybody pick up on the work you had done in the nonmonotonic community? I don't know that literature perfectly, but I don't remember seeing references to your work on pragmatics as being a source of inspiration for the work on nonmonotonic logic.

GG  No, I think it was unknown to any of them. Later one of Reiter's students did essentially recode my analysis using Reiter's logic (Mercer & Reiter 1982).

EJB What about Thomason ? Was Thomason aware of any of your work on pragmatics?

GG  Yes. Thomason was. I made contact with Thomason some time in the mid 1970s and have kept in touch with him intermittently ever since.

EJB Was he in any way influential in you coming up with that formulation or was that something that you really did invent ex nihilo?

GG  I was being supervised by Hans Kamp , of course. I had one of the best logicians in the world looking over my shoulder telling me ``no, that is not going to work'' in the nicest possible way. I wasn't just freelancing in Reading. I was seeing Hans Kamp on a fortnightly basis.

EJB So Hans was in as good a position as anybody to really help you with that at the time?

GG  Absolutely.

EJB There is this nice, very brief statement in the 1979 paper that we need to order the update according to whether we are dealing with implicature, presupposition, general context.

GG  There are four things: general context, two sorts of implicature and presupposition.

EJB Right, and the theory doesn't explain the ordering but the ordering gets you the right results. If you were to revisit that now would you have anything more to say about that?

GG  No. Actually, when I reread it recently, I thought that I had put it just right. I'd confessed, I hadn't pretended that there was some elaborate motive.

EJB I wasn't trying to accuse you of intellectual dishonesty but it's clearly a very deep problem in that whole area and one that is still a major issue amongst people who do nonmonotonic logic. There are certain proposals around which purport to make the ordering follow from a more semantic rather than syntactic element of the theory, but they are not widely accepted as I understand it.

GG  I think that van der Sandt (1988) cast some light on it.

EJB And really between your book (Gazdar 1979a) and Rob van der Sandt's there is a long lacuna when very little was done on this as far as I can see. Is that right?

GG  Well, there was still a certain amount of work on presupposition. There was Scott Soames (1979) who was working at around the same time as me. He independently invented essentially the same theory of presupposition projection. He did some more work after I had stopped working on the topic. And there were various people who elaborated versions of the Karttunen & Peters (1975, 1979) analysis so I think you can find one dissertation a year from the late 1970s onwards on average.

EJB So that there's more there than ...

GG  So that there's more there than you probably know about. My own view of it is not terribly positive. I think a lot of people read what I'd done and what Lauri and Stan had done and thought ``God, it must be possible to do better than this'' which is an entirely rational reaction. But it actually turned out that they weren't able to do better but nevertheless they still had to write a PhD and both traditions thus continued. It wasn't that one tradition defeated the other. People intuitively felt either that Lauri and Stan had it basically right or they felt that I had it basically right and they built on one or the other. From my perspective, van der Sandt was the first thing where I felt ``hey, he's done it better - this looks good''.

EJB And his work does seem to have sparked some more activity and made the topic a little more high profile again, at least to someone like me who is not deeply involved in that stuff. But you've never felt that need to go back to this again, although I think it was an impressive achievement to do a PhD on such an ambitious topic remote from the kind of people who were in the best position to really give you feedback on it.

GG  Yes. I abandoned pragmatics in 1979 although some papers may have dates later due to publication lags. I gave a paper at a conference in Italy on a topic in pragmatics in 1978 or 1979 and the paper was awful and I thought ``this is it, I'm not going to do this again''. It wasn't a paper on presupposition - I can't remember what it was about but it was a paper in the general area of pragmatics. After I had given the paper, at least one member of the audience very politely told me it was awful. Larry Horn , I think, asked some question which revealed the full awfulness of the paper. So I just thought ``well, forget that, I've done pragmatics, I'll do something else''.

EJB So it wasn't that you necessarily felt that you didn't have the right skill set to do it well, just perhaps that you had done it for long enough and you needed to move on to something else to keep your interest engaged.

GG  Well, it was partly that. It was also to do with the problems in pragmatics that I was conscious of at that time, notably issues connected with speech acts. I had tried to make headway on them and I had failed and I didn't think that I wanted to invest any more effort in failing. So I thought that I would work on something easier.

EJB As well as the work on presupposition, there is the work on the semantics/pragmatics interface around issues like the meaning of the true functional connectives and so on, and the more general story about the relationship between semantics and pragmatics. The formalization of the Gricean maxims set up a clear research programme, and the paper that you published in the Journal of Pragmatics (Gazdar 1980b) states that position extremely clearly at a point, I would guess, when nobody else had stated it quite so straightforwardly. So that seems like quite an important contribution to me, although it is clearly not your own unique one. But you were stating that view at a time when most people felt I imagine that the Gricean view was extremely descriptive rather than theoretical.

GG  That paper was essentially pedagogical.

EJB But not necessarily at that time. I would have thought that most people reading Journal of Pragmatics would have found that a useful paper, not just students.

GG  That's quite possible. The editor asked for it because I had already written it. It was lying around. It had been refused by the journal that had originally commissioned it (on the grounds that it was too technical) and it had been published in German in Linguistische Berichte (Gazdar 1978b). The editor of Journal of Pragmatics got in touch with me and said ``I've read this paper - can I have an English version for the Journal of Pragmatics?'' and I said yes.

EJB If you look at Steve Levinson 's text book it is essentially the same.

GG  Yes. Steve and I just had the same view. We had different areas of expertise but our view of pragmatics was the same.

EJB That book (Levinson 1983) has become probably the most successful of all the CUP red series in laying down the foundations, and I would guess that people still read it who want to do PhDs on pragmatics despite the fact that it is now 20 years old. It has had an extraordinarily long shelf life, so I suppose that these days people would be sent to that rather than to the paper in Journal of Pragmatics, but the book came five years later and is essentially putting pragmatics on the same foundation. So in terms of the wider influence on what came after perhaps it is that broadening out from the work on presupposition and the setting up of that framework which has really survived. As far as I can see most people who do pragmatics now would take that as their starting point, and that is now the dominant way of thinking about it.

GG  Well, I don't know as I haven't taken any interest in the area since then. That's an exaggeration as I had to teach it in the early 1980s - I used to use Steve's book for teaching but I haven't been to a pragmatics conference since 1979, or whenever that awful event in Italy was.

EJB Defaults and their use in linguistic description have remained as a kind of theme throughout your career although the kind of defaults that you are dealing with in a system like DATR or the lexical inheritance problem look, at least at first glance, very different. In the work that you did later, you didn't ever explicitly mention the work on pragmatics or go back and look at it from the perspective of the work that you were doing on the lexicon.

GG  Actually that is not quite true because when I did an inaugural lecture at Sussex it dawned on me that the only unifying theme of my work hitherto was defaults so I built the whole lecture round that. The lecture was called Ceteris Paribus and I said what I do basically is defaults and here are three examples. But the lecture wasn't written up.

EJB So you are conscious of that?

GG  When I was very young I learnt a trick and not being able to learn any new tricks, I just kind of reused it later in different contexts.

EJB Well, we'll probably come back to that. I think I've asked all the things that I was prompted to think about from the reading of the pragmatics papers. Let us take a break and then we'll do GPSG and syntax which I guess will be a long chunk.

Next page ..
Copyright © Ted Briscoe & Gerald Gazdar, Wednesday 2 May 2001