Why sequence analysis?

I have been involved with behaviour sequence analysis for about forty years, and on the whole it has been a very positive experience. But I have often been frustrated by the reactions of conventionally-minded colleagues from the "It's different, so it must be wrong" school of thought.

In particular, they seem to think this is an arbitrary and strange way of going about things. "Where on earth did that come from?" they ask. So here is a personal account of just where it did come from, for me at least.

The story begins in about 1972, when I decided to leave behind the subject of my first degree – Medical Sciences – and attempt to do a PhD in a field which was new to me, namely Social Psychology. I had the good fortune to work in Oxford with Michael Argyle, who was then one of the leading figures in British Social Psychology. His research group had earned itself a huge reputation for the study of 'NVC' – non-verbal communication. This meant I had a colossal literature and a lot of experimental know-how to catch up with; or else I had to do something entirely different. Being young, and full of unjustified confidence, I liked the idea of the 'something different' option, so I decided to tackle verbal communication, or conversation, instead.

The timing could not have been more fortunate, as Social Psychology was giving a lot of attention to social interaction, but mainly in the form of NVC rather than the things people said to each other; while linguistics – which was the red-hot discipline of the day – was giving a lot of attention to the things people said or wrote, but mainly in the form of individual sentences, rather than the larger structures of whole texts or conversations. There was a gap waiting to be filled. Before long, linguistic pragmatics, conversation analysis, and discourse analysis would come marching onto the scene as major players, but that was not apparent to me at that stage.

So, what is the interesting thing about conversation? People exchange utterances (not unlike the sentences studied by general linguists, except there are more of them strung together), while exchanging non-verbal signals (not unlike those studied by Social Psychologists).

The key thing, obviously enough, is that conversations – or arguments, debates, ceremonies, interviews, trials, etc – are not just lots of utterances one after another, but lots of appropriately crafted utterances following each other in a systematic pattern. So, what then is the nature of that pattern (I am saving the word 'sequence' for later) that distinguishes a conversation from a random assortment of utterances?

I started shopping around for methods and models that might help me to get a grip on the problem. I did some experiments on the permutations of utterances that people found to be realistic or unrealistic as conversations. I toyed with the idea of a 'grammar' of conversation. This would mean that successions of speech acts would be well-formed as a conversation if and only if they conformed to certain rules (which I would then discover), just as strings of morphemes are well-formed sentences if and only if they conform to certain rules (which Chomsky and others were busily articulating).

Then I came across the work of Robert Bales, a distinguished Social Psychologist at Harvard, who studied the discussions of small problem-solving groups. He divided up their discussions into individual events which were each coded into one of twelve categories, such as 'Asks for opinion, evaluation, analysis, expression of feeling' and 'Gives orientation, information, repeats, clarifies, confirms.' The relative prevalence of these categories could then be compared between groups; between different kinds of problem; and between the stages of a discussion. Most crucially for me, he cast his data into a 12x12 table of 'Reactive Tendencies', which showed for each of the twelve categories, the probability that the next utterance by another speaker would be from each of the twelve possible types. Such 'tables of transitional probability' still form the core of sequence analysis methodology, right up to the present. (I later learned that this approach to the study of event sequences was more generally known as Markov modelling after its inventor, and a sequence of events where each one sets the probability of the next, is called a Markov Chain).

And that was how the milk got in the coconuts, as they say. The PhD on conversation structure was soon behind me, but the method proved to be so versatile, and the job of inventing variations and extensions of it was so fascinating, that it shaped much of my career over the next four decades. My interest in conversation waned, but the approach stayed with me. I became more of an applied psychologist, and started to look at the time-lines affecting some key problems people face. In general, my interest morphed into a fascination with the pathways by which people get into dangerous situations, and – one hopes – out of them (if they make the right choices and decisions). This proved to be a very useful way of looking at the world, and of advising people who have to plan and design for the day when bad news strikes, or at least threatens.

The actual problems I looked at were many and varied, including evacuations from buildings on fire; road traffic collisions; episodes of self-harm; fights, rapes and murders; and a variety of mental disorders. The following example is from a study of beliefs about drink-driving.

 So, I hope there is no longer any doubt about why behaviour sequence analysis might be interesting and useful, but if there is, just read the next sentence.

about analysis and any be behaviour but doubt hope I if interesting is is, just longer might next no read sentence. sequence So, the there there useful, why

Not so sensible, the second time around? Why? Nothing has changed (except the original sequence has been lost). A sentence without the right sequence is nonsense. In the same way, an episode of behaviour without the right sequence is nonsense.

And in just that sense, much of what we study in the behavioural sciences will always be nonsense, unless we record and understand the organised sequences in which events occur.

As John Stuart Mill wrote in 1851:

"Of all truths relating to phenomena, the most valuable to us are those which relate to the order of their succession. On a knowledge of these is founded every reasonable anticipation of future facts, and whatever power we possess of influencing those facts to our advantage."