John M. Gottman

Observing interaction: Assessing observer agreement

Why brother? Imagine that we want to study how objects are used in communication between mothers and their 15-month-old infants, and have made a number of videotapes of mothers and infants playing together. We might focus our attention on those times when either the mother or the infant tries to engage the others interest in some object, and then describe what those attempts look like and what their consequences are. After viewing the videotapes several times, sometimes in slow motion, we might be convinced that we had detected all such episodes, had accurately described them, and were now ready to make statements about how mothers and infants go about attempting to interest each other in objects. We should not be surprised, however, if other investigators do not take our conclusions as seriously as we do. After all, we have done nothing to convince them that others viewing the videotapes would see the same things, much less come to the same conclusions. We are probably all aware how easy it is to see what we want to see, even given the best of intentions. For that reason, we take elaborate precautions in scientific work to insulate measuring procedures from the investigators influence. When measurements are recorded automatically and/or when there is little ambiguity about the measurement (for example, the amount of sediment in a standard sample of seawater), as is often the case in the “hard” sciences, the problem of investigator bias is not so severe.

Observing interaction: Recording behavioral sequences

Recording units: Events versus intervals The title of this chapter, Recording behavioral sequences, suggests two different but somewhat related topics. The chapter could deal primarily with mechanical matters and describe devices used to record data, or it could deal just with conceptual matters and describe different strategies for collecting data about behavior sequences. Actually, this chapter will be a mixture of both, although conceptual and strategic matters will be stressed. One important issue has already been raised by the example of the Bakeman and Brownlee (1980) study of parallel play in chapter 1, and that is the issue of “units.” Before selecting a particular recording strategy, an investigator needs first to decide what “units” are to be used for recording. As we use the term, the recording unit indentifies what prompts the observer to record, and usually is either an interval or an event. For example, an investigator might choose to code time intervals, as Bakeman and Brownlee did, assigning codes to successive time intervals. This is a common strategy, but for many purposes more accurate data result when the events themselves are coded instead. In such cases, observers wait for an event of interest to occur. When one occurs, they code it (i.e., note what kind of event it was) and perhaps record onset and offset times for the event as well. Which is better, to code events or to code intervals? That depends on a number of factors, including the kind and complexity of the coding scheme, the desired accuracy for the data, and the kind of recording equipment available.

Observing interaction: Developing a coding scheme

Introduction The first step in observational research is developing a coding scheme. It is a step that deserves a good deal of time and attention. Put simply, the success of observational studies depends on those distinctions that early on become enshrined in the coding scheme. Later on, it will be the job of observers to note when the behaviors defined in the code catalog occur in the stream of behavior. What the investigator is saying, in effect, is: This is what I think important; this is what I want extracted from the passing stream. Yet sometimes the development of coding schemes is approached almost casually, and so we sometimes hear people ask: Do you have a coding scheme I can borrow? This seems to us a little like wearing someone elses underwear. Developing a coding scheme is very much a theoretical act, one that should begin in the privacy of ones own study, and the coding scheme itself represents an hypothesis, even if it is rarely treated as such. After all, it embodies the behaviors and distinctions that the investigator thinks important for exploring the problem at hand. It is, very simply, the lens with which he or she has chosen to view the world. Now if that lens is thoughtfully constructed and well formed (and aimed in the right direction), a clearer view of the world should emerge. But if not, no amount of corrective action will bring things into focus later. That is, no amount of technical virtuosity, no mathematical geniuses or statistical saviors, can wrest understanding from ill-conceived or wrong-headed coding schemes.

Observing interaction: Representing observational data

Once observers have done their work – that is, once their assignment of codes to events or intervals has been committed to paper or electronic files – it is tempting to think that you can now move directly to analysis of the coded data. Almost always this is premature because it bypasses two important intervening steps. The second step involves reducing sequential data for a session into summary scores for subsequent analysis and is relatively well understood; for details see Chapters 8 and 9. The first step is equally important but often receives less attention. It is the subject of this chapter and involves representing – literally, re-presenting – the data as recorded initially into formats that facilitate subsequent data reduction and analysis. When recording observational data, as described in the preceding chapter, observer ease and accuracy are paramount, and methods and formats for recording data appropriately accommodate these concerns. But when progressing to data preparation, reduction, and analysis, different formats may work better. In this chapter, we consider two levels of data representation. The first is a standard format – that is, a set of conventions – for sequential data that defines five basic data types and reflects the recording strategies described in the previous chapter. The second is more conceptual; it is a way of thinking about sequential data in terms of a universal code-unit grid that applies to all data types and that facilitates data analysis and data modification, as demonstrated in subsequent chapters (especially in Chapter 10).

Analyzing sequential data: First steps

Describing versus modeling If all the steps described in previous chapters – developing coding schemes, recording behavioral sequences reliably, representing the observational data – are in order, then the first fruits of the research should be simple description. Introductory textbooks never tire of telling their readers that the basic tasks of psychology are, one, description, and two, explanation. Similarly, almost all introductory textbooks in statistics distinguish between descriptive statistics, on the one hand, and inferential statistics, on the other. This distinction is important and organizes not just introductory statistics texts but this and the next four chapters as well. Much of the material presented in this and the following four chapters, however, assumes that readers want first to describe their data, and so description is emphasized. Problems of inference and modeling – determining if data fit a particular model, estimating model parameters – are touched on only slightly here. These are important statistical topics and become especially so when one wants to move beyond mere description to a deeper understanding of ones data. That is why so many books and courses, indeed huge specialized literatures, are devoted to such topics. We assume that readers will use scores derived from observing behavioral sequences as input for anything from simple chi-square or analyses of variance, to log-linear modeling, to the modeling approach embodied in programs like LISREL. Our task, fortunately, is not to describe all the modeling possibilities available. Instead, we have set ourselves the more manageable task of discussing how to derive useful descriptive scores from sequential data.