Order this book
With the exception of the Introduction and Epilogue, all the chapters in the book are previously published. Despite this, there is almost no repetition that irritates if the book is read at a sitting. Given the diversity of the individual chapters, it is also clear that Abbott's project has a genuine (rather than contrived) coherence. The chapters can broadly be organised by themes. Firstly, there are those that might be described as tackling the "hidden injuries of statistical methodology". These lay out, in a clear and compelling way, the features and implications of "General Linear Reality" – the projection of the statistical assumptions necessary to regression onto the behaviour of social actors. Secondly, there are papers that use content analysis of "typical" high profile research – work published in the American Journal of Sociology – to show that Abbott's points are not mere niceties but identify endemic tendencies in the literature. Thirdly, there are wide ranging historical surveys of concepts crucial to his project like causality and time. Finally, there are papers that illustrate the equivalent rigour of widely neglected techniques like sequence analysis.
These themes are often combined in interesting ways. The chapter "Seven Types of Ambiguity" begins with a methodological discussion of different ways in which variables can ambiguously reflect the social world. This discussion is supported by a content analysis of a sample of papers using the RELITEN variable in the General Social Survey. Abbott concludes that the positivist approach to sociology does not resolve the issues of ambiguity he raises but simply pushes them into the cracks between studies. As at several points in this book, he casually throws out an interesting research programme, that of establishing whether the progress of statistical sociology is capable of gradually reducing the level of ambiguity or simply shuffles it about. At the same time, he illustrates how such a question might be tackled in a rigorous way.
In a similar vein, the chapter "What Do Cases Do?" illustrates the claim that General Linear Reality is projecting statistical assumptions onto social actors by carrying out a content analysis of the sections of articles devoted to explanations offered for social action. Abbot explores the rhetorical circumstances under which sociologists actually talk about the thought processes, histories and contexts of individuals or simply "re present" (and thus represent) microscopic action as the individual instantiation of aggregate regularities.
In many sociology books published nowadays, ideas like this would be the high point or even the only point. However, the book also contains papers like "Transcending General Linear Reality" which still has "important" written all over it and should be required reading for both social statisticians and qualitative researchers. That chapter also serves as an extremely useful framework for the discussion that follows.
As further evidence for the richness of this book, I have not yet discussed its central idea, the relationship between "narratives" or regular event sequences (like family or occupational "histories") and statistical explanations that rely on various forms of assumptions about individualism and separability. Abbott shows clearly how ostensibly "sequential" techniques in statistics (like event history analysis) do not capture what we understand by sequence as a genuinely causal (and non- reducible) factor in social systems. (To take a simple example, it may be the biographical sequence of your curriculum vitae – neither too narrow, nor too scattered – that makes you an appealing job applicant rather than a mere numerical summary of your achievements.) This "statistical reinterpretation" of sequence is an interesting example of a process also very common in mainstream economics by which ideas "threatening" to normal science – like Bounded Rationality – are reinterpreted in a way that removes their threat but often destroys their coherence (or dramatically increases their implausibility) in the process. In rational choice, for example, making agents rational in their choice of choice algorithms introduces an infinite regress which only extra-theoretical constructs like depth limits can terminate. By contrast, Simon's notion of rationality as – perhaps self-modifying – procedure is logically coherent albeit hard to model mathematically. Abbott further shows how well developed (but largely neglected) techniques like sequence analysis can be used to compare such narratives and identify their regularities. For this, he draws on his substantive work on professionalisation and his occupational role in the American Sociological Association (and attendant journals) during the writing of its history.
This also illustrates another appealing feature of Abbott's work, its rich self-similarity. Modern social theory, like take-away food, often has an initial appeal but leaves no worthwhile residue. We admire for its style but cannot use it outside its own frame of reference. Abbott manages to illustrate the importance of narratives in the lives of social actors using the narratives that sociologists tell about them. His work thus manages to be satisfying both on the rigorous analytical level and to the part of us that approves what mathematicians call "elegance". Neither pure rigour nor pure richness is likely to be sufficient for a compelling account of social life. Many sociologists have pointed this out, but Abbott's book actually demonstrates it by combining the approaches and being compelling!
Interestingly, Abbott acknowledges that his research into sequence had important limits, though not those that the reviewers typically raised. Although studying sequences is a step forward from assuming extreme separability, particularly in social (rather than economic) domains, the far larger step is to show how interaction processes generate these sequences rather than treating them as isolated objects. This "closes the circle" of causation by allowing individual social actors to use sequences in their reasoning, in developing norms or aspirations about "lifestyles" for example. Interesting, and pleasingly for this reader, Abbott is positive about computer simulation, another largely neglected rigorous approach which could cast some light on processes combining sequence and interaction. Chattoe and Gilbert (1997) discuss a (highly simplified) model of this kind.
I do have some regrets (rather than criticisms) regarding the book. However, all but one of these is about what is missing rather than what is present. The regret about the content concerns the Introduction. It is interesting – and provides another example of appealing self-similarity – that Abbott discusses the intellectual development of his ideas and their places in the book. However, I found it too much to take in at the outset despite his clear style. Ideally, this would be the last chapter – with "Transcending General Linear Reality" as the effective Introduction or the reader should avoid reading the Introduction till last when it will make much more sense! The absences are twofold. Firstly, it would have been helpful (and perhaps illustrative) to see whether and how qualitative research "shapes up" against the kind of methodological framework that Abbott presents. Secondly, it would have been interesting (on the same grounds) to have a more detailed discussion of other formal techniques outside the statistical mainstream. Perhaps we can hope for a textbook or reader?
All in all, I can wholeheartedly recommend this book both as a good read and (more importantly) as an important contribution to the project of making sociology properly scientific without emptying it of relevance in the process. I hope it will be read carefully by practitioners of qualitative, quantitative and historical approaches. If dialogue between these groups is possible at all, it will be significantly improved by work like this.