This is a study of dialogues performed in improvisational (“improv”) theater. Improv actors work without scripts, using mime and dialogue to come to a working agreement about cast of characters, setting, and plot, usually within a few interactional turns. Their performances are thus an ideal site for studying how framing works: how interactants come to a mutual understanding of what is going on that is shaped by and subsequently shapes their contributions. Although it does not quite meet its ambitious theoretical goals, the book is well worth reading, both for the descriptions and examples of a fascinating set of language games and for Sawyer's intriguing suggestions about what we can learn from them.

The study is based on videotapes of performances by 15 improv groups and interviews with directors and actors. Improvisation is a 20th-century development influenced by medieval commedia dell'arte and Stanislavskian Method acting. It is used as a rehearsal and warm-up technique and, beginning in Chicago in the 1950s, it has been performed live in cabaret-style settings (most famously by the Second City troupe, the model for TV's Saturday Night Live). An improvisation can begin without any specification of plot, characters, setting, or events, although many improv formats begin by eliciting audience suggestions for the location, the first line of dialogue, the identity of one character, or some other element of the performance. Each format dictates certain other elements of structure, such as the length of the performance, how starting and ending points are determined, and whether or not actors may step out of character to “edit” the action. Actors also orient to “ethnotheory” about what makes effective improvisation work, some of which is explicitly codified in instructional books. For example, actors try to respond to the preceding turn in such a way as to move the scene forward, and actors are expected to relinquish individual control in favor of collaborative shaping of the scene.

Sawyer's theoretical context is the notion of framing. In most research on conversational interaction, interactional frames are described from the perspective of individual interactants, in terms of individuals' actions, intentions, or mental representations. Social forces constrain individuals only via individuals' selective orientation to and interpretation of them; it is thus impossible to formulate laws that predict how such forces will affect individuals' behavior in particular instances. Sawyer proposes a more macro-sociological account in which the object of inquiry is the interactional process, and the goal is to make statistical predictions about participants' behavior over a set of instances. Drawing on theories of sociological emergentism, Sawyer claims that interactional frames arise out of a series of individual turns of dialogue, but are not reducible to them. Frames are the result of a process of “collaborative emergence,” and emergent frames can be shown to affect participants' behavior in predictable ways. In other words, in order to account fully for what goes on in improvisation (and in conversation in general, to the extent that it is improvisatory), we cannot just adduce individuals' actions and others' interpretations of and generalizations about them. Once they begin to emerge, frames constrain participants in a process Sawyer calls “downward causation.”

To test these claims, Sawyer develops hypotheses about how emergent frames will affect improv performances as they proceed. Because the frame does not exist previous to the performance in the form of a script or a set of stage directions, improv actors use metapragmatics (Silverstein 1993) to offer suggestions about what is going on and to respond to these suggestions. With respect to how particular framing strategies constrain subsequent actors' turns, Sawyer predicts that dialogue turns that offer a greater amount of dramatic information or constrain other actors' behavior more narrowly will be stronger, shaping the emergent frame powerfully and influencing the responding actor's dialogue turn in a number of ways. For example, an actor who begins a scene by saying, “I've got to have today's newspaper, do you sell them?” has made an offer about the setting and the other actor's character. She has used stronger metapragmatic strategies than she would have had she started with “Where are the papers?,” which proposes neither a setting nor a role for the other. An actor responding to the first of these might be more likely to accept the proffered setting and characterization and quicker to respond, and it should be interactionally more difficult for him or her to reject the offer.

With respect to the frame's emergence over time, Sawyer predicts that use of more powerful metapragmatic strategies at the beginning of an improvisation will result in the frame's emerging more quickly, that the frame will become more complex and elaborated as the performance proceeds, and that the use of different metapragmatic strategies will result in different kinds of frames. With respect to how the emerging frame (rather than just preceding actors' dialogue turns) then constrains subsequent actors, Sawyer predicts that as the complexity of the frame increases, its interactional power increases, reducing the range of moves that are possible for subsequent speakers on one level and enabling new kinds of interactional moves on another; and that actors use progressively weaker metapragmatic strategies as performances proceed.

Sawyer then compares two short-form improv “games” with somewhat different ground rules to see whether these differences are associated with the predicted metapragmatic contrasts. For example, one of the “freeze games” requires the actors to create a cumulative frame over the course of a number of short scenes. In the other, a new frame has to emerge in each scene. In the former, less metapragmatic work is required as the game progresses. Scenes accordingly become shorter, and actors use less powerful frame-offering metapragmatic strategies. In the latter, scene length stays the same and more powerful strategies are used.

A second empirical chapter consists of short descriptions of a number of other improv formats. Sawyer raises intriguing questions about how the starting rules for each affect the collaborative emergence of its frame. The third data-based chapter returns to the study's specific hypotheses in a comparison of two long-form formats. In one, actors may step out of character to play director, explicitly setting the scene at the beginning of each new episode. In the other, they may not, so scene-setting is accomplished via indirect metapragmatic strategies by actors delivering dialogue in character and simultaneously accomplishing scene-setting edits. This difference has consequences for the nature of the emergent frame – consequences actors do not intend and of which they are unaware. For example, metapragmatic scene-setting is often accomplished by directing character-identifying address forms (e.g., “Sister”) at other actors. Accordingly, in the format in which characterization is indirect, the frame tends to be more focused on characters than on plot.

Sawyer positions the study as “a version of … positivism” (p. 72), in contradistinction to the interpretivist stance of interaction analysis in the ethnomethodological and ethnographic traditions. This makes sense only in the context of Sawyer's narrow definition of interpretivism and correspondingly broad definition of positivism. For Sawyer, an interpretive account is one in which the participants' own explicit, conscious interpretations are the only acceptable explanatory mechanism, and any account in which the analyst makes generalizations that participants do not explicitly orient to is positivist. But many social scientists who would not accept the ontological and epistemological assumptions of positivism would accept accounts invoking generalizations participants are not aware of making. It is far from clear what “awareness” even is, or what it means to say that participants “intend” outcomes or are “conscious” of their intentions. Even the sort of ethnotheory that improv actors do sometimes make explicit (and with which Sawyer contrasts the kinds of explanations he develops) is probably mostly implicit and “unconscious.” The improv teachers and authors who codify this knowledge make analytically useful generalizations just as Sawyer does, and actors may or may not subsequently make use of these generalizations.

Sawyer's approach thus differs from that of conversation analysts not so much in that it offers a competing account of what goes on in conversation rooted in a different philosophy of science, but in that it takes a different perspective, interpreting the social system from the outside rather than through the system-interpreting eyes of particular actors making particular moves. His predictions are necessarily about the behavior of the system, not about the particular behaviors of individual actors. However, Sawyer's more specific methodological claim – that it might be analytically useful in an account of conversation as a social process to adduce interactional frames that emerge in interaction and come to affect the interaction – is interesting and well supported. This is promising early work by someone we will probably hear more from.