In this innovative study of epistemic stance marking in English conversation, Elise Kärkkäinen attempts to cross-fertilize discourse-functional linguistics with conversation analysis, and her effort bears fruit. She moves deftly between quantitative and qualitative methods, juxtaposing frequency tables with conversational transcripts, and comes out with a bold thesis in hand: “Showing commitment to the status of the information that one is providing, i.e. marking epistemic stance … [is] an essentially interactive activity” (p. 183).

The author is no less bold about her professed empiricism, declaring that she “abandons all preconceived notions or theories of function and strives to uncover patterns of epistemic stance in a strictly empirical study” (3). This is, of course, a familiar refrain in the research traditions she favors, and it is best viewed as a rejoinder to those who would impose a “pre-established top-down framework or taxonomy on a set of data” (15), as she charges linguistic politeness theorists with often doing; epistemic markers can't always be explained using putatively universal notions of face and Gricean maxims. Perhaps it is her fervent embrace of empiricism that also explains the lack of sustained theoretical reflection in this book. Readers should not expect lengthy or path-breaking discussions of such notions as “subjectivity,” “commitment,” “speaker” – or even, for that matter, “stance”; minimal, provisional definitions must suffice. But this weakness is also the book's strength, for in limiting its scope and tightening its methodological reins, the author is able to actualize its stated mission: to offer a corpus-based “description” of the interactional functions of epistemic stance in English conversation, as the book's title announces.

Kärkkäinen's primary data consist of five conversational segments, drawn from the Santa Barbara corpus of spoken American English, that total 1 hour, 43 minutes. At times, her modestly sized dataset may seem confining, yet she compensates for this by drawing on her previous research and by referencing larger-scale studies. After three chapters of preliminaries, which include reviews of literature that pass at a brisk pace, we reach the book's core, two data-heavy chapters that make up the bulk of this volume: chap. 4, “Routinization of stance marking at the linguistic and interactional level,” and chap. 5, “Stance-taking as an interactive activity: The case of I think.” In chap. 4, she begins with an overview of the form-types used to signal epistemicity in English conversation, including lexical verbs, adverbs, modal auxiliaries and quasi-auxiliaries. Most prevalent in her data are “epistemic phrases” like I think, which typically occur without the complementizer that and thus have in principle a degree of syntactic mobility (38–45); they don't need to occur clause-initially. The point about mobility is important because Kärkkäinen then wishes to examine patterns of placement – but not placement in respect of clauses or sentences. She finds intonation units (IU) to be more relevant, for “[the] encoding of epistemicity in intonation units exhibits great consistencies and recurrent patterns in the data; by far the most frequent position for speakers to express their epistemic stance is at the beginnings of intonation units” (4). Epistemic phrases (and epistemic adverbs) rarely occur in IU-medial or IU-final position in her data, and although they do occur with some frequency as separate IUs, in such cases they tend to have distinct pragmatic-functional effects.

After considering epistemicity in the compass of single intonation units, Kärkkäinen zooms out, as it were, to explore larger units: conversational turns. As she does this, she redeploys her tripartite framework, exploring epistemic stance markers in terms of “turn-initial,” “turn-medial,” and “turn-final” placement. (Since turns can be comprised of numerous IUs, the “medial” term seems to lose some of its explanatory power, however.) In pursuing what we might term a “multi-scalar” approach to epistemic stance in chaps. 4 and 5, Kärkkäinen often pushes hard against classic conversation analysis (CA). Not only does she grant intonation units and prosodic contextualization cues pride of place, but, more radically, she tries to escape CA's orbit around adjacency sequences: “… if we think of the aspects of sequential organization that are relevant for an interpretation of the functions of epistemic markers in discourse, there is definitely reason to look beyond adjacency pair relations or sequences of two consecutive conversational turns … into longer stretches of discourse and into extended turns by one primary speaker, such as stories or other types of extended turns” (89). Kärkkäinen thus tries to expand our sense of the potentially relevant co(n)text for the study of epistemic stance. For this alone, her work merits attention.

For some readers of Language in Society, however, her construal of context will still feel cramped. She contends that “ethnographic information … proved relevant for the actual analysis only sporadically,” and writes instead that “what was of crucial importance was the sequential location of a given utterance or action, both locally, or in terms of what preceded (and followed), and globally, in terms of where the speech segment fitted in the larger sequence of events in the interaction” (6–7). She concedes, however, that she “was not directly personally involved (nor in many cases was any other researcher) in making the recordings” (6), and hence it is perhaps no surprise that we are provided with only the barest of demographic facts: age, sex, ethnic identification, occupation, and the like – facts that are corralled into a residual category termed the “extra-linguistic.” In fact, we are never in a position to evaluate the relevance of this penumbral category, the extra-linguistic, owing to the limitations of the data. By no means does this derail Kärkkäinen's argument. It only means that she has narrowed somewhat the range of observable epistemic-stance effects. This narrowing also conveniently reinforces her thesis about the primacy of interactional motivation; after all, nearly everything extra-linguistic has already been removed from sight.

Considered in its own terms, however, the book's analyses are well executed. Only in a few places might readers desire more evidence. Consider, in this respect, one of her key observations: that epistemic markers (epistemic phrases like I think) tend to occur IU-initially. In chap. 4, she notes that a “few” (41) epistemic phrases in her data occurred with the that-complementizer. In tables that summarize the placement of epistemic phrases within IUs (56–57), however, the category of “epistemic phrases” is left undifferentiated; though infrequent, it is important to separate out epistemic phrases containing that-complements because they have syntactic constraints that could account for their IU-initial position. Her larger claim is that IU-initial position tends to be interactionally motivated, specifically by recipient design:

It is a testimony to her thoroughness that Kärkkäinen entertains the possibility that IU-initial placement of epistemic phrases like I think might instead be the result of a “grammaticization process [that] has not advanced far enough: these phrases are really not yet like adverbs, of whose mobility in terms of IUs there is in fact more evidence in my data, but have preserved the erstwhile main clause + complement clause syntactic pattern” (99). She dismisses this possibility, however, adding that it is “more plausible” to “look for the interactional motivations behind such patterning” (99). An equally quick dismissal occurs when she discusses how IU-initial markers tended to be “uttered with clearly faster tempo than the rest of the IU” and were “reduced in phonetic form” (66). She uses these observations as (partial) evidence for the “unmarked” status of IU-initial position for epistemicity. Yet in a note (102), she acknowledges that IUs in general often (but not always) begin with anacrusis, and that involves reduction and acceleration too. To be fair, Kärkkäinen relies on multiple sources of evidence, and when the dust clears, I suspect her claims will still be standing. Yet these minor cracks in the edifice reveal the weight of her burden. Like variationist sociolinguists, Kärkkäinen must account and control for multiple sources of motivation – no easy task. And at times, her interest in interactional functions seems to come at the expense of these other sources.

I should note, finally, that the alleged primacy of interactional motivation in epistemic stance-marking is also central to the book's programmatic message, which Kärkkäinen announces in the closing pages. She sees her blend of discourse-functional linguistics and conversation analysis as representative of a new, hybrid field of inquiry: “interactional linguistics.” Not all readers will share the author's enthusiasm for this research program, but they will surely find the book's empirical findings valuable, and its combination of methodologies especially stimulating. As such, this book is a welcome contribution to the burgeoning literature on epistemic stance.