Discourse Processes Questions in Time

Home , Discourse Processes

This article was downloaded by: [University of Wisconsin - Madison] On: 24 July 2014, At: 10:33 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Discourse Processes Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/hdsp20 Questions in Time: Investigating the Structure and Dynamics of Unfolding Classroom Discourse Martin Nystrand , Lawrence L. Wu , Adam Gamoran , Susie Zeiser & Daniel A. Long Published online: 08 Jun 2010.

To cite this article: Martin Nystrand , Lawrence L. Wu , Adam Gamoran , Susie Zeiser & Daniel A. Long (2003) Questions in Time: Investigating the Structure and Dynamics of Unfolding Classroom Discourse, Discourse Processes, 35:2, 135-198, DOI: 10.1207/ S15326950DP3502_3 To link to this article: http://dx.doi.org/10.1207/S15326950DP3502_3

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 DISCOURSE PROCESSES, 35(2), 135–198 Copyright © 2003, Lawrence Erlbaum Associates, Inc.

Questions in Time: Investigating the Structure and Dynamics of Unfolding Classroom Discourse

Martin Nystrand Department of English The University of Wisconsin–Madison National Research Center on English Learning and Achievement

Lawrence L. Wu and Adam Gamoran Department of Sociology The University of Wisconsin–Madison National Research Center on English Learning and Achievement

Susie Zeiser National Research Center on English Learning and Achievement

Daniel A. Long Department of Sociology The University of Wisconsin–Madison National Research Center on English Learning and Achievement

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 In the 1st-ever use of event-history analysis to investigate discourse processes quanti- tatively, this study recasts understanding of discourse in terms of the (a) antecedents and (b) consequences of discourse participant “moves” as they (c) affect the inertia of the discourse and accordingly structure unfolding discourse processes. The method is used to compute the probabilities of the effects of particular discourse moves on subsequent discourse patterns and to measure and systematically contrast static

Correspondence and requests for reprints should be sent to Martin Nystrand, The National Re- search Center on English Learning and Achievement (CELA), Wisconsin Center for Education Re- search, 1025 West Johnson Street, Madison, WI 53706. E-mail: [email protected] 136 NYSTRAND ET AL.

(macrosocial) and dynamic (microsocial) conditions prompting and sustaining dialogic discourse. Theoretically, the authors draw on Russian scholar Mikhail Bakhtin’s epistemological distinctions between monologic and dialogic discourse to identify pedagogically rich sequences of teacher–student interaction as dialogic spells and discussion, which the authors’ previous work has shown to contribute to achievement. Empirically, the authors examine data collected in hundreds of observations of more than 200 8th- and 9th-grade English and social studies classrooms in 25 Midwestern middle and high schools, including detailed coding of more than 33,000 teachers and student questions. Results show that authentic teacher questions, uptake, and student questions function as dialogic bids with student questions showing an especially large effect. Discourse event history analysis is a powerful tool for investigating the structure of unfolding discourse.

Since Saussure (1959/1915), a central problem for discourse theory has been grappling with the heterogeneity of discourse as it unfolds in time. Saussure sought to sidestep the whole problem by bracketing la parole and focusing on la langue, a construct geared, in a seminal move often cited as the foundation of mod- ern linguistics, to extricating language from seemingly intractable problems of time and context. Yet linguistics’focus on la langue has famously done nothing for studies of la parole except to perpetuate Saussure’s characterization of it as hetero- geneous and unsystematic—impossible as a serious problem for any scientific analysis seeking valid propositions and hypotheses about discourse in some general, nomothetic sense. Many initiatives have addressed this issue. Especially noteworthy, of course, have been ethnomethodology and conversation analysis documenting language in interaction co-constructed “on the fly” by the conversants and appropriately understood by the conversants only in the context of its emergence (Goodwin, 1979, 1981; Heritage & Roth, 1995; Jefferson, 1974; Sacks, Schlegloff, & Jefferson, 1974; Schegloff, 1984). Leaving behind efforts to investigate discourse by explicating an underlying structure of the sort that sentence grammarians uncover, sociolinguists and ethnomethodologists, including Cicourel (1973) and Garfinkel (1967), pro- posed concepts like the et cetera principle, reciprocity of phenomenal perspectives, indexicality, and reflexivity to conceptualize situated discourse in everyday interaction (cf. Nystrand, Greene, & Wiemelt, 1993, p. 320). Discourse—language in Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 time—is more adequately understood as structured by the conversants, particularly as they each reciprocally factor the intentions of the other into their interactions. As Schutz (1962) put it, it is “assumed that the sector of the world taken for granted by me is also taken for granted by you, [and] even more, that it is taken for granted by ‘Us’” (1967, p. 12). Or, as Brazil (1985), writing about discourse phonology in conversation, argued, “The most general characterization of the [proclaiming–referring]systemisnot,infact,intextualtermsbutinsocialterms.Inmakingthereferring choice, the speaker invokes the togetherness aspect of the conversational relationship, speaking as it were for the ‘we’ who are the participants” (pp. 67–68). QUESTIONS IN TIME 137

From this perspective, discourse begins as an initial calibration of conversants’ intentions and expectations vis-à-vis the topic and genre of their interaction, and the conversants’ensuing interaction is largely structured by their evolving perspectives on the topic and the discourse itself. Therefore, when one person begins a conversation by asking, “You know that x I’m always talking about?” the purpose of this question is not to quiz the speaker’s conversant but rather and merely to es- tablish a common topic and starting point for the conversation. For Rommetveit (1974), the result of this initial calibration of conversants is a temporarily shared social reality (or TSSR), and the discourse itself is to be understood, he argued, as a progressive modification and/or expansion of this social reality. Each modification and/or expansion, typically accomplished through the introduction and contextualization of new information, defines a subsequent TSSR (roughly speaking, there are about as many TSSRs in a discourse as there are introductions and contextualizations of new information). Whereas such a conceptualization perhaps represents the state of the art in discourse analysis and has done much to account for countless case studies, their nature as case studies makes it difficult to use their methods to investigate the general character and dynamics of discourse, such as how discourse in particular contexts tends to start and end, or the effects of defined contextual factors such as demography, power relations between conversants, institutional settings, and so on as they generally impact discourse processes, not just individual cases. In this article, we report the first-ever discourse event-history analysis to deal with these issues. Event-history analysis is a quantitative methodology used to investigate shifts in events or the status of individuals. It has been used, as we explain later, in sociology to assess the consequences of precipitating events in marriage and divorce, job mobility, teenage pregnancy, and infant mortality; and in political science to assess changes in city government, political timing, and the political effects of economic crises. Its power, demonstrated in our study, resides in its capa- bility to systematically analyze large datasets of interactions—many cases—with attention not only to macro variables, for example, culture and demography that Gee (1999) related to “Big-D” discourses, but also micro variables, for example, particular discourse moves Gee calls “little-d” discourse, as these dynamically shape discourse. Discourse event-history analysis enables the systematic and si- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 multaneous assessment of the relative impacts of both macro and micro factors. Discourse event-history analysis deals with the heterogeneity and time dimensions of discourse head on, recasting understanding of discourse in terms of the (a) antecedents and (b) consequences of discourse participant “moves” as they (c) affect the inertia of the discourse and accordingly structure unfolding discourse processes. In the study reported in this article, we use the method to systematically contrast the effects of particular discourse moves on subsequent discourse patterns. Whereas case studies can describe the nuanced sequences of events in individual encounters, discourse event-history analysis, as a method of quantitative 138 NYSTRAND ET AL.

discourse analysis, complements such studies by providing a more generalized understanding of how genres of discourse unfold, as well as assessing the salient factors, both macro and micro, at work in shaping them. In our study, working from a Bakhtinian account of classroom discourse, we employed these methods to study whole classroom discourse in eighth- and ninth-grade English and social studies instruction while assessing the effects of hypothesized variables on unfolding discourse in these classes. Our previous research (Gamoran & Nystrand, 1991; Nystrand, 1997) documented the prevalence of monologic recitation in such instruction with infrequent but noteworthy episodes of dialogic interactions and open discussion, which, we showed, were closely associated with improvements in student learning. The study reported here sought to understand how such dialogic interactions come about; we sought to model and explain the shifts during lessons from monologic to dialogic discourse patterns, investigating the effects of not only fixed characteristics such as socio- economic status, tracking, and racial and ethnic backgrounds of individual classes, but also dynamic variables such as types of teacher–student interaction enacted through teacher questions and follow up responses to student answers as they shaped subsequent discourse of the classroom. Data consisted of 33,904 question interactions in 872 observations of more than 200 eighth- and ninth-grade English and social studies classes in a wide variety of schools in the Midwest. Our article unfolds in three parts. First, we elaborate theoretical conceptions of monologicanddialogicinstruction.Thenwepresentmethodologicaldetailsrelated, first,tothecollectionandpreparationofdataandtheoperationalizationofrecitation, discussion, and dialogic spells as modes of classroom discourse, and then to the event-history analysis of these data. Finally, we present results concerning the structure of classroom discourse as it is affected by both static macrosocial variables tran- scending the classroom, as well as dynamic variables involved in unfolding classroom discourse. Our results offer insights into the effects on unfolding classroom discourse of race, ethnicity, and demographics; grade, tracking, and subject differences; and characteristics of teachers, including years of experience and gender. We report the surprisingly important role of student questions, as well as conditions promoting both dialogic spells and open discussion. We conclude by reflecting on the generalpotentialofdiscourseevent-historyanalysisasamethodofdiscourseanalysis. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014

MONOLOGIC AND DIALOGIC CLASSROOM DISCOURSE

A large body of empirical work over the last century has documented the predomi- nance of recitation as the principal mode of whole classroom discourse in Ameri- can elementary and secondary classrooms, including studies by Stevens (1912); Colvin (1919); Miller (1922); Thayer (1928); Corey (1940); Bellack (1966); Duffy QUESTIONS IN TIME 139

(1981); Durkin (1978–1979); Hoetker (1967); Hoetker and Ahlbrand (1969); Goodlad (1984); Mehan (1979); Sarason (1983); and Tharp and Gallimore (1988). Recently, Nystrand and Gamoran found that the vast proportion of questions in a large, diverse sample of eighth- and ninth-grade English classrooms are indeed asked by the teacher in recitation, with whole-class discussion (open exchange of ideas) averaging less than 50 s per lesson in the eighth grade and less than 15 s in the ninth grade (Nystrand, 1997; Nystrand & Gamoran, 1991a). The dominant pro- file of whole classroom discourse in these classes involved highly codified test questions, which developed little more than procedural (IRE) reciprocity; moreover, coherence from topic to topic was typically weak or absent (IRE refers to Mehan’s [1979] categories of turntaking in classroom discourse: teacher initiation [question], student response, and teacher evaluation). In all classes, the most common purpose of classroom discourse was to recall and display assigned information, to report on what was already known. Despite the preponderance of recitation, Nystrand and Gamoran’s empirical results, controlled for a range of background and initial performance measures, un- covered a strong and statistically significant association between student achievement and the extent to which classroom discourse moved away from recitation to genres of classroom discourse that recruited and highlighted student ideas and voices, as indicated by (a) proportion of authentic teacher questions, rather than the usual known-answer test questions; (b) extent of uptake, for example, follow up questions; and especially (c) time devoted to discussion. Discussion had a particularly large effect, especially striking because there was so little of it. Russian scholar Mikhail Bakhtin’s (e.g., 1984) epistemological distinctions between monologic and dialogic discourse are useful in conceptualizing the contin- uum of classroom discourse ranging from tightly controlled recitation (in which students demonstrate their recall of assigned information) to open discussion fea- turing an unprescripted exchange of student ideas in the absence of test questions. From this perspective, classroom discourse is monologic to the extent that the main speaker, typically the teacher, operates from a predetermined script; as Bakhtin (1984) put it, “monologism … pretends to possess a ready-made truth [italics in translation]” (p. 110). By contrast, classroom discourse is dialogic to the extent that the participants expand or modify the contributions of the others as one voice Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 “refracts” another. Bakhtin (1984) deplored the typical asymmetrical organization of what he called pedagogical dialogue as a nonproductive monologism:

In an environment of … monologism, the genuine interaction of consciousness is impossible, and thus genuine dialogue is impossible as well. In essence idealism knows only a single mode of cognitive interaction among consciousness: someone who knows and possesses the truth instructs someone who is ignorant of it and in error; that is, it is the interaction of a teacher and a pupil, which, it follows, can only be a pedagogical dialogue. (p. 81) 140 NYSTRAND ET AL.

For Bakhtin, this was a pathological form of discourse akin to a communication disorder. In its most radical form, monologic discourse, which he also referred to as “official discourse,” resists communication: Everyone is compelled to speak the same language, and outer speech is all, seeking to drain the first person pronoun of all its particularity (Holquist, 1990, p. 52). “Monologism, at its extreme,” Bakhtin (1984) wrote,

denies the existence outside itself of another consciousness with equal rights and re- sponsibilities. … Monologue is finalized and deaf to the other’s response, does not expect it and does not acknowledge in it any decisive force. … Monologue pretends to be the ultimate word. (pp. 292–293; italics in original)

Bakhtin contrasts the dialogic means of seeking truth with official monologism, which, he argued, pretends to possess a ready-made truth. From this perspective, the recitation taking place in typical school settings seeks to elicit official answers originating in texts and transmitted only one way—from teachers to students, to be received and recalled intact by students. The resulting monologic discourse, an activity Prawat (1995) called “head-fitting,” is one in which the relationship of teacher and student is restricted to that of evaluator and novice, organized for the transmission of information. This relation forms the basis of a discourse environment in which students have little chance of becoming conversants of consequence, recognized as contributing, producing, or participating actively in the construction of knowledge. For Bakhtin, the pedagogical dialogue of recitation inverts the natural logic of inquiry in the sense that the utterances and interrogatives of normal discourse not only respond to previous utterances but also anticipate response (Bakhtin, 1986). Learning situated in the give and take of dialogic discourse is thus premised not on the recitation of recalled information, but rather on a dynamic transformation of understandings through interaction. In an ideal dialogic learning environment, especially in open discussion as opposed to tightly cast recitation, teachers treat students as potential sources of knowledge and opinion, and in so doing complicate expert–novice hierarchies. By contrast, recitation within typical classrooms is overwhelmingly monologic precisely because the teachers routinely violate this

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 structure. Discourse in the typical classroom does not proceed with teachers responding in dialogic fashion to previous answers or to student remarks. To the con- trary, recitation is typically shaped by those points of information the teacher wants to cover. Consequently, teachers change topics at will, and teacher “uptake” of student questions is often perfunctory (Nystrand & Gamoran, 1991b). When this happens, the “language of schooling” (Kutz, 1997) undertakes

the acquisition of new discourses and expanding frames of references … ironically, not through continuing and evolving discourse, not through kicking ideas around and discussing in depth, or repeated familiarization with them through continuing, sus- QUESTIONS IN TIME 141

tained, or coherent discussions about subtleties of and alternatives to these ideas … but through the memorization of decontextualized information bites whose fixed and unevolving meaning resides within that information. (p. 201)

A particularly powerful insight, due to Mukarovsky, an important theorist in the PragueLinguisticCircle,concernsthedynamicrelationbetweenmonologueanddi- alogue. Makorovsky holds that monologic and dialogic discourse are in “a dynamic polarity in which sometimes dialogue, sometimes monologue, gains the upper hand according to the time and milieu” (1977, p. 85). Lotman (1988, 1990) claims that all language can be treated both dialogically and univocally. When utterances are treatedunivocally,asinrecitation,focusisonthe“accuratetransmissionofinforma- tion”; when they are treated dialogically, as in open discussion, they are used as thinking devices. From this perspective, whereas monologic discourse is useful for establishing topics and conveying information, it is dialogic discourse that opens the floor to discussion and the negotiation of ideas and new understandings. Much instruction involves both modes of discourse, often to good effect. Many teachers, for example, skillfully set discussion up by first reviewing basic material as a way of establishing the topic for discussion, and once this is accomplished they move on to a more probing and interpretive level, in which student ideas and views are elicited and encouraged. In such classrooms, dialogic discourse is a strategic device that teachers can use to foster student engagement and construct a classroom environment conducive to learning. Just how predominant patterns of test-question recitation give way to open discussion conspicuously absent of such questions is the central problem of our study. In this article, we bridge the gap between the parsimony of quantitative analyses and the complexities of classroom interaction with a detailed empirical analysis of the dynamics of classroom discourse in a wide range of classrooms and instructional episodes. Statistically, we use event history methods that permit us to examine several issues about how discourse in classroom instructional settings unfolds over time: What aspects of classroom interaction might shape the dynamics of classroom discourse? Why might certain classes, but not others, shift from monologic to dialogic patterns of discourse? What is the relationship between the social organization of the classroom, for example, tracking, and the emergence of

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 dialogic discourse? How do teachers act to open up dialogic patterns of interaction? Exactly what teacher and student “moves” allow classroom discourse to step up from monologic to dialogic levels of engagements?

DATA

Research investigating the dynamics of classroom discourse requires fine-grained data, not only on the nature and type of classroom discourse—the specific interactions between students and teachers—but also on how these interactions unfold 142 NYSTRAND ET AL.

over time. To test our ideas, we draw on unique data that are highly detailed on the number, types, and sequencing of questions posed by teachers and students within a diverse sample of eighth- and ninth-grade classrooms as gathered by Nystrand and Gamoran relating student achievement in literature and social studies to an extensive list of instructional and classroom discourse variables. At the same time that our analyses examine dynamic aspects of classroom discourse, they also move beyond small-scale studies by examining a large and diverse sample of schools and classrooms to identify both the antecedents and the consequences of classroom discourse unfolding in time.

School Characteristics Data were collected in 16 Midwestern junior high and middle schools in eight Midwestern communities, including rural, urban, and suburban sites, in both public and parochial schools. Six of these communities were public school districts; the other two were Catholic high schools with students from a number of urban and suburban K–8 feeder schools. We control for these fixed characteristics in our event history analyses. Table 1 provides a breakdown of the community and school types that participated in our study.

Class Characteristics We also took account of fixed characteristics of classes. Some of these were aggre- gated from student reports: race, ethnicity, gender, a fall writing test, and SES (an unweighted linear composite based on parents’ education, occupation, and home resources). Others were features of the classes themselves: grade level (eighth or ninth), subject (English or social studies), size, and track level. We also controlled for two self-reported teacher characteristics: gender and years of experience.

Classroom Observation Data In these classes, we collected class observation data from the eighth-grade classes during 1987–1988, and the ninth-grade classes during 1988–1989. Unlike

TABLE 1 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Characteristics of School Sample

Number of Schools

School District Type Total Middle Schools (Grade 8) High School (Grade 9)

Parochial 8 6 2 Public 17 10 7 Small town/rural 6 3 3 Suburban 3 2 1 Urban 8 5 3 QUESTIONS IN TIME 143

ninth-grade classes, which were all called English, eighth-grade classes were vari- ously called language arts, English, reading, communications, literature, and so forth. To deal with this issue, we selected the eighth-grade classes that focused most on reading. We selected approximately four English classes and four social studies classes in each school; in the smaller schools, we observed all the English and social studies classes, but in the larger schools, we selected classes that represented the different ability groups as defined by the school (honors or accelerated, regular or average, basic or remedial). This sampling plan yielded 58 eighth-grade English classes and 57 eighth-grade social studies classes. The next year we moved to the high schools for which the junior high and middle schools served as feeders, selecting about six classes per high school to maintain the same number of classes, again representing the different ability groups as defined by the schools, resulting in 54 ninth-grade English classes and 49 ninth-grade social studies classes (one school did not offer social studies classes for ninth-graders). About 1,500 students participated each year; of all eligible students, about 10% were lost through absence or refusal. About one third of all students participated in both years of the study. Table 2 summarizes these data. A trained observer visited each class four times, twice during fall semester and twice during spring semester, with observations scheduled at the mutual conve- nience of teachers and observers. These visits yielded 872 class observations. Using the Classroom Language Assessment System, CLASS 2.0 (Nystrand, 1988), with audiotape backup, we listed all questions posed by teachers or students during instructional episodes (defined as any coherent classroom activity centering around a particular purpose or topic) occurring within the class observations. In our research, anewepisode starts when the teacher addresses a new purpose or topic. Sometimes episodes will consist of two or more activities. For example, in addressing a particular objective, the teacher may initiate a question-and-answer session, which is then interruptedbyperiodic,brieflectures,andculminateswithahomeworkassignment. Wetreatedsuchepisodepartsassegments,definedasanycoherentpartofanepisode that differs from other activities constituting an episode. We counted as questions all queries for information, including mainly intonation questions and some tag questions, but we did not count either (a) procedural Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 TABLE 2 Scope of Study

Characteristics Grade 8 Grade 9 Totals

Number of English language arts classes 58 54 112 Number of social studies classes 57 49 106 Number of times each class observed 4 4 8 Number of observations 460 412 872 Number of coded questions 12,375 21,529 33,904 144 NYSTRAND ET AL.

questions (e.g., “How many pages do we need to read?” “Does that answer your question?” “Do you have any questions?”), (b) rhetorical questions,or(c)discourse-management questions or repair initiations (e.g., “What?” “Excuse me?” “Did we talk about that?” “Where are we [in the text]?”), which manage classroom discourse, or questions like ”Do you remember our discussion from yesterday?” which initiate discourse topics). Altogether, we listed and coded 35,887 questions from 1,152 instructional episodes across the 872 class observations. Missing data reduced the number of questions to 33,904 for analysis. Our observations also identified when discussion periods occurred during each lesson. As a result, these data provide an unusually rich source of information, not only on teacher and student characteristics for a particular classroom, but on the types, sorts, and sequences of questions occurring within instructional episodes during specific classroom observations. Tables 4 and 5, which juxtapose CLASS lists of questions with transcripts from two lessons, give examples of the kinds of data we collected; we discuss these examples in the following.

Question Event Data The data provide a highly detailed source of information on the sequence and flow of discourse within the classrooms we observed, as captured by the types and sorts of questions and their attendant interactions occurring within instructional episodes. We exploit this element of these data in our event history analyses by examining five indicators (authenticity, uptake, level of evaluation, cognitive level, and question source)foreachofthe33,904questionsinouranalysissample.Thesevariableswere used in our previous work relating quality of classroom discourse and student achievement in English and social studies instruction (Nystrand, 1997). In our studies, we code not questions per se but rather the interactions surround- ing the questions. That is to say, our coding treats questions as sites of interaction. As Heritage and Roth (1995) and Schegloff (1984) contended, the character of any unit of discourse and related interaction is a function of the participants’ understanding, meaning that, unlike the natural sciences, whose focus, Schutz (1962) contended, is on first-order constructs, the proper focus of the social sciences (including psychology, sociology, linguistics, and discourse analysis) is second-order Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 constructs—constructs of the participants’constructs and understandings. Follow- ing this principle, we coded participants’ understandings of their interactions as manifest by their discourse moves. To judge the authenticity of a question, for example, we took cues not only from how students responded to the questions, but also how the teacher evaluated or followed up the students’ responses. As with authenticity, we coded cognitive level according to the level of cognitive functioning the question elicited, not the question by itself. In all cases, we coded not listed questions but rather the character of social interaction elicited and valorized by the questions themselves. QUESTIONS IN TIME 145

Authentic questions. We define an authentic question as one for which the asker has not prespecified an answer. Examples include requests for information as well as open-ended questions with indeterminate answers. As such, an authentic question allows a range of responses unlike more frequently occurring recitation questions, in which a teacher asks a question with a prescripted answer in mind. This distinction is important for our work because authentic questions posed by the teacher signal to students that the teacher is interested in what they think and know, as opposed to whether they can engage in mere recitation by repeating material given in texts or other sources. Moreover, by allowing an indeterminate number of acceptable answers, authentic questions open the floor to students’ ideas. As such, they invite students to contribute something new to the class interaction, which in turn holds the potential for altering the trajectory of discourse in the classroom. By contrast, a test question allows only one possible right answer, a characteristic that Lotman (1988) termed univocal. As a result, test questions concentrate control of classroom discourse in one actor—the teacher—and thus allow students no voice or influence over the flow of classroom discourse. Judging the authenticity of a question ultimately depends on the context of the question; it cannot be determined alone by the text of the question. “What were the causes of the Civil War?” is most likely a test question when part of a review lesson in a high school social studies class, but it could very well be an authentic question in a graduate seminar. The nature of a given instructional episode, that is, the genre of classroom discourse, is the most reliable indicator of authenticity. Hence, when teachers begin a lesson by saying, “Okay, class, let’s check the answers to your study questions,” the questions are invariably test questions (though follow-up discussions of students’ answers can sometimes be authentic). By contrast, when teachers ask about students’ personal experiences as lead-ins, for example, to open-ended discussions of a poem or short story, these questions are almost always authentic. In classroom interactions, the authenticity of a given question is often revealed by how the teacher evaluates or follows up student answers and responses. Whenever the authenticity of a question was unclear or ambiguous to us, we consulted the teacher.

Uptake. We defined uptake as occurring when one conversant, for example, a Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 teacher, asks someone else, for example, a student, about something the other person said previously (Collins, 1982). In an example of uptake, taken from a ninth-grade lesson on The Odyssey, the teacher asks, “What do they have to do to Polyphemus?” A student replies, “Blind him.” The teacher then follows up, asking, “How come the plan is for blinding Cyclops?” This last question is an instance of uptake, since the teacher follows up on the student’s response on “blinding him.” Uptake is often marked by the use of pronouns, for example, “How did it work?” “What caused it?” “What city grew out of this?” In each of these questions, the italicized pronoun refers to a previous answer; linguists call such references 146 NYSTRAND ET AL.

deictic references. Conceptually, uptake is important because it recognizes and en- velops the importance of the student contribution. Following up on student responses makes the response the momentary topic of discourse. As such, uptake may play an important role in facilitating the negotiation of understandings, as conversants listen and respond to each other. Moreover, by building on the voices of others and by establishing intertextual links among speakers, uptake acts to pro- mote coherence within the discourse. To qualify as uptake, a question must incorporate a previous answer, not a previous question; hence, we did not code as uptake teachers making reference to questions or remarks they had previously made or to filmstrips, videos, or texts that had previously been discussed. Nor did we code repeated questions as uptake. The second question in this sequence incorporates the framework of the first question, but since it does not incorporate a student’s answer, it does not involve uptake. In a study of 12th-graders’ discussing and writing about literature, Knoeller (1993) referred to this process as interpersonal voicing, when a speaker frames, expands, or borrows the voice of others in the immediate classroom. Other re- searchers have treated uptake more liberally. Greenleaf and Freedman (1993), for example, treated the teacher’s revision of student responses (perhaps beyond rec- ognition) in furthering the “lesson” as uptake. We are aware that our specific no- tions of uptake could be subject to coding error if explicit referencing to the past speaker were used systematically as a common tool or “technique” to involve students, since “ritual validation nullifies itself” (van Lier, 1997).

Level of evaluation. Typically, teachers’evaluations of student responses are a perfunctory “Right” or “Wrong,”a “Good” or an “Okay,”sometimes merely a nod, sometimes nothing. Sometimes, however, teachers respond more substantially. For example, a teacher might evaluate a student answer by saying, “Good point,” and then ask a follow-up question. If in doing this, teachers certify these contributions and modifications and work students’ answers into the fabric of an unfolding exchange,andtheseanswersmodifythetopicoraffectthecourseofdiscussioninsome way,wecallsuchevaluationshigh-levelevaluation.Inotherwords,whenateacher’s evaluation is high-level, the student really “gets the floor.” Specifically, we operationalized high-level evaluation using two criteria: (a) the teacher’s certifica- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 tion of the response (“Good,” “Interesting,” etc.) and (b) the teacher’s incorporation of the response usually in the form of either an elaboration (or commentary, e.g., “That’s important because …”) or a follow-up question (e.g., “Can you say more aboutthat?”or“Whydoyousaythat?”).Thatis,forlevelofevaluationtobecodedas high, the evaluation had to be more than “Good,” “Good idea,” or a mere repeat of a student’s answer. In all instances of high-level evaluation, the teacher validated the student’s answer so that it affected the subsequent course of the discussion. We did not consider as high-level a teacher’s introduction of new information in response to a student answer unless the teacher incorporated a previous student QUESTIONS IN TIME 147

answer; the criterion was the importance of the student as a source of new information. Also, we applied high-level evaluation only to the evaluation of student responses, not to teachers’ answers to student questions.

Cognitive level. Questions that elicit generalization, analysis, or speculation open up the cognitive field beyond a mere reporting or replication of another’s voice, incorporating the possibility of the speaker’s added perspective and particularity. We therefore coded the level of cognitive functioning that each question sought to elicit, judging it high to the extent that the question could “not be answered through the routine application of prior knowledge” (Newmann, 1988; see also Polanyi’s [1958] distinction between routine perfor- mances and heuristic acts). Like authenticity, the cognitive level of questions cannot be judged altogether from words alone. For example, if the teacher expected students to answer questions by reciting information found in textbooks, we coded questions as reports regardless of their linguistic structure. Hence, though a why-question will normally elicit an analysis, it will elicit a report if the teacher’s focus is the recitation of a textbook’s analysis rather than the class’s reflection or a student’s understanding; then “Why?” really means, “According to your text, why did it happen this way? Do you remember?” In such a case the teacher is seeking only recitation. Factors affecting the cognitive level of any question include the following:

• Knowledgeability of the person answering of the question. The very same question that elicits an analysis from a person who has to figure things out may well elicit a report from another, more knowledgeable individual who already knows and simply needs to explain. For example, “Why did Odys- seus and his men plan deliberately to blind Polyphemus?” may well elicit an analysis from students (assuming, of course, that they have to figure out the answer and not merely recite their textbook on the point), but will most likely elicit a report if a student asks a teacher who already knows the answer. When we were unclear, we asked about it after class. • Experience, ability, and prior knowledge of the person answering the question, including student or teacher. If student answers seemed to require rou- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 tine cognitive operation, we coded questions as eliciting reports. We defined prior knowledge as “prior to the previous night’s homework.” If a teacher asked students about the previous night’s reading, we coded the source of information as the text whereas, if the teacher asked about something learned previously to that, even from a text, we coded the source of information as prior knowledge. We made no distinction between prior knowledge and personal knowledge. • Nature of the instructional activity. When an episode was devoted to review, our normal expectation for responses was a report, even if questions had the 148 NYSTRAND ET AL.

linguistic form of higher level questions (e.g., “What’s the difference between a symbol and an image?” as a study question). • Source of information required by the question, including prior experience, textbooks, and previous teacher lectures.

Level of cognition elicited by questions was measured with a 5-point linear scale calibratedforlevelofabstractionandderivedfromApplebee(1981),Britton,Burgess, Martin, McLeod, and Rosen (1975), and Moffett (1968). Levels were as follows:

Lower 1 Record of an ongoing event: What’s happening? We coded questions as records if they elicited descriptions of what students were observing, feeling, or thinking at the time of the question. Examples include “Any questions on that?” and “What [or why] are you thinking about that?” 2 Recitation and report of old information: What happened? Higher 3 Generalization: What happens? Generalizations display inductive reasoning, building up ideas rather than breaking them down. They address questions such as, What happens? What do I make of what happens? They tie things together and they are not restatements of information. If the question required students to think and not just report something already known or previously thought by someone else, then we scored cognitive level higher than 2. To determine how high involved judging whether the student answering the questions was building up a generalization, in which case we scored it a 3, or breaking down an argument, in which case we coded it as an analysis and rated its cognitive level as 4. 4 Analysis: Why does it happen? Analyses display deductive reasoning, breaking concepts, ideas, and arguments down rather than building up ideas. To be scored as analyses, questions had to require more than restatements of known information. 5 Speculation: What might happen?

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Questions were judged to be lower order (i.e., eliciting records or reports) if they elicited old information or higher order (i.e., eliciting generalizations, analyses, or speculations) if they elicited new information and could not be answered through the routine application of prior knowledge. Superficially a question such as “Do you think that’s important?” might seem to elicit a record (i.e., referring to what the student is thinking at the time of the question), but the question more typically elicits a higher cognitive operation such as an analysis of what is important. Hence, for such preformulated questions (cf. French & MacClure, 1981), we distinguished the preformulators (“Do you think …?”) QUESTIONS IN TIME 149

from their nuclear utterances (the remainder of the question: “Is that important?”), coding only the latter.

Teacher–student source. This variable indicates whether the question was asked by the teacher or by a student.

Reliabilities. All codings were double checked and challenged by at least one other person besides the coder; we consulted tapes whenever we had questions, and a project assistant verified all transcribed questions and challenged codings by listening to these tapes. A small sample of 12 observations involving over 600 questions was observed by two observers to determine coding reliability: Reliabil- ity was computed both at the question level (percent agreement for all questions pooled) and at the observation level (average correlation between raters for the 12 observations). In our subsample of 12 observations, raters agreed perfectly on authenticity for 78% of 619 questions; the Pearson correlation at the observational level was .938. Raters agreed on uptake perfectly for 81.7% of the 619-question subsample used to check for reliability; at the observational level, the interrater correlation was .973. Raters agreed on cognitive level perfectly for 79.0% of the 619-question subsample used to check for reliability; at the observational level, interrater correlation was .965.

METHODS

We present two methods sections. The first concerns the operationalization of dialogic spells; the second concerns event-history analysis.

Operationalizing Dialogic Spells

A key issue in our study was how to operationalize a dialogic spell, including when during an instructional episode there is a shift from monologic to dialogic dis-

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 course, and also when a dialogic spell ends. Our analytic strategy was to proceed in several steps. Our initial efforts were exploratory, with two members of our research team with experience as English language arts teachers reviewing the original audio taped sessions of lessons from selected ninth-grade classes. This work led to tentative identification of active (“minds on” [Langer, 1995]) and passive zones of interaction, which were judged by the level and quality of student engagement. The most obvious such discourse was open-ended discussion in which students freely exchanged ideas and views in conversational modes. In addition to discussion, we focused on positive zones of interaction that often grew out of reci- 150 NYSTRAND ET AL.

tation and sometimes evolved into discussion. These phases of classroom talk were characterized by the following:

1. The interest and enthusiasm of the students appeared to peak. 2. The discourse was built on past discourse as conversants often referred to earlier discourse of other speakers. 3. Students asked questions. 4. Teachers suspended test questions. 5. Teacher questions were answered without repeated prodding or without being assigned to specific students by name.

Although all these characteristics contributed to our impression of increased student investment in the classroom discourse, the most important and consistent index seemed to be the prominence of student questions. When students began to ask questions about what they were studying, the tide of discourse, as it were, often seemed to change into something more symmetrical than the usual classroom interaction dominated by teacher questions. Applebee, Burroughs, and Stevens (2000) report a similar finding from a 12th-grade literature class they studied: “Students noted that discussions were most often initiated by students. One student estimated that a student question initiated discussion ’about 60% of the time’” (p. 416). We were impressed that the emergence of student questions, especially when occurring in clusters, not only seemed to signal student engagement but also the teacher’s loosening of the usual monologic reins of classroom discourse. Our preliminary investigation categorized three distinct modes of whole classroom discourse:

1. Recitation, characterized by IRE patterns and teacher test questions. 2. Discussion, characterized by the open-ended conversational exchanges of ideas largely absent of questions. 3. Dialogic spell, a mode of discourse, somewhere between recitation and discussion, characterized by engaged student questions and an absence of teacher test questions, specified according to the following model:

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Model of Classroom Discourse I. Default structure of whole classroom discourse. • It is the teacher who initiates and manages whole classroom discourse; the most common, default pattern of classroom discourse is monologic. • Whole classroom discourse manifests inertia and tends to continue direction and character until someone, usually the teacher, acts to change it. • To transform monologic classroom discourse into dialogic, the teacher either does something or allows something; both moves are defined as dialogic bids. QUESTIONS IN TIME 151

II. Elements of dialogic bids. • Responding to and taking up ideas and observations that students introduce, for example, through uptake and authentic questions. • Withholding evaluation in such a way as to encourage discussion and conversational interaction. III. Dialogic spell model. • Dialogic bids increase the probability for a subsequent dialogic spell. • A dialogic spell begins with a student question and is followed subsequently though not necessarily immediately, by at least two more; a dialogic spell may include teacher questions as well as student questions. • A dialogic spell is terminated by sustained monologic series of three or more teacher test questions. IV. Monologic bids. • Asking teacher test questions. Any teacher test question is a monologic bid marking the potential start of a monologic spell in much the way that dialogic bids, including authentic questions, uptake, and student questions, increase the probability of a dialogic spell. • Interrupting instructional conversations with low-level evaluation, that is, evaluating a student response (as right or wrong, for example) and, by so doing, signal a shift back to recitation. • Ignoring student comments and questions.

To test and probe this model using the sequences of questions in our data, we proceeded to assign weights to all questions using a 14-point linear scale based on the five coded variables ranging from 0 for radically monologic test questions to 14 for radically dialogic questions (see Dialogic Value Scale, Table 3). Ques- tions were weighted for source (teacher vs. student), authenticity, uptake (whether the question references someone else’s previous utterance), high-level evaluation, and cognitive level. Based on our review of the audiotapes, our scale privileges student questions; thus, all else being equal, student questions are assigned one additional scale value relative to teacher questions. We then proceeded to weight all 33,904 questions in our data set of 1,151 instructional epi- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 sodes according to this scheme.

Plotting Questions in Time To examine changes in the dialogic value of questions as they unfolded over time in each instructional episode, we plotted their values on our dialogic scale over the course of each instructional episode. Studying these plots suggested clusters of questions forming dialogic and monologic spells, and assisted us in identification of shifts from one state to the other. 2 Level Cognitive 2 Level Evaluation high. Neither was given any value Uptake ,L=low,H= 1 response. Value Code Authentic .N=no TABLE 3 2 Dialogic Values Scale Level Cognitive For both level of evaluation and cognitive level 2 2 Level Evaluation , A = authentic. Uptake Student Questions Teacher Questions 1 N = nonauthentic, QA = quasi-authentic 1 unless question was authentic with uptake; then they were equally weighted Value Code Authentic –100N N00N L N ^00N H N L00N N L N 00N L L N2AN H 00N Y H N/A H L N2AN 00N L Y L H H L N L 2AN 00N H Y H H L L L 2AN L L L Y N/A L Y H H 1aN L L 3bQAN 4BQAN L H H H Y L 1aN 3bQAN H H L4BQAN H H L Y H 1aN 3bQAN H H4BQAN H H Y L 1aN L 3bQAN L4BQAN L N L H L L H L 5cQAY 6CQAY L H H 5cQAY H 6CQAY H H H 5cQAY 6CQAY H 5cQAY 6CQAY N/A8DA N8DA L N8DA H N –18DA L L N N 7dA L10 H N L 7dA H12 ^ H N 7dA H12 N L 7dA L E14 H L F N/A H F H G A N/A A A A Y N Y Y Y L L N/A H H L H L H 9 11 11 13 e f f g A A A A Y Y Y Y L H L H L L H H

152 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( TABLE 4 Transcript and Coded Questions From Recitation on Magna Carta (27 Questions in All, 1 From a Student) Transcript With Lines Numbered123 Teacher:45 I’ll tell you what’s6 going to happen today. Shhhh.7 Thank 89 you. You are having a10 test hour Monday. today I on want ummm to …11 spend reviewing together. half Then and an we’re bringing going the to things 12 have minutes. the I last think fifteen you’ll find13 and I’m you going work to hard model on some14 this. way information for that you we to have … bring15 you all copied have the and listened you and have you learned16 have the we last are week going and to—categorize a it put17 half. and your And put feet it down— and in have to I’m groups.18 do going The over to rest the of weekend it yesterday yourself. you I’m19 Now, will not I going know through I thisstart said Questions packet.on20 as But that, Coded before would and I you Numbered get read, Consecutively out and21 by your we Class packets will I fast asked go can you through study22 to the from right that. answers … so democracy, It how you is23 it about started England—the in England. this I when24 did I not gave correct it back like to you25 you. to Individually. do, What is I’d if read I the26 give questions, you give the you right the you answers, right have I’ll answers. it Checkright. if If you from do it. not, So it’s you up read to throughout you thewhat to thingthe learn again,answer and is. you You for have find Monday. to Umm, do now that you as have one a to on practice tell. p. Okay, 192. let’s In do this England the oneand wasthe umm beginning of … democracy, (???) 192. Magna by It Carta. the was Uhh, past (???). in Uhh, yeah, would you like to do the

153 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Normal Normal Normal question: Teacher test question, no What year was the Magna Cartaquestion: signed? Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 What year was the Magna Cartaquestion: signed? Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 Disputes b/w kings and nobles—what ispower? absolute uptake, elicits a report from text,DIALOGIC low VALUE = level 0 evaluation. Table 4 (Continued) power. Okay. Total power. Okay. Number Transcript With Lines Numbered27282930 Student:31 Teacher: I want to do the Okay, first in question. the A it first says question? “Finding Or the do main you idea.” have What a this question? reading is about is that thein Magna England. Carta (???). furtheredThere’s only democracy a few years I want you to 1 Questions as Coded and Numbered Consecutively by Class 3233343536 Student:3738 Teacher: I think it was remember. 1022. Now39 Okay. you (several remember students what talking). (???) Carta Be or was40 No! what signed? (More year (Q1) humming Magna from male41 student). When you (write quiet! it) (another42 student humming). Was it 1066?43 (Q2) you said?44 2 (2) 1215! It istalking muchabout!45 (???)That thisEngland time was the we they started first are 46 kind to of diminish area the that power it of also47 the tell king. about Okay, the and Queen! the Thatpower48 Student: of the the Magna (…). Carta (8) limited between And kings49 Teacher: many and disputes nobles arose in England. the Manyparliament Student: I had and don’t a a know. war war between between king. the And Okay, Teacher: yeah. parliament four and times the the king power. Somebody of Total tell England power. me; tried what absolute is absolute Hmmm?? power? Total (Q3) 3 5051 Student: Unlimited power of kings was brought to to England’s one. by? Read the question.

154 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Normal question: Normal question: Teacher test question, no Normal question: Teacher test question, no Normal question: Teacher test question, no uptake, The following are true about theNormal Middle question: Ages? Teacher test question, noelicits uptake, a report from text, lowDIALOGIC level VALUE = evaluation. 0 Unlimited power of kings was broughtby? to England elicits a report from text, lowDIALOGIC level VALUE = evaluation. 0 The Saxon kings of England werechoice)? … (multiple uptake, elicits a report from text,DIALOGIC low VALUE = level 0 evaluation. The Saxon kings of England werechoice)? … (multiple uptake, elicits a report from text,DIALOGIC low VALUE = level 0 evaluation. #3—William the Conqueror …? Teacher test question, no uptake, elicitstext, a low report level from evaluation. DIALOGIC VALUE = 0 true. Okay? All the following are true about 777879808182 Student: Teacher: to take Students: We the did time that with last umm the week. the following Did You gave (3) details I that p. okay? give to you 198 Let’s us go that let’s right No, do … to … answers no B, umm there. (disagreement). part No B. we We didn’t. have the 73747576 the Middle Ages. The power of great some as nobles that was of almost their as king. people That’shad true. no The power. That’s common true. of There democracy was under a feudalism. great No. deal Alright, I am not going 8 52 (Q4) That’d be William the Conqueror. 4 535455 Teacher:5657 William the Conqueror is right. Okay.58 Student: You want to go umm 59 Teacher:60 Student: Umm … (student reading go quietly) again?61 Ummm Teacher: absolute Number no! rule? 2 Okay? (Q5), (Q6) yeah. answer. Read62 umm the they question shared and power then with63 They …. shared power with the nobles.64 Student: Why don’t you label 65 Teacher:66 Student: Three that? The67 Or (???) number when three that … first the Kathie? came power68 (Q7) Teacher: William in with the theythe Conqueror really nobles. made shared And the number69 nobles four…. give him the land 70 William Student: the Conqueror made the nobles71 give him the land. (quiet)72 Teacher: (student reading question) Which of Student: the following are true Okay, four No. Teacher: ….(Q8) 5 Umm. (Indecipherable). about the That Middle is Ages? not (indecipherable.) 7 6

155 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Rights of English people do notNormal includequestion: what? Teacher test question, noelicits uptake, a report from text, lowDIALOGIC level VALUE = evaluation. 0 A search warrant may be issuedNormal by question: a Teacher test judge question, if no …? elicits uptake, a report from text, lowDIALOGIC level VALUE = evaluation. 0 Table 4 (Continued) modeled after England. Okay, let’s …. 101102 Student:103 Teacher:104 The right105 to work. To work. Okay.106 There’s no written law that everybody has Students: the Yes. right to work. You have have a a right right to for free lawyer, speech, and trial. you you Does have that a sound right familiar? to a quick 9899100 Student: Teacher: The rights of The English rights people of do English not people include do the go not right from include … here. what? Number (Q10) 2. 10 Transcript With Lines Numbered838485 Student: Teacher: Student: B, C, C, D Okay, read it for us. A search warrant may be issued by a judge only (Q9) … B 9 Questions as Coded and Numbered Consecutively by Class 8687888990 Teacher:9192 Okay! A search warrant— now if we’re there93 coming is into a the reason Bill tosomething of believeillegal. 94 Student: the search will turn up 95 Teacher: Rights of96 Student: Yes. England— Like A a search search warrantonly warrant may if97 Where here. Teacher: be there do issue is you by a think a reasonsomething we judge to illegal. get Magna Does believe some Carta. the this of search sound these will familiar? ideas turn from? And up we are going to get into more and more of that. How do we get our Bill of Rights?Constitution? And It howwas do so we set up our

156 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Normal Normal Normal question: Teacher Normal question: Teacher test Normal question: Teacher test question: Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 What is the other house? Since late 1600s House of Lords had …? test question, no uptake, elicits alevel evaluation. report DIALOGIC from VALUE text, = low 0 Who has more power? question, no uptake, elicits a reportevaluation. DIALOGIC from VALUE text, = low 0 level Where does prime minister come from? question: Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 The magna carta …? question, no uptake, elicits a reportevaluation. DIALOGIC from VALUE text, = low 0 level ould like to finish the o…Iw 108109 Students:110111 Since the Teacher: late112 1600s the House of Lords of England had Okay. Student: House of Lords. What are the … What is (indecipherable) the (Q11) House other of Commons. house? (Q12) 12 107 Teacher: Okay, C uhh 3.121122 Student: What was number 4? complete the sentence … that’s where 11 we stopped yesterday. 14 113114 Teacher:115116 Okay, Students: so what117 does it say? Teacher: That the118 House of Commons. Commons or the 119 House Students: of120 Commons! Okay. Has House more of Teacher: power Lords than has the more House power (???) of today? (Q13) Okay, he is123 the leader Lords. of Where … does of the124 the prime dominant minister party. come (2) Teacher: from? (Q14)125 Student:126 Alright, I’m Umm… Teacher: not it’s D. going127 to d I’ll do128 that. Okay, Student: give 13 me129 the first one. The Teacher: Magna130 Carta? (Q15) Loud Student: The Magna131 Carta made the Help Teacher: … him (4)132 out! and clear. Monarch (4). … 133 Monarch Students: ask134 the Great Council for Teacher: its135 consent for any Taxes. new 136 Taxes. What it said— Magna what? Carta said… you, monarch you … okay … cannot gotaxes. You out have and to tell come the and people that ask to we us giveare in you in the 15 turn parliament to and tell you if we will give you taxes.

157 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Normal Normal question: Teacher Authentic teacher question, Authentic student question, uptake, How did they get the braverycouldn’t to do tell that? king he elicits a report requiring prior knowledge,analysis, elicits low level evaluation. DIALOGIC VALUE12 = How did they get brave? test question, no uptake, elicits alevel evaluation. report DIALOGIC from VALUE text, = low 0 What is happening when most peoplecountry indon’t your like what ruler does? question: Teacher test question, no uptake,analysis, elicits low level evaluation. DIALOGIC VALUE = 0 Have you ever revolted? uptake, elicits a report requiring priorlevel evaluation. knowledge, DIALOGIC low VALUE = 11 Table 4 (Continued) Transcript With Lines Numbered137 Student: How did they get the bravery to tell the king that he can’t 16 Questions as Coded and Numbered Consecutively by Class 138139140 Teacher:141142 Okay. Good question.143 Good question!! Whatever happened 144 do that? Student: (Q16) 145146 that people Teacher: The could reason get147 they enough did guts? together. is I’m (Q17) because putting Okay. the To in148 get umm that. the They It judicial have has Okay, to system that’s to a get be149 little together. more later than than one. she Okay? said Yeah. 150 that umm (???) the they Student: Great had a151 fair trial? Teacher:152 Council … But yeah, … the153 (indecipherable) Great Councilsupported wasthe Okay, the the king, king one and154 can that wanted say to there that fight are if them. enough he But people155 has when that enough do people not to like the156 law … 157 protect him. Support him.158 But hedidn’t. There (???). were At more that159 noble time men they uhhh and … Student: more more people of160 … the uhh uhhpeople … that Teacher: uhhwant it …161 the in other ais way feudalist the Student: They around. system area revolt. (???). more we162 And will this answerpeople next. in Anyone? Teacher: What your is country163 happening do when notthe most likeruler They what does? revolt. the (Q18) 164 monarch or They Student: revolt! (3) Okay, 17 now you kids Student: 18 (???) have you ever Yeah. No. revolted? (Q19) 19

158 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Normal Authentic teacher Normal question: What did we have in Milwaukee? Teacher test question, no uptake, elicitsrequiring a prior report knowledge, low level evaluation. DIALOGIC VALUE = 0 Who has more power in thequestion, plant? uptake, elicits analysis, low levelDIALOGIC evaluation. VALUE = 13 The bill of rights—what did itquestion: offer? Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 m … I don’t know if we have a winner yet, 165166 Student: Teacher: People can strike … like What that is one the … big strike we have had in … in Milwaukee (Q20) 20 167168 Student:169 Teacher:170 Student: (Petrochemical)? 171 Okay, Teacher: how long172 has it been going A on? long173 time. A Student: long174 time? Um Teacher:175 I don’t176 know. Still locked in177 uhhh … do still we? Student: are meeting178 … Umm do Student: the179 Teacher: Yes 180 people that Students: The work people for181 who the own plant person the have that The Teacher: plant. more owns owner power the of182 than plant? the the (Q21) plant… (several responding at once) 183 21 OK, it184 happened to the teachers (???).185 They went on a 186 strike. Okay? They hired187 new teachers. uhh It control happened workers to in188 the airlines, didn’t strike? it? And They they went just189 on rehired new on ones. both So sides there’s and190 power (???) hashave more more people, than and the191 they other. (People) take the chances. (war) This starts. is Okay?192 really Okay, let’s how go through Student: here because I have so many things want I to The want give Bill to you of do some 22 Rights today. time The I guarantees to bill still people (look of the at rights— right the it to notes). talksOkay? have Umm. about Number a thetwo. (Q22) Bill of Rights. lawyer.

159 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Normal question: Normal question: Teacher Normal question: Teacher test Normal question: Teacher test question, no If you can’t afford a lawyerhappens? here in US, what uptake, elicits a report requiring priorlevel evaluation. knowledge, DIALOGIC low VALUE = 0 Differences with king led to what? Teacher test question, no uptake, elicitstext, a low report level from evaluation. DIALOGIC VALUE = 0 What is a republic? question, no uptake, elicits a reportevaluation. DIALOGIC from VALUE text, = low 0 level What is difference b/w republic andNormal a question: monarchy? Teacher test question, noelicits uptake, a report from text, lowDIALOGIC level VALUE = evaluation. 0 What happened in 1649? test question, no uptake, elicits alevel evaluation. report DIALOGIC from VALUE text, = low 0 Table 4 (Continued) n a … tell me what is the Transcript With Lines Numbered193 Teacher: That sound familiar? Questions as Coded 23 and Numbered Consecutively by Class 194195 Students:196 Teacher:197 Umm-hmm. Yeah. 198 Okay, Student: if you199 can’t afford a Student: lawyer in200 the United States Student: what You’ll be201 uhh—- Teacher: —appointed one. 202 do you Appointed. do? (Q23) 203 You Student: get an204 appointed one. Okay. Alright, number three. Teacher: After many years the merchants were allowed to sit Where as did Okay, after I many leave years off? the merchants were allowed to sit members as of the House of Commons. 24 205206207 Student:208209 Teacher: Differences with210 the king led Student: parliament to211 (set up) members a of That’s Teacher: right. the What House212 is of a Commons. republic? Alright, Umm (Q25) four. …213 (2) republic in What Student: 1649? is (Q24) 214 the difference betwee Teacher:215 Student: Umm …216 (4) In a Anyone? republic you’re217 (2) umm difference (8) between … Something a where republic218 the and government a is monarchy? Teacher: ruled (Q26) by219 people who are220 Yah Student: (3) It221 is not elected a by Teacher: king other in people? other 26 A words. republicpeople? Okay. is So Student: They… elected what threw by out the the … They (kind of) threw out happened the do Yeah. king! you think in 1649? (Q27) 25 27

160 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded QUESTIONS IN TIME 161

To explain our analytic strategy, we now consider two contrasting examples. In each case, we begin with (a) a transcript of class interaction (these transcripts are for illustrative purposes in this article; we did not routinely transcribe the lessons we observed). Displayed to the right of the transcript, we display (b) the resulting list of coded questions we collected from the observation comprising our event-history data (all our observations involving lessons with questions generated such lists); in this list, readers will find the codings for each question, along with its dialogic value. And finally, we display (c) question plots (“Qplots”), which sequentially track the dialogic values for all questions asked during the instructional episode.

An example of a monologic spell. Our first example is a monologic sequence from our data set with no dialogic spells, consisting only of recitation test questions about the Magna Carta: the year it was signed, the names of the two houses of parliament, and so on. The entire lesson was devoted to teacher test questions and the transmission and recitation of known information. In Table 4, we juxtapose a transcript of the lesson with the numbered list of questions as coded by CLASS. The question codings provided the basis for our event-history analysis. Figure 1 sequentially plots the dialogic values of the questions in Table 4 over the course of the lesson as captured and coded by CLASS. By computing the mean dialogic value of the question sequence, defined as the dialogic density of the instructional episode, we find a near absence of dialogic value (density = 0.83), dra- matically depicting the plodding character of a lesson without a dialogic pulse geared entirely to review of previously learned information. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014

FIGURE 1 Question plot for recitation about Magna Carta: lesson organized as a monologic spell. Questions are noted with codes (here E, e, and f) based on the Dialogic Value Scale laid out in Table 4; student question noted in upper case. Teacher Why wouldn’t it Normal question: Authentic teacher question, uptake, elicits How does Bob Ewell get killed? Teacher test question, no uptake, elicitstext, a low report level from evaluation. DIALOGIC VALUE =How did 0 you figure out thattest Boo question, killed uptake, him? elicits analysis, lowevaluation. DIALOGIC level VALUE = 1 Boo has the kitchen knife, right? make sense? Authentic teacher question, uptake,analysis, elicits low level evaluation. DIALOGIC VALUE11 = Why not? analysis, low level evaluation. DIALOGIC VALUE11 = With a Dialogic Shift (33 questions in All, 7 From Students) TABLE 5

To Kill a Mockingbird Questions From Highly Interactive Lesson on Transcript With Line Numbers123 Teacher:4 Student:5 How Teacher: does Bob Ewell6 get killed? Student: Boo (Q1) Radley. 7 How did you figure8 out that Well, Boo yeah. killed But him?9 I (Q2) guess that Teacher: I thought10 that the knife …11 There are two knives12 Students: involved. I There’s really a didn’t switchblade understand and this13 Teacher: (part). Boo I at thought the it beginning, Yeah. (Multiple was but14 Student: overlapping then short I comments). was not sure. I15 a think kitchen … 2 knife. Boo has16 Student: It the Wouldn’t that said kitchen make that knife, 1 sense? he right? doesn’t (Q3)17 Student: want to reveal it to18 the Teacher: Right. (???) 19 Student: …even if he totally because (did (???) Shhhhh! 20 Student: it) it …. would ruin Questions you as know21 Coded Well Boo’s and he’s life … Numbered … Consecutively 3 by Class 22 Student: …he’d get all this attention23 and Student: he couldn’t 24 No. But he does say25 Teacher: thatNo, (that he was)wouldn’t be the obviously able time … 26 Student: to (???). again if they all Why found not? out (Q4) Well, he’s going to that have to he go did to it. trial, and uhh … all this stuff and everyone will know (???). about that he has 4

162 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Authentic teacher Normal question: Teacher test So you think Heck Tate isquestion, wrong? uptake, elicits analysis, low levelDIALOGIC evaluation. VALUE = 11 What do they hear? question, no uptake, elicits a reportevaluation. DIALOGIC from VALUE text, = low 0 level 272829 Student:30 I think it’s worth it. No, Teacher: it don’t mean (necessarily) So you think that obstruction Heck of Tate was justice, wrong just in because covering theydie) up?(Q5) don’t in want front (to of a house. 5 313233 Student:3435 Yeah! Well, Heck Tate said that36 anyway, it’s gonna be 3738 self-defense anyway,39 Teacher: however it comes up because on you40 the can trial really argue it to that go Somebody41 Student: way through reconstruct so the this you whole scene. just trialthe The have andjury kids42 Teacher: then and are it stuff walking just would to be already get up know43 Student: Her toabout. to what? some [uncoded answer procedural that question] you She44 Teacher: has a … home in45 Student: Ham. the pitch black. Scout has …her46 Teacher: a ham ham outfit on. outfit on. 47 Ham? Ham,48 Students: right. Yeah, it’s a pageant49 and Student: she’s dressed like (laughing) 50 Students:51 Teacher: Oh my God! That’s a (Laughing) some ham. 52 really (???). And53 she can’t see much, because54 Student: her view is really 55 Teacher:56 (laughing) limited by … Students: the feel ham that costume. they are And she under itcan Teacher: the is… oak pitch tree. black, And and (multiple what overlapping brief answers) They hear (???) cat swishing, does they … hear what somebody do kind they of hear? (Q6) 6

163 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Normal Teacher test What happens next after she calls[uncoded Cecil repaired aquestion] … What happens under that big old oak tree? question, no uptake, elicits a reportlevel evaluation: from That’s text, a high good reconstructionDIALOGIC …. VALUE = 0 question: Teacher test question, no uptake,report elicits from a text, low level evaluation.VALUE = DIALOGIC 0 [She hears a crunch] which is what? Table 5 (Continued) Transcript With Line Numbers575859 dragging his feet. Somebody’s following them. thinks Scout immediately it’s Cec … Cecil trick Jacobs on playing them a again. What happens next after she calls 7 Questions as Coded and Numbered Consecutively by Class 606162636465 Students: out Cecil Teacher: Jacobs has a newreconstruct (Laughter) (???) in on? your (Q7) mind Can … you thisis is… Can confusing Ham! anybody (???) because (laughter) morning. ham Scout What iscostume do stuck… you there think in happens her ham 8 808182 Teacher: Student: Which is what? …(Jem’s hand) … she hears a crunch and … (Q9) 66676869 Student:7071 Well umm like Jem stops72 to kind of like under look that73 in big the old oak tree? (Q8) 74 (???). Scout75 Teacher: thinks that Jem’s (friends) he’s hitreally76 Student: her,not but and then … and suddenly umm Uhh-huh. 77 shows then up she in yells (???) andtries starts to78 Student: And tobut umm she like she falls. run. falls, And Then big but … she ham,79 Student: she because and tries she’s it’s to (in kind run this) of away. hard And (Laughter). to then run. Yeah. And umm … umm— and she’s hear this like … shout (from the tree) Jem and (???) stuffcomes about down (???).to And her then(???) Jem (is totally again) and then 9

164 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Teacher test question, no Teacher test question, no uptake, Normal question: Teacher test question, Then what? no uptake, elicits a report fromevaluation. DIALOGIC text, VALUE low = level 0 We can assume that Boo heardand what then was did going what? on elicits analysis, low level evaluation. DIALOGIC VALUE = 0 Why would he stab him? uptake, elicits analysis, low level evaluation. DIALOGIC VALUE = 0 83 Teacher: Then what? (Q10) 10 848586 Student:8788 And then there’s … there’s89 nothing. And then suddenly 9091 she’s being92 Teacher: grabbed (that someone’s ready up to) and93 umm run sneak out of there,pulled somethingaway You94 Student: might, from like when her that. you umm, And watch and he’s really the … nothing.95 Student: movie and And next … then week, umm she you … triesgirl might and tois.96 Student: We’re then … But watching find she’s the (???) where movie? with the [uncoded Jem. procedural97 Students: Yes! (2) question] Sweet! want to98 Teacher: Awesome. watch carefully how they (overlapping reconstruct chatting)99 … (4) That’s100 a good reconstruction Arlene. It’s101 very Student:102 Teacher:103 She’s confusing. (???) It’s104 … on page 276 Yes! She’s and figure frightened it’s out105 about really because it, confusing we’re but to … she’s uhh an not … obstructed106 sure Scout view. has (at least) 107108 what’s going Student: on. Actually,109 I think yeah, I it’s Teacher: have 264. the Our110 wrong company … shuffled feet. and That’s dragged Grabbed (???) his the reconstruction111 butcher for knife you. nice … You job. Yeah!! did Good We a can old112 assume Boo! that Heard on Boo what and has was then heard going did what’s on, what? going grabbed (Q11) Student: the butcher He knife, was ran trying out, to and Why kill stabbed would … Bob he Ewell. stab 11 (2) him, do seem you sort suppose? of (Q12) … Doesn’t that 12

165 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Normal [S. referring to Teacher test How do you know he hasquestion, a no knife? uptake, elicits analysis, lowevaluation. DIALOGIC level VALUE = 0 Why? [Why does the ham costumeAuthentic save student her?] question, uptake, elicits analysis. DIALOGIC VALUE = 12 I didn’t understand one thing …say “What about doesthe this relationship of thedon’t children?” understand … what I this means question from the study guide] Authenticquestion, student uptake, elicits analysis. DIALOGIC VALUE =12 Who does Atticus think killed Bobquestion: Ewell? Teacher test question, no uptake,report elicits from text, low level evaluation.VALUE = DIALOGIC 0 Trixie, why are you looking soAuthentic incredulous? teacher question, uptake, elicits analysis,level evaluation. low DIALOGIC VALUE = 11 Table 5 (Continued) Transcript With Line Numbers113 Teacher: How do you know he has a knife? (Q13) 13 Questions as Coded and Numbered Consecutively by Class 114115 Student:116 Students:117 Teacher: Because her118 ham costume (Laughing) has a long119 (arm) … If Student: she hadn’t120 been in that Teacher: ham costume121 she would have Why? (Q14)122 Was she … Because Student: he was123 … he been hadn’t killed. Teacher: put And his that124 (???) costume in saved her. the Student: Ohhhhh! ham 125 Kim, you had126 your costume hand it up. Okay, could I have127 didn’t been understand in one … thing. Teacher: in Umm, Scout. 128 wait (2). Student:129 Which … which130 question Why are … you Teacher: It’s What on? like, does131 it’s this umm say children? (5) about (Q15) I the I don’t relationship don’t132 know understand. of it’s 1, the 2, Are 3, you 4 on133 chapters 28 to 15 31? Okay,134 everybody, let’s lines down. 135136 take a 14 look at137 your study we’ll guide.get And,a we’ll freebie138 get out a of pageant … Kim how Scout: does here. ham “On139 describe the … way the how to mockingbird, Teacher: does the that’s the Scout140 first describe one. think Who killed does ….” Atticus Who does Atticus Jem. Ewell? think He (Q16) Jem. uhh thinks Get Yeah. When killed Jem. your Jem Bob notebook was out,you attacked Andy.can (from) You get can self-defense, 16 some … free notes out of this. he stabbed Bob Ewell. Trixie, why incredulous? are (Q17) you looking 17

166 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Teacher test Authentic student Why does Atticus think that JemNormal killed question: him? Teacher test question, noelicits uptake, a report from text, lowDIALOGIC level VALUE = evaluation. 0 You think Atticus doesn’t know whoNormal Boo question: Radley Teacher is? test question, uptake,a elicits report from text, low levelVALUE = evaluation. 0 DIALOGIC But how does he know aboutbetween the Boo relationship Radley and the kids? question, uptake, elicits analysis. DIALOGIC VALUE =12 Why doesn’t Atticus want to hushquestion, up? no uptake, elicits analysis, lowevaluation. DIALOGIC level VALUE = 0 141142 Student: Teacher: I think that’s pretty weird. But I why (don’t does know Atticus if think umm) that … Jem would kill … (Q18) 18 143144 Student:145 Student:146 Because … 147 He doesn’t148 know that Boo Teacher: Radley was149 even there. He Student:150 Atticus doesn’t know151 who Boo thinks Radley that Student: No, is? Boo he (Q19) Radley152 knows is who some was he type (in) is. of (???). He (sniper)153 just guy doesn’t who think Teacher: that But he how154 does he know about Students: the155 relationship between like Boo goes Yeah, Teacher: that’s outside. true. 156 (brief overlapping Student: comments).157 Radley and Shhhh! the kids?158 (Q20) 19 Teacher: (???) Boo159 Radley [very quiet] nervous …160 he’s helping Who? Jem is161 unconscious because of Student: the terrible162 break (because in he’s Student: unconscious) … 163 Boo carries— 164 his arm Teacher: And and then, so165 wait, Boo the hasthinks next tothat section carryBoo was him is166 he home.(taking wanted AndJem to Atticus home) —Why Student: hush doesn’t …. up he167 want … Why Teacher: doesn’t Atticus168 want him who … … Atticus. (???) 169 20 Why doesn’t Atticus170 want to hush up (Q21)171 … see Heck Tate is saying, “Well, Ewell fell on his that knife! killed There him.” was And nobody perhaps … says, that’s “But where Jem Atticus did it incover up self-defense for and that. I I don’t don’t want want want to people it to hushed think up. that I Atticus don’t Finch’s son 21

167 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Teacher Teacher test Authentic teacher question, Teacher test question, uptake, elicits So he wants to protect Jem’smake reputation: sense Does to that everybody? uptake, elicits analysis, low level evaluation. DIALOGIC VALUE = 11 What does that say about Atticus’his relationship children? with analysis, low level evaluation. DIALOGIC VALUE = 1 (does) the sheriff knows that BooAuthentic Radley student did question, it? uptake, elicits analysis. DIALOGIC VALUE = 12 Do we know that the sheriffquestion, knows? uptake, elicits analysis, low levelDIALOGIC evaluation. VALUE = 1 How do you know that the sheriff knows? test question, uptake, elicits analysis, lowevaluation. DIALOGIC level VALUE = 1 Table 5 (Continued) Transcript With Line Numbers172 was protected by the sheriff. So he doesn’t want it hushed 22 Questions as Coded and Numbered Consecutively by Class 173174175 Student:176 Teacher:177 Jem (???)178 out of curiosity So Teacher: (???) he [mumbled wants179 comment]. to protect up. Jem’s reputation. Does180 that … Now, Student: what does181 that say about Atticus’ relationship182 with does that That make like sense183 if to he everybody? said (3) Teacher: something (Q22) different184 in town his than children? he (Q23) 185 He Student: (won’t) be186 23 different at said home at Teacher: than home he then187 is his (out children [mumbled] of wouldn’t Student: [very town). trust low him188 mumbled anymore comment] (out Well Teacher: of how town) about189 … if I He said won’t Student: something (Does) be about the less190 (is sheriff honest? knows that) Okay. that … (2) Boo Teacher: Well, Radley do did we it? know (Q24) the Student: sheriff That 24 knows? Boo (Q25) (???) How do you know that the And sheriff how knows? do (Q26) they know for sure that Heck Tate knows? Yes. 25 26 191192193194195196197 He says, Teacher: umm, when198 Atticus saysup (???)… his Heck menTate199 says covered we’re then not he Let’s talking go says about back that Jem to [mumbled] and that he’sit’s sortimpossible) passage. of(???) It’s on implying and 278, (that then 279. he [mumbled]. And talks I about Boo Radley think Keera’s right that this is Tate knows where what we happened. know Got that your Heck procedural book question] Ryan? [uncoded

168 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded ) continued ( Teacher test question, uptake, elicits What does this say about hischildren? relationship with his analysis, low level evaluation. DIALOGIC VALUE = 1 Does that seem to indicate thatTeacher test Heck question, Tate knows? uptake, elicits analysis,evaluation. low DIALOGIC level VALUE = 1 200201 Student:202 Student:203 Teacher: (???) 204 Student: We haven’t— 205 278–279. 206 Student: We haven’t,207 we’ve never answered the Teacher: question about how Yeah! Hold on a minute. it Now, what changed does the this relationship say … about his relationship with his children? (Q27) 27 208209 Student:210 Teacher:211 (???) 212 Is that the213 … it doesn’t say anything214 about changing. 215216 It says, What does217 this say his about children? hisBut relationshipyou’re218 right, with we yet, haven’t but it Holly answered asked219 about Heck and Tate then so we’ll let’s come220 do back. (that) (1) bottom Umm of … 278 (3) it221 On says, the “I the very neverlaw heard for tell a222 it’s citizen against to from do being Student: his committed, utmost which223 to is prevent exactly maybe a what you’ll Teacher: crime he say did. it’s224 my But duty it, to not Where’s tell hush the the it225 (???). town up. all ButAll about youthe Oh, knowladies I’m what’d in sorry226 happen Macomb … then? includingknocking “To my myon way wife his of woulddoor227 thinking, be bringing Mr. angel Finch, food cakes.” 228229 taking the one man230 who hasservice doneand you dragging andhim thiswith townhislimelight, a shyto Student: great ways me into that’s a the sin. about It’s to a have sin it and on I’m my it’d not be head. Yeah different. If But it not was this any do man, other you Mr. man, think Finch.” [name]? What Does thatmay seemknow … to (Q28) indicate that he 28

169 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Authentic student Authentic student question, Teacher test question, no uptake, elicits Who says that now? uptake, elicits a report from text.=10 DIALOGIC VALUE So what is he referring to? question, uptake, elicits a report fromVALUE = text. 10 DIALOGIC How does the fact that Atticussay won’t something be about dishonest his relationship withchildren? his analysis, low level evaluation. DIALOGIC VALUE = 0 Table 5 (Continued) Transcript With Line Numbers231232 Teacher:233234 And he says235 it would be a sin236 to drag him into the 237 limelight, and Student: then Atticus says to understand Scout, why you “Do can’t you (ever) tell (???).” And that Wait, she you who says, were says “Well, his that it’d shooting sort now? a of (Q29) mockingbird.” be (2). like Remember that saying (Bancroft) it’s a sin to kill a mockingbird? Questions as Coded and Numbered Consecutively by Class 29 238239 Teacher:240241 Scout says that.242 “Atticus disengaged himself at looked 243 Student:244 Teacher:245 at me”. Well I’m what on246 is 279. he “What “Well, referring it’d do Scouts to? be Student: you referring (Q30) sort mean?” to247 of And anybody like in Scout That’s shooting … town says, a Student: knowing mockingbird.” that248 Boo had Teacher: Ohhh. 249 Oh, okay. 250 killed Bob Now, we Ewell. still251 have not rankled with this252 one enough yet. 253 Student: 30 254 Atticus won’t Teacher: be dishonest.255 How … his How re— Umm does … (laughs). that his256 explain relationship withyou hisstarted Here’s the children?that question. Maureen, off,257 Who and does we that. Atticus never Something really think (to to might that) the have … point on Student: killed Bob That Ewell? (???) Jem. … Why doesn’t up? he Because want he it doesn’t hushed want to this be say dishonest. about What his does relationship with his children? (Q31) 31

170 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded Authentic teacher question, Authentic student question, uptake, elicits Why do you say that? uptake, elicits analysis, low level evaluation. DIALOGIC VALUE = 11 What’s going to happen to Misskids Stephanie now? and her speculation. DIALOGIC VALUE = 12 258 Teacher: Why do you say that, Kim? (Q32) 32 259260 Student:261262 Because umm263 … because I mean his264 son … people would 265266 just … Student: like …267 (???) would that, say, publicize well, it you or268 know, whatever, just and really just, (close Teacher: Now you it what’s know, off)269 going and to say, Atticus happen “My wants with kid everybody Miss (???).” to270 Stephanie But know and that He her his doesn’t I Student: kid want, don’t has you know271 done … know, his … that’s reputation a Students: to good be question!272 ruined. I don’t kids what’s now? Teacher: Oh, (Q33) no.273 (???) call (Laughing, her. overlapping brief comments). 274 33 going to I happen don’t know. to275 It’s all … those it’s (cute interesting little) speculation. kids276 We … Students:277 Teacher:278 don’t (multiple know brief what overlapping happens comments). 279 with the course Ewell now (Laughing) family they’re Who all because knows,280 orphans. of somebody As might they’ll Anna get want says, (Stephanie’s a maybe good281 father for a change. 282283 kids). (3) I want284 to hold Monday those because two I last want285 questions to for start Now I’m you going in to your286 put small you groups. the in big small issues groups that287 to will talk pullthan about thegiving you book a together,288 rather test on responsible it. for So one each big289 group question, will andmaybe be I’lltalk ask about you that to question inresponses youron groups. an Putoverhead transparency, your and what then you explain came up with tobig theissue, rest this oftopic usthat onyougroup Monday.discusswill This inbe your what small you writewell youras essay you on.can Soin doyourmore as groups,material becauseto then go you’llon have later (on the) essay.

171 Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 2014 July 24 10:33 at Madison] - Wisconsin of [University by Downloaded 172 NYSTRAND ET AL.

An example of a dialogic spell. In contrast to this recitation on the Magna Carta, we now examine a dialogic spell in the context of a ninth-grade English lesson on To Kill a Mockingbird. Table 5 juxtaposes a full transcript with the complete list of questions coded using CLASS; Figure 2 shows the questions plot, which sequentially plots the dialogic values of the questions over the course of the session. In this lesson, students have just completed a quiz, and the teacher opens the floor to student reactions and questions. The session clearly distinguishes itself from the review of the Magna Carta in that it is prompted by a student’s curiosity. He is quickly joined by several other students, frequently without teacher prompting. The teacher’s role is mainly one of directing conversational “traffic,” focusing issues, and guiding students through the text to answer their own questions. The exchange quickly evolves into a discussion lasting more than 15 min. The teacher wraps up the discussion by instructing them on their next tasks in small groups where, as in the preceding whole-class discussion, they are to grapple with big issues, figuring out an interpretation of the novel that can form the basis for an essay assignment. The focus of this dialogic exchange, starting with homework reading and extending through discussion and essay writing, is high-level comprehension. If, in Lotman’s (1988, 1990) terms, the emphasis of the monologic lesson on the Magna Carta is “accurate transmission of information,” the ori- entation here is toward the text as a “thinking device.” Our initial explorations suggested that some of the ways teachers seek to kindle dialogic interaction include (a) asking authentic questions, which value and elicit student ideas and not just mastery of information, (b) practicing uptake, in which teachers ask students follow-up questions to pursue points and lines of inquiry in- troduced by students, and (c) using high-level evaluation to valorize students’ responses (cf. Rex & McEachen, 1999), allowing their ideas and responses to influence the direction of discussion. Also important is the development of long-term “ethos” (Christoph & Nystrand, 2001; Kachur & Prendergast, 1997; Langer, 1999): If the majority of lessons are in recitation mode, it is difficult to initiate a discussion. Conversely, where dialogue is more the norm, discussion may erupt despite the teacher’s plans. As we note earlier, discourse has inertia, and the conversants quickly develop expectations about how to comport themselves.

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Construct validity. As the lesson on To Kill a Mockingbird shows, engaged student questions can be pivotal to the character and course of classroom discourse, especially when the teacher responds by opening the floor to other students’ comments and questions. When this happens, classroom discourse is trans- formed into a dialogic spell. Is not the use of the same variables (authentic questions, uptake, etc.) to specify both our key construct (dialogic spell) and our predictors of the construct (dialogic bids) tautologic? The important point here is that dialogic spells represent concentrations or concatenations of dialogic variables, and that these clusters are qualitatively different from the individual predic- QUESTIONS IN TIME 173

FIGURE 2 Question plot for highly interactive lesson on To Kill a Mockingbird: lesson with dialogic spell. Questions are noted with letter codes based on the Dialogic Value Scale shown in Table 3; student questions are noted in upper case.

tors: If dialogic spells represent an “ignition” of teacher–student interaction, dialogic bids may be thought of as “kindling” generating “sparks” that increase the odds of ignition; but the sparks, until ignition, are not “the fire.” To examine this issue, we provide two sets of analyses to assess the validity of our measure of dialogic spells. In order to test the internal consistencies of our scale and measurement of dialogic spells, we first conducted a simple analysis of variance (ANOVA) contrasting mean dialogic densities within and outside dialogic spells. As a second check, we contrast results of event history analyses of dialogic shifts with results from parallel analyses of two empirically related, albeit theoretical distinct, outcomes—discussions and student questions. To support our construct distinguishing dialogic and monologic spells, we Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 tested their respective dialogic densities in our entire data set using ANOVA. The mean density inside all dialogic spells was 6.35 inside, 2.35 outside, a ratio of nearly 3:1, consistent with our expectations. A simple ANOVA yielded a statistically significant difference in mean densities for dialogic and monologic spells (F = 4,307; df = 33,967). It is also interesting to find that far more questions occurred outside dialogic spells than within them: 30,381 questions occurred outside dialogic spells, and 3,588 questions occurred inside dialogic spells. These results clearly support the case that our dialogic spell construct results in distinctions that are inherently meaningful. 174 NYSTRAND ET AL.

Event History Models We examine three outcomes in our event history analyses: dialogic spells, student questions, and open discussion. As noted, our theoretical arguments lead us to expect that the general classroom conditions promoting dialogic spells, student questions, and discussion will be similar; nevertheless, these three outcomes differ in important ways. Thus, we expect open discussion to be less predictably structured than dialogic spells, and we expect dialogic spells to occur much less frequently than student questions. We defined the onset of a dialogic spell in the manner described earlier, when we observe a student question followed subsequently (although not necessarily immediately) by two more student questions. We defined discussion operationally as the free (unprescripted) exchange of information among at least three students and the teacher that lasted at least a half-minute during a classroom instructional episode. Student questions were identified similarly using a dummy variable that indicated whether a question was posed by the teacher or by a student. Our definition of a dialogic spell is more complicated and is given later after we review the substantive content provided by the questions in our data. All three outcomes (dialogic spells, student questions, and discussion) are thus defined at the level of instructional episodes. Our event history analyses focus on the timing of these three outcomes within an instructional episodes. We define timing in terms of the number of questions that have gone by at a given point in an episode, not in terms of the time elapsed since the start of the episode. Thus, in a classroom lesson, the teacher might pose three questions (interspersed with three student answers), at which point a student might ask a question. For our student question outcome, the relevant “timing” of classroom events thus consists of three teacher questions, followed by the outcome of interest—a student question. The timing of our other outcomes is defined similarly. Our main analyses employ event history techniques (Cox & Oakes, 1984; Tuma & Hannan, 1984) to uncover the conditions that lead to the emergence of dialogic spells, student questions, and discussion. Event history analysis is a recently developed quantitative technique for investigating the causes (antecedents) and results (consequences) of events. Known by a variety of names, it is an important methodology in many disciplines, including Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 1. Survival analysis in biostatistics (to study demographic changes, especially death). 2. Failure time analysis in engineering (to study product failure). 3. Durationanalysisineconomics(toinvestigatewhatleadstounemployment). 4. Event history analysis in sociology (to assess the causes of marriage and divorce [Hannan et al., 1977]; job mobility [Felmlee, 1982; Sørensen & Tuma, 1981; Tuma, 1976]; childbirth; infant mortality [Russell & Hammerslough, 1983]; rate of premarital birth [Wu & Martinson, 1993; Wu, 1996]; contex- QUESTIONS IN TIME 175

tual effects on student attention [Felmlee & Eder, 1983]); law enforcement (to assess arrests, convictions, jail sentences, and recidivism [Rossi, Berk, & Lenihan, 1980]); political science (to assess changes in city government [Knoke,1982];politicaltiming[Box-Steffensmeier,Arnold,&Zorn,1997]; and the political effects of economic crises (Gasiorowski, 1995]); international relations (to analyze the causes of revolutions, wars, and international conflicts).

In each case, longitudinal data are investigated using multivariate, regression-like methods to specify and estimate the sources of heterogeneity affecting hypothesized changes. Sociologists studying education often examine continuous outcomes such as years of schooling completed. Given such an outcome, statistical methods such as ordinary least-squares regression are commonly used to examine the effects of covariates on variation across individuals in such an outcome. By contrast, the outcomes in this article are binary in nature—whether (and when) a dialogic spell, student question, or discussion spell has occurred in a given classroom episode. If our research had sought merely to determine whether a dialogic spell, student question, or discussion is observed in a given classroom episode, statistical techniques such as logistic or probit regression methods would be appropriate. But as noted previously, our analytic focus here concerns not only whether particular outcomes are observed to have emerged within a given classroom episode, but also the timing of such outcomes. Indeed, our theoretical discussion in some sense privileges the latter issue—just when dialogic discourse, student questions, or discussion emerges—given our theoretical focus on how discourse unfolds within the classroom. Our focus on these two aspects—both whether and when a binary outcome is observed—requires the extension of more familiar methods such as logistic regression techniques. Consider the analogous issue of human mortality, in which the researcher is interested in both whether and when the event (i.e., death) is observed. In principle, death can occur at any given instant; hence, though the outcome of interest can be represented by a binary variable (for example, 0 representing the status “alive” and 1 representing the status “dead”), the timing of this event Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 will typically be captured by a continuous variable such as age in years, months, days, and so on. Similarly for our data, a dialogic spell, student question, or discussion may emerge at any point in a given instructional episode. Thus, in formal terms, the data needed to represent these outcomes consist of two pieces of information—a binary variable indicating whether or not the event of interest has occurred during the period of observation, and a continuous variable either giving the timing of the event or the so-called “censoring” time for cases for which the event has not occurred, where the censoring time records the amount of time elapsed. Thus, in the case of human mortality, some individuals will be observed to have 176 NYSTRAND ET AL.

died during the period of observation, in which case the event (say, coded as “1”) and the age at death will be recorded; for other individuals, death will not yet have occurred, in which case the lack of the event (coded, say, as “0”) and age at last observation will be recorded. In a logistic regression framework, the statistical outcome of interest is the so-called log odds, log(p/1 – p), giving the natural logarithm of the probability of the event divided by the probability that the event has not occurred. To model human mortality, it is necessary to extend this notion to all possible ages at which death can occur—in effect, an infinite sequence of log odds. The result for mortality is what demographers and statisticians term the age-specific mortality rate—a rate describing the force of mortality at any age t, where age is understood as a continuous variable taking positive values. A typical empirical result in human popu- lations is that the mortality rate is high during infancy; declines through childhood, adolescence, and early adulthood; and rises again during the later adult years. Note, in addition, that the mortality rate can only take nonnegative values—the mortality rate is typically positive, although it could in principle decline to zero at some ages. In our study, the outcomes are formally similar to mortality in that we are interested in the timing and occurrence of a binary outcome—when and whether a dialogic spell, student question, or discussion occurs in a given classroom episode. Our analyses employ continuous time hazard regression methods, which assume that the underlying time dimension is a continuous quantity (see Appendix).

STATIC AND DYNAMIC VARIABLES

A particular advantage of these models is that one can investigate covariates that capture any aspect of a classroom’s history prior to time t; in particular, covariates need not be static, but may themselves vary with time. We exploit this aspect by employing a number of variables that capture ebbs and flows in classroom discourse. We thus investigated two general categories of covariate effects: (a) static variables, related to both teacher and school characteristics and such background factors as SES, gender, race, ethnicity, and class size, as well as average levels of Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 student achievement within the classroom; and (b) dynamic variables, which vary both within and across lessons within our classroom observations. We investigated the effects of class and teacher characteristics for two reasons. First, to obtain unbiased estimates of dynamic variables, it is necessary to control for potentially confounding static variables. Second, because structural and compositional variables such as track, SES, gender, race, and ethnicity have been shown globally to affect classroom discourse (Gamoran & Nystrand, 1991; Nystrand, 1997), we sought to explicate the mechanism of these effects as we focused on unfolding classroom discourse. QUESTIONS IN TIME 177

Weinvestigatedthedynamiceffectsofdiscoursebecauseofitstendenciestoward coherence and inertia: Once in motion, discourse tends to maintain both genre and topic unless the speaker or writer somehow signals a shift (Nystrand, 1986). These shifts can range from the explicit (e.g., “Now I’d like to talk about something a little different”) to the implicit (e.g., indenting to indicate a new paragraph [cf. Rodgers, 1966]). In classroom discourse, one might well expect that, the longer recitation is underway, the more difficult it might be to shift to more dialogic patterns of interaction; and in this study we examined the inertia of classroom discourse in two ways: by measuring the effects of immediately previous discourse on subsequent discourse in the same instructional episode, and by modeling (via the baseline hazard) the propensity for the outcome to occur as the episode unfolds.

RESULTS

The Classroom as a Context for Discourse Are dialogic spells a common occurrence? Prior research, including previous analyses of our own data, led us to suggest that dialogic discourse is rare, and our new analyses confirm this expectation. In 1,151 instructional episodes that we observed in eighth- and ninth-grade English and social studies classes, only 66 episodes (6.69%) had even one dialogic spell—1,074 episodes (93.31%) involved no dialogic spells, whereas 7 episodes had two, and 4 episodes had three. Table 6 summarizes these findings. How are differences among classrooms related to the occurrence of dialogic spells?

Grade and Subject Differences The proportion of episodes with at least one dialogic spell was about 50% higher in social studies than in English classes, as indicated in Table 6. The proportion of dialogic spells per episode was about twice as high in eighth-grade classes as in ninth-grade classes. Table 7 summarizes this finding. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 TABLE 6 Number of Dialogic Spells per Episode by Subject

Spells per Episode Total Episodes Total % With at Least One Number of Dialogic Subject 0 1 2 3 Dialogic Spell Episodes Spells

Social studies 520 39 6 4 49 569 8.61% English 554 27 1 0 28 582 4.81% Total 1,074 66 7 4 77 1,151 6.69% 178 NYSTRAND ET AL.

TABLE 7 Number of Dialogic Spells per Episode by Grade

Spells Per Episode Total Episodes Total % With at Least One Number of Dialogic Grade 0 1 2 3 Dialogic Spell Episodes Spells

Eighth 464 42 4 3 49 513 9.55% Ninth 610 24 3 1 28 638 4.39% Total 1,074 66 7 4 77 1,151 6.69%

Dialogic Spells and Tracking Although monologic instruction was evident in urban, suburban, and rural public and parochial schools, the most monologic lessons by far were in low track classes. Indeed, the most striking finding in our study is the virtual absence of dialogic spells in low track classes: only 2 dialogic spells in the 197 instructional episodes we observed (see Table 8), no doubt a result of emphasis on skill development and test questions about prior reading, both negative predictors of dialogic discourse.

Dialogic Spells Versus Discussion Despite considerable lip service among teachers to “discussion,” we found little discussion in any classes in the sense of an open and in-depth exchange of ideas. This is most unfortunate given that our previous studies found strong associations between discussion and student achievement (Nystrand, 1997; Nystrand & Gamoran, 1991a). What most teachers in our study called “discussion” was, in the words of one teacher, “question–answer discussion”—that is, some version of recitation. By any standard, discussion was infrequent, only slightly more frequent than dialogic spells. In social studies, for example, whereas 91.39% of all episodes involved no dialogic spells, 90.33% had no discussion. In English, the figures were similar: Whereas 95.15% of all episodes had no dialogic spells, 91.24% had no discussion (see Table 9). On average, discussion in English took 50 s per class in eighth grade and less than 15 s in ninth grade. Average time for discussion in social studies was 42 s in Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 eighth grade and 31.2 s in ninth grade. In eighth-grade English classes, 20.7% spent 1 min or more on average; only 2 classes of the 58 regularly involved 7 min or more. In ninth-grade English classes, only 5.6% had more than a minute daily; only 1 class of the 58 averaged more than 2 min. Discussion occupied nearly twice as much time in high-track classes than low. To examine the relationship between discussion (open-ended conversational exchange of ideas largely absent of questions) and dialogic spells (characterized by engaged student questions and an absence of teacher test questions), we cross-tabulated their occurrences in our data set (see Table 10). Of 1,561 episodes QUESTIONS IN TIME 179

TABLE 8 Number of Dialogic Spells per Episode by Track

Spells Per Episode Total Episodes With Total % at Least One Number of Dialogic Track 0 1 2 3 Dialogic Spell Episodes Spells

Low 195 2 0 0 2 197 1.02% Regular 425 20 2 2 24 449 5.35% High 161 12 1 0 13 174 7.47% Other 293 32 4 2 38 331 11.48% Total 1,074 66 7 4 77 1,151 6.69%

TABLE 9 Discussion and Dialogic Spells (Descriptive Statistics)

Total Episodes With Episodes % Episodes % Lessons Number of No Dialogic With No With No With No Subject Episodes Spells Discussion Dialogic Spells Discussion

Social studies 569 520 514 91.39% 90.33% English 582 554 531 95.19% 91.24% Total 1,151 1,074 1,045 93.31% 90.79%

we observed, about 79% had neither discussion nor dialogic shifts, and only about 2% had both. Among episodes with dialogic spells, 14.38%1 also had discussion, whereas, among episodes without dialogic spells, the occurrence of discussion was 7.51%.2 Among episodes that had discussion, 24.77%3 also had dialogic spells; among those without discussion, only 13.73%4 had dialogic spells. Thus, discussion and dialogic spells tend to co-occur, though the overlap is not complete; the results capture distinct, though overlapping, phenomena.

The Evolving Nature of Classroom Discourse To examine the unfolding of classroom discourse, we used event history techniques to model the emergence of three types of sequences embedded in classroom Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 discourse: dialogic spells, student questions, and discussion. In our event-history

1This percentage was computed by dividing the percentage of dialogic spells with discussion (2.11%) by the total percentage of dialogic spells (12.56% + 2.11%). 2This percentage was computed by dividing the percentage of episodes with discussion but without dialogic spells (6.41%) by the total percentage of episodes without dialogic spells (78.92% + 6.41%). 3This percentage was computed by dividing the percentage of episodes with dialogic spells with discussion (6.41%) by the total percentage of episodes with discussion (6.41% + 2.11%). 4This percentage was computed by dividing the percentage of episodes with dialogic shifts but without discussion (12.56%) by the total percentage of episodes without discussion (78.92% + 12.56%). 180 NYSTRAND ET AL.

TABLE 10 Cross Tabulation of Dialogic Spells and Discussion

Discussion

Dialogic Spells No Yes Total episodes

No 78.92% 6.41% 1,332 Yes 12.56% 2.11% 229 Total episodes 1,428 133 1,561

Note. Percentages are computed as a proportion of 1,561 total episodes observed.

models, we examined the effects of both fixed attributes of classrooms, teachers, and schools (e.g., class mean SES, teacher experience, school urbanicity) and dynamic elements of classroom discourse (e.g., authentic questions, uptake, high-level evaluation, cognitive level of questions). Models that included our dynamic measures of classroom discourse were estimated twice: once with our dynamic measures of classroom discourse specified in terms of the cumulative number of each type of question prior to that point in time (e.g., in assessing the rate of the outcome at question t, how many authentic questions have occurred prior to this question), and a second time with the dynamic predictors specified as moving averages over the five questions prior to this point (e.g., what proportion of the previous five questions were authentic?). Since our theory is unclear on whether the episode as a whole leads up to the events in which we are interested, or whether events reflect only the most recent past, we adopted both approaches in our analyses. The goal of these analyses is to identify the conditions—both static and dynamic—under which dialogic spells, student questions, and discussion are most likely to occur.

What Promotes Dialogic Spells? The first column of Table 11 examines associations between the static attributes of classes, teachers, and schools, and the rate of transitions into a dialogic spell. Dialogic spells are more likely to occur in smaller classes, those of higher average

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 SES, and in social studies as opposed to English classes. Consistent with the bivariate descriptive results, dialogic spells are least likely to occur in low-track classes.5 Unexpectedly, dialogic spells are also less likely in classes with higher

5Dialogic spells are also more likely to occur in “other” tracks than in regular tracks (the omitted category). Although it is tempting to ascribe the effect of “other” track levels to heterogeneity of students in such classes, unfortunately this category reflects a mixed bag of classes that included mixed-ability classes, school-within-school classes, and classes in smaller schools (which were more likely to be mixed for their schools but relatively homogeneous compared to the sample as a whole), so it is difficult to feel confident in such an interpretation. QUESTIONS IN TIME 181

TABLE 11 Event History Analysis of Transitions Into a Dialogic Spell Using a Cox Model With Fixed, Cumulative Time Varying Covariates, and Moving Average Time Varying Covariates

Model 1: Fixed Model 2: Fixed + Model 3: Fixed + Variables Cumulative Variables Moving Averages

Static variables Class characteristics Class SES 1.01*** (.27) .52 (.28) .56* (.28) Class size –.03* (.01) –.03* (.01) –.02 (.01) Grade (1 = ninth) –.20 (.18) –.57** (.19) –.30 (.20) Subject (1 = English) –.47** (.15) –.51** (.16) –.47** (.15) Track Low track –.71* (.29) –.68* (.29) –.40 (.29) High track .38 (.23) .31 (.21) .24 (.21) Other track .56* (.20) .16 (.20) .08 (.21) % African American .85 (.94) .12 (.91) –.99 (1.00) % Hispanic .01 (.87) –.25 (.87) 1.01 (.87) % Female .70 (.52) 1.71** (.54) .94 (.56) Fall writing score –.46** (.16) –.40* (.16) –.38* (.17) Teacher characteristics Sex (1 = female) .10 (.15) .20 (.15) .34* (.15) Years teaching –.01 (.01) –.01 (.01) .00 (.01) School characteristics Catholic –.26 (.23) –.43 (.24) –.06 (.24) Urban .14 (.10) .05 (.10) –.01 (.10) Rural –.25 (.36) –.52 (.35) –.24 (.37) Dynamic variables Authentic questions .04* (.01) .19 (.13) Student questions .54*** (.04) 5.49*** (.33) Uptake .06* (.02) .39 (.26) High-level evaluation .08 (.10) –.77 (.87) High cognitive level .00 (.02) .09 (.11)

Note. Standard errors in parentheses. *p < .05. **p < .005. ***p < .0005 (two-tailed).

average fall writing scores. We scrutinized this association for possible collinearity Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 with tracking, but found that the negative association between average fall writing scores and dialogic spells held for all tracks. Thus, dialogic spells occur less often in lower tracks; but within tracks, classes with higher average writing performance exhibit fewer dialogic spells. This may indicate a tradeoff between an emphasis on written and oral performance. Thenegativeassociationbetweenlow-trackclassesanddialogicspellsisparticu- larlysignificantinlightofearlieranalysesofthesedata,whichshowedthatauthentic questions and uptake (two key features of dialogic discourse) were about equally common in high- and low-track classes (Gamoran et al., 1995; Nystrand, 1997). In 182 NYSTRAND ET AL.

additional analyses of variances we conducted (results not shown), we found, not surprisingly,thatthedensityofauthenticquestionsanduptakeissignificantlyhigher in dialogic spells than in monologic spells. This suggests that, although high- and low-track classes have roughly similar aggregate distributions of authentic ques- tionsanduptake,therearesignificantdifferencesbytrackwithrespecttothecluster- ing of such questions within dialogic spells, with authentic questions and uptake tending to be relatively dispersed in low-track classes, and more likely to be structured systematically within dialogic spells in high-track classes. In the second column of Table 11, we introduce our dynamic measures of instruction into the analysis. Here the effects of the dynamic variables are cumulative, so that we are asking whether the cumulative number of authentic questions, student questions, uptake, high-level evaluations, and questions with high cognitive level at any given moment in an instructional episode is associated with the subsequent onset of a dialogic spell. Thus, if dialogic spells emerge as a consequence of dialogic bids by teachers or students, we should find positive associations between cumulative features of previous questions and the subsequent emergence of dialogic discourse. Conversely, if dialogic spells are random with re- spect to the cumulative features of previous questions, no such associations should appear. The results of this model indicate that our cumulative measures of authentic questions, student questions, and uptake are in indeed associated with the subsequent onset of a dialogic spell. The effect of student questions is particularly noteworthy, with one student question raising the rate of onset of a dialogic spell by 72%6 relative to instructional episodes in which no student question is observed. Note, moreover, that these models specify multiplicative (and hence non- linear) effects of our cumulative measures; hence, our estimates imply, ceteris pari- bus, that the rate of a dialogic spell is 194% higher7 in classroom episodes in which two student questions are observed relative to those in which no such questions are observed. Two other of our dialogic bid measures, authentic questions and uptake, have significant effects in Model 2. Thus, the effect of one prior authentic question raises the rate of onset of a dialogic spell by 4%,8 whereas the effect of, for example, five prior authentic questions raises the rate of onset by 22%.9 Similarly, the effect of one prior instance of uptake raises the rate of onset of a dialogic spell by 6%,10 whereas the effect of, for example, four prior instances of uptake raises the Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 rate of onset by 27%.11 In the third column, we estimate a similar model except that the dynamic predictors consist of our moving average measures, defined over the previous five ques-

6100*[exp(.54) – 1] = 71.6. 7100*[exp(2 * .54) – 1] = 194.5. 8100*[exp(.04) – 1] = 4.08. 9100*[exp(5 * .04) – 1] = 22.14. 10100*[exp(.06) – 1] = 6.18. 11100*[exp(4 * .06) – 1] = 27.12. QUESTIONS IN TIME 183

tions, instead of cumulatively over the entire episode. Note that our moving average measures can assume only six possible values (0, .2, .4, .6, .8, and 1.0) in contrast with our cumulative measures, which can assume any nonnegative integer value. These differences are reflected in the parameter estimates, which differ by an order of magnitude for our dynamic measures in the last two columns of Table 11. Two findings in Model 3 are particularly salient. First, once again student questions are strongly associated with the onset of dialogic spells, with the occurrence of one student question in the last five questions raising the rate of a dialogic spell by 200%12 relative to an instructional episode in which no student question is observed in the last five questions. Authentic questions and uptake have positive coefficients but, unlike Model 2, they are not statistically significant. Second, the coefficient for low track, which was –.71 and –.68 in Models 1 and 2, declines in magnitude to –.40 in Model 3 and is no longer statistically significant. This finding provides an important clue about why dialogic discourse is so rare in low-track classes, suggesting that clusters of student questions (i.e., the occurrence of student questions within a five-question sequence), which, according to Model 3, show a large and significant association with the onset of a dialogic spell, are rarely found in low-track classes.

When Do Students Ask Questions? If student questions are important for spurring dialogic discourse, what factors might, in turn, lead students to ask questions within a classroom setting? We address this issue in Table 12, which estimates models similar to those of Table 11 but with student questions as the dependent variable. Model 1 yields a similar pattern of findings as in the analogous model in Table 11, including negative coefficients for both low-track classes and average fall writing score. Student questions are also less likely to occur in classes with more experienced teachers, although the effect is small, with one additional year of experience decreasing the rate of student questions by about 1%.13 In Model 2, we see that cumulative rates of authentic questions, uptake, and high-level evaluation are all powerfully associated with student questions. Note that theeffectforquestionsofhighcognitivelevel,whichwasclosetozeroinTable11for Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 dialogic spells, is negative and statistically significant for student questions in Table 12. These findings point to a potential tension within classrooms: While student questions appear to spur the onset of dialogic discourse within instructional settings, challengingteacherquestions(thatis,thoseposedathighcognitivelevels)maypose a barrier to active student participation in classroom discourse (via student questions) by reinforcing the voice of the instructor as dominant and authoritative,

12100*[exp(.2 * 5.49) – 1] = 99.8. 13exp(–.01) = .99. 184 NYSTRAND ET AL.

though just how this might play out in any given case would seem to depend on how the teacher handles student responses to such questions. In Model 2 in Table 12, the effect of an authentic question is .03; hence, one prior authentic question raises the rate of onset of a student question by 3%,14 whereas, for example, six prior authentic questions would raise the rate of onset by 20%.15 The effect of uptake is .05; hence, one prior instance of uptake raises the rate of onset by 5%,16 whereas 10 prior instances of uptake would raise the rate by 65%.17 We observe the largest magnitude effect (.12) for high-level evaluation; hence, one prior instance of high-level evaluation raises the rate of onset by 13%,18 whereas three prior instances of high-level evaluation would raise the rate by 43%.19 Finally, high cognitive level has a negative effect (–.03); hence, one prior instance of high cognitive level lowers the rate of onset by 3%,20 whereas, for example, six prior instances of high cognitive level would lower the rate by 16%.21 Model 3 yields similar results for effects on student questions when the predictors are specified as moving averages rather than in a cumulative fashion. As in Ta- ble 11, the most notable difference between Models 2 and 3 is that the negative coefficient for low track declines in magnitude (from –.37 to –.22) and is not statistically significant in Model 3. This pattern is consistent with our hypothesis that classroom discourse is more responsive to local, as opposed to more aggregate, aspects of the evolving character of classroom discourse, and that these differences may better capture observed differences between low- and regular-track classes. In Table 12, Model 3, the estimated coefficient for high-level evaluation is .51; a single instance would increase the probability of eliciting a student question in the subsequent five questions by 11%,22 two instances of high-level evaluation would increase the probability by 23%,23 three instances by 36%,24 and so forth.

What Brings on Discussion? Finally, we ask what features of schools, classes, teachers, and classroom interaction precipitate discussion during classroom instruction. Recall that we em-

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 14100*[exp(.03) – 1] = 3.05. 15100*[exp(6 * .03) – 1] = 19.72. 16100*[exp(.05) – 1] = 5.13. 17100*[exp(10 * .05) – 1] = 64.87. 18100*[exp(.12) – 1] = 12.75. 19100*[exp(3 * .12) – 1] = 43.33. 20100*[exp(-.03) – 1] = –2.96. 21100*[exp(6 * –.03) – 1] = 16.5. 22100*[exp(.2 * .51) – 1] = 10.74. 23100*[exp(.4 * .51) – 1] = 22.63. 24100*[exp(.6 * .51) – 1] = 35.80. QUESTIONS IN TIME 185

TABLE 12 Event History Analysis of Transitions Into Students Asking Questions, Using a Cox Model With Fixed, Cumulative Time-Varying Covariates, and Moving Average Time-Varying Covariates

Model 1: Fixed Model 2: Fixed + Model 3: Fixed + Variables Cumulative Variables Moving Averages

Static variables Class characteristics Class SES .54*** (.12) .51*** (.12) .42*** (.12) Class size .02** (.01) –.02** (.01) –.02*** (.01) Grade (1 = ninth) .02 (.08) –.08 (.08) –.27** (.08) Subject (1 = English) .19** (.06) –.15* (.06) –.13* (.06) Track Low track –.39** (.12) –.37** (.12) –.22 (.12) High track .11 (.10) .15 (.09) .18 (.09) Other track .32*** (.08) .33*** (.08) .27** (.08) % African American –.02 (.40) .10 (.40) .50 (.39) % Hispanic .46 (.36) .39 (.36) –.12 (.36) % Female .22 (.23) .10 (.23) –.38 (.23) Fall writing score –.21** (.07) .00 (1.00) .00 (1.00) Teacher characteristics Sex (1 = female) .06 (.06) .04 (.06) .04 (.06) Years teaching –.01* (.00) –.01* (.00) .00 (.00) School characteristics Catholic –.16 (.10) –.20* (.10) –.21* (.10) Urban .05 (.04) .07 (.04) .15** (.05) Rural –.16 (.15) –.11 (.15) .06 (.15) Dynamic variables Authentic questions .03*** (.01) .96*** (.04) Uptake .05*** (.01) 1.05*** (.09) High-level evaluation .12** (.04) .51* (.24) High cognitive level –.03*** (.01) –.37*** (.04)

Note. Standard errors in parentheses. *p < .05. **p < .005. ***p < .0005 (two-tailed).

ploy a rather specific operational definition of discussion—the unprescripted exchange of ideas among students and the teacher—as opposed to more com- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 mon question–answer recitation. The models portrayed in Table 13 are essen- tially the same as those in Tables 11 and 12, with our dynamic measures of classroom instruction calculated in terms of the questions that occurred prior to when discussion began, rather than prior to the question that signaled the beginning of a dialogic spell (in Table 11) or a question asked by a student (in Table 12). In contrast to the results of Tables 11 and 12, none of the fixed attributes of classes, teachers, or schools in Table 13 is significantly associated with the onset of a discussion spell. The coefficient for low track, which has the largest association 186 NYSTRAND ET AL.

TABLE 13 Event History Analysis of Transitions Into a Discussion Spell, Using a Cox Model With Fixed, Cumulative Time-Varying Covariates, and Moving Average Time-Varying Covariates

Model 2: Fixed + Model 3: Fixed Model 1: Fixed Cumulative Variables + Moving Variables Variables Averages

Static variables Class characteristics Class SES –.18 (.75) –.45 (.78) –.62 (.79) Class size –.03 (.03) –.04 (.03) –.05 (.03) Grade (1 = ninth) .51 (.48) .41 (.51) –.01 (.51) Subject (1 = English) –.25 (.36) –.52 (.38) –.42 (.37) Track Low track –1.97 (1.10) –1.99 (1.11) –2.05 (1.11) High track –.14 (.59) –.10 (.52) –.08 (.53) Other track .02 (.50) –.09 (.49) –.20 (.50) % African American 1.28 (2.19) 1.64 (2.16) 2.11 (2.20) % Hispanic .01 (2.05) –.07 (2.08) –.22 (2.04) % Female 1.04 (1.36) .91 (1.43) .18 (1.48) Fall writing score .39 (.45) .52 (.46) .61 (.45) Teacher characteristics Sex (1 = female) .23 (.36) .23 (.38) .23 (.38) Years teaching –.01 (.02) –.01 (.02) –.01 (.02) School characteristics Catholic .17 (.57) .35 (.59) .21 (.57) Urban .32 (.30) .30 (.31) .37 (.31) Rural –.50 (1.00) –.32 (1.03) –.19 (1.03) Dynamic variables Authentic questions .01 (.02) .43 (.29) Student questions .08 (.05) 1.38* (.70) Uptake .04 (.03) 2.02*** (.55) High-level evaluation .06 (.15) –.37 (2.02) High cognitive level .04 (.03) .60* (.24)

Note. Standard errors in parentheses. *p < .05. **p < .005. ***p < .0005 (two-tailed).

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 with discussion is (–1.97), not statistically significant, probably reflecting the greater variability of discussion within low-track classes as well as the relative dif- ficulty of estimating this effect given that discussion occurs relatively infrequently in instructional settings, let alone those within low track classes. The results from Model 2, which adds cumulative indicators of dynamic variables as additional predictors, show effects that are generally in the expected direction, but which are not significant coefficients. Only in Model 3, which focuses on whether or not a discussion occurs subsequent to each successive set of five questions, do we find sig- QUESTIONS IN TIME 187

nificant predictors. Here, as in previous tables, the results imply that discussion occurs more in response to proximal features of discourse—more likely to occur when immediately preceded by high proportions of student questions, uptake, and questions with high cognitive demands. Cognitive level thus exerts two competing influences on discussion, with a positive direct impact counterbalanced by a negative indirect effect via its negative impact on student questions (see Table 12), which is to say, challenging teacher questions often stimulate discussion at the same time they tend to discourage student questions.

A Dynamic Model of Unfolding Classroom Discourse In Bakhtin’s (1981) terms, dialogically organized instruction makes public space for student responses, eliciting and accommodating the differing values, beliefs, and perspectives of the conversants, and ideally including the voices of different classes, races, ages, and genders. Dialogically organized instruction is fueled by such pluralism and heteroglossia, and the extent of social interaction clearly shapes both instruction and learning. To the extent that classroom interaction is dialogic, the balance of discourse is symmetrical so that the teacher’s voice is but one voice among many, albeit a critically important one. Yet repeated empirical findings that monologic discourse is prevalent across American classrooms—our previous studies, for example, found that 85% of the classes we observed were devoted to some combination of preplanned lecture, recitation, and seatwork (Nystrand, 1997; Nystrand & Gamoran, 1991a)—clearly suggest that monologic discourse dominates within classrooms. Indeed, the force of monologic discourse within the classroom exerts considerable inertia so that classroom interaction tends to continue in direction and character unless someone, invariably the teacher, counteracts this force, either directly by doing something, or indirectly by allowing a shift to occur. Sequences of dialogic whole classroom discourse take three essen- tial forms distinguished by the role played by the teacher.

Teacher moves: Priming the possibilities for dialogic discourse. Teachers have a number of options for overcoming the monologic inertia of normal classroom discourse. They can act in direct ways to shift classroom discourse from Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 monologic to dialogic patterns with little or no transition, for example, by posing a particularly timely or provocative question. But a shift from monologic to dialogic discourse can also be foreshadowed in indirect ways, for example by the teacher’s repeated efforts to elicit student contributions that lay the ground for such a shift. The dialogic bids include actively welcoming and soliciting student ideas and observations by following up their responses, and opening the floor to students by asking questions that do not have predetermined answers. Teachers may also break the mold by deliberately withholding or skipping their evaluation of students’ re- 188 NYSTRAND ET AL.

sponses, for example, allowing a student remark to substitute for their evaluation, and encouraging students to respond to each other. All these moves “kindle” the possibilities of discussion and substantive interaction. All such efforts are bids for dialogicality, and each such bid, our study shows, increases the probability of dialogic interaction cumulatively.

Student moves: Engaged response. Recitation patterns can also break down when students take up these bids by freely voicing their own ideas and asking engaged questions. If appropriate teacher moves provide the “kindling” for dialogic interaction, such key student moves provide the “spark.” Student questions about the content of instruction, for example, heighten the dialogic potential of classroom discourse, and they are an important source of dialogic bids in their own right.25 Unlike teachers, students don’t ask questions when they already know the answer, but instead typically pose questions eliciting additional information and/or clarifying something the teacher has said. And because students’ questions follow up something someone else has said, they especially have the power to enhance the dialogicality of classroom discourse when the teacher, rather than answering the question and quickly getting back to the lesson plan, allows classroom talk to move in directions prompted by the question. As a consequence, we might expect students, as well as the teacher, to play key roles in initiating sequences of dialogic discourse. Clusters of student questions especially, we find, carry dialogic weight, even when interrupted by teacher questions, indicating movement away from recitation and test questions to more substantively interactive dialogic discourse.

Ignition: Discussion. When conditions are right, especially including student uptake of dialogic bids offered by the teacher, the result is a critical discourse mass that yields open discussion in which teachers and their students work out understandings face-to-face—the quintessential form of dialogic interaction. When this happens, the teacher’s role is limited mainly to getting and then keeping the ball rolling, leaving it to students to make most of the substantive observations. Discussion tends to be marked by the absence of questions, either by teacher or students, except for purposes of clarification. Discussion is not about the transmis- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 sion and recitation of information; the teacher’s role is not mainly that of asking

25Wells (personal communication, December 24, 2000) notes that a “fairly common type of student bid in the DICEP (Developing Inquiring Communities in Education Project) data takes the form of the student first bidding for a turn and then, when given the floor, giving some information—experience, opinion, conjecture—that s/he thinks is relevant to the topic under discussion. The onus is then on the teacher to respond with uptake and encouragement of further development of the student’s sub-topic.” For more on DICEP, see http://webcat.library.wisc.edu:8000/WebZ/html/homeframe.html:sessionid= 0?style=rss:next=html/homeframe.html:bad=html/homeframe.html QUESTIONS IN TIME 189

questions to see how much students knew and to go over the points they did not yet understand. Instead, discussion is about figuring things out—in face-to-face discussion, teacher and students together. The following brief, transcribed excerpt from a ninth-grade English lesson on Mark Twain’s The Adventures of Huckleberry Finn, illustrates the three phases of this model of dialogic interaction. The teacher (a) primes the possibilities dialogically by inviting students to speculate about possible examples of racism in the novel. Soon after, (b) an engaged student wonders if Twain was merely being historically accurate, not racist, when he referred to “niggers”: “Isn’t [Twain] being historically accurate when he says ‘those niggers’?” asks Sam. “So why is it racist?” This in turn gives way to (c) an open discussion of racism in The Adven- tures of Huckleberry Finn: (a) Teacher move dialogically priming the possibility for discussion Ms. Turner: Can you recall things from Huck Finn that, um, seemed racist to you? Tasha: Miss Watson’s, that guy she’s always calling ‘Miss Watson’s nigger.’ Ms. Turner: Ok. Jim? Jim: Well, they sell the slaves. Also, they said in one part, ‘fetch in the nigger.’ Ms. Turner: [Yeah Jim: [and it’s like, you know, it’s like you’re saying to a dog, ‘Here, boy.’ Ms. Turner: Right—‘We fetched in the niggers to have prayers.’ Yeah, that’s in probably the first couple of pages. Good. Sam? (b) Engaged student response Sam: Isn’t he being historically accurate when he says ‘those niggers’? Ms. Turner: Oh, yes, absolutely. Sam: So why is it racist? Ms. Turner: Well this—that’s kind of what I was trying to bring out on the first day—is that Twain is really just trying to mirror the society, and especially the society of Missouri at the time, but Twain is using the word rather sarcastically. I mean, you’re right, he’s be-

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 ing historically accurate, but he’s also trying to make a point, um, about the different people who are saying things like that. How did that make you guys feel, I mean what was your gut reaction to all that? Linda? Linda: Ashamed. Ms. Turner: In what way? (c) Ignition: Discussion Linda: That the one that it was for was wanted to believe that it was something else. 190 NYSTRAND ET AL.

Cassie: Everyone claims it’s so historical, you can find that anywhere ‘nigger,’you know, you just hear that and people always think it’s so historical. Ms. Turner: Like, oh, we wouldn’t do that anymore. Cassie: Yeah, like oh, we’re not primitive. You know, and it’s not, I mean, everybody does that, all the time. Well, not everybody, but people, people do that people can’t get in[to] apartment buildings because they’re black. Ms. Turner: Um-hm. Jim: They can’t go to certain stores because they’re black, or they’re arrested because they’re black. You know, it’s just, I mean, everybody is always saying how historical it is, and it’s right here, and it’s right now.

Here we can see that, metaphorically, getting a discussion going is a little like building a fire: With enough kindling of the right sort, accompanied by patience, and along with the spark of student engagement, ignition is possible, though perhaps not on teachers’first or second try (see Marshall, Smagorinsky, & Smith, 1995, for examination of the challenges teachers must overcome to get discussion going).

DISCUSSION AND CONCLUSION

Current understanding of instructional processes and classroom discourse has been shaped, on the one hand, by quantitative studies that have isolated global characteristics of effective classroom discourse and, on the other, by case studies that have provided accounts and analyses of numerous individual lessons. Each kind of research has idiosyncratic limitations. On the one hand, the individual nature of case studies makes it difficult to generalize beyond the individual cases, which consequently limits the use of such studies to investigate the general effects of classroom discourse on student learning. On the other hand, the abstracting and decontextualizing character of statistical studies identifying (and reifying) global characteristics of effective classroom discourse, even when associations with achievement are demonstrated, makes it difficult to understand the structure of the Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 event and to know how to implement the findings. Nevertheless, traditional approaches to the study of schools have, as a result, tended to neglect what is intu- itively important about the classroom environment, which is that skilled teachers enliven the learning experience by engaging their students in active inquiry, in contrast to the more typical teaching environment, which is characterized by rote memorization and recitation of instructional materials. We argue that our findings provide tentative but intriguing clues about what ed- ucators might take as factors important to incorporate into their teaching. In our own previous research, for example, we have concluded that teachers will do well QUESTIONS IN TIME 191

to ask their students authentic questions. Yet this kind of advice, sound as it may be, has limitations for teachers who must figure out how and when best to do it. What has been missing is a general understanding of how effective classroom discourse unfolds, how it is initiated, and just what the teacher’s role is in shaping it. By applying methods of event history analysis to a rich and extensive data set, we both address these issues and demonstrate the value of time-sensitive quantitative methods for research on both discourse and instruction. As an example of what our dynamic analysis revealed that the static analysis obscured, consider the case of instruction in low-track classes. Previously, we reported that authentic questions and uptake occurred with similar frequency in high- and low-track classes. This study does not contradict that finding, but it shows that the pattern and clustering of authentic questions, uptake, and other elements of dialogic instruction are much less likely to occur as a sequence in low-track classes. Moreover, student questions, which help precipitate dialogic spells and which are themselves important elements of dialogic spells, occur infrequently in low-track classes. Examination of global characteristics of discourse obscured these crucial findings, since it is the patterning and sequencing of these elements, not their global averages, which appear important to classroom discourse. These findings are consistent with our prior theoretical conceptualization of classroominstruction,whichisthatinstructionisbestthoughtofnotintermsofwhat teachers do to students, but what teachers and students do together (Nystrand & Gamoran, 1991). Our findings in this article reaffirm this perspective. In the past, we supported our view by showing that rates of off-task behavior were higher, and rates of completing assignments were lower in low-track classes; this was part of the reason for achievement disadvantages of low-track students. This article complements those findings by showing that rates of student questions are also lower in low-track classes. Yet we are emphatically not saying that low-track students are “at fault” for their own disadvantages, nor are we “blaming” teachers for the low achievement of low-track students. Nevertheless, our results emphasize the importance of understanding teacher and student behaviors, as reflected in the questions they pose, as a dynamic process. Our event history analyses show how the absence of student ques- tionsinlow-trackclassesfailstoprecipitatedialogicspells,andatthesametimehow Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 the absence of clusters of dialogic teacher questions inhibits the emergence of student questions. In low-track classes, therefore, neither teachers nor students tend to offer dialogic bids, and hence monologic forms of discourse predominate in these tracks to an even greater degree than in other tracks. Although classroom discourse is the principal medium of learning in school—Cazden (1988) calls it “the language of learning”—teachers rarely pay attention to how they structure it. Understandably, they focus mainly on what they are teaching and what their students are learning; at best, awareness of classroom discourse itself is subsidiary. Our findings suggests that this is unfor- 192 NYSTRAND ET AL.

tunate since the structure, quality, and flow of classroom discourse are all likely to affect what students learn and how well they learn it. Yet, at present, teachers possess little systematic information about these connections, in part because past research has often been inattentive to the role of discourse in learning. Our work addresses this issue in several ways, by modeling the pulse of classroom discourse and assessing the role of authentic questions, student questions, and uptake, for example, in terms of their cumulative and proximal role in structur- ing the foundation for dialogic zones of interaction. Student–teacher interaction—indeed even individual teacher questions—have their roots in previous interactions, with current interactions carrying implications for subsequent ones. Understanding how classroom discourse unfolds and their constitutive role in the process, thus, may help teachers gain informed control over how they interact with students and how they might create instructional settings that both engage students and foster learning. Since its inception in the 1970s, discourse analysis has been an exclusive zone of qualitative research. Indeed, for many, “quantitative discourse analysis” is an oxymoron; a recent review of qualitative and quantitative methods in discourse research (Lazaraton, 2002) reports, “Remarkably, it was impossible to locate a published applied linguistics study which claims to be both discourse analytic and quantitative in nature” (p. 34). We believe that discourse event history analysis, using naturalistic, observational data, may change this state of affairs. Though perhaps less nuanced than conversation analysis (CA), event history analysis is well suited, like CA, to investigating the dynamics of unfolding discourse. Its promise, demonstrated in our study examining discourse as an evolving process while assessing the effects of hypothesized variables on this unfolding process, resides in its power to analyze large numbers of conversant interactions with attention to contextual factors, on the one hand, and dynamic patterning of unfolding discourse, on the other, systematically contrasting both macro- and microsocial effects.

ACKNOWLEDGMENTS

The research in this study was supported by two awards from the U.S. Depart- Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 ment of Education’s Office of Educational Research and Improvement: Data were collected during 1986 through 1989 at the National Research Center on Ef- fective Secondary Schools, and were supported by a grant from the Office of Ed- ucational Research and Improvement (Award G–008690007–89); the study reported here is based partly on results from that study and were reported in Nystrand (1997), Nystrand and Gamoran (1991), and Gamoran and Nystrand (1991). Data analysis for the study reported here was conducted during 1997 through 1999 at the National Research Center on English Learning and Achieve- ment and was supported by a grant from the Office of Educational Research and QUESTIONS IN TIME 193

Improvement (Award R305A60005). However, the views expressed herein are those of the authors and do not necessarily represent the views of the supporting agencies. James Fehrenbacher provided computer programming used to weight questions and display question plots. The authors appreciate the comments of Arthur Applebee, David Bloome, Samantha Caughlan, Cecilia Ford, George Hillocks, Douglas Maynard, and Gordon Wells. Wu was a National Research Center on English Learning and Achievement (CELA) researcher at the time of the study. Both Long and Zeiser were research assistants at the time of the study.

REFERENCES

Applebee, A. N. (1981). Writing in the secondary school (Research Report No. 21). Urbana, IL: Na- tional Council of Teachers of English. Applebee, A., Burroughs, R., & Stevens, A. (2000). Creating continuity and coherence in high school literature curricula. Research in the Teaching of English, 34, 396–428. Bakhtin, M. (1981). The dialogic imagination. Austin: University of Texas Press. Bakhtin, M. (1984). Problems of Doestoevsky’s poetics (C. Emerson, Trans.). Minneapolis: University of Minnesota Press. Bakhtin, M. (1986). Speech genres & other late essays (V. W. McGee, Trans.; C. Emerson & M. Holquist, Eds.). Austin: University of Texas. Bellack, A., Kliebard, H., Hyman, R., & Smith, F. (1966). The language of the classroom. New York: Teachers College Press. Box-Steffensmeier, J., Arnold, L., & Zorn, C. (1997, June). The strategic timing of position taking in congress: A study of the North American Free Trade Agreement. American Political Science Review, 91, 324–338. Brazil, D. (1985). Phonology: Intonation in discourse. In T. van Dijk (Ed.), Handbook of discourse analysis. Vol. 2: Dimensions of discourse (pp. 57–75). London: Academic. Britton, J., Burgess, T., Martin, N., McLeod, A., & Rosen, H. (1975). The development of writing abili- ties: 11–18. London: Macmillan. Cazden, C. (1988). Classroom discourse: The language of teaching and learning. Portsmouth, NH: Heinemann. Christoph, J. N., & Nystrand, M. (2001). Eliciting and sustaining discussion in a low-achieving ninth-grade inner-city English class (CELA Research Report). Albany, NY: The National Research Center on English Learning and Achievement. Cicourel, A. (1973). Cognitive sociology. Harmondsworth, England: Penguin.

Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Collins, J. (1982). Discourse style, classroom interaction and differential treatment. Journal of Reading Behavior, 14, 429–437. Colvin, S. (1919). The most common faults of beginning high school teachers. The professional prepa- ration of high school teachers. In G. M. Whipple & H. L. Miller (Eds.), Eighteenth Yearbook of the National Society of Education, Part I (pp. 262–272). Bloomington, IL: Public School Publishing Co. Corey, S. (1940). Some implications of verbatim records of high school classroom talk. NEA Proceed- ings, 371–372. Cox, D. R., & Oakes, D. (1984). Analysis of survival data. London: Chapman and Hall. Duffy, G. (1981). Teacher effectiveness research: Implications for the reading profession. In M. Kamil (Ed.), Directions in reading: Research and instruction: 30th yearbook of the National Reading Con- ference (pp. 113–136). Rochester, NY: National Reading Conference. 194 NYSTRAND ET AL.

Durkin, D. (1978–1979). What classroom observations reveal about reading comprehension instruction. Reading Research Quarterly, 14, 481–533. Felmlee, D., & Eder, D. (1983). Contextual effects in the classroom: The impact of ability groups on student attention. Sociology of Education, 56, 77–87. French, P., & Maclure, M. (1981). Teachers’ questions, pupils’ answers: An investigation of questions and answers in the infant classroom. First Language, 2, 31–45. Gamoran, A., & Nystrand, M. (1991). Background and instructional effects on achievement in eighth-grade English and social studies. Journal of Research on Adolescence, 1, 277–300. Gamoran, A., Nystrand, M., Berends, M., & LePore, P. (1995, Winter). An organizational analysis of the effects of ability grouping. American Educational Research Journal, 32, 687–716. Garfinkel, H. (1967). Studies in ethnomethodology. Englewood Cliffs, NJ: Prentice Hall. Gasiorowski, M. (1995, December). Economic crisis and political regime change: An event history analysis. American Political Science Review, 89, 882–897. Gee, J. (1999). An introduction to discourse analysis: Theory and method. London: Routledge. Goodlad, J. I. (1984). A place called school: Prospects for the future. New York: McGraw-Hill. Goodwin, C. (1979). The interactive construction of a sentence in natural conversation. In G. Psathas (Ed.), Everyday language. New York: John Wiley. Goodwin, C. (1981). Conversational organization: Interaction between speakers and hearers.New York: Academic. Greenleaf, C., & Freedman, S. (1993). Linking classroom discourse and classroom content: Following the trail of intellectual work in a writing lesson. Discourse Processes, 16, 465–505. Hannan, M., Tuma, N., & Groeneveld, L. (1977). Income and marital events: Evidence from an income-maintenance experiment. American Journal of Sociology, 82, 1186–1211. Heritage, J., & Roth, A. (1995). Grammar and institution: Questions and questioning in the broadcast news interview. Research on Language and Social Interaction, 28, 1–60. Hoetker, J., & Ahlbrand, W. (1969). The persistence of the recitation. American Education Research Journal, 6, 145–167. Hoetker, W. (1967). Analyses of the subject matter related verbal behavior in nine junior high school English classes. Unpublished Ed.D. dissertation, Washington University. Holquist, M. (1990). Dialogism: Bakhtin and his world. London: Routledge & Kegan Paul. Jefferson, G. (1974). Error correction as an interactional resource. Language and Society, 2, 181–199. Kachur, R., & Prendergast, C. (1997). A closer look at authentic interaction: Profiles of teacher–student talk in two classrooms. In M. Nystrand (Ed.), Opening dialogue: Understanding the dynamics of language and learning in the English classroom. New York: Teachers College Press. Knoeller, C. (1993). How talking enters writing: A study of 12th graders discussing and writing about literature. Unpublished doctoral dissertation, University of California, Berkeley. Knoke, D. (1982). The spread of municipal reform: Temporal, spatial, and social dynamics. American Journal of Sociology, 87, 1314–1339. Kutz, E. (1997). Language and literacy: Studying discourse in communities and classrooms. Portsmouth, NH: Boynton Cook. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 Langer, J. (1995). Envisioning literature: Literary understanding and literature instruction. New York: Teachers College Press. Langer, J. (1999). Excellence in English in middle and high school: How teachers’ professional lives support student achievement (CELA Research Report 12002). Albany, NY: The National Research Center on English Learning and Achievement. Lazaraton, A. (2002). Quantitative and qualitative approaches to discourse analysis. Annual Review of Applied Linguistics, 22, 32–51. Lotman, Y. (1988). Text within a text. Soviet Psychology, 26, 32–51. Lotman, Y. (1990). Universe of the mind: A semiotic theory of culture (Ann Shulman, Trans.). Bloomington: Indiana University Press. QUESTIONS IN TIME 195

Marshall, J., Smagorinsky, P., Smith, M., & Dale, H. (1995). The language of interpretation: Patterns of discourse in discussions of literature. Urbana, IL: National Council of Teachers of English. Mehan, H. (1979). Learning lessons. Cambridge, MA: Harvard University Press. Miller, W. (1922). The administrative use of intelligence tests in the high school. In G. Whipple (Ed.), The administrative use of intelligence tests.Twenty-first yearbook of the National Society for the Study of Education, Part II (pp. 189–222). Bloomington, IL: Public School Publishing Co. Moffett, J. (1968). Teaching the universe of discourse. Boston: Houghton Mifflin. Mukarovsky, J. (1977). Two studies of dialogue. In J. Burbank & P. Steiner (Eds.), The word and verbal art. New Haven, CT: Yale University Press. Newmann, F. M. (1988). Can depth replace coverage in the high school curriculum? Phi Delta Kappan (January), 69, 345–348. Nystrand, M. (1986). The structure of written communication: Studies in reciprocity between writers and readers. Orlando, FL: Academic. Nystrand, M. (1988). CLASS (Classroom language assessment system) 2.0: A Windows laptop computer system for the in-class analysis of classroom discourse. Madison: Wisconsin Center for Educa- tion Research. Nystrand, M. (1997). Opening dialogue: Understanding the dynamics of language and learning in the English classroom. New York: Teachers College Press. Nystrand, M., & Gamoran, A. (1991a). Instructional discourse, student engagement, and literature achievement. Research in the Teaching of English, 25, 261–290. Nystrand, M., & Gamoran, A. (1991b). Student engagement: When recitation becomes conversation. In H. Waxman & H. Walberg (Eds.), Contemporary research on teaching. Berkeley, CA: McCutchan. Nystrand, M., Greene, S., & Wiemelt, J. (1993, July). Where did composition studies come from? An intellectual history. Written Communication, 10, 267–333. Polanyi, M. (1958). Personal knowledge. London: Routledge & Kegan Paul. Prawat, R. S. (1995). Misleading Dewey: Reform, projects, and the language game. Educational Re- search, 24, 13–27. Rex, L., & McEachen, D. (1999, August). “If anything is odd, inappropriate, confusing, or boring, it’s probably important”: The emergence of inclusive academic literacy through English classroom discussion practices. Research in the Teaching of English, 34, 65–129. Rodgers, P. (1966). A discourse-centered rhetoric of the paragraph. College Composition and Commu- nication, 17, 2–11. Rommetveit, R. (1974). On message structure: A framework for the study of language and communication. London: John Wiley. Rossi, P., Berk, R., & Lenihan, K. (1980). Money, work, and crime: Some experimental results.New York: Academic. Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization of turn-taking in conversation. Language, 50, 696–735. Sarason, S. (1983). Schooling in America: Scapegoat and salvation. New York: Free Press. Saussure, F. (1959/1915). [Course in general linguistics] (C. Bally & A. Sechelaye in collaboration Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 with A. Riedlinger, Eds.; Wade Baskin, Trans.). New York: Philosophical Library. Schegloff, E. A. (1984). On some questions and ambiguities in conversations. In J. M. Atkinson & J. Heritage (Eds.), Structures of social action: Studies in conversation analysis (pp. 28–52). Cam- bridge: Cambridge University Press. Schutz, A. (1962). Collected papers Vol. 1: The problem of social reality. The Hague, The Netherlands: M. Nijhoff. Stevens, R. (1912). The question as a measure of efficiency in instruction: A critical study of classroom practice (Contributions to Education No. 48). New York: Teachers College, Columbia University. Tharp, R., & Gallimore, R. (1988). Rousing minds to life: Teaching, learning, and schooling in social context. New York: Cambridge University Press. 196 NYSTRAND ET AL.

Thayer, V. (1928). The passing of the recitation. Boston: D. C. Heath. Tuma, N. B., & Hannan, M. T. (1984). Social dynamics: Models and methods. Orlando, FL: Academic. Trussell, J., & Hammerslough, C. (1983). A hazards-model analysis of the covariates of infant and child mortality in Sri Lanka. Demography, 20, 1–26. Van Lier, L. (1997). Observation from an ecological perspective. TESOL Quarterly, 31, 783–787. Wu, L. (1996). Effects of family instability, income, and income instability on the risk of premarital birth. American Sociological Review, 61, 386–406. Wu, L., & Martinson, B. (1993). Family structure and the risk of a premarital birth. American Sociolog- ical Review, 58, 210–232.

APPENDIX

Our analyses employ continuous time even history regression methods, which assume that the underlying time dimension is a continuous quantity. We define key quantities of interest under this assumption. Let T denote the random variable giving the time of the event for a given case and let t denote the range of times during which the case might be observed. Note then that the period corresponding to t ) λ(t) = lim (1) ε¯0 ε

whereεistakentobeasmallpositivequantity.ThenumeratorinEquation1givesthe probability that the event occurs for a given case during the interval between t and t + ε, given that the case has not yet experienced the event of interest, that is, that T, the timeoftheevent,isgreaterthant,thetimeatwhichweevaluatethevalueofλ(t).This probability in Equation 1 is divided by ε, which is taken to be a positive quantity; hence, a hazard rate refers to a probability per unit time. Because time is assumed to be a continuous quantity, it follows that both t and λ(t) can take an infinite number of values. As a result, a necessary aspect of Equation 1 is the limit over ε, which allows λ(t) to be defined over all possible values of t. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 The quantity λ(t) functions as the dependent variable in our analyses, in much the same way as the quantity log(p/1 – p) is the outcome of interest in a logistic regression. As noted earlier, we operationalize ”time” in our analyses by noting where in a sequence of questions a dialogic spell, student question, or discussion emerges. Then, under these assumptions, our main analytic object is to identify the antecedents of dialogic spells, student questions, and discussions, examining, for example, how teacher questions are involved in the emergence of these outcomes. To do so, we introduce covariates that capture the role of teacher questions, where the effect of teacher questions is allowed to increase (or decrease) the rate at which QUESTIONS IN TIME 197

a dialogic spell, student question, or discussion emerges. We employ a standard proportional hazard specification to incorporate the effects of our covariates. More formally, let t =1,…J index questions within an instructional episode and let λ denote the rate at which dialogic spells, discussion spells, or spells of consecutive student questions are generated. Then a proportional hazard specification specifies the rate at which such spells are generated as:

λ(t) = λ 0(t) exp(β 1 x1(t) + β 2 x2 + … + β k(t)xk) (2)

The model in Equation 2 contains a baseline rate, λ 0(t), analogous to the mortality rate for a reference group of individuals or to the intercept in a linear or logistic regression model. (Recall that the intercept in a linear or logistic regression refers to the group of individuals for whom all values of the covariates are zero—for example, White males born between 1940 and 1944 with 12 years of schooling completed. This is true as well for the baseline hazard rate, λ 0(t), except that this quantity is permitted to vary with time—for example, the age-specific mortality rate for White males born between 1940 and 1945 with 12 years of schooling completed.) Because we lack theory on how λ 0(t) might vary across an instructional episode, we use a model developed by Cox (1972), which makes only weak assumptions about the functional form for the quantity λ 0(t), and use Cox’s method of partial likelihood (Cox, 1972; Cox & Oakes, 1984) to estimate these models. Note also that in Equation 2, we have written x1(t) to emphasize that the predictor covariates can be time-varying. The specification in Equation 2 is termed a proportional hazard model because covariates have so-called proportional effects on the hazard rate λ(t). To see this, suppose that x1 is a dummy variable equal to 1 for ninth-grade classes and 0 for eighth-grade classes. Then holding constant all other covariates in the model, λ(t) for ninth-grade classes will be given by

λ (9th grade)(t) = λ 0 exp(β1*1) whereas that for eighth-grade classes will be given by

λ(8th grade)(t) = λ0 exp(β1*0) Note that the ratio of these two terms is given by Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014 λ(9th grade)(t ) = exp(β1 ) λ(8th grade)(t )

hence, under this specification, ninth-grade classes will differ from eighth-grade classes by the quantity exp(β1). In particular, note negative values of β1 imply that exp(β1) < 1, whereas positive values of β1 imply that exp(β1) > 1; similarly, β1 =0 implies that exp(β1) = 1, because the corresponding rates for the two groups are equal. 198 NYSTRAND ET AL.

To see this concretely, suppose that the outcome is the rate for the transition to a dialogic spell and that β1 has a positive effect estimated as 0.2. Then β1 = 0.2 corresponds to exp(0.2) = 1.22, which translated into words states that the rate of dialogic spells is 22% higher in ninth-grade classes than in eighth-grade classes, in which 22% = 100% × [exp(0.2) – 1]. Conversely, suppose that β1 has a negative effect estimated as –0.2. Then β1 = –0.2 corresponds to exp(–0.2) = 0.82, which translated into words states that the rate of dialogic spells is 18% lower in ninth-grade classes than in eighth-grade classes, in which –18% = 100% × [exp(–0.2) – 1]. Downloaded by [University of Wisconsin - Madison] at 10:33 24 July 2014