Effects of a Visible Author in Statistical Texts

2 downloads 0 Views 2MB Size Report
Karen Q. Jones, Alan Klockars, and Richard Paxton for their invaluable ..... 17). Gragson and. Selzer (1990) suggest that writers of expository text fiction-.
Copyright 1995 by the American Psychological Association, Inc. 0022-0663/95/S3.00

Journal of Educational Psychology 1995, Vol. 87, No. 1, 47-65

Effects of a Visible Author in Statistical Texts Susan Bobbitt Nolen University of Washington The effects of a visible author (one who reveals aspects of him- or herself) on women's experience reading statistical texts were examined among 47 female college students who read texts that differed in the extent to which the author revealed attitudes and personality. Data included "think-and-feel-aloud" protocols, measures of concentration, mood, level of perceived challenge, and readers' images of the author. Women reading the visible author text interacted with the author while reading; this relationship appeared to influence the relations among comprehension, motivation, and affective response. For these women, author image and initial self-efficacy for statistics were related to cognitive engagement, feelings of accomplishment, and intrinsic motivation. Implications for text construction and methodology in research on the interaction of cognition and affect during learning tasks are discussed.

passive construction, use of the third rather than the first person, and by not directly discussing their personal history, views, or personality. Although there is evidence that some expert readers consider the author when reading some kinds of texts (Shanahan, 1992; Wineburg, 1991), the invisibility of textbook authors probably makes this less likely. What might happen if the author of a mathematics text were clearly visible to readers? If made visible and personable, might an author be seen as a potential ally—a personal guide through an otherwise difficult domain? Midgley and her colleagues (Midgley, Eccles, & Feldlaufer, 1991) found that teacher supportiveness was positively related to the extent to which students see math as valuable, useful, and interesting. Might authors, too, play a role as supportive instructors? If the author is visible and supportive rather than remote and uncaring, might readers be more motivated to learn from the text, expend more effort, have more confidence in their ability to understand? My primary goal in this study was to explore the effects of author visibility and the nature of the image of the author that the readers constructed as they read, on the relationship among reader, author, and text. Specifically, I was interested in how that relationship influences readers' cognitive and affective experiences of the text. Expecting that these effects might depend in part on individual differences among readers, I examined the relationship among prior experience with statistics, entering self-efficacy and interest in statistics, and the readers' cognitive and affective experiences. Recent interest in the relationships between motivation and learning, generally, and between affect and cognition has spawned numerous studies of students' learning strategies and motivation (e.g., Ames & Archer, 1988; Graham & Golan, 1991; Nolen, 1988; Nolen & Haladyna 1990a, 1990b; Pintrich & De Groot, 1990). However, few researchers have gone beyond surveys of students' goals, selfefficacy, and study strategies. Although models that are based on these data have posited reciprocal effects of learn-

As students advance through the educational system, they spend thousands of hours studying various kinds of texts. Much research has gone into writing cognitively considerate textbooks, or texts that facilitate understanding, learning, and remembering (Armbruster, 1984; Armbruster & Anderson, 1985; Beck, McKeown, & Gromoll, 1989; Garner, Brown, Sanders, & Menke, 1992). The term considerate text suggests that the most important relationship in studying is between the textbook and the reader, but the term tells us nothing about the relationship between the reader of the text and the author. Even in considerate textbooks authors are likely to keep their distance from readers by taking a stance that Crismore (1984) calls the "anonymous authoritative 'author'" (p. 279). In some domains, years of reading texts by "anonymous authoritative authors" might contribute to students' beliefs about the nature of the subject matter. In mathematics, for example, Schoenfeld (1988) reports that most high school and college students believe that "mathematics is studied passively, with students accepting what is passed down 'from above'" (p. 151). Crismore's authoritative author is common in math textbooks, which is consistent with and perhaps furthers this view of mathematics as a corpus of established and immutable knowledge. In this study I investigated an aspect of this stance that I call "visibility." Anonymous authors remain invisible to readers through the use of rhetorical devices, including

This research was supported in part by a Graduate School Research Fund grant and by the College of Education Research Fund, University of Washington. Many of the ideas on which this study is built were developed in collaboration with Samuel S. Wineburg. Thanks also to Nia Johnson-Crowley, Pam Grossman, Karen Q. Jones, Alan Klockars, and Richard Paxton for their invaluable assistance with this project. Correspondence concerning this article should be addressed to Susan Bobbitt Nolen, Department of Educational Psychology, 322 Miller Hall DQ-12, University of Washington, Seattle, Washington 98195. Electronic mail may be sent via Internet to [email protected]. 47

48

SUSAN BOBBITT NOLEN

ing or understanding and motivation and interest, these relationships have not been examined on-line during studying. Nor have features of the texts studied been systematically included in this work. Although recently some researchers have begun to look at affect and cognition in mathematics and science classrooms (Meece, Blumenfeld, & Hoyle, 1988; Lee & Anderson, 1993), no one has studied readers' thoughts and feelings while the readers were studying difficult text. Therefore, my second goal in this study was to develop ways to examine the dynamic and interactive relationship among comprehension, motivation, and affect, as learners study difficult text. I focused on women reading statistics texts for three reasons. First, the underrepresentation of women in fields requiring higher level mathematics has been well documented (cf. Meece & Eccles, 1993). Although the causes of this underrepresentation remain elusive, many theorists have suggested that affective factors, especially interest, value, and confidence play a decisive role in course and career selection. This hypothesis is based on models developed primarily through statistical analysis of large groups (Fennema & Peterson, 1985; Leder & Fennema, 1990). However, as Leder and Fennema (1990) state, until we get inside the learner's "black box," (p. 198) we will not understand how thoughts and feelings affect performance and participation in mathematics. Thus, in this study, I used interview and verbal protocol methods to capture some of these interactions as they happened. Second, I focused on women reading challenging statistics texts to induce protocols rich in thoughts and feelings about comprehension. Many motivational researchers have documented that challenging materials are most likely to engender both feelings of intrinsic motivation and frustration (see Nicholls, 1989, for a review). Learners studying difficult material (like statistics text) are likely to monitor their level of understanding and to make attributions for it; these attributions lead to various emotional responses (Weiner, 1992). In this study, the presence of a visible author could provide someone for the reader to blame for her lack of understanding. Thus, the nature of participants' attributions for their understanding and feelings was explored. Finally, as a branch of mathematics, statistics was likely to produce anxiety in at least some of the participants (Meece, Wigfield, & Eccles, 1990). This anxiety, coupled with an expectation of a standard remote authorial stance, should lead to low initial efficacy expectations. If anxious readers encounter a visible and supportive author, they might become more confident while studying. Anxious students reading a text with a standard anonymous authoritative author on the other hand, would be more likely to maintain their low efficacy judgments. Thus, the main questions guiding this research were: (1) How does the visibility of the author of a statistics text influence readers' cognitive and affective experiences while reading? and (2) Can the relationship between cognition and affect while reading text be assessed on-line?

The Texts For this study, then, I chose two challenging instructional texts similar in all respects except for the visibility of the author.1 A passage on Pearson's product—moment correlation, taken from a standard introductory statistics textbook, Statistical Methods in Education and Psychology (Glass & Hopkins, 1984), served as the invisible author text. Glass and Hopkins give a fairly conceptual (rather than mathematical) presentation of statistics, and the book is, in general, a considerate text: coherent, cohesive, and written at an appropriate level of difficulty. The visible author text was taken from The Mismeasure of Man (Gould, 1981), a New York Times best-seller recounting the checkered history of intelligence testing and criticizing the reification of IQ scores. Understanding this history requires a general grasp of the concepts of correlation and factor analysis, and so Gould takes time out from his thesis to teach these concepts to his readers. The two texts present the same topics in a similar order and are of roughly the same length and level of difficulty. Gould (1981) uses a number of rhetorical techniques to become visible to his readers, whereas Glass and Hopkins (1984) use techniques that maintain their anonymity. These strategies include self-disclosures, choice of agent, means of addressing the reader, and use of what Crismore (1984) calls "attitudinal metadiscourse" (p. 281). Crismore defines metadiscourse as "discoursing about the discourse; it is the author's intrusion into the discourse . . . to direct... readers . . . so that they will know how to 'take' the author" (p. 280). Attitudinal metadiscourse is used to "implicitly signal . .. attitude toward the content structure . . . and toward readers." (p. 281) An expert judge (with a doctorate in English Language and Literature), who was unaware both of the hypotheses of this study and of the authors of the two texts, coded the texts for the authors' use of these rhetorical strategies. Table 1 shows the frequencies with which these strategies appear in the two texts. How do these strategies increase an author's visibility? The most straightforward way to introduce oneself into a text is to disclose something about oneself to the reader. In this passage, Gould (1981) tells the reader about his childhood personality ("I was a bratty little kid like that"; p. 243), his academic background ("I know something of the biology of the situation"; p. 242), and his passion for baseball (the fact that he could "tell you the year Rogers Hornsby hit .424"; p. 242-243). Gould's opening paragraph ("The spirit of Plato dies hard. We have been unable to escape the philosophical tradition"; p. 239) tells readers he is familiar with Western philosophy and sees it as relevant to statistics. Glass and Hopkins (1984) tell the reader nothing about themselves. Another way to become more visible to readers is to appear as a major actor. In clauses with a specified action, Gould appears as agent 21 times, sometimes singly ("I") and 1

Special thanks to Jill Fitzgerald for challenging me to more clearly articulate and defend this distinction and for the care and thoughtfulness that she showed throughout her reviews.

49

VISIBLE AUTHOR Table 1

Frequency of Rhetorical Strategies That Enhance or Reduce the Author's Visibility Frequency

Strategy Self-disclosure Agency in clauses with an action I We You (implied in imperatives) One Things (e.g., variables correlate) Other persons Agentless clauses Means of addressing the reader Imperatives (e.g., Consider . . .) You Inclusives {We or Our) One ( . . . one can presume . . . ) Rhetorical questions Attitudinal metadiscourse Emphatics Saliency Evaluative Hedges

3

Glass & Hopkins (1984) (1,043 words) 0

13 8

0 0

5 1

0 2

39 2 11

27 5 25

5 1 11

0 0 0

1 0

2 5

20 2 15 24

11 4

Gould (1981) (1,099 words)

6

29

sometimes in collaboration with the reader ("We"). In contrast, Glass and Hopkins (1984) write exclusively in the third person. Each text contains clauses in which an action is specified but in which the agent is not (e.g., "Failure to recognize that correlation may not mean causation is"; Glass & Hopkins, 1984, p. 106), but there are more than twice as many of these agentless clauses in Glass and Hopkins. By taking on the role of agent, Gould ensures that he remains prominent in the reader's eye. In common with most textbook writers, Glass and Hopkins use the third person to keep themselves out of the picture. Gould (1981) further draws attention to himself by addressing the reader directly. He tells the reader what to do ("consider the relationship between my age and the price of gas"; p. 242) or how to feel ("never trust a gut feeling"; p. 239). In addition to using the implied "you" in imperatives, Gould uses "you" as an object, thus highlighting the fact that he and the reader are interacting directly. Glass and Hopkins (1984) address the reader only obliquely, using the third person impersonal (e.g., "scatterplot enables one to surmise"; p. 81) and rhetorical questions ("Would the percentage of high school graduates entering college rise if the pay of teachers were increased?"; p. 105), which are answered by unspecified persons ("Certainly affirmative answers to these questions are not justified"; p. 105). Authors signal their attitude toward the content through the use of attitudinal metadiscourse. Crismore (1984) de-

scribes four types of attitudinal metadiscourse: (a) emphatics, which indicates the degree of certainty of an assertion; (b) saliency, which indicates the importance of an idea; (c) evaluative, which indicates the author's attitude toward a fact or idea; and (d) hedges, which indicates the degree of uncertainty. Table 1 shows how the authors differ in their use of metadiscourse. Gould (1981) uses it frequently; in his opening paragraphs, he employs emphatics ^correlation has been particularly subject to"; p. 239), an evaluative stance ("the fascination of statistics . . . never trust a gut feeling"; p. 239), and hedges ("indeed it does, sometimes—but only sometimes."; p. 239). Although Glass and Hopkins (1984) use informational metadiscourse ("In this chapter the meaning, use and computation of measures of relationship are studied"; p. 79), their use of attitudinal metadiscourse is much more circumscribed. Like Gould, their most frequent form of metadiscourse is the hedge, another common tool of academic writers. Although indicators of saliency were used sparingly in both texts, Glass and Hopkins used emphatics and evaluative stance much less frequently than did Gould. Authors may also reveal themselves through the assumptions projected about their readers. As Ong (1975) states "all [writers] fictionalize their audiences" (p. 17). Gragson and Selzer (1990) suggest that writers of expository text fictionalize their audiences much as writers of fiction do. Although both excerpts that I used in this study are clearly instructional, the rhetorical choices made by the authors cast their readers into different roles. Unlike Glass and Hopkins (1984), who maintain the traditional lecturer-student relationship with their dispassionate and straightforward presentation of information, Gould (1981) casts his readers as fellow-critics of psychometrics. Gould's use of attitudinal metadiscourse, especially the evaluative, highlights the role of critical thinker, whereas the scarcity of such strategies in the Glass and Hopkins passage project an image of the reader as receiver of knowledge. Another question, then, arose from these rhetorical differences: Would these fictionalized roles affect the affective or cognitive responses of the readers in the present study?

Think-and-Feel-Aloud Task To capture the ongoing interaction between readers' cognitive and affective responses to these passages on-line, my colleagues and I have created the think-and-feel-aloud task (Nolen, Johnson-Crowley, & Wineburg, 1994). This task extends the immediate retrospective think-aloud task to noncognitive aspects of verbal behaviors, as called for by Ericsson and Simon (1981). The technique is unique in that it prompts individuals to describe both thoughts and emotions while reading and allows them to describe connections between thoughts and feelings. Participants read a short but cohesive segment of an intact text and are then prompted to tell the interviewer what they are thinking and feeling. The interviewer prompts the reader to refer to specific aspects of the text, including particular words and phrases, that triggered various thoughts and

50

SUSAN BOBBITT NOLEN

emotions. Because participants are not limited to describing their understanding of the content, other cognitions, including thoughts about themselves, the author, and the setting, are free to surface. By describing these thoughts and related emotions, the reader provides a fuller picture of her or his internal experience while studying. Researchers have found that asking for a dump of activated information following immediate retrospection is an effective way of capturing information to which the participant attended during a task (Ericsson & Simon, 1981). Method

Overview of Procedure The participants (all female) completed initial surveys assessing their background in statistics, age, interest in statistics, and selfefficacy for statistics. The women were randomly assigned to read one of the two statistical texts. Before reading, as well as after reading each third of the text, women completed experience sampling questionnaires that tapped their level of concentration, perception of skill and challenge in reading the text, intrinsic motivation (wish to be there rather than elsewhere), and mood while reading. Twenty women performed two additional tasks: After reading a section of text, they generated think-and-feel-aloud protocols, and at the end of the passage they imagined aloud what it would be like to share a cab with the author of the text. All women

completed items that asked what the author would be like as a statistics instructor (Image of Author questionnaire). Finally, all women responded in writing to two open-ended questions requiring them to apply what they had learned from the passage.

Participants A female graduate student recruited volunteers from an introductory educational psychology course and from a women's studies course. The recruiter represented the study as an investigation of students' responses to different kinds of texts. Eleven female students from an introductory educational psychology course and nine from a women's studies course volunteered to participate in 1-hr interviews. Twenty-seven women from another section of the same introductory educational psychology course were recruited to complete the paper-and-pencil portions of the study. The women ranged in age from 21 years to 43 years, with a mean of 28.6 years. Participants were randomly assigned to text; in the Gould group 13 women had taken statistics previously, and 11 had not. In the Glass and Hopkins group 11 had previously taken statistices, and 12 had not.

Measures, Tasks, and Materials Means, standard deviations, and reliability coefficients for all measures and tasks are displayed in Table 2. Initial questionnaires, filled out at the time of recruitment, included requests for infor-

Table 2 Means, Standard Deviations, and Reliability Coefficients (Standardized Item Alpha) for Questionnaire Scales and Application Task Glass & Hopkins (1984) Gould (1981) (« = 23) (« = 24) Measure 2

M

SD

M

SD

Initial questionnaire 0.99 2.78 0.63 2.99 .76 Interest in statistics (3) 0.73 3.09 0.95 3.06 .76 Self-efficacy for statistics (2) Experience sampling questionnaire 1.24 5.67 1.25 5.55 .77 Cognitive engagement l b (3) 1.08 5.22 1.64 5.35 .83 Cognitive engagement 2 1.13 1.61 5.66 5.51 Cognitive engagement 3 .72 1.13 1.46 4.45 Accomplishment l b (3) 4.24 .73 1.17 1.51 4.25 Accomplishment 2 4.18 .75 1.13 1.57 5.00 Accomplishment 3 4.69 .80 1.76 2.39 6.28 Wish to be here (pre) (1) 5.63 2.08 2.40 Wish to be here 1° 6.37 6.31 2.13 2.42 Wish to be here 2 6.17 6.21 2.18 2.58 Wish to be here 3 6.80 6.13 2.19 2.85 Wish to be here (app) 5.46 4.98 d 3.53 3.24 Skill minus challenge (pre) -2.30 -2.35 3.37 3.51 Skill minus challenge 1 1.72 0.75 4.62 2.88 Skill minus challenge 2 0.76 0.00 4.03 4.42 Skill minus challenge 3 3.87 0.44 3.88 3.56 Skill minus challenge (app) -2.07 -2.15 Image of the author6 0.74 2.46 0.86 2.86 .82 Contextualized instruction (6) 0.80 2.52 1.08 2.29 .86 Decontextualized instruction (5) 0.75 2.42 0.90 2.78 .89 Caring instructor (6) 2.74 2.22 3.25 .95 2.44 Application task (2) Note. The number of items is in parentheses. Dashes indicate single items. e a 4-point scale. 5-point scale. b 7-point scale. c 10-point scale- d Range from - 9 to 9.

VISIBLE AUTHOR mation about participants' interest and performance in three kinds of courses: math and statistics, art history, and foreign language. Only the responses to questions about statistics were analyzed. Three interest in statistics items asked participants to rate, on a scale ranging from very (1) to not at all (5), how interesting they would find an introductory statistics course and a statistics text and to rate how useful knowing basic statistics would be. Two items examining self-efficacy for statistics asked the women to rate how likely they would be to understand (a) the material in an introductory statistics class and (b) an introductory statistics text. Participants' scores for each of these scales were computed by calculating the mean response for the set of items. Experience sampling questionnaires. Items developed for the Experience Sampling Method (Csikszentmihalyi & Larson, 1984) were adapted for use here. Four times during the interview participants completed questionnaires asking them to report their level of concentration, their perceived level of skill and challenge in reading the text, and their intrinsic motivation. Perceived challenge and skill for reading statistics text and concentration were rated on 10-point scales, ranging from not at all (1) to very much (10). A single item used in a number of studies to measure intrinsic motivation (Csikszentmihalyi & Csikszentmihalyi, 1988) asked participants "Do you wish you were doing something else?" Ratings were reversed so that high ratings indicate a wish to be here. Feelings during reading were measured using 7-point scales anchored in semantic differential fashion by pairs of opposite feelings (e.g., active-passive, detached-involved, confused-clear). For participants completing verbal protocols, a prompt at the end of each questionnaire directed participants to tell the researcher what they had been thinking and feeling while they were reading the text. In addition to the wish to be here, three multi-item variables were derived from the Experience Sampling Questionnaire (ESQ.). The first was the skill minus challenge, which was the difference between woman's perceived skill and her perception of the challenge of the task. Csikszentmihalyi and his colleagues (Csikszentmihalyi & Csikszentmihalyi, 1988; Csikszentmihalyi & Larson, 1984) argue that for tasks perceived as relatively difficult, a balance between skill and challenge (indicated by a score close to zero) is an indicator of the.flowstate. In American students, Carli, Delle Fave, and Massimini (1988) reported that positive experience tended to be associated with slightly higher levels of skill than of challenge (indicated by scores greater than zero). Cognitive engagement consisted of three items: level of concentration, feeling involved (vs. detached), and active (vs. passive). Accomplishment was an ongoing measure of subjective accomplishment in the reading task, with three items: feeling clear (vs. confused), strong (vs. weak), and satisfied (vs. unsatisfied). Image of the Author questionnaire. Participants were asked to imagine what it would be like to take a statistics course from the author of the text. The Image of the Author questionnaire2 asked the participants to rate the likelihood that the author would use various teaching techniques. Items were created to tap three related dimensions: (a) contextualized instruction (6 items), which attempts to help students relate statistical concepts to their own lives and to the broader world (e.g., "have you investigate uses and misuses of statistics in the newspaper," "encourage you to come up with your own ideas," "use small group activities in class"); (b) relatively decontextualized instruction (5 items) ("focus more on computation than meaning," "write lots of formulas on the chalkboard," "leave you behind"); and (c) caring (6 items) ("be willing to adjust syllabus when many students are struggling," "offer encouraging comments to students having difficulties.") Application task. At the end of the session, students read two

51

brief descriptions of correlational studies, including results and causal interpretations. In the conclusions of both studies, the researchers drew spurious causal conclusions. Two open-ended questions after each description asked students to judge the adequacy of the interpretation. Item two also required participants to correctly interpret negative correlations and differences in the magnitude of two negative correlations. One item appeared on each page, leaving ample room to write the response below each item. At the top of the second page (before the second item) were items asking participants to rate their skill and the challenge of answering the questions, and the extent to which they wished to be somewhere else. A coder who was unaware of condition scored the application task responses using a three-part rubric. For the first item, responses received from zero points (incorrect or no relevant response) to three points (two points if the participant included a statement that one cannot infer cause from correlation and one point for a reference to a plausible alternative explanation). The first part of Item 2 was scored with the same rubric, with the additional criterion that to receive any points, the negative correlation coefficient had to be correctly understood as showing an inverse relationship. For the second part of Item 2 participants received one point for showing understanding of the relative size of the coefficient and one point if that understanding was correctly applied. A second coder scored a randomly selected 40% of the sample. Interrater agreement was high, with 95% of the total scores falling within one raw score point (58% exact matches); the rank-order correlation was .95. Texts. Participants read one of two passages on correlation: Gould (1981) or Glass and Hopkins (1984). Both texts are shown in Appendix A. The Glass and Hopkins selection was edited slightly for length and to ensure comparability of content. Both passages contained the following topics in similar order: introduction and definition of correlation, errors in interpretation of correlation as causal, definition of types of correlation (positive, negative, zero), explanation of Pearson's product-moment correlation, scatterplots of example correlations, a lengthy example illustrating multiple possible causal relationships for the same correlation, and a concluding statement that one cannot infer causation from correlation. Length was comparable (Gould had 1,099 words, and Glass and Hopkins used 1,043 words), and both texts were rated as college-level with the Flesch-Kincaid formula, with the Gould text somewhat more difficult (Grades 16 and 14, respectively.) Each text was divided among three pages. For subjects completing think-and-feel-aloud protocols (N = 20), each page was divided into two segments. At the end of the first segment on each page (Segments 1, 3, and 5) was a written prompt asking participants to tell the researcher what they were thinking and feeling. All participants saw a prompt at the end of each page directing them to fill out one of the ESQs.

Procedure Volunteers filled out the initial survey and consent forms in the classes from which they were recruited. All volunteers in the women's studies course and one educational psychology section were contacted by phone to arrange individual interviews; all but one participated. Women in the other educational psychology course were contacted to arrange small-group sessions. Interviews. The same graduate student who recruited volun2 The Image of the Author Questionnaire and the "imagine yourself sharing a taxi" prompt were developed in collaboration with Samuel S. Wineburg.

52

SUSAN BOBBITT NOLEN

teers conducted all interviews, which were tape-recorded and later transcribed. The interview schedule was identical for both text conditions. The interviewer explained the procedure as follows: I will be giving you some text to read. The text is divided into sections. After reading a section of text, you will see a prompt asking you to tell me what you are feeling and thinking. At various times you will fill out a brief questionnaire about your feelings while reading. You are free to go back over the previous sections in the booklets if you want. You are free to mark on the text pages in any way. Here is a pencil for you to use. After you have filled out the last questionnaire I will collect the text pages and ask you some questions related to your image of the author of the text. When you are through answering the questions the last thing I will ask you to do is complete some comprehension problems in which you apply what you have learned from the passage to a real life situation. After answering any procedural questions she introduced the topic by saying, "The text I would like you to read is about a topic in statistics." The interviewer then explained the ESQ form, and the participant filled it out. The interviewer then collected the ESQ and handed the first text page to the participant. When she reached the prompt at the end of the first, third, and fifth sections of text, the participant generated a think-and-feel-aloud protocol, describing her thoughts and feelings while reading. The interviewer probed, if necessary, by asking for clarification or elaboration, or by prompting the participant to describe her feelings (or thoughts). At the end of the first thinkand-feel-aloud, the participant was reminded that she could review any portion of the text already read, as needed. Following the second, fourth, and sixth sections, the participant first completed an ESQ and then described her thoughts and feelings. Following the last think-and-feel-aloud, the interviewer collected all text pages and asked the participant to imagine the following scenario: "Suppose you're downtown and it's raining. You and the author of the text hail the same cab. You offer to share the cab. What would the author be like—look like?" After the participant had described the author, the interviewer administered the Image of the Author questionnaire, with the following directions: "Now suppose you took an introductory statistics class from the author. Here's a list of things the author might do. How likely would it be for the author to do each of these?" Finally, the participant completed the application task. The participant was then thanked, allowed to ask questions about the study, and paid $10. Small-group administration of paper-and-pencil tasks. The remaining 27 women completed the paper-and-pencil measures in groups of 4 to 7, mixed with respect to texts, conducted by either the graduate student or myself. After explaining the ESQ forms and answering any questions, the participants were told that the text they were reading was on a topic in statistics. The participants completed the first packet (initial questionnaire, four ESQs, text, and image items), then exchanged it for the second packet (application task). When the participants were finished, they turned in the second packet and were thanked and paid $5.

naire responses, I began analyses of all questionnaire data with Procedure X Text analyses of variance (ANOVAs). There were no significant main effects or interactions involving procedure (all ps > .05), so the data were collapsed. All analyses of questionnaire and application task data are based on 47 cases. To assess the effects of text and prior coursework in statistics on the affective variables across time, I performed separate 2 (text) X 2 (statistics experience) X 3 (time) repeated measures analyses of variance, with the unique sums of squares method, on ratings for engagement, accomplishment, wish to be here, and skill minus challenge. Occasionally a participant skipped an item; data from participants who did not respond at all three times were dropped on a by-analysis basis. Means are shown in Table 2. There was a main effect for time on accomplishment, F(2, 88) = 6.14, p < .01, MSE = .83; Tukey's honestly significant difference (HSD) showed Time 3 ratings to be higher than Time 1 (q = 3.85, p < .05) and Time 2 (q = 4.85, p < .01), which were not different from each other. There was a main effect for time on skill minus challenge, F(2, 86) = 4.63, p < .05, MSE = 7.94, but also a Text X Time interaction, F(2, 86) = 3.25, p < .05. Three 2 (text) X 2 (time) repeated measures ANOVAs were conducted, and each interaction was tested separately, with a joint Bonferroni test that divides the alpha level among the three interactions. Only the interaction of Text X Time (1 vs. 3) was significant, F(l, 45) = 6.71, p = .013, suggesting that the time effect was due primarily to the Glass and Hopkins group rating their skills nearly four points higher than the challenge of the text at Time 3 (M = 3.87, SD = 4.03 ). The other means were closer to 0 (ranging from 0 to 1.22); 0 indicates a balance between perceived skills and challenge. No other main effects or interactions were significant (all ps > .05). Application task scores ranged from 0 to 8, with a mean of 3.12 and a standard deviation of 2.29. A 2 (text) by 2 (statistics experience) ANOVA was computed, and it yielded a significant main effect for Experience, F(l, 43) = 4.69, p < .05, MSE = 4.84. Those who had taken at least one statistics course scored higher (M = 3.83, SD = 2.46) than did those who had not (M = 2.39, SD = 1.88). There were no main effects or interactions involving text (both Fs < 1). In summary, readers of both texts felt a somewhat greater sense of accomplishment at the end of the task than in the midst. The means of the ESQ variables and the application task scores did not vary by text, except that readers of the Glass and Hopkins (1984) text perceived their skills to be significantly greater than the challenge posed by the final section.

Results Relationships Between Individual Difference Variables and Cognitive and Affective Experience To determine whether differences in procedure (interviews vs. paper-and-pencil groups) influenced the question-

Image of the Author and Experience of the Text To examine the relationship between readers' image of the author and their experience of the text, I computed correlations for each text condition (see Table 3). The pattern of correlations for the two groups differed clearly.

53

VISIBLE AUTHOR

g i n en eN

I-H 00

00 00

en en

en o

tM T-H en

eN

I r— o\ O\ NNH

t

rr vO i

O

CN in en 1-c

f cn O t--

•*

i n i n •-H

OV 00 •* H

m en

~.

*

28

m I-H

a

5

ent

.5

_