Journal of Educational Psychology1993, Vol. 85, No. 3, 424-436 Copyright 1993 by the American Psychological Association, Inc

0022-0663/93/53.00

Situational Interest: Its Multifaceted Structurein the Secondary School Mathematics Classroom

Mathew Mitchell

The present study attempted to tackle a hurdle that continues to plague the research on interest: thelack of an adequate theoretical model. In particular, the tenability of a hypothetical construct ofinterest as it applies to the secondary mathematics classroom was proposed and empiricallyassessed. Building on previous theoretical work, the study used qualitative and quantitative meth-odologies to first develop a model and then assess its construct validity. The results indicate thatit is useful to distinguish between personal and situational interest. Furthermore, the results indicatethat the structure of situational interest is multifaceted, clarifying 5 subfacets of situational interestin the high school mathematics classroom.

Classroom boredom has been well documented in the re-search literature for several decades (e.g., Buxton, 1973;Csikszentmihalyi & Larson, 1984; Jackson, 1968), espe-cially in the high school mathematics classroom (Kline,1977; Papert, 1980; Whitehead, 1912/1959). As Sarason(1983) observed, "schools are not very interesting places formost of the people in them. That is indisputably the case forhigh and middle or junior high schools" (p. 64). Classroomboredom, though, may really be an indication of a biggerschooling problem, namely the lack of motivation to learn.Because disinterest in learning is one primary manifestationof this, one way to attack classroom boredom is from theperspective of an intrinsic motivational variable called in-terestingness (Hidi, 1990; Schank, 1979). The present studyattempts to tackle a hurdle that continues to plague the re-search on interest: the lack of an adequate theoretical modelof interest. In this study I propose and empirically assess thetenability of a hypothetical construct of interest. In addition,the study focused on mathematics in the secondary mathe-matics classroom because it seems to be the source of bothmuch boredom and low achievement at the secondary level(Dossey, Mullis, Lindquist, & Chambers, 1988; Putnam,Lampert, & Peterson, 1990). Indeed, as noted in the recentreport Everybody Counts: A Report to the Nation on the Fu-ture of Mathematics Education (National Research Council,1989), "Mathematics is the worst curricular villain in drivingstudents to failure in school. When mathematics acts as afilter, it not only filters students out of careers, but frequentlyout of school itself (p. 7). Perhaps even more to the point,Davis (1992) observed that

With the best of intentions, we have created a curriculum ofmathematics that has been severed from the real world. Itconsists of meaningless bits and pieces, and we ask students to

I thank James H. Block, Suzanne Hidi, Hsiu-Zu Ho, Richard J.Shavelson, and three anonymous reviewers for their critical feed-back and comments on drafts of this article.

Correspondence concerning this article should be addressed toMathew Mitchell, Curriculum & Instruction Program, School ofEducation, University of San Francisco, 2130 Fulton Street, SanFrancisco, California 94117. Electronic mail may be sent tomitchellm@alm.admin.usfca.edu.

learn it as a large collection of meaningless bits and pieces,(p. 730)

One could hardly expect such a curriculum to be consideredinteresting by most students. Thus, taking steps toward un-derstanding how to enhance interest in the mathematicsclassroom may prove to be one of the most direct ways toapproach the problem of effective mathematics instruction.

Interest is a hypothetical construct the usefulness of whichmust be demonstrated by investigations into its constructvalidity (Shavelson, Hubner, & Stanton, 1976). The first stepin construct validity is the identification of theoretically con-sistent and distinguishable facets of interest (within-networkstudies). The next step is the study of the ways in which theseinterest facets are related to other constructs (between-networks studies). Within-network studies focus on the de-velopment of theoretical models of the construct and of mea-surement instruments that are consistent with these models.Because the research on interest is at a relatively early stage(Hidi, 1990), this particular study focuses only on the within-network aspect of construct validation.

Definition of Interest

Dewey (1913) provided what seems to be the best workingdefinition of interest in the following, "Genuine interest is theaccompaniment of the identification, through action, of theself with some object or idea, because of the necessity of thatobject or idea for the maintenance of a self-initiated activity"(p. 14). Note that Dewey's definition distinguishes betweenactivity in general and maintenance of self-initiated activity,which is an essential component of interested behavior. Fur-thermore, Dewey proposed that there are two key factors inthe interest construct: identification and absorption. Withregard to identification, Dewey stated,

The genuine principle of interest is the principle of the rec-ognized identity of the fact to be learned or the action pro-posed with the growing self; that it lies in the direction of theagent's own growth, and is, therefore, imperiously demandedif the agent is to be himself, (p. 7)

Furthermore Dewey (1913) proposed that "interest is notsome one thing: it is a name for the fact that a course ofaction, an occupation, or pursuit absorbs the powers of an

424

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 425

individual in a thorough-going way" (p. 65). Dewey's con-ception of interest seems particularly useful as the relation-ship between identification, absorption, and the maintenanceof a self-initiated activity offers a straightforward way toanalyze classroom activities.

Schiefele (1991) recently proposed some additional fea-tures of interest. Schiefele's proposals are important in thatthey direct further research on interest toward specific con-tent domains, rather than treating interest as a relatively gen-eral construct. In particular, Schiefele (1991) proposed thefollowing:

Interest is a content-specific concept. It is always related tospecific topics, tasks, or activities.

When understood as a content-specific concept, interest fitswell with modern cognitive theories of knowledge acquisition,in that new information is always acquired in particulardomains.

Subject-matter-specific interest is probably more amenable toinstructional influence than are general motives or motiva-tional orientations, (p. 301)

The term interest as used in this article refers to an interestdirectly tied to the content of instruction. Too often educatorstry to make their classes more interesting, but their "appealis simply made to the child's love of something else" (Dewey,1913, p. 12). For instance, the research on seductive details(Garner, Alexander, Gillingham, Kulikowich, & Brown,1991; Garner, Gillingham, & White, 1989) indicates that textmaterial can be made more interesting, but if one is not care-ful, the enhanced interest will have little to do with the par-ticular content being taught. As an example, Garner et al.(1991) found that text written about Stephen Hawking, whichincluded details about his personal life, was rated as moreinteresting than text that excluded this material. However,when subjects read the "personal life" text, they recalledfewer of the important ideas contained in the text than thosewho had read the regular text.

Development of the Model of Interest

To develop a model of interest, three stages, described indetail in the following sections, were undertaken. In thisstudy, I adopted the social ecological research orientation ofMoos (1976, 1979), who emphasized the importance of so-cial perceptions as the key to manipulating an environment.Thus, model development proceeded under the assumptionthat student perceptions of interest were the key factor to beassessed. The first stage involved the development of a pre-liminary model according to the current research literature oninterest. The second stage involved elaborating on the initialmodel by using naturalistic, or qualitative, techniques to bet-ter understand student perceptions of interest in their math-ematics classrooms. Finally, in the third stage, I developeda survey instrument and collected data that were then quan-titatively analyzed, to assess the tenability of the model de-veloped through the first two stages. The anticipated resultof these three stages of development was that psychologicaltheory and qualitative results could be combined to developand test the viability of an extended conceptual framework

of interest in the mathematics classroom through a surveyinstrument.

Stage 1: Theoretical Underpinnings of the Model

This article builds on previous research in constructing amodel of interest and empirically testing that model in thesecondary mathematics classroom. The primary distinctionabout the structure of interest was set forth by Hidi and Baird(1988) and by Krapp (1989) in their distinction between per-sonal and situational interest. A personal interest refers to aninterest that people bring to some environment or context.For example, typically some students will come to a math-ematics classroom already interested (or disinterested) in thesubject. On the other hand, situational interest refers to aninterest that people acquire by participating in an environ-ment or context. Thus, the personal interest approach em-phasizes working with individual differences, whereas thesituational interest approach emphasizes the importance ofcreating an appropriate environmental setting.

From an educational point of view, situational interest isthe real topic of concern as teachers have no influence overstudents' incoming personal interests. It should be noted thatalthough Krapp proposed this fundamental distinction be-tween personal interest and situational interest, there is atpresent no research that directly speaks to the tenability ofthis proposal. Whereas personal interest and situational in-terest can be thought of as distinct, they also have been hy-pothesized to be related. In particular, from an educationalperspective, one would hope that if a classroom is high insituational interest, that environment would change an in-dividual's personal interest level regarding the subject overtime. In other words, whereas a teacher may have no controlover students' incoming personal interests, that same teachermay be capable of having a noticeable influence on the stu-dents' outgoing personal interests by the end of the schoolyear.

Concentrating on situational interest, I next proposed thatsituational interest itself is multifaceted. Figure 1 presents thehypothesized multifaceted model of situational interest pro-posed in this study.

The second level of the model distinguishes betweencatching interest and holding interest. As Hidi and Baird(1986) noted, "interest has a durational aspect—there aretriggering conditions and there are conditions which ensurethe continuation of interest" (p. 191). Indeed, Malone andLepper (1987) pointed out that it is relatively simple to designcomputer programs with technical gimmicks that will at-tract students' interest, but many of those same programs

PI SI

CATCH

, -V

HOLD

Figure 1. Theoretical construct of interest. (PI = Personal inter-est; SI = Situational interest.)

426 MATHEW MITCHELL

fail to maintain the user's interest over time. It is hypothe-sized that a hold situational interest facet will be morestrongly related to overall situational interest than anycatch facet as a hold facet is a condition that maintains in-terest over a period of time. Alternatively, as John Deweymight have put it, hold facets will be more closely alignedto "genuine interest" because they will tend to maintainself-initiated activity. In particular it is posited that condi-tions that hold interest over the course of a high school se-mester will have a greater impact on students than condi-tions that simply catch or trigger interest. However,whereas Hidi and Baird have proposed a reasonable dis-tinction between catching and holding, there is no researchthat directly addresses the tenability of their proposal.

In addition, a review of the relevant literature has led meto propose that the essence of catching lies in finding variousways to stimulate students, whereas the essence of holdinglies in finding variables that empower students. For instance,a stimulant is commonly defined as a variable that tempo-rarily increases the activity of an organism. The work ofBerlyne (1960, 1966) and Malone and Lepper (1987) hashelped to guide a theoretical rationale for determining thekinds of interest facets that should be particularly effectivefor catching interest. In particular, Malone and Lepper pro-posed that there exist at least two kinds of stimulation: cog-nitive and sensory. They defined sensory stimulation as "theattention-attracting value of variations and changes in thelight, sound, or other sensory stimuli of an environment"(Malone & Lepper, 1987, p. 235). Similarly, Malone andLepper (1987) defined cognitive stimulation as occurring be-cause "people have a cognitive drive to bring 'good form' totheir cognitive structures and that instructional environmentscan stimulate curiosity by making people believe that theirexisting knowledge structures are not well formed" (p. 236).Thinking of stimulation, in one form or another, as the un-derlying variable in catching interest fits well with the ideathat some variables may be particularly good catalysts forself-initiated behavior.

Holding interest, on the other hand, can be viewed as ef-fective when a variable is used that empowers the student.The term empower is used in its traditional sense of bestow-ing power for an end or purpose. For instance, making thecontent of learning meaningful for students tends to empowerthem because such a variable gives students more power toachieve their personal ends. Such a variable tends to be long-lasting because an individual's personal goals were estab-lished before the content was taught and will likely continue

after the teaching of the meaningful content. Indeed, Papert(1980) proposed that an "appropriable" mathematics mustincorporate the "power principle," that is, "it must empowerthe learner to perform personally meaningful projects thatcould not be done without it" (p. 54). I propose that variablesthat 'empower' students tend to hold interest because evenwhen the source of empowerment is removed, the studentwill likely still find the subject empowering to them. Em-powerment as the underlying variable behind holding interestfits well with Dewey's two primary characteristics of inter-est: identification and absorption. First, if one is identifiedwith the content of an activity, it will be experienced as per-sonally meaningful and thus empowering. Second, if one isabsorbed in the process of an activity, it will be experiencedas moving one toward achieving a personal end—and thusbe empowering.

For example, a wonderful color illustration in a book maycatch someone's interest, but when the page is turned theinterest in the illustration is likely to disappear also. On theother hand, if reading per se is perceived as empowering toa student, then his or her interest in reading is likely to remain(i.e., hold) once a classroom reading session is finished.

Stage 2: Naturalistic Development of the Model

The next stage in model development derives from theresults of focus groups and open-ended questionnaires usedwith high school mathematics students (Mitchell, 1992). Thestudents were all from first-year algebra or geometry class-rooms, with an age range between 14 and 16 years. A focusgroup consists of a small number of participants, usually 5-9people (Krueger, 1988; Morgan, 1988). Focus groups wereconducted with a small number of randomly selected vol-unteer students from five different classes in December,1991. Next, open-ended questionnaires were used to gatherinformation from all the classes used in the study in earlyJanuary 1992. In the open-ended questionnaire, studentslisted aspects of their class that were interesting (or boring)and then stated why each aspect was interesting (or boring)to them. The qualitative data were analyzed according toguidelines proposed by Krueger (1988).

The expanded conceptual model built from these twosources of qualitative data is shown in Figure 2. In particular,the focus groups and open-ended questionnaires identifiedfive distinct facets that contribute to students' perceptions ofclassroom interest. Using Dewey's (1913) definition of "gen-uine" interest, I decided that facets that only sparked interest

--GroupWork

CATCH

-»

Computers Puzzles Meaning-fulness

Involve-ment

Figure 2. Hypothesized construct of situational interest (SI) in the mathematics classroom.

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 427

but did not seem to maintain student activity would be clas-sified as catch facets. On the other hand, facets that appearedto be successful at maintaining student activity would beclassified as hold facets. The results of the qualitative anal-ysis indicated that three of those facets (computers, groupwork, and puzzles) seemed to function mainly as stimulationvariables and thus are categorized as catch facets. The twoother facets (meaningfulness and involvement) seemed tofunction mainly as empowerment variables and thus are cat-egorized as hold facets. The rationale for including each facetin the model is described below.

Overall, the primary reason students gave for mentioningany of the catch facets (puzzles, group work, or computers)was that they provided a change of pace and variety to theusual state of affairs in secondary classrooms. Thus, the catchfacets seemed to function as three different ways of tempo-rarily stimulating student activity. None of the catch facetsin and of themselves, however, appeared to serve as an ef-fective way to maintain student activity.

First, puzzles appeared to be effective primarily becausethey function as a form of cognitive stimulation that catchesinterest. Puzzles is a generic term covering a variety of toolsused for arousing students' curiosity. These puzzles werereferred to as either logic puzzles, mind-teasers, or starters.In general, these puzzles were not perceived as being "reg-ular" mathematics but rather as being "weird" or "unusual"problems that demanded that students think more clearly orcreatively. A number of students noted how these puzzlesmade them really use their brains—something these studentsfelt rarely happened when studying regular mathematics. Thegeneral idea of puzzles as a tool for sparking curiosity hasa rather lengthy history in the research literature, startingwith Berlyne (1960,1966) and continuing to the present withsuch researchers as Lepper (1988), Malone (1981), and Vossand Keller (1983).

Second, group work by itself appeared to be effective pri-marily because it seems to act as a form of social stimulationthat catches interest. Typically adolescents are very social,and group work provides a form of stimulation by encour-aging students to talk to one another about subject-matterconcerns and ideas. For instance, the data indicated that useof group work in mathematics classes added to the interest-ingness of the classroom environment because it allowed thestudents to discuss problems with one another. The processof communication seems to be crucial. For instance, the newMathematics Frameworkfor California Public Schools (Cal-ifornia Mathematics Framework Committee, 1992) consis-tently emphasizes the importance of mathematical commu-nication; group work appears to provide a direct way toaccomplish this goal. Second, students noted that they weremuch more willing to ask questions of other students than torisk looking dumb when asking questions of the teacher infront of the entire class. Students also related that their fellowstudents were often able to explain concepts better than theteacher because these students spoke in a "language" theycould readily understand. The benefits of small-group workhave been well documented in the research literature (Good,Mulryan, & McCaslin, 1992; Johnson & Johnson, 1975;Noddings, 1985; Slavin, 1980), and the students' comments

reaffirmed many of the conclusions of this research base.Third, computers appeared to be effective primarily be-

cause they seem to act as a form of cognitive stimulation thatcatches interest. Lepper and Malone (1987) provided an ex-cellent overview of the ways in which computers effectivelycatch the interest of students. The qualitative data indicatedthat computers seemed to be perceived as interesting pri-marily because students could explore and test conjectures,for example, by using the program Geometric Supposer(Yerushalmy & Chazan, 1990). It was the exploratory use ofcomputers that seemed to be of interest to students becausethey could more readily learn things on their own. However,it should be noted that not all classes that used computers didso in an exploratory fashion. Some classes simply had stu-dents go through prescribed steps in an activity sheet. None-theless, even in these cases computer use stood out becauseit offered a different way of learning the material and createdsome variety in the classroom.

Should a practitioner reading about the above catch facetsconclude that puzzles, group work, and computers offer threesurefire ways to get students' interest? The answer appearsto be yes and no. The catch facets do seem to offer quick fixesto gaining student interest in the short run. However, it needsto be emphasized that it is doubtful that these three facets,in and of themselves, will work in the long run. In particular,what seemed to separate the "interesting" classrooms fromthe others was the ways in which the teachers used the threecatch facets. For example, Good et al. (1992) noted thatgroup work used in mathematics classrooms tends to resultin more interesting mathematics because the structure ofgroup work pushed teachers to design lessons directed to-ward developing higher order thinking skills rather than com-putational skills. Although the nature of the data from thisstudy could not verify it, Good et al. observed that the bestteachers tended to use the catch facets as vehicles for im-plementing their versions of meaningful or engaging lessonplans. Therefore we now need to take a closer look at the twohold facets.

The first hold facet, meaningfulness, appeared effectiveprimarily because content that is perceived as being person-ally meaningful to students is a direct way to empower stu-dents and thus hold their interest. Meaningfulness refers tostudents' perception of the topics under study in their math-ematics class as meaningful to them in their present lives. Inpoint of fact, the data indicate that many students made state-ments about the meaninglessness of much of their work. Itis worthwhile to note that for elementary-level studentsmeaningfulness seems to be synonymous with their ability tounderstand or do mathematics. However, for secondary stu-dents this capacity did not make a subject or topic meaningfulfor them.

Typically students did not see the content of mathematicsas important or related to their daily lives irrespective of howwell they were doing in their mathematics class. On the otherhand, students could point to other subjects that dealt withtopics that concerned them (e.g., environmental issues in abiology class or human life issues presented through liter-ature in an English class). Although it is unclear from thestudy's data what the complete range of meaningful topics is

428 MATHEW MITCHELL

for adolescents, it is clear that mathematics classrooms rarelytouch on any of the topics within that range.

The role of meaningfulness, particularly in mathematicseducation, is receiving increasing attention from policy re-ports (California Mathematics Framework Committee, 1992;National Research Council, 1989) as well as researchers(Anderson, Shirey, Wilson, & Fielding, 1987; Kaput, 1989;Lepper & Malone, 1987; Mellin-Olsen, 1987; Papert, 1980;Schank, 1979). Although these documents use slightly dif-ferent vocabulary, they all complement the basic point madeby Papert (1980) that a good mathematical curriculum is onethat allows the learner to perform meaningful projects thatmake cultural sense, or fit, for the larger culture within whichthe student lives.

Finally, involvement appeared effective primarily becausewhen the process of learning is experienced as absorbing,then that process is perceived as empowering to students andwill therefore tend to hold their interest. Involvement refersto the degree to which students feel they are active partic-ipants in the learning process. The data indicate that involve-ment has a strong inverse relationship with lecturing. Themore a particular teacher tended to lecture, by default thefewer opportunities students felt they had to actually learn thematerial themselves. Basically students felt involved whenthey got to do activities in order to learn new material, ratherthan sitting and listening. Doing in this context does not referto mechanical work, such as drill and practice. Instead, itrefers to a process in which students are being active par-ticipants in the learning of new material. The notion of in-volvement fits with some of the research implications ofconstructivism (e.g., Carey, 1985; Mellin-Olsen, 1987; Na-tional Council of Teachers of Mathematics, 1991; Schoen-feld, 1987) as well as with the importance Dewey placed onabsorption (Dewey, 1913).

It should be noted that very few teachers seemed to besuccessful at involving their students. Those teachers whowere successful differed in what they specifically did in theclassroom, but their commonality was in having students"doing" things (e.g., working on a project, writing bookchapters, learning new concepts with carefully structuredgroup work) for the great majority of the class time. In otherwords, lecturing was kept to a minimum (usually 10 min orless) while a variety of student activities were used over thecourse of a semester to impart information and develop skills.

The initial, qualitative development of the multifacetedmodel of situational interest seems to mirror many of theimportant features of the latest statewide document attempt-ing to influence the direction of future mathematics instruc-tion. The Mathematics Framework for California PublicSchools (California Framework; California MathematicsFramework Committee, 1992) in particular emphasizes theimportance of both meaningful mathematics and involve-ment. The California Framework tacitly recognizes theseissues in stating that "Students construct their understandingof mathematics by learning to use mathematics to make senseof their own experience. This understanding of mathematicsbecomes more powerful when students use it to achieve pur-poses that are meaningful to them" (California MathematicsFramework Committee, p. 33). The California Framework

also addresses the value of computers, puzzles, and groupwork in the mathematics classroom.

After Stages 1 and 2 of model development, I developeda relatively rich model of the multidimensional nature ofsituational interests in the high school mathematics class-room. The model distinguishes between personal and situ-ational interests, as well as proposing that situational interestitself has five subfacets (groups, puzzles, computers, mean-ingfulness, and involvement). Previous theoretical work andqualitative research went hand-in-hand in the development ofthis model as based on student perceptions of classroom in-terest. However, at this point, I had no way of empiricallyassessing the construct validity of this proposed model. InStage 3, therefore, I developed a survey instrument based onthe model and quantitatively analyzed the tenability of themodel. This stage is described in detail in the next section.

Stage 3: Survey Development and Data Analysis

The third stage represents the core of the study. A survey,termed the interest survey, was developed, pilot tested, andgiven to a sample of high school mathematics students. Theresulting data were then analyzed to test the feasibility of themodel of interest that had been developed through the firsttwo stages. The details of the sample used, instrument de-velopment, and analysis strategy are presented in the Method

section.

Method

Sample and Procedure

The sample consisted of 350 high school students from threedifferent high schools in the Southern California area. The studentsample was predominantly White and was composed of 147 boysand 188 girls (with 15 nonresponses to the gender item). All of thestudents were attending college preparatory classes in either algebraor geometry. Thirteen different classes were sampled, incorporatingseven different teachers. All of the teachers used were "master"teachers in their school district, roughly indicating that these teach-ers were among the top 40% of mathematics teachers in their area.Teachers in this particular school district are given "master" statusif (a) they had been observed and requested as a mentor for a studentteacher by the subject-matter teacher education supervisor at thelocal university and (b) the principal of the school also thinks theteacher is accomplished enough to serve as a mentor for a studentteacher. Two classes from each teacher were surveyed, with theexception that one teacher only taught one of the target subjectcourses.

I administered the interest survey to the students for 20 minwithin a scheduled 50-min mathematics class session in January1992.

Measurement Instrument

The interest survey is a measure of adolescent interest in themathematics classroom designed to measure two general areas ofinterest (personal and situational) and five specific components ofsituational interest: meaningfulness, involvement, computers,groups, and puzzles. Because of the study's emphasis on student

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 429

perceptions, the interest survey followed a self-report surveyformat.

Student statements made in either the focus groups or the open-ended questionnaire were used for the stems. The main criterionused for selecting stems was to find statements that were simple andunambiguous. Second, I sought statements that were both positiveand negative regarding a particular facet, for instance, a positivestem such as "Math class is fun" and a negative stem such as "Mathclass is boring this year." Transcripts were reviewed to find state-ments, either positive or negative, that clearly related to one of thesix situational interest scales: general situational interest, mean-ingfulness, involvement, group work, puzzles, or computers. Thescale for personal interest contained items I had developed and usedin a previous study (Mitchell, 1991).

An initial bank of 85 items was pilot tested. These items rep-resented items from seven scales (personal interest, situational in-terest, meaningfulness, involvement, group work, computers, andpuzzles). Students responded on a 6-point Likert scale ranging fromstrongly disagree (1) to strongly agree (6). The first stage of pilottesting involved having two students do the survey with a talk-aloudprotocol. Ambiguous items (as denoted by students placing a ques-tion mark beside any stem that was unclear to them) and difficult-to-understand items (as inferred from a subsequent listening to thetape of the talk-aloud session by the researcher) were dropped at thispoint. A total of 18 items were eliminated, leaving 67 items for thenext stage of pilot testing. The second stage of pilot testing involvedusing the remaining items for in-class pilot testing using two class-rooms. The internal consistency of the items within each scale wascomputed, and weak items (i.e., generally those items with a Pear-son product-moment correlation coefficient between the item andthe sum of scores on the remaining scale items below .50) weredropped. Twenty-two items were eliminated in this manner, leaving45 items for use on the final survey. The results from the two pilotclassrooms were included in the final analysis as no new items weresubsequently included in the final survey. Each of the scales on thefinal survey was represented by either 5 or 7 items. It should benoted that although there was a balance of positive and negativeitems within each scale initially, the manner in which items wereeliminated led to an imbalance of some scales. For the purposes ofthis study, I decided to allow the empirical data to guide the se-lection of final items rather than try to force a balance of positiveand negative items.

Examples of items used from each of the seven scales are pro-vided below. The Appendix contains the complete set of items usedin the final set of analyses.

I have always enjoyed studying mathe-matics in school.I actually look forward to going to mathclass this year.The stuff we learn in this class will neverbe used in real life.We learn the material ourselves insteadof being preached at.The mind-teasers or logic puzzles we doare fun.We try to discover things on the com-puter in our class.I like the groups in our class becauselearning is more fun when things can bediscussed.

Personal interest:

Situational interest:

Meaningfulness:

Involvement:

Puzzles:

Computers:

Group work:

Analysis Strategy

Responses to negatively worded items were reversed so that forall items a response of 6 represented the highest possible positive

rating of an item. There were a few missing responses (less than 2%)for each interest survey item. The missing responses were replacedby the mean response for that individual to other items on the samescale.

The analysis strategy consisted of seven steps. First, the internalconsistency of all the scales was checked (scales were to be usedonly if their internal consistency was greater than .70). The next twosteps involved factor analyses, first on the general personal andsituational interest items, and then on the items representing eachof the five situational interest subfacet scales. The factor analysesinvolved the maximum likelihood extraction technique, with anoblique rotation (with delta set to 0). If there were any items thatdid not clearly load onto either factor, they were dropped from thescale, and the resultant new scale was reanalyzed for its reliability.

Once the first three steps were completed, the fourth step in-volved making sure that each final scale had an even number ofitems as item pairing was to be used for the subsequent factor andLISREL analyses. The procedure of using item pairs instead ofsingle items was preferable for the following reasons: (a) By thisprocedure, the ratio of the number of subjects to the number ofvariables is increased (a minimum ratio of 10:1 was desired), (b)each item-pair variable is more reliable than a single-item variableand should have a smaller unique component, and (c) the factorloadings should be less affected by the idiosyncratic working ofindividual items. The major disadvantages of this procedure are that(a) information about individual items is lost, (b) items must bereasonably homogeneous with respect to the dimension that theywere designed to measure so that it is important to perform a pre-liminary item analysis such as the one described above before thisprocedure is used, and (c) the parameter estimates and the factorscores resulting from the factor analysis are likely to vary some-what, depending on the particular pairing of items used. The prac-tical effects of each of these possible disadvantages was subse-quently assessed. If a single item needed to be dropped at this point,the item with the lowest item-total correlation was excluded. Be-cause each scale started with an odd number of items (either fiveor seven), one item was dropped from each of the seven scales atsome point during the first three steps. For only one scale (mean-ingfulness) did an item drop because of a low corrected item-totalcorrelation. From the other scales, either one item was randomlyeliminated or one item was eliminated if it had a relatively highloading (>.30) on more than one factor. The resulting scales werethen divided into two (or three) item pairs.

Factor analysis and a set of LISREL analyses were used to assessthe hypothesis of a multifaceted structure to interest. The item pair-ing was done such that the first two items in a scale on the surveywere assigned to the first pair, the next two items were assigned tothe second, and so forth. The factor analyses were then conductedusing these item pairs. To check for variability in results due to aparticular item pairing, another set of factor analyses was conductedwith a second set of random item pairs. Next, a LISREL analysiswas conducted; this step is described in detail below. The final stepwas to assess the correlational relationships of the latent variablesfrom the LISREL analysis. The primary purpose of this analysis wasto assess the tenability of the distinction between the catch and holdfacets of situational interest.

Evaluation of Alternative Models

An analysis of the covariance structure of the data, computedwith LISREL 7.15 (Joreskog & Sorbom, 1989), was conducted toexamine the multifaceted structure of situational interest. Figure 3gives one diagram of how a LISREL model was conceptualizedbefore an analysis was conducted. The boxes in Figure 3 represent

430 MATHEW MITCHELL

e18

e19

Figure 3. LISREL path diagram for Model 5. (PI = Personalinterest. SI = Situational interest. Mean = Meaningfulness. In-volve = Involvement. Group = Group work. Puzzle = Puzzles.Comp = Computers. PI, PI and PI, P2 = Personal interest, Pairs1 and 2, respectively. SI, PI; SI, P2; and SI, P3 = Situationalinterest, Pairs 1, 2, and 3, respectively. M, PI and M, P2 =Meaningfulness, Pairs 1 and 2, respectively. I, PI; I, P2; and I, P3= Involvement, Pairs 1, 2, and 3. G, PI; G, P2; andG, P3 = GroupWork, Pairs 1, 2, and 3, respectively. P, PI; P, P2; and P, P3 =Puzzles, Pairs 1, 2, and 3, respectively. C, PI; C, P2; and C, P3 =Computers, Pairs 1, 2, and 3, respectively, e = error term.)

the observed measurements; in this study each item pair representedone such measurement, yielding 19 boxes total. The circles repre-sent the unobserved constructs underlying the measurements andthe e's represent residuals (or errors) estimated in the data analysis.Notice that the straight arrows indicate that both the underlyingconstructs and errors give rise to the performance observed on the

measurements. Also, as Figure 3 indicates, each observed mea-surement is expected to load on only one facet of interest (all otherpossible loadings for a particular measurement are constrained tozero). The curved arrows indicate that the facets of situational in-terest (SI) are correlated; these correlations are estimated in the dataanalysis. For simplicity, not all curved arrows between all latentvariables have been drawn. However, in the LISREL models used,the program was allowed to find covariances between all possiblepairs of the latent variables. Also, the abbreviations used in theboxes indicate a particular item pair in all cases. For instance, PI,PI indicates personal interest pair 1. The first one (or two) letters(e.g., PI) indicate the construct on which the item pair is hypoth-esized to load. The last two symbols (e.g., P2) simply indicate whichitem pair the box represents.

Covariance structure analysis has traditionally relied on a chi-square significance test to determine the degree to which a proposedmodel fits the observed data. However, as many researchers havepointed out (e.g., Bentler & Bonnett, 1980; Byrne, 1989; Marsh,Balla, & McDonald, 1988), chi-square goodness-of-fit tests are of-ten inadequate for model evaluation because they are contingent onsample size. Various alternatives have been proposed. The two in-dices used in the current study include the comparative fit index(CFI; Bentler, 1990) and the relative noncentrality index (RNI; Mc-Donald & Marsh, 1990). Currently, these two measures seem tooffer the best indications of fit without being unduly affected bysample size. Both scales give results on a scale between 0 and 1.As a rule of thumb, any model with an index of less than .90 needsimprovement. Ratios greater than .90 are considered adequate, andthe closer one gets to 1, the more impressive the fit becomes.

Results and Discussion

Reliability

The internal consistency coefficients (Cronbach's alphas)for the scales ranged from .77 to .93 (in particular, .92 forpersonal interest, .90 for situational interest, .77 for mean-ingfulness, .86 for involvement, .93 for groups, .88 for puz-zles, and .92 for computer). For purposes of a psychometricinstrument, alpha coefficients of at least .70 would be desired(Nunnally, 1978). As the results indicate, all of the scaleshave a more than satisfactory coefficient. In addition, itempairing is not advisable if the reliability of the scales is notsatisfactory. However, the data in this study indicate that allof the individual items within a scale are reasonably homo-geneous.

Factor Analysis of the Interest Survey Responses

The factor analyses conducted on item pairs (see Tables 1and 2) identified the two general interest factors and the fivesituational interest subfactors with great clarity. The resultspresented are based on a factor analysis conducted with themaximum likelihood extraction technique with an obliquerotation.1 In assessing the best solution for a factor analysis,

1 Two matters should be noted about the factor analysis results.First, results from factor analyses obtained with a variety of dif-ferent extraction techniques and rotations resulted in the samepattern of results as presented in Table 1. For example, using aprincipal axis extraction with orthogonal rotation provided results

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 431

Table 1Factor Analyses of the Interest Survey Item Pairs

Item/pair

PI, PIPI, P2SI, P2SI, PISI, P3

M, P2M, PIC, P2C, PIC, P3G, P3G, PIG, P2P, P2P, P3P, PII, P3I, PII, P2

Factor 1

Situational0.980.85

-0.090.030.08

Factor 2 Factor 3

interest and personal interest general-0.06

0.080.970.860.77

Situational interest1.020.66

-0.030.020.01

-0.01-0.02

0.040.04

-0.020.000.08

-0.030.00

0.02-0.01

0.940.870.82

-0.03-0.03

0.070.040.07

-0.12-0.05

0.060.04

—

subfacet item pairs0.000.000.00

-0.030.020.960.880.820.060.03

-0.060.020.010.02

Factor 4

item pairs

—

0.02-0.01-0.01-0.09

0.080.00

-0.020.040.890.850.79

-0.010.08

-0.02

Factor 5

—

-0.060.060.050.10

-0.07-0.03

0.030.04

-0.08-0.01

0.150.870.780.74

Note. PI = Personal interest;= Group work; P = Puzzles;indicate not applicable.

; SI = Situational interest; M = Meaningfulness; C = Computers; GI = Involvement; PI = Pair 1; P2 = Pair 2; P3 = Pair 3. Dashes

the rule of thumb used was to retain factors with eigenvaluesgreater than 1. For the factor analysis in Table 1, the secondeigenvalue was 1.19 and the third was 0.30, indicating thata two-factor solution is reasonable. Similarly, for the factoranalysis in Table 1, the fifth eigenvalue was 1.09 and the sixthwas 0.41, indicating that the five-factor solution was optimal.The parsimony postulate (Kim & Mueller, 1978) for factoranalysis states that if both an N factor and an TV - 1 factorsolution are consistent with the data, then the N - 1 modelshould be accepted as more accurate because of its greaterparsimony. Thus, factor analyses were also conducted withless than 2 (or 5) rotated factors. For the solutions with lessthan the targeted number of factors, at least one of the interestsurvey factors was not adequately represented in that thevariables designed to measure it did not load substantially onany factor. Therefore, the original 2- (and 5-) factor solutionswere interpreted to be the most parsimonious solutions.

A second set of factor analyses was then conducted witha set of random item pairs. The intent of these factor analyseswas to check for variability in results attributable to a par-ticular set of item pairs. As the results in Table 2 indicate, thealternative set of item pairs resulted in an identical pattern ofresults, although the percentage of variance accounted forby each factor varied somewhat from the first set of factoranalyses.

virtually identical to those given in Table 1. Second, when one usesan oblique rotation (which assumes that there may be correlationsbetween factors) it is possible to get factor loadings that are greaterthan 1 (e.g., Norusis, 1990, p. 335). Thus, the one factor loading of1.02 (for meaningfulness, Pair 2) is possible given that an obliquerotation was used in the factor analysis.

A factor analysis of items from all seven scales indicatedthat the involvement items loaded on the same factor asdid the general situational interest items. The rest of theitems, however, loaded onto factors representing the sepa-rate scales that they represented. At this point, however, itwas unclear whether it would be sensible to think of in-volvement and situational interest as one factor (implyingthat involvement was synonymous with situational inter-est). The LISREL analysis (described below) was used tohelp assess this possibility.

LISREL Analysis

With an exploratory factor analysis, a particular structurecannot be statistically tested against alternative structures(Byrne, 1989). In addition, the structure of the model cannotbe controlled (beyond setting the number of factors and thelevel of correlation between factors). For both of these rea-sons, a confirmatory factor analysis conducted with LISRELwas performed. The main objective was to compare the vi-ability of a variety of different models to explain the observedmeasures. Construct validity often proceeds by disconfirm-ing alternative models. Thus, I hoped that the primary model(based on seven scales) would describe the data substantiallybetter than any reasonable countermodel.

A nested set of six models was examined: Null model—anull model of complete independence of all observedmeasurements—this provides a measure of the total covar-iation in the data; Model 1—a single-facet, general interestmodel; Model 2—a two-facet, personal and situational in-terest model; Model 3—a four-facet model including per-sonal and situational interest with catch and hold subfacets

432 MATHEW MITCHELL

Table 2Factor Analyses Computed With an Alternative Pairing of the Interest Survey Items

Item/pair Factor 1 Factor 2 Factor 3 Factor 4 Factor 5

Situational interest and personal interest general item pairsPI, P2 0.99 -0.03 — — —PI, PI 0.86 0.03 — — —SI, PI 0.02 0.92 — — —SI, P3 0.07 0.91 — — —SI, P2 -0.06 0.82 — — —

G, P2G, PIG, P3C, P3C, PIC, P2M, PIM, P2P, P2P, PIP, P3I, P3I, P2I, PI

0.950.900.880.05

-0.01-0.01

0.05-0.06-0.03

0.000.090.020.000.03

Situational interest-0.03

0.050.010.920.870.86

-0.010.030.040.05

-0.12-0.01

0.050.00

subfacet item pairs-0.04

0.050.000.050.03

-0.060.960.630.00

-0.010.040.01

-0.080.12

-0.010.000.030.05

-0.080.02

-0.060.110.910.890.70

-0.050.070.00

-0.020.030.03

-0.090.090.04

-0.070.140.02

-0.060.080.860.780.77

Note. PI = Personal interest; SI = Situational interest; M = Meaningfulness; C = Computers; G= Group work; P = Puzzles; I = Involvement. PI = Pair 1; P2 = Pair 2; P3 = Pair 3. Dashesindicate not applicable.

of situational interest; Model A—a six-facet model includingpersonal and situational interest (with the involvement itemsincorporated into the situational interest latent variable) andthe four other subfacets of situational interest (meaningful-ness, puzzles, group work, and computer subfacets of situ-ational interest); and Model 5—a seven-facet model thatadded the involvement subfacet of situational interest toModel 4. Model 5 is the full hypothesized model and is pic-torially represented by Figure 3. The factor analyses con-ducted previously indicated that Models 4 and 5 should endup being the real competitors, although Models 1 through 3represent theoretically plausible alternatives given that noempirical evidence that addresses the plausibility of the sit-uational interest/personal interest or the catch/hold distinc-tions had heretofore been collected.

The results of fitting each of the six models to the datacollected are presented in Table 3. We interpret the data inTable 3 to mean that Model 5 provides the best fit to the data.Each indicator of goodness of fit shows that Model 5 is betterthan Model 4, that Model 4 is better than Model 3, and soforth. Tests of statistical significance, conducted with theproportional reduction in the chi-square measure, showedthat the differences between each of these models were sta-tistically significant {p < .01). Whereas Model 4 providedan adequate model for describing the data, Model 5 describedthe data better both statistically and substantively. Substan-tively, the theory predicted involvement to be a subfacet ofsituational interest. In addition, it is more helpful for edu-cators to know that increasing student involvement is a keytool for enhancing situational interest.

Table 3LISREL Results

Model

Null12345

df171152151146140131

x2

4,9952,9432,5521,878

496326

N

350350350350350350

CFI

0.420.500.640.930.96

RNI

0.420.510.640.930.96

Model description

Total covariationGeneral interest onlyPI and SI onlyPI, SI, catch and holdPI, SI,a meaning plus three catch facetsFull hypothesized model

Note. CFI = Comparative fit index; RNI = Relative noncentrality index; PI = Personal interest;SI = Situational interest. Dashes indicate not applicable.a Here SI indicates that both the SI item pairs and the involvement item pairs were hypothesized tobe part of the SI latent variable.

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 433

Correlational Analysis

Although the LISREL analysis provided further evidencethat the hypothesized multifaceted structure of situationalinterest was reasonable, as of yet there is little evidence tosupport the distinction made between those facets that catchand those that hold situational interest. Toward this end, Iconducted an analysis using the correlation matrix of factorsresulting from the factor analysis obtained with a maximumlikelihood extraction with an oblique rotation. In assessingthese correlational results, I considered correlation coeffi-cients greater than .70 as strong and those around .50 asmoderate. The results are presented in Table 4. These dataessentially support the conceptual distinction made betweencatching and holding. For instance, the two hypothesizedhold facets, meaningfulness and involvement, have a mod-erate and a strong correlation with situational interest, re-spectively. The three catch facets (groups, puzzles, and com-puters), however, have weak correlations with situationalinterest, and each is weakly correlated with the other facets.

Furthermore, I conducted significance tests to see whetherthe difference between the hold and catch facets was statis-tically confirmed. Toward this end, I tested the differencesbetween two dependent correlation coefficients (Glass &Hopkins, 1984, p. 310). In particular, I examined whether thehold/situational interest correlations were significantlyhigher than the highest catch/situational interest correlation(i.e., whether rs of .75 and .50 are significantly higher than.37): rs (347) = 9.59 and 2.35, respectively, bothps < .025,one-tailed.

Summary and Conclusion

The purpose of this study was to assess the tenability ofa hypothesized multifaceted construct of situational interestwithin the secondary mathematics classroom. The study hadfour major purposes: (a) to assess whether the distinctionbetween personal interest and situational interest was rea-sonable, (b) to assess whether the situational interest struc-ture was multifaceted, (c) to determine whether the full hy-pothesized situational interest model was the best fittingmultifaceted model of situational interest, and (d) to assessthe general distinction made between catch and hold facetswithin situational interest. The results of the various analysesclearly identified two general scales (personal and situationalinterest) and five subscales related to situational interest. Fur-

thermore, the LISREL analyses indicated that this full modelwas a better fitting multifaceted model than any of a seriesof reasonable alternative models. In addition, the correla-tional analyses lent support to the conceptual distinctionmade between catching and holding interest.

One potential limitation to this study was the use of itempairs rather than single items. However, because each scalewas very homogeneous (a > .77) and because a second ran-dom pairing of items provided an identical pattern of results,item pairing seemed reasonable for the purposes of this study.In addition, although the study developed an instrument thatmeets some valued construct validity criteria, the interestsurvey needs to be used in a variety of other settings (in-cluding different locations, non-college-prep classes, studentpopulations with different ethnic mixtures, etc.). Nonethe-less, the interest survey has a well-developed factor structure,and it measures dimensions that are reliable and based on atheoretical model. At this point in time, one of the majorweaknesses of research into interest is the lack of instrumentsused to measure it. Although the interest survey is at a pre-liminary stage, further development of the instrument is war-ranted, as the interest survey appears to offer a measurementtool of great practicality for further research purposes. Oneadditional limitation of this study is that it has relied solelyon survey data. In this regard, it is important that future stud-ies concerning a model of interest incorporate behavioralmeasures as well.

In addition, this study adopted Schiefele's (1991) recentproposal that interest is a construct specific to particular sub-ject matter. However, it is far from clear to what degreeSchiefele's proposal holds. For instance, it is certainly ofinterest to find out to what degree the present model accu-rately describes student perceptions across subject matters.In particular, it seems reasonable that facets such as mean-ingfulness and involvement may hold students' interest re-gardless of the subject area. Whereas the present study fo-cused on interest in the mathematics classroom, futurestudies need to address the subject-specific assumptionsabout models of interest.

Finally, although the five subfacets of situational interestthat emerged in this study helped to elaborate a model ofinterest, future studies will need to address more fully the roleof each of the subfacets. For instance, are all forms of groupwork related to situational interest or are only some formsrelated? If only some, under what conditions is group workperceived as interesting by students?

Table 4Correlations Between Factors

Measure 1

1. SI2. Meaning3. Involve4. Group5. Computers6. Puzzles

0.500.750.320.280.37

0.390.140.100.22

0.330.340.36

0.200.36 0.07

Note. SI = Situational interest; Meaning = Meaningfulness; Involve = Involvement; GroupGroup work. Dashes indicate not applicable.

434 MATHEW MITCHELL

Given the above limitations, the results of this study haveimplications for further theoretical work in interest research.Previously, interest has been treated as a general concept.Although important distinctions have been made betweenpersonal and situational interest, and also between catchingand holding interest, empirical evidence has not been gath-ered to support these theoretical distinctions. Toward thatend, this study provides initial quantitative evidence sup-porting the tenability of those distinctions. Furthermore, thisstudy sets forth a theoretical rationale for distinguishing be-tween the catch and hold facets of interest by using the pro-posed contrast between stimulation and empowerment.

This study also contributes to a potentially powerful ap-proach to improving mathematics education. As Dewey(1933) succinctly phrased it, "Teaching may be compared toselling commodities. No one can sell unless someone buys"(p. 35). Clearly, in many mathematics classrooms, studentshave not been "buying." Recent studies note that as studentsget older, fewer of them report liking mathematics (Dosseyet al., 1988; Stodolsky, Salk, & Glaessner, 1991). In addition,it appears that as students get older, their performance inmathematics gets worse (Kantrowitz & Wingert, 1991).However, enhancing classroom interest may offer the mostdirect way to improve this educational morass. As Csikszent-mihalyi (1990) stated, "if intrigued by the opportunities ofthe domain, most students will make sure to develop theskills they need to operate within it" (p. 126). The results ofthis study begin to define a conceptual model of ways inwhich to enhance classroom environments, by specifyingfive instructional tools that contribute to students' percep-tions of classroom interest.

To appreciate the potential usefulness of the situationalinterest approach in mathematics education, it is importantto emphasize the state of affairs in most high school class-rooms today. For instance, Weiss (1990) found that only 31 %of secondary mathematics teachers state that they give aheavy emphasis to getting students more interested in math-ematics. One reason for this lack of emphasis from practi-tioners may be a lack of knowledge about how to system-atically develop situational interest in their classrooms.Furthermore, there is some evidence that few secondarymathematics classrooms use either of the hold facets. Forexample, consider a recent survey of llth-grade students(Dossey et al., 1988) in which it was found that 82% of thestudents reported never doing mathematics laboratory ac-tivities, and 87% reported never making reports or doingprojects in their mathematics class. Although the categoriesused by Dossey et al. do not strictly conform to the holdfacets in this study, their results nonetheless make it clear thatimplementation of the situational interest hold facets is likelya rare phenomenon in many high school mathematics classes.

However, at least at a policy level, our approach tomathematics education may be changing. For example,both the Curriculum and Evaluation Standards for SchoolMathematics of the National Council of Teachers of Math-ematics (1989) and the California Framework (CaliforniaMathematics Framework Committee, 1992) emphasizedthe need for students to do meaningful mathematical activ-ities and to be actively involved in the learning process.

These policy committees seem to understand the criticalimportance of observations, such as the one made byKaput (1989) in writing,

Few now deny that school mathematics as experienced bymost students is compartmentalized into meaningless piecesthat are isolated from one another and from the students' widerworld. . . .This experienced meaninglessness of school math-ematics devastates the motivation to learn or use mathematics.(p. 99)

The results concerning the hold facets indicate that if ed-ucators and educational psychologists wish to enhance thesituational interest of mathematics classrooms, they need toradically rethink how they typically conduct mathematicslessons. In particular, involvement stands out as the facetwith the strongest relationship with overall situational in-terest. This result is not very surprising given the work ofStodolsky et al. (1991) who concluded that "the usual in-structional pattern in math classes establishes an expectationon the part of students that they will be 'told' math" (p. 111).The "telling" of mathematics is a deeply ingrained instruc-tional orientation. However, as the study's results clearly in-dicate, the more students perceive themselves as active learn-ers rather than as passive absorbers of knowledge, the morea classroom environment will persist as a mathematics in-terest holder.

References

Anderson, R. C , Shirey, L. L., Wilson, P. T., & Fielding, L. G.(1987). Interestingness of children's reading material. In R. E.Snow & M. J. Farr (Eds.), Aptitude, learning and instruction:Vol. 3. Conative and affective process analyses (pp. 287-299).Hillsdale, NJ: Erlbaum.

Bentler, P. M. (1990). Comparative fit indices in structural models.Psychological Bulletin, 107, 238-246.

Bentler, P. M., & Bonnett, D. G. (1980). Significance tests andgoodness of fit in the analysis of covariance structures. Psycho-logical Bulletin, 88, 588-606.

Berlyne, D. E. (1960). Conflict, arousal, and curiosity. New York:McGraw-Hill.

Berlyne, D. E. (1966). Curiosity and exploration. Science, 153, 25-33.

Buxton, C. (1973). Adolescents in schools. New Haven, CT: YaleUniversity Press.

Byrne, B. (1989). A primer of LISREL. New York: Springer-Verlag.California Mathematics Framework Committee. (1992). Mathemat-

ics framework for California public schools. Sacramento, CA:California State Department of Education.

Carey, S. (1985). Conceptual change in childhood. Cambridge,MA: MIT Press.

Csikszentmihalyi, M. (1990). Literacy and intrinsic motivation.Daedalus, 119, 115-140.

Csikszentmihalyi, M., & Larson, R. (1984). Being adolescent. NewYork: Basic Books.

Davis, R. B. (1992). Reflections on where mathematics educationnow stands and on where it may be going. In D. Grouws (Ed.),Handbook of research on mathematics teaching and learning(pp. 724-734). New York: Macmillan.

Dewey, J. (1913). Interest and effort in education. Cambridge, MA:Riverside Press.

Dewey, J. (1933). How we think (Rev.). New York: Heath.Dossey, J. S., Mullis, I. V. S., Lindquist, M. M., & Chambers, D.

SITUATIONAL INTEREST IN THE MATHEMATICS CLASSROOM 435

L. (1988). The mathematics report card: Are we measuring uplPrinceton, NJ: National Assessment of Educational Progress atEducational Testing Service.

Garner, R., Alexander, P. A., Gillingham, M. G., Kulikowich, J. M.,& Brown, R. (1991). Interest and learning from text. AmericanEducational Research Journal, 28, 643-659.

Garner, R., Gillingham, M. G., & White, C. S. (1989). Effects of '"seductive details" on macroprocessing and microprocessing inadults and children. Cognition and Instruction, 6, 41-57.

Glass, G., & Hopkins, D. (1984). Statistical methods in educationand psychology. Englewood Cliffs, NJ: Prentice Hall.

Good, T., Mulryan, C , & McCaslin, M. (1992). Grouping for in-struction in mathematics: A call for programmatic research onsmall-group processes. In D. Grouws (Ed.), Handbook of re-search on mathematics teaching and learning (pp. 165-196).New York: Macmillan.

Hidi, S. (1990). Interest and its contribution as a mental resourcefor learning. Review of Educational Research, 60, 549-571.

Hidi, S., & Baird, W. (1988). Strategies for increasing text-basedinterest and students' recall of expository texts. Reading ResearchQuarterly, 23, 465-483.

Jackson, P. W. (1968). Life in classrooms. New York: Holt.Johnson, D., & Johnson, R. (1975). Learning together and alone.

Englewood Cliffs, NJ: Prentice Hall.Joreskog, K. J., & Sorbom, D. (1989). LISREL: Analysis of linear

structural relationships by the method of maximum likelihood.Mooresville, IN: Scientific Software.

Kantrowitz, B., & Wingert, P. (1991, June 17). Adismal report card.Newsweek, 64-66.

Kaput, J. J. (1989). Information technologies and affect in math-ematical experience. In D. B. McLeod & V. M. Adams (Eds.),Affect and mathematical problem solving (pp. 89-103). NewYork: Springer-Verlag.

Kim, J., & Mueller, C. W. (1978). Factor analysis: Statistical meth-ods and practical issues. Beverly Hills, CA: Sage.

Kline, M. (1977). Why the professor can't teach. New York: St.Martin's Press.

Krapp, A. (1989). Interest, learning and academic achievement. InP. Nenniger (Chair.), Task motivation by interest. Symposiumconducted at the meeting of the Third European Conference ofLearning and Instruction, Madrid, Spain.

Krueger, R. A. (1988). Focus groups: A practical guide for appliedresearch. Newbury Park, CA: Sage.

Lepper, M. R. (1988). Motivational considerations in the study ofinstruction. Cognition and Instruction, 5, 289-309.

Lepper, M. R., & Malone, T. W. (1987). Intrinsic motivation andinstructional effectiveness in computer-based education. In R. E.Snow, & M. J. Farr (Eds.), Aptitudes, learning, and instruction:Vol 3. Conative and affective process analyses (pp. 255-286).Hillsdale, NJ: Erlbaum.

Malone, T. W. (1981). Toward a theory of intrinsically motivatinginstruction. Cognitive Science, 4, 333-369.

Malone, T. W., & Lepper, M. L. (1987). Making learning fun: Ataxonomy of intrinsic motivations for learning. In R. E. Snow, &M. J. Farr (Eds.), Aptitudes, learning, and instruction: Vol 3.Conative and affective process analyses (pp. 223-254). Hillsdale,NJ: Erlbaum.

Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness-of-fit indexes in confirmatory factor analysis: The effect of sam-ple size. Psychological Bulletin, 103, 391-410.

McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariatemodel: Noncentrality and goodness of fit. Psychological Bulletin,107, 247-255.

Mellin-Olsen, S. (1987). The politics of mathematics education.Dordrecht, The Netherlands: D. Reidel.

Mitchell, M. T. (1991). Formative evaluation report of the Math-ematics Achievement Program (MAP). (Tech. Rep. No. 91-01).Santa Barbara: University of California at Santa Barbara, Officeof Instructional Consultation.

Mitchell, M. T. (1992). A multifaceted model of situational interestin the secondary mathematics classroom. Unpublished doctoraldissertation, University of California at Santa Barbara, CA.

Moos, R. H. (1976). The human context: Environmental determi-nants of behavior. New York: Wiley.

Moos, R. H. (1979). Evaluating educational environments. SanFrancisco: Jossey-Bass.

Morgan, D. L. (1988). Focus groups as qualitative research. New-bury Park, CA: Sage.

National Council of Teachers of Mathematics. (1989). Curriculumand evaluation standards for school mathematics. Reston, VA:Author.

National Council of Teachers of Mathematics. (1991). Professionalstandards for teaching mathematics. Reston, VA: Author.

National Research Council. (1989). Everybody counts: A report tothe nation on the future of mathematics education. Washington,DC: National Academy Press.

Noddings, N. (1985). Small groups as a setting for research onmathematical problem solving. In E. Silver (Ed.), Teaching andlearning mathematical problem solving: Multiple research per-spectives (pp. 345-360). Hillsdale, NJ: Erlbaum.

Norusis, M. J. (1990). SPSS base system user's guide. Chicago:SPSS.

Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York:McGraw-Hill.

Papert, S. (1980). Mindstorms: Children, computers, and powerfulideas. New York: Basic Books.

Putnam, R., Lampert, M., & Peterson, P. (1990). Alternative per-spectives on knowing mathematics in elementary schools. Reviewof Research in Education, 16, 57-152.

Sarason, S. (1983). Schooling in America: Scapegoat and salvation.New York: Free Press.

Schank, R. C. (1979). Interestingness: Controlling inferences. Ar-tificial Intelligence, 12, 21'3-297'.

Schiefele, U. (1991). Interest, learning, and motivation. Educa-tional Psychologist, 26, 299-323.

Schoenfeld, A. H. (1987). Cognitive science and mathematics ed-ucation: An overview. In A. H. Schoenfeld (Ed.), Cognitive sci-ence and mathematics education (pp. 1-31). Hillsdale, NJ: Erl-baum.

Shavelson, R. J., Hubner, J. J., & Stanton, G. C. (1976). Self-concept: Validation of construct interpretations. Review of Edu-cational Research, 46, 407^441.

Slavin, R. (1980). Cooperative learning. Review of Educational Re-search, 50, 315-342.

Stodolsky, S. S., Salk, S., & Glaessner, B. (1991). Student viewsabout learning math and social studies. American EducationalResearch Journal, 28, 89-116.

Voss, H. G., & Keller, H. (1983). Curiosity and exploration: The-ories and results. San Diego, CA: Academic Press.

Weiss, I. (1990). Mathematics teachers in the United States. Inter-national Journal of Educational Research, 14, 139-155.

Whitehead, A. N. (1959). The aims of education and other essays.New York: Macmillan. (Original work published 1912)

Yerushalmy, M., & Chazan, D. (1990). Overcoming visual obstacleswith the aid of the supposer. Educational Studies in Mathematics,21, 199-219.

(Appendix follows on next page)

436 MATHEW MITCHELL

Appendix

Items Used in the Final Survey

Notes on the Survey Format

All of the personal interest items were put together as one groupof items at the beginning of the survey. These items were prefacedwith the statement: "For the first 4 questions, think about how youfelt about mathematics before the school year began."

Ail of the other items were randomly mixed in the rest of thesurvey.

All of the items had the following response format:

Our class is fun.

strongly agree agree slightly agree slightly disagreedisagree strongly disagree

Personal Interest

1. Mathematics is enjoyable to me.2. I have always enjoyed studying mathematics in school.3. Compared to other subjects, I feel relaxed studying mathematics.4. Compared to other subjects, mathematics is exciting to me.

Situational Interest

1. Our class is fun.2. I actually look forward to going to math class this year.3. Our math class is dull.4. This year I like math.5. I don't find anything interesting about math this year.6. My other classes are more interesting than math.

Meaningfulness

1. The stuff we learn in this class will never be used in real life.2. Class would be better if the math problems were more related

to life problems.3. I see the math we've learned as important in life.4. I will never use the info in this class again, so I don't need it.

Involvement

1. Our teacher has fun activities to learn the stuff that we need toknow.

2. We just come in, take notes, go home, do homework, and it's thesame thing every day.

3. We learn the material ourselves instead of being preached at.4. We usually sit and listen to the teacher talk.5. We often do something instead of the teacher just talking.6. We often hear long, long explanations and I quickly lose interest.

Group Work

1. I like the groups in our class because learning is more fun whenthings can be discussed.

2. I like working in groups because I can ask one of the people inthe group and they'll explain things on a level I can understand.

3. When we work in groups we exchange ideas.4. When we do group work it's like working as a team.5. The groups we use in class make work easier.6. Working in groups makes our class more enjoyable.

Puzzles

1. I enjoy doing starters or mind-teasers.2. We do starters which warm me up and get my head going.3. I like how we do logic puzzles which exercise my brain.4. It's good to have a starter in math to get us thinking.5. The mind-teasers or logic puzzles we do are fun.6. The mind-teasers or logic puzzles we do make me think.

Computers

1. We try to discover things on the computer in our class.2. We use computers which let me experiment with what's being

taught.3. In our class we often work on computers where I discover things

on my own.4. We work on computers to actually put problems together our-

selves and see them in reality.5. We work on computers more than in a book.6. Using computers in our class is fun.

Received May 11, 1992Revision received March 4, 1993

Accepted March 4, 1993 •