INDIVIDUAL DIFFERENCES IN TEXT ...

P.cychologicaf Reports, 1999, 84, 167-177. O Psychological Reports 1999

INDIVIDUAL DIFFERENCES IN TEXT COMPREHENSION AS A FUNCTION O F TEST ANXIETY AND PRIOR KNOWLEDGE ' ALEXANDER E. MINNAERT

Leiden Universily Center /or the Sttrdy of Edrrcation and Insmrction

Surnmay.-This study investigated che relationship between reading comprehension and comprehension monitoring with undergrddu~res(223 women, 69 men). Further, the effect of test anxiety and of prior knou~letlgeon reading comprehension and on comprehension monitoring was examined in groups of students of equal intellecbetter on tual ability. Students with high scores on reading comprehension a comprehension monitoring task as well. Individual differences in reading comprehension with a muldplc-choice response format emerged as a function of the interaction berween test anxiety and prior knowledge. Students with low prior knowledge and high test anxiety performed worst of all. We found a far Less detrimental effect of test anxiety and prior knowledge on monitoring comprehension than on reading comprehension.

In the 1980s and 1990s an augmented research output in the field of classic reading comprehension versus comprehension monitoring has been observed (Wagoner, 1983; Baker, 1989; Kinnunen & Vauras, 1995). Comprehension monltormg refers to cognitive monitoring limited to the comprehension of connected discourse (Wagoner, 1983). It stems from the more metacognitive approach in learning and instruction and refers to students' differentiation between what they know and do not know. It is regarded as a precursor of self-regulated learning (Boekaerts, 1997). As stipulated by Tobias and Everson (1997), students can hardly be expected to exercise effective regulation of their learning and to select appropriate strategies to attain their goals, if they cannot distinguish between what they know and do not know. Besides, comprehension monitoring appeared to have a substantial relationship with academic achievement in high school for both boys and girls (Otero, Companario, & Hopkins, 1992). In general, outcomes of assessment are strongly influenced by test anxiety (Crocker & Schmitt, 1987; Snow, 1993). In practice, the role of test anxiety in reading comprehension and comprehension monitoring is, however, often forgotten. The confhcting effects of fachtating and debhtating anxiety on academic performance indicated their value in predicting goal-directed behavior (Rand, Lens, & Decock, 1991; Birenbaum & Pinku, 1997). Sarason (1984) ~ o i n t e dout that test anxiety may impede ~erformanceand act upon 'Address correspondence to Dr. Alexander Minnaert, Leiden University, Faculty of Social and Behavlor.d Sciences, Center for [he Study of Education and Instruction, Wassenaarseweg 52, P.O. Box 9555, 2300 RE Leiden, The NetherIands or e-mail ([email protected]).

168

A. E. MINNAERT

cognitive and affective regulatory strategies; however, reports on the effects of these motivational tendencies on reading comprehension and comprehension monitoring activities in higher education are scarce. Wittmaier (1972) showed that highly test-anxious students had significantly lower competence in study skllls than low test-anxious students. Similarly, research lent support to the fact that low test-anxious undergraduate students enrolled at university reported higher study s l d s than high test-anxious students (Culler & Holahan, 1980; Naveh-Benjamin, McKeachie, Lin, & Holmger, 1981). In secondary schools, the empirical findings about the relationship between test anxiety, reading comprehension, and monitoring skds (as measured by selfreport measures) are ambiguous and indeterminant. Pintrich and De Groot (1990) found no significant ( h e a r or n o h e a r ) relationship for seventhgrade students between test anxiety and self-regulation ( r = -.13) or between test anxiety and reading comprehension strategies (r = .04). The study that took account of a causal effect of test anxiety on reading comprehension strategies and on self-regulation, reported interesting, although quite similar findings for seventh-grade students (Pintrich, Roeser, & De Groot, 1994): scores on test anxiety were moderately correlated with self-regulation at the first (r = -.25) and the second (r = -.29) measurement; however, the relationship between test anxiety at Time 1 and self-regulation at Time 2 became statistically insignificant. Nevertheless, there seems to be a gap in the literature about the relationship between test anxiety and on task measurements of skds for monitoring comprehension. It is well known that prior knowledge, i.e., the knowledge available before a certain learning task, has a strong influence on reading comprehension and on academic performance (Glaser, 1984; Dochy & Alexander, 1995; Minnaert & Janssen, 1996). Besides, evidence was found that comprehension monitoring is partially related to domain-specific and partially to general knowledge characteristics (Korkel & Schneider, 1991; Minnaert, 1996). So, the domain-specificity hypothesis of comprehension monitoring should be kept in mind. The investigation of the interaction between test anxiety and prior knowledge in reading comprehension and comprehension monitoring is, however, almost completely absent in the research literature. Moreover, research conducted on the same target group for both reading comprehension and comprehension monitoring is also very sparse. In t h ~ sstudy, individual differences in reading comprehension performance and comprehension monitoring in relation to university students' test anxiety and prior knowledge are focused. The effects of test anxiety and prior knowledge level are hypothesized to be of importance in reading comprehension and comprehension monitoring outcomes. Evidence is found in the literature that reading comprehension performance and comprehension monitoring outcomes in primary education were

INDIVIDUAL DIFFERENCES IN TEXT COMPREHENSION

169

most often positively related to each other (see Utsuda, 1990; Kinnunen & Vauras, 1995); however, do university students with better reading comprehension have better comprehension monitoring too? In other words, can one generalize the findmgs of Mitsuda (1990) and of Kinnunen and Vauras (1995) to higher education? Baker (1989) stipulated that adults who have more expertise are better readers and those who are more successful students seem to have greater awareness and control of their own cognitive activities while reading. Therefore, Hypothesis one is that university students with better reading comprehension \MLU also have better comprehension monitoring. We hypothesized that both readmg comprehension and comprehension monitoring are functions of both test anxiety and domain-specific prior knowledge. Evidence for this hypothesis can be found in Naveh-Benjamin, et al. (1981), who suggested that highly anxious students report more worry due to their inadequate prior knowledge state or their inadequate encoding of information about the subject matter. Hypothesis two states that both test anxiety and prior knowledge influence readmg comprehension. Hypothesis three states that both test anxiety and prior knowledge influence comprehension monitoring. METHOD Subjects Participants were undergraduates in psychology who volunteered for the study. Within the first week of the academic year, 337 students were assessed on test anxiety, prior knowledge, and reading comprehension. A few weeks later, 313 of these freshmen were also assessed on comprehension monitoring. Ln the final sample complete data were available for 223 women and 69 men whose average age was 18 to 19 years.

Measurement of Test Anxiety and Prior Knowledge We used Mehrabian's questionnaire as a measure of individual differences in achieving tendency (Mehrabian & Bank, 1978). Theoretical and empirical analyses showed conceptual similarity between test anxiety and the achievement motivation factor 'fear of failure' (see Rand, et a/., 1991; Hagtvet & Benson, 1997). Therefore, an interchangeable use of test-anxiety measures is quite legitimate as has been done before by others (see Rand, et al., 1991). The main reason to refer achievement motivation was given by the research results of Kuhl (1983, 1994) and Boekaerts (1994). Discriminant validity of achievement motivation compared to more action-controlled metacognitive mediation was clearly demonstraced by their research. Principal €actor analysis on the data of the translated version (with permission by Prof. W. Lens, University of Leuven) of Mehrabian's questionnaire yielded two principal factors corresponding to the tendencies to achieve success and to

170

A. E. MINNAEKT

avoid failure. The score on the second factor was used in this study. The internal consistency of the scores on latter items was fairly good, a =.81. Domain-specific prior knowledge in psychology was measured by a general information test with multiple-choice questions (see Minnaert & Janssen, 1996). Recent media information about psychology, domain-related information in the literature suggested as preparation before the start of the academic year, and curriculum information at students' disposal were the sources to construct content-valid, domain-specific prior knowledge questions. The available domain-specific and curriculum-related knowledge state of students was the intended focal point of this test. Every correct answer (there was only one correct answer per item) added one point to the subject's score. The sum of scores on these 30 items of prior knowledge was used for further research. The alpha-reliabhty of the items in the prior knowledge test was .74. In a previous study (see Minnaert, 1996), this measure showed predctive vali&ty (r =.33, N = 161) with the percentage of points at the end of the first academic year.

Tasks and Measure7nent of Reading Comprehension and Comprehension Monitoring The reading comprehension cask within a conventional multiple-choice format took [he form of a silent reading test. Ten texts on topics in the psychology study program plus appropriate multiple-choice questions to be answered immediately after reading each text were designed to measure the accuracy skdls in deep-level comprehension of text. This type of silent reading test possessed not only high content, face, and nomological validity but also predictive validity for success and progress in higher education (Minnaert & Janssen, 1996). Students were allowed two hours of study time for the silent reading test with a total of 80 multiple-choice questions. Every correct answer added one point to the subject's score. The sum of scores on all questions was used in this research as a measure of text comprehension. The alpha-reliability of text comprehension was 3 4 . In a previous study (see Minnaert, 1996), this measure of text comprehension correlated .44 (N=161) with the percentage of points at the end of the first academic year. In the replication study, the predictive validity was virtually the same (r = .43, N = 176).

Comprehension monitormg was assessed with problem-solving tasks. Students were given 20 inferenr~alstatements about four different texts to measure this kind of metacognition: two about exact sciences (the speech production system, and classification of the animal kingdom) and two about human sciences (the family system of suicidal children, and youth protection organizations). After reading the text, subjects had to check for each statement whether it was in correspondence with the content of the text (right)

INDIVIDUAL DIFFERENCES IN TEXT COMPREHENSION

171

or not (wrong) or whether it was not possible to decide on the basis of the text content (neither right nor wrong). After each inferential statement, subjects had to evaluate the certainty of their decision on a 7-point scale anchored by 'uncertain' and 'certain.' The number of correct decisions was used for further research purposes. Therefore, four academic staff members collaborated in a preparatory study on the evaluation of the 'correctness' of the decisions, statement by statement. All staff members made the same decision on each of the inferential statements. Consequently, the interscorer agreement for these data was perfect.

Item Response Theoy: Dimensionality of Comprehension Monitoring To evaluate the internal consistency and to answer the question, "do we measure one dimension or one slull," we ran an Item Response Theory analysis using the One Parameter Logistic Model (OPLM) program (Verhelst, 1992). The RI, test, i.e., an asymptotically x2-distributed test statistic, was used to evaluate the dimensionahty of the response patterns. The conditional maximum kehhood procedure was used to estimate the item parameters; the weighted maximum hkel~hoodprocedure was used as default to estimate the subject parameters. Testing the partial domain-specificity hypothesis of comprehension monitoring (Korkel & Schneider, 1991; Minnaert, 1996), we successfully constructed two different measures of text comprehension monitoring skills, monitoring of exact-scientific contents (R1,= 27.6, df = 27, p = ,431 and of human-scientific contents (Rr,=23.5, df= 27, p = .66). The appropriate psychometric m o d e h g method uthzed here was Item Response Theory analysis that accounted for polytomous items, i.e., one- and two-parameter partial credit models [for a taxonomy of unidin~ensionalmodels for -polytomous items, see Thissen and Steinberg (1986)l. We combined . two essential components of comprehension monitoring, namely, an inferential decision-malung component and an evaluation component. In this way, the data were modeled as polytomous items reflecting a specific combination of both components. With polytomous items (scored as 0, 1 , . . . , m;), one assumes that the higher the score on the item, the higher the ability of the individual. Preliminary estimations of the items involved suggested to model the item weights as following: two points for each certain and correct decision, one point for each uncertain but correct decision, no points for each certain but incorrect decision, and one point for each uncertain and incorrect decision. From a psychological point of view, it seems relevant to acknowledge that decisions taken in uncertainty refer to an intermediate position of comprehension monitoring even insensitive of the outcomes of the decision taken. The evaluation component seems to prevail over the inferential decision-making component in case of uncertainty. Consequently, we test-

172

A. E. M I N N A E R T

ed the hypothesis that the inferential decision-malung component and the .evaluation component are based on the same, general comprehension monitoring abhty. Testing for unidimensional data m o d e h g was successful: for comprehension monitoring of exact-scientific contents Rl, = 59.2 (df= 57, p = ,39) and of human-scientific contents R1,=64.5 (df=57, p = 2 2 ) . All continuous scores were standardized into z scores to improve the transparency in the comparison of scoring methods.

Have the Better Text Cornprehenders a Better Comprehension Monitoring Level? The scores on readmg comprehension, as measured by the silent reading test, were a posteriori categorized into low, medlum, and high groups with below and above .66 standard deviation as the cutting point. Analyses of variance for the criteria 'comprehension monitoring' (for human- and exact-scientific contents in combination with the two scoring methods), with text comprehension as the independent variable, indicated significant effects of readmg comprehension (Table 1). The group with high text cornprehension surpassed significantly the lower group on comprehension monitoring, regardless of the content of monitoring. The medurn group was positioned in between the lower and higher groups on text comprehension. The better text comprehenders seemed to have better comprehension monitoring. This finclmg is in h e with the statements of Baker (1989). TABLE 1 MEANSAND STANDARD DEVIATIONS O F COMPREHENSION MONITOR~NG BY TkXT COMPREHENSION (N=292) Comprehension Monitoring Content

(12

M

FZ2,,

Text Comprehension Medium

Low = 92)

SD

(n=lll)

M

SD

Tukey HSD

High ( n = 90)

M

SD

Is Reading Comprehension a Function of Text Anxiety and Prior Knowledge? Due to nonlinear trends (y = -0.39 + 0 . 4 2 -~ 0.089x2+ 0.0048x', r = -0.59) between test anxiety and reading comprehension, and comprehension monitoring, we trichotomized the scores on test anxiety into low, medum, and high groups. The same was done for domain-specific prior knowledge. One could argue that the groups of test anxiety are confounded by differences in general ability. In a p r e h i n a r y research analysis, no statistically sig-

173

INDIVIDUAL DIFFERENCES I N TEXT COMPREHENSION

nificant differences in verbal, numerical, and diagrammatic intelhgence were found between the three test-anxiety groups (Wdks' h=.988; F4,,,=0.84, p = S O ) . The results of the analysis of variance with test anxiety and prior knowledge as independent variables and the standardized score on the silent reading test as the dependent variable (see Table 21, show a statistically significant main effect of prior knowledge (F,,,, = 17.60, p < .OOl), a nonstatistically significant main effect of test anxiety (F2,,,=0.94, p = .39), and a statistically significant interaction between prior knowledge and test anxiety (F,,,, = 2.41, pmedium, low) and that students with low prior knowledge and high test anxiety performed the worst of all. The latter finding is in h e with the suggestion of Naveh-Benjamin, et al. (1981) that highly anxious students reported more worry due to their inadequate knowledge of the subject matter. Notice the dysordinal interaction between prior knowledge (low and medium group) and test anxiety (low and medurn group). The reading comprehension score of students with low prior knowledge becomes lower with increasing test anxiety, marlung the inhibiting effect of test anxiety in case of a poor prior knowledge base. In summary, text comprehension was dependent on prior knowledge and on the interaction between test anxiety and prior knowledge. TABLE 2 MEANREADING COMPREHENSION SCOREA N D STANDARD DEWATIONS BY PRIORKNOWLEDGE A N D T m ANXIETY (N=292) -

Kno\vledge LOW

Medium High

-

Test Anxiety Medium

Prior Low

H~gh

M

SD

M

SD

M

SD

-.07 -.I9 .47

1.11 1.02 0.82

-.26 -.22 .35

0.96 0.77 0.86

-.66 .07 .40

1.09 0.83 0.78

Is Comprehension Monitoring a Function of Test Anxiety and Prior Knowledge? Analysis of variance with test anxiety and prior knowledge as independent variables and the standardized score on the polytomous scoring method for comprehension monitoring as the dependent variable (see Table 3), gave neither statistically significant main effects of test anxiety (F,,,= 1.50, ns), of prior knowledge (F,,, = 1.80, ns), or their interaction (F,,, = 0.34, ns). This finding meant that the comprehension monitoring score was not influenced by test anxiety and prior knowledge. This finding seems to be right irrespective of the content of comprehension monitoring.

174

A. E. MINNAERT TABLE 3

Comprehension Prior Monitoring Content Knowledge Exact-scientific

Human-scientific

Low Medium High Low Medium Hieh

Low

M

SD

.31 .02 3 -.09 .23 -.01

1.00 0.93 0.97 1.08 1.03 0.94

Test Anxiety llledium M SD

-.I0 -.23 -.08 -.09 .19 .21

1.05 1.03 1.25 1.09 0.78 1.00

High

M

SD

-.I4 .07 -.09 -.28 .17 -.06

0.78 0.87 1.08 0.96 0.86 1.11

DISCUSSION Students with high reading comprehension also performed better on a comprehension monitoring task (in h e with Hypothesis one). The group of university students with high reading comprehension surpassed significantly the lower group on comprehension monitoring of exact-scientific and of human-scientific contents. Individual differences in reading comprehension with a multiple-choice response format emerged as a function of prior knowledge and of the interaction between test anxiety and prior knowledge (in line with Hypothesis two). The main effect of test anxiety was, however, not significant (not in line with Hypothesis two). Comparing the low, medium, and high groups on test anxiety and on prior knowledge, students with low prior knowledge and high test anxiety performed the worst of all (z score of -.66). Students with high prior knowledge seemed not influenced by test anxiety. Students' scores on comprehension monitoring were far less influenced by test anxiety and prior knowledge (not completely in h e with Hypothesis three). The group of students with low prior knowledge and high test anxiety performed stdl the worst of all (with z scores of -.I4 and -.28 on exactscientific, and human-scientific contents, respectively). We noticed a far less detrimental effect of test anxiety and prior knowledge on comprehension monitoring than on reading comprehension, talung into account that the low, mechum, and high test-anxious groups did not differ on verbal, numerical, and diagrammatic intelligence. In summary, students with high scores on test anxiety seemed disadvantaged by the classical assessment procedure of reading comprehension in a multiple-choice format, while the alternative assessment outcomes with the comprehension monitoring task were far less sensitive to test anxiety and prior knowledge. According to Shepard (19921, classical standardized testing often leads to different outcomes and conclusions than alternative, performance-based assessment. Present results support the statement of Shepard. The polytomous scoring ~ r o c e d u r egenerated more adequate information because it can

INDMDUAL DIFFERENCES IN TEXT COMPREHENSION

175

provide test anxious and 'novice' students more 'true' information about their learning comprehension processes and progress. The outstanding results with the polytomous scoring procedure (even regardless of the content of comprehension monitoring) can be ascribed to the fact that both the inferential decision making component and the evaluation component ('feelmg of knowing') are taken into account. Allowing a 'feeling of knowing' component into the assessment of knowing seems to neutrabze the interaction between (debhtating) anxiety and (low) prior knowledge. Students of equal intellectual ability seem not so penalized or worried by this metacognitiveoriented assessment of reading comprehension than by- the classical reading comprehension with a multiple-choice response format. Replication and cross-validation of this findmg is, however, strongly requested. The outcomes on a classical readmg comprehension task were affected by prior knowledge and by the interaction of test anxiety and prior knowledge; contrary, the outcomes on a comprehension monitoring task were not substantially affected by prior knowledge and test anxiety. These findings seem congruent with traditional versus metacognitive training outcomes. Hasselhorn and Korkel (1986) emphasized that metacognitive training a m d o rates comprehension performance of texts, especially when texts of unknown and unfamihar contents are presented. The more traditional training procedures fostered the activation and reflective use of prior knowledge relevant to the content of the text. Their causal modeling results indicated that the metacognitive reading instruction led to considerable improvement in comprehension for novices. On the contrary, experts benefited the most from traditional reading instructions. The findings also call attention to the fact that a classical assessment of text comprehension may not be equally beneficial for all students. Especially for students with a combination of low prior knowledge and high test anxiety, disguising or detrimental effects of these students' characteristics on the actual test performance should be taken into account. These observed test scores might differ substantially from their true abhty scores. Uncritical assessment and classification of students' text comprehension performance might further decrease their performance. Comprehension monitoring tasks that account for both the inferential decision-making component and the f e e h g of knowing component seem to add to assessment procedures in educational practice. The latter kind of assessment might reduce the gap between assessment, learning, and instruction (see Dochy, Moerkerke, & Martens, 1996) because it bears resemblance to ordmary learning situations (ecological validity), it respects and enhances students' self-regulated readingcomprehension behavior, and it provides feedback on what students know and do not know. Besides, those who counsel test-anxious students on testtaking behavior should be especially cautious in identifying their prior knowl-

176

A. E. MINNAERT

edge state to differentiate appropriately between observed and true scores. Furthermore, the polytomous scoring technique unveils overconfident, unmindful students who are too certain of the answers they presumed to be correct (but in fact were not). Early confrontation with this type of assessment in classroom practice might lead to a more appropriate self-estimation of one's own capabdities and to a more efficient way of monitoring one's own learning process and progress. It is demonstrated that highly anxious students may spend more time studying to reduce the negative effect of evaluative anxiety on performance (Naveh-Benjamin, et a/., 1981; Schneider, 1989). The question we address here is whether researchers, assessment specialists, and counselors take domain-specific prior knowledge into consideration in addition to test anxiety. Researchers already pointed out that a large, well-organized, and flexibly accessible domain-specific knowledge base has a powerful influence on learning processes and products (see, e.g., De Corte, 1995; Dochy & Alexander, 1995). Nevertheless, the effect of the interaction between prior knowledge and test anxiety on learning processes and assessment outcomes seems often forgotten or unintentionally neglected. REFERENCES BAKER. L. (19893 Metacognition, comprehension monitoring, and the adult reader. Edztcatio~tal Psychology Review, 1, 3-38. BIRENBAUM, M., & P I N K UP.. (1997) Effects of rest anxiety, information org'~n~zation, and testing situation on performance on two test formats. Conte~nporayEdrlta/ro~zalPsychology, 22, 23-38. BOEKAERTS, M. (1994) Action control: how relevant is it for classroom learning? In J. Kuhl & J. Beckmann (Eds.), Volitiort a~zdperso~zality:action versars state orientation. Seattle, WA: 427 435. Hogrefe & Huber. BOEKAERTS, M. (1997) Seu.rerul~tedlearning: a new concept embraced by researchers, policy makers, educators, reacher\, and students. Learrzirtg and Instn,ction, 7 , 161-186. CROCKER, L., &SCHMIT, A. (1987) Improving multiple-choice test performance for examinees with diFferent levels of test anxiety. Jozrrnal of Experirneizfal Edtrcation, 55, 201-205. CULLER. R. E., & HOLAHAN, C. J. (1980) Test anxiety and academic performance: the effects of study-related behaviors. Jozrrnal of Edz~catiortalPsychology, 72, 16-20. DE CORTE.E. (1995) Fostering cognitive growth: a perspective from research on mathematics learning and instruction. Ed~rcationalP~ychologist,30, 37-46. DOCHY,F. J. R. C., &ALEXANDER, P. A. (1995) Mapping p r ~ o rknowledge: a framework for discussion among researchers. Etrropearz Jortrnal of Prycholog~'oJEdzrca/io~~, 10, 225-242. DOCHY,F. J. R. C., MOERKERKE, G., &MARTENS, R. (1996) htegr,~tingassessmenr, learning and instruction: assessment of domain-specific and doma~n-transcendingprior knowledge and progress. S~trdiesin Edzrcatio~zalEvalzratiort. 22, 309-339. GLASER. R. (1984) Education and thinking: the role of knowledge. A~nericanPsychologist, 39, 93-104. HAGNET, K. H., & BENSON. J. (1997) The motive to avoid failure and test anxiety responses: empirical support for integration of two research traditions. Anxiety, Stress and Coping: a~zI~ztenzational]ozrrnal,10, 35-57. 1. (1986) Metacognitive versus traditional reading instructions: the HASSELHORN. 114 & KORKEL, m e d ~ ~ t ~role n y of domain-specific knowledge on children's text-processing. Huiizat~Learning: jorirttoi of Practical Research and Applzcations, 5, 75-90. KINNUNEN, R., &VAURAS,M. (1995) Comprehension monitoring and the level of comprehension in high- and low-achieving primary school children's reading. Leanzi~zgand 117strtrction, 5, 143-165.

INDIVIDUAL DEFERENCES IN TEXT COMPREHENSION

177

K ~ R K EJL.,, & SCHNEIDER. W (1991) Domain-specific versus meracognitive knowledge effects on text recall and comprehension. In M. Carretero, M., Pope, R-J. Simons, & J. Pozo (Eds.), Learning and instnrction: Ezrropean research in art international context. Vol. 3. Oxford, UK: Pergamon. Pp. 311-323. KUHL,J. (1983) Motiuatio~t,Kortflik~und Handlzrrrgskontrolle. Heidelberg Springer-Verlag. KUHL,1. (1994) Action versus state orientation: psychometric properties of the Action Conrrol Scale (ACS-90). In J. Kuhl & J. Beckmann (Eds.), Volition and personality: acfiott verszis state orientation. Seattle, WA: Hogrefe & Huber. Pp. 47-59. MEHRABIAN, A,, &BANK,L. (1978) A questionnaire measure of individual differences in achieving tendency. Edz,cational and Psychological Measurement, 38, 475-478. MINNAERT, A. (1996) Academic performance, co nition, metacognition and motivation. Assessing freshmen characteristics on task: a valifation and replication study in Higher Education. Unpublished doctoral dissertation, Univer. of Leuven, Belgium. MINNAERT, A. (1996) Can metacognition compensate for intelligence in the first year of Belgian higher education? Psychologica Belgica, 36, 227-244. MINNAERT, A,, & JANSSEN, P J. (1996) How general are the effects of don~ain.spec~hc prior knowledge on study expertise as compared to eneral chinkm s k d s ) In M 131renbaum & F. ~ o c h y(EGIS.), Alternatives in assessment opachieuernents, L'eortrlng procerler arid prior knowledge. Boston, M A : Kluwer Academic. Pp. 265-281. MITSUDA. M. (1990) The development of text comprehension monitoring activides. Japanese Psychological Research, 32, 55-65. NAVEH-BENJAMIN, M., M C ~ C H I W. E , J.. LIN,Y-L., &HOLINCER, D. P. (1981) Test anxiety: deficits in information processing. Jortrrral of Edrrcatior~alPsychology, 73, 816-824. OTERO. CAMPANARIO. J. M.. &HOPKINS.K. D. (1992) The relationship between academic ac levement and metacognitive com rehension monitoring ability of Spanish secondary Measurement, 52, 419-430. school students. Edrtcational and ~r~%ological PINTRICH, I? R., & DE GROOT,E. V. (1990) Motivational and self-re ulated learning components of classroom academic performance. ]ozrrrtal of ~ d r r c a t i ~ r ~ o l % s ~ c82, h o l33-40. o~ PINTRICH, P. R,ROE~ER, R. W., & D E GROOT.E. A. M. (1994) Classroom and individual differences in early adolescents' motivation and self-regulated learning. Jourrzal of Early Adolescence, 14, 139-161. RAND, I?, LENS,W , & DECOCK,B. (1991) Negadve motivation is half the sto achievement motivation combines positive and negative motivation. Scandinauian Journa7~f~drrcationol Research, 35, 13-30. SARASON, 1. G. 11984) Stress, anxiety and cognitive intederence: reaction to tests. ]ozrmal of Persotzalitv a17d Social Psychology, 46, 929-938. SCHNEIDER, K. (19891 Motivation and performance. Paper presented at the First European Congress ot Psychology, Amsterdam, The Netherlands. SHEPARD, L. A. (1992) What policy makers who mandate tests should know about the new ps chology of intellectual ability and learning. In B. R. Gifford & M. C. O'Connor (Eds.), clanging assessments: alternatiue viezus of aptitzrde, achievement and L~s~rrrction. Boston, M A : Kluwer Academic. Pp. 301-328. SNOW,R. E. (1993) Construct val~dttyand consrmcted-response tests. In R. E. Bennett & W. C. Ward (Eds.), Constnrction versus choice in cognitive measurement: isszres in constrzrcted response, performance testing, and porl/olio assessment. Hillsdale, NJ: Erlbaum. Pp. 45-60. THISSEN, D.,&STEINBERG, L. (1986) A taxonomy of item response models. Psychometrico, 51,

k:

-567-577 - . . .. TOBIAS.S., &EVERSON. H. T. (1997) Stud ing the relationship benveen affective and meracognitive variables. Anxiety, Stress and A p i n g : An lnterna~ionalJournal, 10, 59-81. VERHELST, N. D. (1992) Het eenparameter logistische model IOPWM): eetr theoretische inleiding en een harrdleidittg bij he&computerprogratnma [The one-parameter lo istic model (OPLM): a theoretical introdrrction and a manzral to the computer program]. ~ O P DMemorandum 92-3) Arnhem, The Netherlands: The National Insticure for Educational Measurement. WAGONER, S. A. (1983) Comprehension monitoring: what ir is and what we know about it? Reading Research Quarterly, 18, 328-346. WI-IER, B. (1972) Test anxiety and study habits. The Jozrntal of Edzrcatio~~al Research, 65, 352-354.

Accepted December 23, 1998.