Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Learning and Individual Differences 24 (2013) 198–203

Contents lists available at SciVerse ScienceDirect

Learning and Individual Differences


journal homepage: www.elsevier.com/locate/lindif

Test anxiety in written and oral examinations


Jörn R. Sparfeldt a,⁎, Detlef H. Rost b, 1, Ulrike M. Baumeister b, 1, Oliver Christ b, 1
a
Saarland University, Department of Educational Science, Campus A 5.4, D-66123 Saarbrücken, Germany
b
Philipps-Universität, Marburg, Germany, Fachbereich Psychologie, Gutenbergstraße 18, 35032 Marburg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The distinction of different test anxiety reactions (e.g., worry, emotionality) is well established. Recently, ad-
Received 26 August 2010 ditional relevance has been given to school-subject-specific test anxiety factors. The present study explored a
Received in revised form 6 October 2012 further aspect concerning the structure of test anxiety experiences, specifically oral versus written examina-
Accepted 22 December 2012
tion modes. A questionnaire was administered to 682 high school students (grades 9 to 10) in which three
aspects were systematically combined: different test anxiety factors (worry, emotionality), school-subjects
Keywords:
Test anxiety
(mathematics, German), and examination modes (written, oral). Confirmatory factor analyses supported a
School subjects structure with eight school-subject-specific and examination-mode-specific test anxiety factors. Differential
Written examination relationships revealed evidence for convergent and discriminant validity of these eight specific test anxiety
Oral examination factors with two test anxiety initiating conditions (recitation situations, lack of knowledge).
Construct validity © 2012 Elsevier Inc. All rights reserved.

1. Introduction Marsh, 1988; Marsh & Yeung, 1996). Most of these studies display
problems: They assessed general anxiety instead of TA, and/or measured
Test anxiety (TA) is a frequently researched topic (Cizek & Burg, TA with only one item, and/or did not take the multidimensionality of TA
2006) with ample evidence for its multidimensionality (Zeidner, into account. As an exception Sparfeldt et al. (2005) assessed three
1998). Whereas TA was usually assessed without referencing the corre- TA-factors in four school-subjects: worry, emotionality, and TA-tied
sponding content area, recently school-subject-specific assessments impairments of information processing. Model comparisons document-
have been emphasized (e.g., Marsh & Yeung, 1996; Sparfeldt, ed the importance of considering both aspects (TA-factor, school-
Schilling, Rost, Stelzl, & Peipert, 2005). One important differentiation— subject) simultaneously. Differential relationships between school-
focusing on examination modes—has been neglected. Therefore, the subject-specific TA-factors and grades evidenced convergent/
present study investigates school-subject-specific TA-factors in written discriminant validity. The various school-subjects comprise classes
(TA-written) and oral (TA-oral) examinations. of situations requiring partially different competencies; these differ-
Liebert and Morris (1967) distinguished worry (ruminating about ences seem to structure the TA-experience. Moreover, at least some
self-threatening consequences of failure) from emotionality (perceived antecedents and correlates of TA are tied to school-subjects (as compe-
physiological stress reactions). This two-dimensional TA-structure has tence beliefs; e.g., Marsh & Yeung, 1996), others are structured more
repeatedly been shown with correlations of .55≤r≤.76 (cf., Keith, general (e.g., maladaptive coping). Focusing on TA-assessment Zeidner
Hodapp, Schermelleh-Engel, & Moosbrugger, 2003). Nevertheless, worry (1998, p. 122) suggested a combination of—among others—response
and emotionality are phenomenologically distinct, correlate differentially modes (TA-factors) and evaluation situations (examination modes).
with achievement, and display differential growth-patterns depending on Written examinations and oral examinations are widespread examina-
the examination's temporal distance (Hembree, 1988; Seipp & Schwarzer, tion modes, not yet systematically considered in TA-research. Despite
1991). Additional TA-dimensions (e.g., self-preoccupation, interference) different subtypes, oral examinations in schools are usually more adap-
were suggested, not denying the distinction and usefulness of these tive, more communicative, less restricted, and with a stronger reference
two factors (cf. Zeidner, 1998, 2007). to the individual examinee than written examinations (Jäger, 2004).
TA has been usually assessed without referencing a specific From the examinee's perspective, oral examinations are characterized
content. Recently different school-subject-related TA were analyzed by the potential evaluation of an audience (at least one examiner).
simultaneously (e.g., Goetz, Frenzel, Pekrun, Hall, & Lüdtke, 2007; Some studies have investigated different aspects, correlates and
coping modes of TA-oral (e.g., Buchwald & Schwarzer, 2003; Huwe,
Hennig, & Netter, 1998; Krumpholz, 1993), but not considering
⁎ Corresponding author. Tel.: +49 681 30257490; fax: +49 681 30257488.
E-mail addresses: j.sparfeldt@mx.uni-saarland.de (J.R. Sparfeldt),
TA-written and TA-oral. Only Lukesch (1982) combined these two ex-
rost@staff.uni-marburg.de (D.H. Rost). amination modes and three school-subjects systematically, but with
1
Tel.: +49 6421 2821727; fax: +49 6421 2823910. only one item per school-subject-specific and examination-mode-

1041-6080/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.lindif.2012.12.010
J.R. Sparfeldt et al. / Learning and Individual Differences 24 (2013) 198–203 199

specific TA-reaction, not considering different TA-factors. Unfortunate- lack of knowledge)? Recitation situations refer particularly to oral ex-
ly, the corresponding intercorrelations were not reported. Czeschlik aminations. We expected higher correlations in recitation situations
(2008) focused on the relative frequencies in fourth-graders with 81% with TA-oral than TA-written factors. Furthermore, recitation situa-
being “not test-anxious”, 5% “anxious in written”, 5% “in oral”, and 9%
tions should correlate higher with emotionality than with worry. Re-
“in written and oral examinations”. In a high school sample, TA-oral
garding the correlation pattern of the lack of knowledge factors and
and TA-written correlated r = .58.
the recitation situations factors with the TA-factors, we expected nu-
Rost and Schermer (1992, 2007) differentiated classes of situations
that initiate TA in many students. Lack of knowledge refers to TA as merically higher correlations if the two factors referred to an identical
“caused by the realization that achievement demands cannot be met” examination-mode and/or school-subject than if one or both of these
(Rost & Schermer, 1989, p. 42), being closely related to academic aspects differed.
self-beliefs as self-concept and self-efficacy, thereby being tied to specific
context areas and potentially to examination modes. Lack of knowledge 2. Method
correlated r=.51 with worry and r=.48 with emotionality (Rost &
Schermer, 2007, p. 92). Recitation situations refer to TA-initiating situa- 2.1. Participants and Sampling
tions with a social focus, specifically the experience of TA as a “conse-
quence when the achievement has to be presented to other people and Participants were German academic-tracked high school ninth
when reactions which pose a threat to self-esteem are anticipated” and tenth graders (Gymnasium: prep school for university) attending
(Rost & Schermer, 1989, p. 43). Recitation situations correlated differen- 29 classes in five schools. Only 0.5% of the parents did not allow their
tially with worry (r=.18) and emotionality (r=.34; Rost & Schermer, child to participate; 14.5% of the students were absent due to reasons
2007, p. 92). The formulations of the corresponding items refer to oral unrelated to the study (e.g., illness). Data from N = 682 participants
presentations/contributions/talks in class. Therefore, particularly oral (mean age: 15.4 years, SD= 0.71; 406 female) were collected in April
examinations with a stronger social component (compared to written 2008 during regular lessons by a trained experimenter.
examinations) represent such TA-initiating situations.
The multidimensionality and the content-specificity of TA are well 2.2. Instruments
documented, yet examination modes have been neglected. Therefore,
studies that analyze systematically school-subject-specific TA-factors Emotionality and worry in written examinations were measured
in oral and written examinations are needed. This study pursues this with a German adaptation (TAI-G; Hodapp, 1991) of the well-
approach: established Test-Anxiety-Inventory (Spielberger, 1980). Each item
was complemented by a specification of the school-subjects “mathe-
(1) Factorial validity. Is it possible to separate the two central and matics” (numerical domain) and “German” (verbal domain; native lan-
sample-selected TA-factors (worry, emotionality) in mathematics guage) that differed in demands on learners, didactics, and testing
and German (native language) related to oral and written examina- methods. The items were arranged in a table form (grid). Each item
tions using confirmatory factor analyses (CFA)? We expected an ade- formed one row and was presented with a placeholder (…) for the
quate model fit with eight different TA-factors resulting from a school-subject specification. Each school-subject formed one column.
The students were expected to fill out the placeholder (…) mentally
systematic combination of two TA-factors, two school-subjects, and
with the school-subject from the corresponding column. Regarding
two examination modes. This model should fit better than more par-
the presentation mode, an experimental study revealed very similar
simonious models. The correlations of the school-subject-specific and psychometric properties of a multidimensional academic self-concept
examination-mode-specific TA-factors are expected to be numerically inventory with the items presented (a) in a grid and (b) in the usual
higher if two of the three dimensions of the corresponding factors are questionnaire presentation mode (Sparfeldt, Schilling, Rost, & Thiel,
identical (e.g., worry in mathematics in written examinations and worry 2006). TA-oral was assessed with identical items presented on a sepa-
in mathematics in oral examinations refer to the identical TA-factor and rate page in an analogue manner. Thereby, each participant answered
the same school-subject, but to different examination modes) than if each item of each TA-factor separately for both school-subjects and
only one of the three dimensions is identical; the correlations should both examination-modes (Table 1). The use of identical item stems
also be numerically higher if one rather than none of these dimensions minimized variance due to different operationalizations.
The items to assess TA-initiating conditions were taken from anoth-
is identical. According to Sparfeldt et al. (2005), the correlations of the
er well-established German TA-inventory (DAI; Rost & Schermer, 2007)
worry-factors should be similar or slightly higher than those of the
and adapted to measure TA-initiating conditions in mathematics and
emotionality-factors. For the other two dimensions (school-subject,
German. They were also presented in a grid. Regarding recitation situa-
examination-mode), no such deduction can be made. tions, the questionnaire contained the eight slightly adapted items of
(2) Criterion-related validity. Are there differential relationships to be the DAI-short form which explicitly formulate different types of oral
found among school-subject-specific and examination-mode-specific presentations and examinations. Regarding lack of knowledge, oral
TA-factors with specific TA-initiating conditions (recitation situations, and written examinations in mathematics and German could refer

Table 1
Instruments used to assess school-subject-specific and examination mode-specific test anxiety in oral and written examinations.

Written examinations Oral examinations

Introductory phrase “In written examinations in the school subject … “In oral examinations in the school subject …
[Mathematics]/[German] …” [Mathematics]/[German] …”
Worry: item examples (item number); source “… I worry about my results”, “… I find myself thinking about the consequences of failing” (7 items)a; TAI-G (Hodapp, 1991)
Emotionality: item examples (item number); source “… I feel uneasy”, “… my muscles are very tight” (8 items); TAI-G (Hodapp, 1991)

Notes. All items were answered on a six-point scale ranging from 1 (not at all true) to 6 (perfectly true). Pilot studies had shown that six-point scales were appropriate.
a
In accordance with the greater heterogeneity of worry (e.g., Hodapp, 1991), preliminary analyses with the ten more general worry items from the TAI-G revealed that three
items showed inconsistent loadings and were therefore dropped.
200 J.R. Sparfeldt et al. / Learning and Individual Differences 24 (2013) 198–203

theoretically to such situations. Therefore, the questionnaire contained Table 3


those five of the eight item stems that were semantically adaptable to Fit indices of the different models analyzing the structure of test anxiety (see text
for details).
both examination-modes (Table 2).
χ2 df TLI CFI RMSEA SRMR
2.3. Analyses M1 8 school subject-specific 3285.90 1592 .943 .949 .039 .047
and examination mode-
A maximum-likelihood (ML)-CFA (program Mplus 5; Muthén specific test anxiety facet
factors
& Muthén, 1998–2007) was run to test an eight-factor-model
M2 1 broad test anxiety factor 13,631.46 1620 .601 .635 .104 .271
corresponding with the eight school-subject-specific and examination- M3 2 test anxiety facet factors 12,661.37 1619 .633 .664 .100 .197
mode-specific TA-factors (M 1). Each TA-factor was indicated by the cor- M4 2 subject-specific factors 9709.98 1619 .731 .754 .086 .137
responding items. Measurement errors of the items with identical stems M5 2 examination 11,791.85 1619 .662 .691 .096 .227
were allowed to correlate. M 1 was compared to more parsimonious mode-specific factors
M6 4 subject-specific test 7600.93 1614 .800 .818 .074 .088
alternative models (Table 3). To describe the model fit, Chi2 with df
anxiety facet factors
were supplemented by TLI, CFI, RMSEA, and SRMR. RMSEA values M7 4 examination 10,193.13 1614 .714 .739 .088 .127
below .08 (.05) typically signify an acceptable (good) fit (e.g., Byrne, mode-specific test anxiety
1998). An acceptable model fit is indicated if TLI exceeds .90 (Marsh, facet factors
M8 4 subject-specific and 6017.98 1614 .853 .866 .063 .122
Craven, & Debus, 1999). Similar values were recommended for the CFI.
examination mode-specific
Admittedly, these values are more or less accepted guidelines. The corre- factors
lational pattern of the eight factors was inspected for evidence of conver-
Notes. Comparisons of the fit of the different models with the chi-square difference test
gent/divergent validity. for nested model comparisons always revealed a statistically significant better fit of
Investigating the second question, an analogous CFA was run in M 1 than M 2 to M 8 (pb .01).
which M 1 was supplemented by six factors corresponding to the six ini-
tiating conditions (two school-subject-specific reciting situations fac-
tors, four school-subject-specific and examination-mode-specific lack distribution, although the bias in the fit indices (RMSEA, SRMR)
of knowledge factors). The pattern of the correlations was inspected. was much smaller (Appendix 2). The bias in the estimates for the cor-
The Full-Information-ML-Estimation-algorithm was used to handle relations (Mpercentage of bias = −0.01, SD= 0.12, Min= −0.23, Max=
the few missing values (maximum per variable: 4%). Univariate normal- 0.29) and their standard errors (Mpercentage of bias = −0.98, SD= 0.65,
ity was not severely violated (skewness: –0.99 to 1.45, Median=0.29; Min= −2.03, Max= 0.00) was small. Overall, the model size effect is
kurtosis: –1.23 to 1.61; Median= −0.81). Such values are considered small in size indicating the relative robustness of our conclusions.
not to undermine the robustness of ML-estimation (e.g., Finney & The correlations of the eight M 1-factors (Table 4) were positive
DiStefano, 2006; Muthén & Kaplan, 1985). To control for potential ef- and mainly of small effect size (cf. Cohen, 1992), if neither the
fects due to the school class affiliation and the clustering of the data, TA-facet, nor the school-subject, nor the examination mode were
analyses were run using the method Complex in Mplus (with group- identical (.18 ≤ r ≤ .32). The correlations were mostly numerically
mean-centered variables). higher and mainly of medium effect size when at least one of the
three aspects was identical (.28 ≤ r ≤ .61). Most correlations were nu-
3. Results merically even higher and mainly of large effect size when two of the
three aspects were identical (.48 ≤ r ≤ .81).
3.1. School-subject-specific and examination-mode-specific TA In order to evaluate the relevance of these three aspects, we kept
two of these factors constant and inspected the correlation coefficients
M 1 revealed an acceptable/good fit (Table 3; Appendix 1). M 2 of the third varying aspect. Regarding worry and emotionality, the cor-
(one broad TA-factor) showed a poor fit. Considering just one of the relations reached .48 ≤ r ≤ .59 (medium/large effect) if both TA-factors
three aspects (M 3 to M 5)—neglecting the other two aspects— were related to the same school-subject and to the same
resulted in poor fits. M 6 to M 8 (considering two of the three examination-mode, but to different TA-factors. Regarding the two
aspects) revealed a substantial misfit. Thus a structure that comprised school-subjects, the coefficients were numerically higher (.59≤ r ≤ .81;
all three aspects was supported. large effect) when both factors were related to the same TA-facet and
The complexity of these models might cause a bias in fit sta- to the same examination-mode, but to different school-subjects.
tistics, parameter estimates and standard errors (e.g., Herzog, Concerning the two examination-modes, the correlations were in
Boomsma, & Reinecke, 2007). Therefore, we conducted a Monte- about the same range (.63 ≤ r ≤ .73; large effect) if both factors
Carlo-simulation for M 1 (see Muthén & Muthén, 2002) with the pa- were related to the same school-subject and to the same TA-facet,
rameter values estimated from our data serving as population pa- but to different examination-modes.
rameter values. Based on 10,000 replications, we inspected the The six correlation coefficients related to worry (.52 ≤ r ≤ .81;
approximation of the observed with the theoretical chi-square dis- large effect) were numerically slightly higher than those related to
tribution (as for RMSEA and SRMR) and calculated the percentage of emotionality (.38 ≤ r ≤ .76; medium/large effect). The coefficients
bias in the parameter estimates and standard errors for the factor for mathematics (.39 ≤ r ≤ .72) were either very similar to those for
correlations. Results showed a bias in the observed chi-square German (.31 ≤ r ≤ .73) or slightly higher (medium/large effect),

Table 2
Instruments used to assess initiating conditions of test anxiety.

Introductory phrase “I am afraid in… [Mathematics]/[German] …”


Recitation situations: “… to recite something in front of the class” (8 items); DAI (Rost & Schermer, 2007)
item examples
(item number); source
Lack of knowledge: item examples Oral: “… if I remember only parts of the subject matter while Written: “…if I remember only parts of the subject matter
(item number); source giving a presentation” (5 items); DAI (Rost & Schermer, 2007) while taking a test” (5 items); DAI (Rost & Schermer, 2007)

Notes. All items were answered on a six-point scale ranging from 1 (not at all true) to 6 (perfectly true). Pilot studies had shown that six-point scales were appropriate.
J.R. Sparfeldt et al. / Learning and Individual Differences 24 (2013) 198–203 201

Table 4 Table 6
Latent correlations of the school-subject-specific (mathematics, German) and examination- Latent correlations of the school-subject-specific (mathematics, German) and examination-
mode-specific (oral, written) test anxiety factors (worry [WOR], emotionality [EMO]) of M 1 mode-specific (oral, written) test anxiety factors (worry [WOR], emotionality [EMO]) with
(see Table 3). the initiating conditions of test anxiety (recitation situations, lack of knowledge)
(see Table 5).
Mathematics German
Mathematics German
Oral Written Oral Written
Oral Written Oral Written
WOR EMO WOR EMO WOR EMO WOR EMO
WOR EMO WOR EMO WOR EMO WOR EMO
Mathematics
Oral EMO .59 Recitation situations
Written WOR .72 .39 Mathematics .53 .78 .38 .50 .41 .67 .30 .44
Written EMO .49 .65 .58 German .38 .60 .23 .29 .50 .80 .36 .51
German Lack of knowledge
Oral WOR .81 .40 .52 .28 Mathematics oral .68 .66 .57 .55 .56 .50 .46 .39
Oral EMO .37 .76 .18 .38 .53 Mathematics written .58 .45 .70 .61 .41 .27 .47 .36
Written WOR .61 .22 .74 .30 .73 .31 German oral .53 .47 .38 .32 .67 .61 .55 .49
Written EMO .32 .46 .28 .59 .46 .63 .48 German written .46 .30 .48 .36 .57 .42 .62 .49

Notes. All coefficients are statistically significant (p b .001). Notes. All coefficients are statistically significant (p b .001).
Correlation of the two recitation situations factors (mathematics–German): r = .84.
Correlations of the four lack of knowledge factors: r (German oral–German written)=.88; r
and the coefficients revealed similar to numerically slightly higher (German oral–mathematics written)=.62; r (German oral–mathematics oral)=.83; r
values for oral (.37 ≤ r ≤ .81) over written (.28 ≤ r ≤ .74) examina- (mathematics oral–German written)=.73; r (mathematics oral–mathematics written)=
tions (medium/large effect). .85; r (mathematics written–German written)=.74.
Correlations of the two recitation situations factors (mathematics/German) with the four
lack of knowledge factors: r (German oral)=.57/.67; r (German written)=.38/.46; r
3.2. Correlations with TA initiating conditions (mathematics oral)=.69/.56; r (mathematics written)=.48/.33.

The ML-CFA revealed an acceptable model fit (Table 5; Appendix 3).


In a Monte-Carlo-simulation (10,000 replications) the bias in the subject-specific and examination-mode-specific TA-factors can be sep-
chi-square fit statistic was substantial, but again much smaller for arated. Differential correlations of these TA-factors to the TA-initiating
RMSEA and SRMR (Appendix 2). The bias in parameter estimates of conditions displayed criterion-related validity. The results provide
the factor correlations (Mpercentage of bias = − 0.13, SD = 0.09, empirical support to a multi-facet diagnostic approach considering
Min = − 0.35, Max = 0.12) and standard errors (Mpercentage of bias = school-subject-specific and examination-mode-specific TA-factors. The
− 0.01, SD = 0.56, Min = − 1.23, Max = 1.40) was small. Overall, established multidimensionality of TA-reactions holds also true for a
the model size effect was small in size. content-specific perspective, such as different school-subjects. Elabo-
The factor-correlations of this model showed a systematic pattern rating on these former results and supporting our first expectation,
(Table 6): The convergent correlations of a TA-facet with the recitation sit- the CFA showed—in line with Zeidner (1998)—the additional separa-
uations in the same school-subject were numerically higher than the cor- tion of different examination modes.
responding divergent correlations (i.e., the correlations of a TA-facet with Concerning school-subject-specific and examination-mode-specific
the recitation situations in the other school-subject). Recitation situations TA-factors the findings are in line with expectations. All correlations
in one school-subject correlated numerically higher with TA-factors were positive and the magnitude of the coefficients varied systematical-
corresponding to oral examinations of the same school-subject than ly in a psychologically meaningful way: The more identical the dimen-
with TA-factors corresponding to written examinations of the same sions, the higher the correlations. However, the magnitude of some
school-subject. Comparing the correlations of recitation situations with coefficients was substantial (up to r = .81); latent coefficients of this
(a) the worry factor and (b) the corresponding emotionality factor, the size do not rule out that the different factors are distinct empirical
correlations were numerically higher for emotionality than for worry. constructs.
The correlations of the TA-factors with lack of knowledge revealed Concerning the relations within both school-subjects, the pattern of
also a specific pattern: The convergent correlations corresponding to the school-subject-specific and examination-mode-specific TA-factors
the same school-subject were numerically higher than the correspond- is very similar: (a) These six coefficients within mathematics and
ing divergent correlations corresponding to different school-subjects. German correlated substantially (r = .97), (b) the means of these six co-
The convergent correlations corresponding to the same school-subject efficients in mathematics and German were similar (r = .58 and r = .54,
and the same examination-mode were numerically higher than the cor- respectively). Therefore, examination-mode-specific TA-factors are dif-
responding divergent coefficients corresponding to different school- ferentiated along the boundaries of different school-subjects, but the
subjects and/or different examination-modes. The relationships were pattern of the relations seems to be similar. Whether this applies to
mostly numerically slightly higher for worry than emotionality. other school-subjects, it remains to be investigated. The pattern of the
correlations with initiating conditions of test anxiety is also in line
4. Discussion with the expectations, evidencing criterion-related validity and, there-
by, the theoretical and practical relevance of our approach: The correla-
Our study considered different examination modes (oral, written) tions are substantial and numerically higher if they refer to identical
while assessing TA. CFA-results evidenced that the eight school- school-subjects (recitation situations), or identical school-subjects and
examination-modes (lack of knowledge) than if they refer to different
Table 5 examination-modes and/or school-subjects. TA-initiating conditions—
Fit indices of the model analyzing the structure of test anxiety (see text for details).
albeit close to TA—exhibit some conceptual distance to TA. Future re-
χ2 df TLI CFI RMSEA SRMR search should add further criterion-related evidence.
8 school-subject-specific and 7073.27 4245 .946 .950 .031 .045 The results are restricted to the chosen operationalizations and
examination mode-specific test sample. Replications are needed. In future studies, additional school-
anxiety facet factors supplemented subjects should be added to investigate a potential domain-formation
by the 6 factors of initiating of TA (comparable to academic self-concepts; see e.g. Marsh, 1992)
conditions of test anxiety
or content-specific factors like “numerical” or “verbal”—analogous to
202 J.R. Sparfeldt et al. / Learning and Individual Differences 24 (2013) 198–203

intelligence factors. Similarly, the sample-selected TA-factors and ex- Appendix 3. Factor loadings (standardized regression coefficients)
amination modes should be extended. The genesis of the different of the school-subject-specific (mathematics [MAT], German [GER]),
TA-factors (and, e.g., the interplay with other variables) cannot be an- examination-mode-specific (oral, written) test anxiety factor
swered by cross-sectional studies. Longitudinal and experimental stud- items (worry, emotionality) and of the test anxiety initiating con-
ies are needed. Additionally, a person-centered approach could dition items (recitation situations, lack of knowledge; Table 5)
supplement the realized variable-centered approach.
In our study, a multi-facet structure of the TA-experience was
shown. General TA-measures seem to mask specific, but systematic
and relevant aspects of the experience of TA. As students' TA differs
according to the examination-mode the same holds probably true
Worry Emotionality Lack of Recitation
for educational/psychological interventions: Whereas some strategies knowledge situations
are largely independent of the examination-mode (e.g., content-
Oral Written Oral Written Oral Written
specific knowledge), others should vary. For example, a behavioral
approach in which oral examinations are simulated in advance in MAT/ MAT/ MAT/ MAT/ MAT/ MAT/ MAT/
GER GER GER GER GER GER GER
order to optimize communication skills should be beneficial for oral
examinations, whereas writing skills should be optimized prior to Item 1 .53/.49 .72/.67 .77/.76 .83/.80 .77/.77 .75/.74 .67/.77
Item 2 .46/.45 .53/.47 .79/.79 .80/.72 . 79/.77 .79/.79 .68/.71
written examinations. It would be interesting to connect the pro-
Item 3 .60/.58 .72/.67 .80/.80 .84/.79 .78/.80 .87/.86 .85/.85
posed diagnostic approach more tightly to interventions. But al- Item 4 .57/.54 .71/.66 .81/.79 .80/.73 .84/.85 .80/.77 .77/.78
though TA-interventions were quite effective in general—especially Item 5 .85/.84 .87/.83 .82/.79 .80/.73 .60/.60 .79/.76 .70/.70
those, that combine cognitive or behavioral approaches with skill- Item 6 .78/.73 .83/.76 .88/.84 .86/.83 .72/.71
focused approaches (cf. Ergene, 2003; Hembree, 1988; Zeidner, Item 7 .80/.77 .86/.80 .83/.80 .83/.78 .81/.82
Item 8 .87/.84 .89/.81 .78/.76
1998)—further research is needed to support the suggested differen-
tial treatment outcomes, especially in school-aged students.

Appendix 1. Factor loadings (standardized regression coefficients) References


of the school-subject-specific (mathematics [MAT], German [GER])
Buchwald, P., & Schwarzer, C. (2003). The exam-specific approach to coping scale and
and examination-mode-specific (oral, written) test anxiety factor interpersonal resources. Anxiety, Stress, and Coping, 16, 281–291.
items (worry, emotionality) in M 1 (Table 3) Byrne, B. M. (1998). Structural equation modeling with LISREL, PRELIS, and SIMPLIS: Basic
concepts, applications and programming. Mahwah, NJ: Erlbaum.
Cizek, G. J., & Burg, S. S. (2006). Addressing test anxiety in a high-stakes environment:
Strategies for classrooms and schools. Thousand Oaks, CA: Corwin Press.
Worry Emotionality Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.
Czeschlik, T. (2008). Umgang mit ängstlichen Schülern [Handling of anxious students].
Oral Written Oral Written
In M. K. W. Schweer (Ed.), Lehrer-Schüler-Interaktion [Teacher student interaction]
MAT/GER MAT/GER MAT/GER MAT/GER (pp. 343–360) (2nd. ed.). Wiesbaden, Germany: VS Verlag für 394 Sozialwissenschaften.
Ergene, T. (2003). Effective interventions on test anxiety reduction. School Psychology
Item 1 .53/.49 .72/.67 .77/.76 .83/.80 International, 24, 313–328.
Item 2 .46/.46 .53/.47 .79/.79 .80/.72 Finney, S. J., & DiStefano, C. (2006). Non-normal and categorical data in structural
Item 3 .60/.58 .72/.67 .80/.80 .84/.78 equation modeling. In G. R. Hancock, & R. O. Mueller (Eds.), Structural equation
Item 4 .57/.54 .72/.67 .81/.79 .80/.73 modeling: A second course (pp. 269–315). Greenwich, CT, NJ: Information Age
Item 5 .85/.83 .87/.83 .82/.79 .80/.73 Publishing.
Item 6 .78/.73 .83/.76 .87/.84 .86/.83 Goetz, T., Frenzel, A. C., Pekrun, R., Hall, N. C., & Lüdtke, O. (2007). Between- and within-
Item 7 .80/.77 .86/.80 .82/.79 .84/.78 domain relations of students' academic emotions. Journal of Educational Psychology,
Item 8 .87/.84 .89/.81 99, 715–733.
Hembree, R. (1988). Correlates, causes, effects, and treatment of test anxiety. Review of
Educational Research, 58, 47–77.
Herzog, W., Boomsma, A., & Reinecke, S. (2007). The model-size effect on traditional
Appendix 2. Expected and observed proportions for chi-square, and modified tests of covariance structures. Structural Equation Modeling, 14,
RMSEA, and SRMR for M 1 (model with 8 school subject-specific and 361–390.
Hodapp, V. (1991). Das Prüfungsängstlichkeitsinventar TAI-G: Eine erweiterte und
examination mode-specific test anxiety facet factors; Tables 3, 4) and modifizierte Version mit vier Komponenten [The Test Anxiety Inventory TAI-G: An
M-initiating (M 1 supplemented by the 6 factors of initiating condi- expanded and modified version with four components]. Zeitschrift für Pädagogische
tions of test anxiety; Tables 5, 6) based on Monte-Carlo-simulations Psychologie, 5, 121–130.
Huwe, S., Hennig, J., & Netter, P. (1998). Biological, emotional, behavioral, and coping
(see text for details)
reactions to examination stress in high and low stated anxious subjects. Anxiety,
Stress, and Coping, 11, 47–65.
Jäger, R. S. (2004). Von der Beobachtung zur Notengebung. Ein Lehrbuch (5. Aufl.)
[From observation to grades. A textbook. Fifth Ed.]. Landau, Germany: Empirische
Pädagogik.
Expected proportions M1 M-initiating Keith, N., Hodapp, V., Schermelleh-Engel, K., & Moosbrugger, H. (2003). Cross-sectional
χ2 RMSEA SRMR χ2 RMSEA SRMR and longitudinal confirmatory factor models for the German Test Anxiety Inventory:
A construct validation. Anxiety, Stress, and Coping, 16, 251–270.
Observed proportions Krumpholz, D. (1993). Kognitionen von Studierenden während einer mündlichen
Prüfung in Abhängigkeit von Prüfungsängstlichkeit, Geschlecht und Studienfach
0.990 0.999 1.000 0.996 1.000 0.976 0.994
[Cognitions of students during oral examinations depending on test anxiety, sex
0.980 0.998 1.000 0.988 1.000 0.966 0.986 and subjects]. Zeitschrift für Differentielle und Diagnostische Psychologie, 14, 177–188.
0.950 0.996 0.843 0.963 1.000 0.943 0.957 Liebert, R. M., & Morris, L. W. (1967). Cognitive and emotional components of test anx-
0.900 0.990 0.831 0.907 1.000 0.905 0.908 iety. Psychological Reports, 20, 975–978.
0.800 0.968 0.789 0.797 1.000 0.829 0.798 Lukesch, H. (1982). Fachspezifische Prüfungsängste. Eine deskriptive Analyse der
0.700 0.941 0.729 0.687 0.999 0.741 0.692 schulsystem- und schulartbezogenen Verbreitung der Ängste von Schülern vor
0.500 0.847 0.570 0.478 0.994 0.547 0.486 mündlichen und schriftlichen Prüfungen [Subject-specific examination anxieties:
0.300 0.690 0.348 0.286 0.977 0.312 0.288 A descriptive analysis of the school system and school type related dissemination
0.200 0.568 0.216 0.193 0.955 0.189 0.196 of anxieties in the presence of oral and written examinations]. Psychologie in
0.100 0.395 0.076 0.104 0.897 0.074 0.102 Erziehung und Unterricht, 29, 257–267.
0.050 0.264 0.021 0.056 0.819 0.022 0.056 Marsh, H. W. (1988). The content specificity of math and English anxieties: The High
0.020 0.150 0.002 0.026 0.693 0.004 0.028 School and Beyond study. Anxiety Research, 1, 137–149.
Marsh, H. W. (1992). Content specificity of relations between academic achievement
0.010 0.096 0.000 0.017 0.599 0.001 0.016
and academic self-concept. Journal of Educational Psychology, 84, 35–42.
J.R. Sparfeldt et al. / Learning and Individual Differences 24 (2013) 198–203 203

Marsh, H. W., Craven, R., & Debus, R. (1999). Separation of competency and affect Johnsen (Eds.), Advances in Test Anxiety Research, Vol. 7. (pp. 114–129). Amsterdam/
components of multiple dimensions of academic self-concept: A developmental Lisse: Swets & Zeitlinger.
perspective. Merrill-Palmer Quarterly, 45, 567–601. Rost, D. H., & Schermer, F. J. (2007). Differentielles Leistungsangst Inventar (DAI) [Differen-
Marsh, H. W., & Yeung, A. S. (1996). The distinctiveness of affect in specific school tial test anxiety inventory] (2nd ed.). Frankfurt a.M., Germany: Harcourt Test Services.
subjects: An application of confirmatory factor analysis with the National Seipp, B., & Schwarzer, C. (1991). Angst und Leistung – Eine Meta-Analyse empirischer
Educational Longitudinal Study of 1988. American Educational Research Journal, 33, Befunde [Anxiety and academic achievement: A meta-analysis of empirical results].
665–689. Zeitschrift für Pädagogische Psychologie, 5, 85–97.
Muthén, B., & Kaplan, D. (1985). A comparison of some methodologies for the factor Sparfeldt, J. R., Schilling, S. R., Rost, D. H., Stelzl, I., & Peipert, D. (2005).
analysis of non-normal Likert variables. British Journal of Mathematical and Statistical Leistungsängstlichkeit: Facetten, Fächer, Fachfacetten? – Zur Trennbarkeit nach
Psychology, 38, 171–189. Angstfacette und Inhaltsbereich [Test anxiety: The relevance of anxiety factors as
Muthén, L. K., & Muthén, B. O. (1998–2007). Mplus user's guide (5th ed.). Los Angeles, well as school subjects]. Zeitschrift für Pädagogische Psychologie, 19, 225–236.
CA: Muthén & Muthén. Sparfeldt, J. R., Schilling, S. R., Rost, D. H., & Thiel, A. (2006). Blocked versus random-
Muthén, L. K., & Muthén, B. O. (2002). How to use a Monte Carlo study to decide on ized format of questionnaires: A confirmatory multiple group analysis. Educational
sample size and determine power. Structural Equation Modeling, 9, 599–620. and Psychological Measurement, 66, 961–974.
Rost, D. H., & Schermer, F. J. (1989). The various factors of test anxiety: A subcomponent Spielberger, C. D. (1980). Test anxiety inventory (preliminary professional manual). Palo
model of test anxiety measurement. In R. Schwarzer, H. M. van der Ploeg, & C. D. Alto, CA: Consulting Psychologists Press.
Spielberger (Eds.), Advances in Test Anxiety Research, Vol. 6. (pp. 37–52). Amsterdam/ Zeidner, M. (1998). Test anxiety: The state of the art. New York: Plenum Press.
Berwyn, PA: Swets & Zeitlinger/Swets North America. Zeidner, M. (2007). Test anxiety in educational contexts: Concepts, findings, and
Rost, D. H., & Schermer, F. J. (1992). “Reaction to tests” (RTT) and “Manifestations future directions. In P. A. Schutz, & R. Pekrun (Eds.), Emotion in education
of test-anxiety” (DAI-MAN): Same or different concepts? In K. A. Hagtvet, & T. Backer (pp. 165–184). Burlington, MA: Elsevier.

You might also like