Educ Stud Math (2013) 82:145–164

DOI 10.1007/s10649-012-9414-x

Development of a short form of the attitudes

toward mathematics inventory

Siew Yee Lim & Elaine Chapman

Published online: 7 June 2012

# Springer Science+Business Media B.V. 2012

Abstract Existing instruments designed to measure mathematics attitudes were too long, dated,
or assessed with only western samples. To address this issue, a shortened version of the Attitudes
Toward Mathematics Inventory (short ATMI) which measures four subscales—enjoyment of
mathematics, motivation to do mathematics, self-confidence in mathematics, and perceived
value of mathematics—was created. Its factor structure, reliability, and validity were assessed
with 1,601 participants from Singapore. Confirmatory factor analyses supported the original
four-factor structure. Within this structure, however, several items were found to correlate highly
with others. Their removal either improved or did not impact the properties of the instrument. As
a result, these items were removed to produce the short ATMI. Furthermore, a very high
correlation (r0.96) was found between the enjoyment and motivation subscales. Results of
further analysis suggested the removal of the motivation subscale. The short ATMI exhibited
strong correlations with the original scale (mean r0.96), good overall internal consistencies, both
for the full short version (α0.93) and for the individual subscales (mean α0.87), and satisfactory
test–retest reliability over a 1-month period (mean rxx 0.75). The validity of the short ATMI was
further demonstrated through inter-correlations between its subscales, and through correlations
with mathematics anxiety and achievement test scores. Participants were able to complete the
short ATMI in less than 10 min, making it a viable option when survey administration time is
limited. This time would reduce further with the removal of the motivation subscale.

Keywords Attitudes Toward Mathematics Inventory . Mathematics attitudes . Confirmatory

factor analysis . Pre-tertiary

S. Y. Lim : E. Chapman
The University of Western Australia, Graduate School of Education, 35 Stirling Highway, Crawley, WA
6009 Perth, Australia

S. Y. Lim
Hwa Chong Institution, Singapore, 661, Bukit Timah Road, Singapore 269734

Present Address:
S. Y. Lim (*)
285, Ghim Moh Road, Singapore 279622
e-mail: lim_siew_yee@moe.edu.sg
146 S.Y. Lim, E. Chapman

Over the past 40 years, there has been considerable interest in the study of students’ attitudes
toward mathematics. In the introduction for a special issue on the discussion of mathematics
attitudes in Educational Studies in Mathematics, Zan, Brown, Evans and Hannula (2006)
suggested that the most important research on mathematics attitudes is on the understanding
of the interrelationships between affect and cognition, and stated that “Studies of attitude
were based on two beliefs: attitude toward mathematics is related to achievement, and
affective outcomes (such as liking mathematics) are significant per se” (p. 113).
Indeed, studies have shown that strong relationships exist between various domains of
attitudes (e.g., enjoyment of mathematics; motivation to do mathematics; self-confidence in
mathematics; and perceived value of mathematics) and mathematics achievement (Antonnen,
1969; Atkinson & Raynor, 1974; Ball, 1977; Bouchey & Harter, 2005; Foire, 1999; Minato,
1983; Minato & Yanase, 1984; Samuelsson & Granstrom, 2007). Students’ dropout rates in
mathematics courses have also been found to relate closely to their mathematics attitudes
(Barkoukis, Tsorbatzoudis, Grouios, & Sideridis, 2008). This relationship suggests that stu-
dents’ attitudes toward mathematics have economic implications on their future, as students
who dropped out of mathematics courses are reported to earn less in the labor market than those
who graduated from the courses (Altonji, 1995; Joensen & Nielson, 2009; Rose & Betts, 2004).
Studies on mathematics attitudes and their relationships with various constructs are
thus important. The integrity of the results of these studies, however, is contingent on
the psychometric properties of the instruments used.
Various researchers (e.g., Ma & Kishor, 1997; Zan et al., 2006) have lamented on the lack
of proper instruments to measure mathematics attitudes. Indeed, existing instruments
designed to measure students’ attitudes toward mathematics are either dated, or provide
data on validity and reliability only with western samples. Many of the existing instruments
also take a long time to administer. Hence, the aim of this study was to develop and assess
the psychometric properties of a short version of the Attitudes Toward Mathematics
Inventory (short ATMI; Tapia & Marsh, 2004) on an Asian sample.
The ATMI is one of the latest instruments to appear which measures students’ attitudes toward
mathematics (Chamberlin, 2010). It was developed to provide a parsimonious alternative to
existing mathematics attitudes scales (Tapia & Marsh, 2004). The original ATMI was written in
English and comprised 49 items that measured six domains: confidence (Goolsby, 1988; Linn
& Hyde, 1989; Randhawa, Beamer, & Lundberg, 1993), anxiety (Hauge, 1991; Terwilliger &
Titus, 1995), value (Longitudinal Study of American Youth, 1990), enjoyment (Ma & Kishor,
1997; Thorndike-Christ, 1991), motivation (Singh, Granville, & Dika, 2002; Thorndike-Christ,
1991), and parent/teacher expectations (Dossey, 1992; Kenschaft, 1991). Scoring was done
with a five-point Likert Scale, with response options ranging from “strongly disagree” to
“strongly agree” (Tapia, 1996).
After conducting an exploratory factor analysis using data from a high school sample, the
confidence and anxiety subscales were combined to form a single factor (Tapia, 1996). In
addition, items on the parent/teacher expectation subscale were dropped due to extremely
low item-to-total correlation. The resulting scale comprises four subscales and a total of 40
items measuring enjoyment (ENJ), motivation (MOT), self-confidence (SC), and value (VAL).
The ENJ subscale comprises 10 items, and measures “the degree to which students enjoy
working (on) mathematics” (Tapia & Marsh, 2004, p. 17). The MOT subscale comprises five
items, and measures students’ “interest in mathematics and (their) desire to pursue further
studies in mathematics” (Tapia & Marsh, 2004, p. 17). The SC subscale comprises 15 items,
and measures students’ “confidence and self-concept of (their) performance in mathematics”
(Tapia & Marsh, 2004, p. 17). Finally, the VAL subscale comprises ten items, and measures
“students’ beliefs on the usefulness, relevance and worth of mathematics to their lives”
Short form of the attitudes toward mathematics inventory 147

(Tapia & Marsh, 2004, p. 17). A confirmatory factor analysis (CFA) conducted by Tapia and
Marsh (2002) confirmed the four-factor structure on a pre-tertiary sample from the USA.
The psychometric properties of the ATMI have also been established. Specifically, the ATMI
displayed satisfactory item-to-total correlations which ranged from .50 to .82 (Tapia & Marsh,
2004). Good reliability and internal consistencies were also found, with alpha reliability
coefficients that ranged from .95 to .97 (Tapia, 1996; Tapia & Marsh, 2000, 2004). High levels
of internal consistency within each subscale were reported, with Cronbach alpha coefficients of
.89, .88, .95, and .89 for the ENJ, MOT, SC, and VAL subscales, respectively. Finally, satisfactory
test–retest reliabilities over a 4-month period were established, with Pearson correlations of .89
for the full scale, and .84, .78, .99, and .70 for the ENJ, MOT, SC, and VAL subscales respectively
(Tapia & Marsh, 2004).
Interestingly, despite its more stable factor structure and stronger psychometric proper-
ties, the ATMI is not yet as popular as the Fennema-Sherman Mathematics Attitudes Scale
(FSMAS) (Fennema & Sherman, 1976) (Chamberlin, 2010). One possible explanation for the
popularity of the FSMAS over the ATMI, is that the FSMAS was developed more than 30 years
ago, and between the development of the FSMAS and ATMI, no other instrument designed to
measure mathematics attitudes was satisfactorily assessed for validity and reliability. Hence
the ATMI may need more time to gain popularity over the FSMAS. A second explanation is that
the ATMI was assessed only with samples from Mexico or the USA. Thus, there is a need to
investigate the properties of the ATMI with a culturally different sample. Finally, the popu-
larity of the ATMI may increase if an even shorter form can be established to measure
mathematics attitudes accurately, as the short form will require less time for administration.
The primary aims of this study were to develop a short form of the ATMI and
investigate the psychometric properties of this short form on a predominantly Chi-
nese sample from Singapore by (1) examining its factor structure; (2) estimating its
overall internal consistency and internal consistencies of individual subscales; (3)
evaluating its test–retest reliability over a 1-month period; and (4) investigating
relationships between scores on its subscales and scores from measures of theoreti-
cally related constructs (mathematics anxiety and mathematics achievement). The use
of the English language was retained in the short form.

1 Method

This study was conducted in three phases. Phase 1 (n012) was a pilot study to
preliminarily assess the suitability of the ATMI items for the participants of this study.
Phase 2 (n01,601) was the main study, which provided data to assess the psycho-
metric properties of the ATMI. Finally, phase 3 (n0208) assessed test–retest reliability
over a 1-month period. All items were randomly sequenced prior to administration in
all three phases.

2 Participants

All participants were mathematics students enrolled in six out of the 21 pre-tertiary institutions
in Singapore, which prepare students for the General Certificate of Education Advanced Level
(GCE “A” level) Mathematics Higher 2 (Syllabus 9740) examination administered by the
University of Cambridge-London Examination Syndicate (UCLES) in the English language. In
terms of academic achievement at a national mathematics examination conducted annually,
148 S.Y. Lim, E. Chapman

these six institutions comprised two top, two average, and two bottom institutions, which were
ranked top five, between 8th and 13th, and bottom five respectively, for the past 5 years.
Phase 1 participants had an average age of 17.9. Males made up 41.7 % of the sample. All
the participants were Chinese. Phase 2 participants had an average age of 17.8 years. This
sample comprised 42.2 % males, 49.2 % females, and 8.6 % who did not indicate their
gender. Additionally, 98.9 % were Chinese, 0.62 % were Indians and the rest were Malays.
Phase 3 participants had an average age of 17.7. Males made up 51.4 % of the sample, with
the ethnic composition being 99.5 % Chinese and 0.5 % Indians.

3 Instruments used for validation of the ATMI

To validate the ATMI, mathematics anxiety and achievement were respectively measured with
the Fennema-Sherman Mathematics Anxiety Scale-Revised (FSMAS-R) (Lim & Chapman,
2011), and a pen-and-paper mathematics test.

3.1 Mathematics anxiety

The FSMAS-R consists of two subscales measuring anxiety toward mathematics (FS-ANX), and
ease with mathematics (FS-EASE), on a five-point Likert scale with response options ranging
from “strongly disagree” to “strongly agree”. The FS-ANX and FS-EASE subscales comprise
four and five items, respectively. FS-ANX measures “feelings of anxiety, dread, nervousness,
and associated body symptoms related to doing mathematics” (Fennema & Sherman, 1976,
p. 326), while FS-EASE measures participants’ comfort level when they are engaging in
mathematics activities (Lim & Chapman, 2011).
The factor structure, reliability, and validity of the FSMAS-R were assessed with 1,601 pre-
tertiary students in Singapore (Lim & Chapman, 2011). Exploratory and confirmatory factor
analyses supported a two-factor model. Good internal consistencies (full scale α0.91;
αs0.92 and .82 for the FS-ANX and FS-EASE subscales, respectively) and 1-month test–retest
reliabilities (rxx 0.87 and .77, p<.01 for the FS-ANX and FS-EASE subscales, respectively) were
obtained. The construct validity of the FSMAS-R was also demonstrated through inter-
correlations between subscales and correlations with self-confidence in mathematics and
mathematics achievement test scores.

3.2 Mathematics achievement

A 3-h pen-and-paper test was used to assess the mathematics ability of the participants. A
panel of five teachers, each with at least 5 years of mathematics teaching experience in a
local pre-tertiary institution, set the test items and marking schemes, which were similar to
past years’ GCE “A” level mathematics questions and marking schemes respectively. To
assess the content validity of the achievement test, a setter of the GCE “A” level mathemat-
ics paper from UCLES, was invited to review the test items and marking schemes, and to
provide feedback on their validity. Test scripts were marked using the same marking scheme
developed by the five teachers. To ensure scorer reliability (Houser, 1998), each test item
was marked by the same teacher for all scripts. Additionally, 10 % of the scripts were
randomly chosen and evaluated again by two senior teachers with at least 10 years of
teaching experience. Inter-rater reliability between the five main markers and the two senior
teachers ranged from 0.98 to 1.
Short form of the attitudes toward mathematics inventory 149

4 Procedures

In phase 1, 12 participants responded to the ATMI. They were also asked to respond to
open-ended questions, adapted from Bell (1987), on whether the instructions and
items in the modified scale were clear, and how the scale could be improved. Besides
obtaining written responses to these open-ended questions, a discussion based on the
same questions was held with every participant to solicit feedback that participants
might otherwise fail to report from the written survey. Results from this pilot phase
indicated that the instructions and items in the ATMI were clear and appropriate. The
participants in phase 1 took at most 15 min to complete the scale.
In phase 2, the ATMI was administered to 1,618 participants by nine teachers in
classroom or lecture settings. All participating teachers attended a briefing before they
administered the scale. Participants were told explicitly that the results of the survey
would not affect their school grades in any way, and that they could choose to remain
anonymous or opt out of the survey. Codes were given to all 1,618 participants to
trace their mathematics achievement test scores. After analyzing the data collected in
this phase, a short form of the ATMI (short ATMI) was created by removing items that
did not contribute significantly to enhancing the properties of the instrument. Data
collected in this phase was also used to assess the inter-consistencies of the final ATMI
subscales, and to investigate correlations between scores on the short ATMI subscales,
and with scores on the FSMAS-R and mathematics achievement test.
In phase 3, the short ATMI was administered twice over a 1-month period to another 208
participants to assess test–retest reliability. All participants were able to complete the short ATMI
within 10 min.

5 Results

Prior to data analysis, screening was conducted for missing data. Any completed ATMI forms
with more than two missing responses were excluded from the final data pool. This reduced
the initial sample of 1,618 participants to 1,601. Any missing responses on the 1,601 forms
appeared to be random oversights (rather than systematic exclusions) of individual items
within the survey. Median scores were used to replace the missing individual responses in
these cases. No missing responses were located in the FSMAS-R forms of these 1,601
participants. However, out of these 1,601 participants, achievement data were available for
only 1,027; thus, all analyses involving achievement test scores relied on this reduced
sample. All responses to negatively worded statements were reversed prior to the data
analysis. The remainder of this section is organized according to the four aims stipulated
for this study.

6 Factor structure of the ATMI

The factor structure of the ATMI was assessed using CFAs on two subsamples, which were
formed by splitting the participants randomly (stratifying for gender and school) into two
groups. This resulted in 338 males, 394 females, and 68 students with unknown gender in
subsample 1 (n0800); and 338 males, 394 females, and 69 students with unknown gender in
subsample 2 (n0801).
150 S.Y. Lim, E. Chapman

Prior to conducting these analyses, screening tests for conformity to underlying CFA
assumptions were conducted. These tests generally produced satisfactory results. Mahala-
nobis distances and z scores (using a conservative α level owing to the large sample size)
indicated no notable univariate or multivariate outliers within the dataset. There was also no
evidence of multicollinearity amongst the item scores. Skewness and kurtosis coefficients
were then examined to detect any departures from normality in the item distributions. This
analysis indicated significant levels of skew across many items within the scale. As a result,
the PRELIS Normal Scores module in LISREL 8.7 was used to improve the score distribu-
tions before conducting the factor analyses. Table 1 contains item-level descriptive statistics
for both subsamples.
The two separate CFAs were conducted using LISREL 8.7. Bivariate correlations
between items for subsamples 1 and 2 are presented in Tables 2 and 3, respectively.
Several indices were used to assess model fit. As the chi-squared statistic (χ2) is strongly
dependent on sample size (Hu & Bentler, 1995), χ2/df ratios instead of probability values are
presented for each model. In general, χ2/df ratios ranging from two to five are considered to
represent adequate model fit (Byrne, 1989; Tanaka, 1993). As the various models subjected
to CFA were nested designs, the chi-squared change (Δχ2) statistic (Hu & Bentler, 1995)
was used to test for differences in fit between the models. Two absolute fit indices (the
standardized root mean residual, SRMR, and the goodness-of-fit index, GFI) and two incre-
mental fit indices (the non-normed fit index, NNFI, and the comparative fit index, CFI) are also
presented. Various authors (e.g., Browne and Cudeck (1993) and Byrne (1989)) suggest that
good model fit is indicated by indices of less than .08 for the SRMR, and greater than .90 for
the GFI, NNFI, and CFI. Table 4 presents the fit indices for the four-factor model on both
subsamples.
Due to the large sample size, all the χ2 values were significant. The NNFI and CFI
supported the four-factor model. However, the χ2/df ratios and GFI were outside the
recommended ranges, and modification indices suggested that modifications were
necessary.
To assess internal consistency, Cronbach’s alpha coefficients for each subscale were
estimated using SPSS 19. Table 5 presents the results for both subsamples.
Nunnally and Bernstein (1994) recommend internal consistency scores to range
from .70 to .90, and suggest a high level of item redundancy, if internal consistency
scores fall outside this range. Very high Cronbach’s alpha values were obtained for
the overall scale and all the individual subscales, except for the MOT subscale. Hence
a short form of the ATMI (short ATMI) was explored by deleting items using the
classical item analysis approach. Three criteria were used to delete items: (1) item
loadings; (2) improvement in fit indices; and (3) internal consistency. Specifically,
items that had low factor loadings or high cross-factor loadings were shortlisted and
removed one by one. Fit indices and internal consistencies were examined after the
removal of each item. The removed item was returned to the scale, if fit indices
worsened or internal consistencies fell outside the recommended range of .70 to .90.
Consequently, five ENJ, one MOT, ten SC, and five VAL items were removed. Table 6
presents the fit indices for the modified four-factor model.
Very similar results were obtained from the two subsamples. Specifically, the
modified four-factor model produced a significant reduction in χ2 over the original
four-factor model (Δχ2(588) 04,738.78, p < .01 for subsample 1, and Δχ2(588) 0
6,116.74, p<.01 for subsample 2).
In addition, after combining data from both subsamples, the Pearson correlations between
the ATMI and its short form for all the subscales were high (ENJ: r0.96, p<.01; MOT: r0.98,
Short form of the attitudes toward mathematics inventory 151

Table 1 Descriptive statistics for the ATMI

Label Full item statement Subsample Subsample

1 2

M SD M SD

ENJ1 I get a great deal of satisfaction out of solving a mathematics problem. 3.63 1.09 3.74 0.97
ENJ2 I have usually enjoyed studying mathematics in school. 3.44 1.04 3.47 1.02
ENJ3 Mathematics is dull and boring. 3.51 1.15 3.53 1.12
ENJ4 I like to solve new problems in mathematics. 3.30 1.03 3.40 1.00
ENJ5 I would prefer to do an assignment in mathematics than to write an essay. 3.66 1.33 3.62 1.33
ENJ6 I really like mathematics. 3.14 1.10 3.18 1.07
ENJ7 I am happier in a mathematics class than in any other class. 2.72 1.06 2.82 1.06
ENJ8 Mathematics is a very interesting subject. 3.28 1.00 3.31 0.96
ENJ9 I am comfortable expressing my own ideas on how to look for solutions 3.04 0.97 3.12 0.93
to a difficult problem in mathematics.
ENJ10 I am comfortable answering questions in mathematics class. 3.13 0.95 3.19 0.91
MOT1 I am confident that I could learn advanced mathematics. 3.34 1.08 3.32 1.05
MOT2 I would like to avoid using mathematics in university. 3.32 1.20 3.27 1.21
MOT3 I am willing to take more than the required amount of mathematics. 2.89 1.07 2.93 1.07
MOT4 I plan to take as much mathematics as I can during my education. 2.76 1.03 2.78 1.07
MOT5 The challenge of mathematics appeals to me. 3.05 1.11 3.12 1.06
SC1 Mathematics is one of my most dreaded subjects. 3.59 1.31 3.56 1.30
SC2 My mind goes blank and I am unable to think clearly when working 3.69 1.09 3.69 1.08
with mathematics.
SC3 Studying mathematics makes me feel nervous. 3.62 1.06 3.59 1.03
SC4 Mathematics makes me feel uncomfortable. 3.61 1.05 3.60 1.07
SC5 I am always under a terrible strain in a mathematics class. 3.57 1.03 3.57 1.03
SC6 When I hear the word mathematics, I have a feeling of dislike. 3.72 1.06 3.67 1.09
SC7 It makes me nervous to even think about having to do a mathematics problem. 3.63 0.98 3.65 0.96
SC8 Mathematics does not scare me at all. 2.81 1.07 2.83 1.03
SC9 I expect to do fairly well in any mathematics class I take. 3.24 1.03 3.27 1.03
SC10 I am always confused in my mathematics class. 3.51 0.98 3.53 0.97
SC11 I have a lot of self-confidence when it comes to mathematics. 2.83 0.99 2.90 0.98
SC12 I am able to solve mathematics problems without too much difficulty. 2.97 0.95 2.97 0.91
SC13 I feel a sense of insecurity when attempting mathematics. 3.38 1.07 3.40 1.05
SC14 I learn mathematics easily. 3.02 0.92 3.04 0.92
SC15 I believe I am good at solving mathematics problems. 3.00 0.96 3.07 0.94
VAL1 Mathematics is a very worthwhile and necessary subject. 3.64 0.96 3.73 0.92
VAL2 I want to develop my mathematical skills. 3.59 0.93 3.67 0.88
VAL3 Mathematics helps to develop the mind and teaches a person to think. 3.66 0.93 3.72 0.92
VAL4 Mathematics is important in everyday life. 3.46 0.98 3.52 0.97
VAL5 Mathematics is one of the most important subjects for people to study. 3.54 0.95 3.65 0.94
VAL6 College mathematics lessons would be very helpful no matter what I decide 3.43 0.96 3.47 1.00
to study in future.
VAL7 I can think of many ways that I use mathematics outside of school. 2.89 1.04 2.92 1.03
VAL8 I think studying advanced mathematics is useful. 3.28 1.03 3.29 1.02
VAL9 I believe studying mathematics helps me with problem solving in other areas. 3.37 0.95 3.42 0.93
VAL10 A strong mathematics background could help me in my professional life. 3.54 0.95 3.57 0.94
Table 2 Bivariate correlations for items in the ATMI for subsample 1
152

ENJ1 ENJ2 ENJ3 ENJ4 ENJ5 ENJ6 ENJ7 ENJ8 ENJ9 ENJ10 MOT1 MOT2 MOT3 MOT4 MOT5 SC1 SC2 SC3 SC4 SC5

ENJ1 – .47 .42 .59 .39 .51 .33 .48 .30 .33 .44 .33 .40 .41 .52 .39 .25 .23 .31 .25
ENJ2 – .53 .61 .48 .69 .53 .60 .46 .40 .51 .46 .55 .52 .59 .53 .39 .35 .43 .36
ENJ3 – .55 .45 .61 .47 .56 .34 .30 .42 .57 .50 .50 .51 .62 .46 .53 .59 .45
ENJ4 – .53 .69 .48 .64 .48 .46 .51 .49 .62 .57 .67 .53 .38 .36 .47 .36
ENJ5 – .53 .49 .43 .31 .33 .41 .48 .50 .45 .45 .55 .39 .35 .46 .35
ENJ6 – .60 .70 .47 .47 .54 .57 .66 .64 .67 .63 .43 .40 .51 .39
ENJ7 – .51 .40 .40 .43 .45 .54 .55 .52 .52 .35 .35 .39 .32
ENJ8 – .49 .46 .46 .46 .64 .62 .63 .52 .38 .34 .43 .37
ENJ9 – .50 .43 .37 .48 .51 .55 .38 .37 .34 .36 .34
ENJ10 – .41 .35 .48 .45 .53 .42 .41 .37 .42 .42
MOT1 – .41 .51 .46 .51 .47 .41 .37 .43 .38
MOT2 – .55 .54 .47 .59 .45 .43 .54 .47
MOT3 – .68 .61 .55 .38 .35 .44 .37
MOT4 – .65 .46 .36 .32 .40 .33
MOT5 – .54 .39 .40 .47 .41
SC1 – .58 .61 .67 .57
SC2 – .57 .60 .55
SC3 – .72 .61
SC4 – .65
SC5 –
SC6
SC7
SC8
SC9
SC10
SC11
SC12
SC13
S.Y. Lim, E. Chapman
Table 2 (continued)
ENJ1 ENJ2 ENJ3 ENJ4 ENJ5 ENJ6 ENJ7 ENJ8 ENJ9 ENJ10 MOT1 MOT2 MOT3 MOT4 MOT5 SC1 SC2 SC3 SC4 SC5

SC14
SC15
VAL1
VAL2
VAL3
VAL4
VAL5
VAL6
VAL7
VAL8
VAL9
VAL10

SC6 SC7 SC8 SC9 SC10 SC11 SC12 SC13 SC14 SC15 VAL1 VAL2 VAL3 VAL4 VAL5 VAL6 VAL7 VAL8 VAL9 VAL10
Short form of the attitudes toward mathematics inventory

ENJ1 .33 .32 .15 .32 .25 .31 .26 .26 .30 .36 .44 .46 .38 .34 .32 .36 .25 .35 .31 .29
ENJ2 .48 .38 .36 .47 .38 .47 .42 .43 .50 .52 .44 .45 .34 .31 .40 .35 .35 .47 .36 .34
ENJ3 .59 .49 .30 .37 .48 .41 .32 .50 .39 .43 .43 .42 .29 .34 .35 .33 .35 .42 .36 .31
ENJ4 .48 .38 .37 .39 .42 .48 .47 .43 .51 .54 .50 .53 .40 .47 .42 .42 .42 .52 .43 .39
ENJ5 .46 .37 .34 .38 .38 .40 .36 .39 .38 .43 .35 .34 .22 .25 .36 .27 .25 .36 .28 .28
ENJ6 .54 .46 .41 .47 .39 .56 .48 .48 .55 .57 .45 .49 .37 .35 .41 .38 .42 .53 .38 .37
ENJ7 .43 .35 .36 .38 .36 .50 .40 .41 .47 .45 .32 .27 .21 .29 .27 .27 .32 .38 .30 .28
ENJ8 .48 .39 .35 .45 .35 .48 .42 .39 .50 .51 .52 .45 .42 .39 .42 .38 .44 .56 .44 .38
ENJ9 .38 .33 .42 .39 .39 .48 .43 .40 .51 .48 .34 .29 .26 .26 .32 .32 .37 .42 .36 .29
ENJ10 .38 .44 .42 .50 .44 .54 .54 .44 .61 .59 .33 .27 .26 .25 .25 .22 .33 .41 .32 .30
MOT1 .47 .43 .38 .47 .42 .51 .44 .45 .49 .54 .42 .39 .30 .25 .34 .33 .29 .49 .34 .32
MOT2 .56 .47 .31 .39 .48 .43 .36 .51 .40 .47 .38 .36 .23 .29 .37 .31 .36 .45 .34 .38
MOT3 .46 .38 .39 .45 .39 .52 .46 .44 .50 .55 .40 .43 .36 .32 .39 .35 .45 .55 .36 .35
MOT4 .41 .33 .35 .39 .33 .49 .42 .38 .45 .48 .46 .42 .31 .36 .43 .39 .45 .57 .43 .43
153
Table 2 (continued)
154

SC6 SC7 SC8 SC9 SC10 SC11 SC12 SC13 SC14 SC15 VAL1 VAL2 VAL3 VAL4 VAL5 VAL6 VAL7 VAL8 VAL9 VAL10

MOT5 .49 .40 .40 .44 .40 .53 .48 .44 .53 .55 .41 .43 .29 .30 .32 .32 .41 .51 .39 .30
SC1 .64 .60 .40 .49 .58 .56 .46 .61 .49 .56 .38 .30 .26 .25 .31 .29 .31 .42 .30 .27
SC2 .53 .55 .37 .40 .56 .45 .38 .61 .46 .47 .31 .23 .19 .18 .25 .22 .26 .34 .28 .23
SC3 .59 .58 .39 .35 .57 .45 .42 .64 .43 .47 .22 .16 .12 .14 .18 .13 .24 .30 .21 .18
SC4 .67 .64 .42 .42 .61 .49 .47 .67 .51 .52 .35 .28 .18 .21 .26 .23 .31 .38 .26 .26
SC5 .65 .63 .35 .40 .64 .42 .45 .63 .46 .50 .31 .21 .15 .16 .18 .18 .24 .33 .26 .19
SC6 – .60 .32 .43 .58 .44 .39 .60 .47 .50 .38 .31 .22 .25 .32 .27 .32 .39 .31 .30
SC7 – .37 .40 .63 .44 .43 .64 .46 .49 .33 .23 .22 .21 .22 .21 .26 .36 .34 .25
SC8 – .36 .38 .50 .54 .50 .51 .50 .18 .18 .11 .12 .18 .17 .30 .31 .27 .16
SC9 – .36 .56 .45 .42 .52 .60 .37 .31 .31 .27 .34 .30 .29 .38 .32 .33
SC10 – .43 .46 .68 .48 .47 .34 .24 .19 .24 .26 .22 .33 .34 .32 .23
SC11 – .63 .53 .63 .68 .31 .27 .24 .23 .30 .29 .38 .43 .31 .28
SC12 – .49 .62 .62 .27 .25 .18 .23 .25 .22 .40 .39 .28 .30
SC13 – .52 .54 .29 .21 .15 .20 .21 .18 .31 .34 .25 .17
SC14 – .68 .32 .26 .23 .22 .29 .25 .34 .40 .30 .28
SC15 – .36 .35 .25 .27 .34 .30 .38 .47 .34 .31
VAL1 – .58 .54 .56 .59 .53 .42 .54 .54 .53
VAL2 – .50 .47 .47 .48 .37 .49 .43 .49
VAL3 – .48 .46 .47 .31 .43 .45 .41
VAL4 – .56 .52 .47 .45 .49 .51
VAL5 – .53 .42 .51 .48 .48
VAL6 – .38 .50 .50 .54
VAL7 – .46 .50 .40
VAL8 – .50 .54
VAL9 – .51
VAL10 –
S.Y. Lim, E. Chapman
Table 3 Bivariate correlations for items in the ATMI for subsample 2

ENJ1 ENJ2 ENJ3 ENJ4 ENJ5 ENJ6 ENJ7 ENJ8 ENJ9 ENJ10 MOT1 MOT2 MOT3 MOT4 MOT5 SC1 SC2 SC3 SC4 SC5

ENJ1 – .46 .41 .60 .39 .54 .39 .50 .32 .31 .43 .34 .40 .38 .50 .35 .23 .22 .35 .31
ENJ2 – .52 .58 .38 .67 .51 .60 .41 .44 .56 .47 .57 .52 .52 .53 .38 .40 .52 .45
ENJ3 – .52 .37 .61 .40 .58 .29 .36 .45 .58 .47 .44 .51 .64 .50 .55 .65 .55
ENJ4 – .45 .64 .46 .63 .47 .42 .56 .44 .59 .52 .66 .46 .37 .36 .48 .39
ENJ5 – .50 .52 .40 .29 .35 .41 .45 .39 .42 .39 .48 .31 .34 .43 .35
ENJ6 – .59 .72 .42 .51 .57 .58 .67 .63 .64 .59 .41 .39 .55 .47
ENJ7 – .53 .35 .38 .43 .45 .52 .54 .47 .45 .30 .29 .42 .35
ENJ8 – .45 .44 .53 .50 .63 .57 .65 .49 .37 .37 .51 .43
ENJ9 – .47 .46 .36 .44 .49 .48 .29 .33 .27 .35 .28
ENJ10 – .51 .36 .48 .43 .50 .44 .40 .41 .47 .44
MOT1 – .49 .56 .51 .56 .49 .43 .41 .50 .45
MOT2 – .53 .54 .47 .59 .48 .47 .60 .53
MOT3 – .64 .60 .45 .37 .32 .46 .39
Short form of the attitudes toward mathematics inventory

MOT4 – .61 .42 .27 .24 .40 .31

MOT5 – .49 .35 .35 .46 .45
SC1 – .58 .60 .69 .63
SC2 – .58 .60 .54
SC3 – .70 .62
SC4 – .70
SC5 –
SC6
SC7
SC8
SC9
SC10
SC11
SC12
SC13
155
Table 3 (continued)
156

ENJ1 ENJ2 ENJ3 ENJ4 ENJ5 ENJ6 ENJ7 ENJ8 ENJ9 ENJ10 MOT1 MOT2 MOT3 MOT4 MOT5 SC1 SC2 SC3 SC4 SC5

SC14
SC15
VAL1
VAL2
VAL3
VAL4
VAL5
VAL6
VAL7
VAL8
VAL9
VAL10

SC6 SC7 SC8 SC9 SC10 SC11 SC12 SC13 SC14 SC15 VAL1 VAL2 VAL3 VAL4 VAL5 VAL6 VAL7 VAL8 VAL9 VAL10

ENJ1 .41 .25 .16 .29 .28 .32 .30 .28 .31 .36 .41 .44 .38 .31 .32 .34 .26 .34 .27 .22
ENJ2 .55 .40 .34 .44 .43 .45 .45 .44 .49 .51 .49 .44 .34 .38 .40 .35 .35 .48 .31 .35
ENJ3 .70 .48 .24 .34 .50 .38 .37 .50 .38 .42 .45 .37 .28 .34 .35 .33 .35 .41 .35 .31
ENJ4 .49 .34 .29 .44 .40 .45 .45 .37 .47 .52 .45 .49 .45 .47 .43 .36 .37 .46 .38 .37
ENJ5 .43 .35 .30 .41 .31 .37 .35 .36 .35 .40 .32 .25 .28 .28 .33 .25 .27 .32 .25 .31
ENJ6 .59 .38 .35 .47 .42 .50 .49 .44 .52 .58 .52 .47 .39 .39 .45 .44 .46 .54 .36 .38
ENJ7 .43 .30 .30 .38 .28 .41 .38 .31 .38 .42 .34 .33 .29 .28 .32 .29 .32 .37 .27 .27
ENJ8 .57 .40 .28 .44 .39 .43 .45 .42 .47 .50 .52 .48 .43 .45 .44 .44 .43 .53 .44 .37
ENJ9 .27 .27 .34 .40 .28 .52 .46 .36 .47 .51 .29 .29 .33 .28 .29 .29 .38 .37 .35 .32
ENJ10 .39 .42 .45 .55 .45 .58 .55 .46 .67 .67 .31 .26 .29 .24 .27 .25 .33 .39 .31 .30
MOT1 .49 .40 .40 .53 .41 .57 .53 .45 .56 .59 .46 .44 .40 .36 .42 .38 .35 .55 .40 .41
MOT2 .62 .47 .29 .37 .43 .46 .40 .50 .42 .44 .46 .37 .28 .34 .39 .35 .37 .46 .35 .37
MOT3 .46 .34 .37 .45 .35 .52 .49 .40 .51 .51 .46 .46 .34 .40 .41 .39 .45 .56 .37 .38
MOT4 .38 .25 .33 .42 .26 .47 .41 .33 .45 .48 .48 .46 .35 .38 .43 .38 .46 .55 .40 .41
S.Y. Lim, E. Chapman
Table 3 (continued)
SC6 SC7 SC8 SC9 SC10 SC11 SC12 SC13 SC14 SC15 VAL1 VAL2 VAL3 VAL4 VAL5 VAL6 VAL7 VAL8 VAL9 VAL10

MOT5 .49 .41 .35 .46 .40 .52 .50 .43 .51 .56 .43 .44 .39 .34 .35 .34 .39 .50 .41 .35
SC1 .71 .55 .40 .46 .55 .49 .45 .61 .50 .53 .35 .25 .19 .28 .25 .24 .30 .33 .25 .27
SC2 .54 .52 .33 .38 .53 .44 .45 .56 .45 .47 .27 .16 .20 .22 .26 .22 .23 .28 .24 .25
SC3 .58 .61 .39 .39 .57 .43 .41 .63 .44 .46 .28 .16 .15 .22 .19 .17 .20 .27 .27 .19
SC4 .71 .63 .39 .41 .60 .50 .47 .70 .51 .53 .38 .25 .25 .27 .30 .29 .29 .38 .33 .29
SC5 .70 .67 .37 .43 .67 .47 .47 .67 .46 .50 .32 .21 .18 .21 .24 .21 .24 .31 .27 .20
SC6 – .61 .31 .41 .58 .46 .43 .62 .45 .46 .42 .33 .26 .30 .35 .33 .29 .39 .31 .29
SC7 – .38 .41 .61 .41 .42 .66 .44 .49 .26 .14 .23 .23 .23 .19 .17 .27 .27 .18
SC8 – .43 .31 .54 .51 .43 .50 .54 .18 .10 .12 .18 .19 .18 .23 .28 .24 .22
SC9 – .44 .54 .49 .44 .57 .63 .31 .31 .28 .30 .31 .30 .26 .36 .38 .34
SC10 – .40 .42 .67 .45 .53 .29 .21 .21 .25 .22 .20 .21 .28 .26 .22
SC11 – .66 .54 .65 .68 .30 .25 .25 .24 .30 .27 .36 .38 .30 .34
SC12 – .49 .62 .63 .33 .23 .23 .21 .29 .27 .33 .37 .25 .31
SC13 – .51 .56 .32 .20 .23 .23 .21 .25 .24 .29 .26 .23
Short form of the attitudes toward mathematics inventory

SC14 – .70 .31 .28 .25 .26 .27 .23 .32 .42 .30 .28
SC15 – .33 .29 .29 .27 .31 .27 .32 .41 .36 .36
VAL1 – .53 .51 .55 .66 .56 .41 .58 .43 .51
VAL2 – .43 .41 .47 .46 .37 .46 .34 .45
VAL3 – .46 .53 .41 .28 .39 .44 .41
VAL4 – .55 .45 .46 .47 .46 .44
VAL5 – .55 .37 .59 .47 .57
VAL6 – .42 .56 .47 .50
VAL7 – .49 .41 .45
VAL8 – .48 .60
VAL9 – .52
VAL10 –
157
158 S.Y. Lim, E. Chapman

Table 4 Fit indices for the original four-factor model

χ2 df χ2/df SRMR GFI NNFI CFI

Subsample 1 5,216.64* 734 7.11 .07 .75 .97 .97

Subsample 2 6,552.47* 734 8.93 .07 .71 .97 .97

*α=.001, level of significance

p<.01; SC: r0.92, p<.01; and VAL: r0.96, p<.01; mean r0.96). Figure 1 shows standardized
path coefficients and standard errors (in parentheses) in the modified four-factor model using
data from subsample 2. All paths in the model were positive and significant.
From Fig. 1, it is clear that the correlation between ENJ and MOT was extremely high in
subsample 2, suggesting that, in effect, these two entire subscales were redundant with one
another. CFAs were then performed on two models (Table 7), the first of which excluded the
entire ENJ subscale (model 1), the second which removed the entire MOT subscale (model 2).
Both models fit the data very well. In fact, the overall fit of the two models was somewhat
better than that of the original four-factor. Similar results were obtained using data from
subsample 1.

7 Internal consistency and test–retest reliability

The second and third aims of this study were to assess the internal consistency and test–retest
reliability of the short ATMI using SPSS 19. Table 8 presents the internal consistency scores
for both subsamples, and test–retest reliability over a 1-month period.
As indicated in Table 8, internal consistency scores were within the recommended range
for both subsamples (overall α for the full scale 0.93, and mean α for individual
subscales0.87).
According to Cohen’s (1988) guideline of at least .70 for good test–retest reliabil-
ity, test–retest reliability of the short ATMI assessed with participants in phase 3 (n0
208) over a 1-month period for all subscales was deemed satisfactory (mean rxx across
all subscales 0.75).

8 Relationships with theoretically related constructs

The last aim of this study was to investigate the pattern of correlations between
subscales of the short ATMI and theoretically related constructs—mathematics anxiety

Table 5 Internal consistency scores for both subsamples

Overall Individual subscales

ENJ MOT SC VAL Mean α

Subsample 1 (n=800) .97 .90 .86 .94 .91 .90

Subsample 2 (n=801) .96 .90 .85 .94 .90 .90
Short form of the attitudes toward mathematics inventory 159

Table 6 Fit indices for the modified four-factor model

χ2 df χ2/df SRMR GFI NNFI CFI

Subsample 1 477.86* 146 3.27 .04 .94 .98 .99

Subsample 2 435.73* 146 2.98 .04 .95 .99 .99

*α0.001 level of significance

(measured with the FS-ANX and FS-EASE subscales) and mathematics achievement
(measured with the mathematics achievement test described above). As all the four
subscales on the short ATMI measure positive affective domains, they were expected to
correlate positively with one another. These positive relationships amongst the four
subscales were reported by Tapia and Marsh (2002). Additionally, scores on all four
subscales on the short ATMI were expected to correlate positively with scores on the FS-EASE
subscale and mathematics achievement test scores, and negatively with scores on the FS-ANX
subscale. These relationships have been reported in numerous studies (e.g., Aiken,
1974; Antonnen, 1969; Fairbanks, 1992; Guay, Marsh, & Boivin, 2003; Hackett,
1985; Sherman & Fennema, 1977; Thorndike-Christ, 1991). In particular, the most
negative correlation was expected between SC and FS-ANX, as the SC subscale was
formed by collapsing two originally separate subscales measuring mathematics confi-
dence and anxiety (Tapia, 1996). On the other hand, the most positive correlation was
expected between SC and mathematics achievement (Gerardi, 1990; Witherspoon,
Speight, & Thormas, 1997).
To investigate the relationships mentioned above, the two subsamples were com-
bined and average scores for all items in each subscale were calculated. Pearson’s
product-moment correlation coefficients were then calculated using SPSS 19 and
presented in Table 9. Prior to calculating the coefficients, the distributions of all
summed variables were examined. No notable deviations from normality were
detected.
As expected, the four subscales on the short ATMI correlated positively with one another.
Furthermore, all four short ATMI subscales correlated negatively with the FS-ANX subscale, with
the strongest negative correlation observed between the SC and FS-ANX subscales (r0
−.90, p<.01). The four short ATMI subscales also correlated positively with the FS-EASE
subscale and mathematics achievement test scores. Again as predicted, amongst the
four short ATMI subscales, the SC subscale gave the strongest correlation with math-
ematics achievement test scores.

9 Discussion

The main aims of this study were to develop a short form of the ATMI (short ATMI), and to
assess the reliability and validity of this short form on a predominantly Chinese sample from
Singapore. Results of two CFAs conducted in this study supported a modified four-factor
model, in which five ENJ, one MOT, 10 SC, and five VAL items were removed. This resulted in
only five ENJ, four MOT, five SC, and five VAL items remaining in the short ATMI. All four
subscales in the short ATMI displayed very high correlations with the corresponding
original subscales in the ATMI. In addition, the short ATMI showed good internal
consistencies, and displayed similar Cronbach alpha values to those obtained for the
160 S.Y. Lim, E. Chapman

Fig. 1 Modified four-factor model of the short ATMI based on data from subsample 2
Short form of the attitudes toward mathematics inventory 161

Table 7 Fit indices for CFAs with one factor removed using data from subsample 2

χ2 df χ2/df SRMR GFI NNFI CFI

Model 1: ENJ removed 193.67* 74 2.62 .04 .97 .99 .99

Model 2: MOT removed 191.85* 87 2.21 .03 .97 .99 .99

*α0.001 level of significance

original ATMI. These results suggested that the 21 removed items were effectively
redundant.
Further to the above, a very high correlation was obtained between two of the
resulting subscales: ENJ and MOT. Two subsequent CFAs, in which each factor was
excluded in turn, produced fit indices equivalent to those of the full four-factor model.
Thus, these results suggest that in addition to using the shorter forms of the subscales
tested in this study, future users of the instrument could remove either one of the ENJ
or the MOT subscales altogether. Given that the ENJ subscale produced somewhat better
internal consistency values, and that the fit of the CFA with MOT removed was slightly
better than that produced with ENJ removed, it is recommended here that future
administrations exclude the MOT subscale. This would reduce the total number of
items in the instrument to 15, from an original total of 40.
Based on data collected from a predominantly Chinese sample from Singapore at the pre-
tertiary level, other psychometric properties of the short ATMI were highly favorable. The
short ATMI subscales displayed good internal consistencies and satisfactory test–retest
reliability over a 1-month period.
The correlations between the ATMI subscales, and their correlations with theoreti-
cally related constructs (mathematics anxiety, ease, and achievement), were all in line
with the results of numerous other empirical studies and theoretical reasoning, and
supported the construct validity of the instrument. In particular, the strong relation-
ships found between the various domains of mathematics attitudes and achievement
further attested to the robust relationships amongst these constructs that were reported
in various past studies. Since there is increasing concern about the excessive amount
of time that students are spending on surveys and questionnaires (Porter, 2004), future
studies may consider the use of multiple regression analysis to identify the key
domains of mathematics attitudes that educators should focus upon when selecting
learning process variables for mathematics achievement. However, the relationships
reported between the ATMI subscales and the other constructs were specific to the

Table 8 Internal consistency scores for both subsamples

Overall Individual subscales

ENJ MOT SC VAL M

Subsample 1 (n=800) .93 .88 .85 .90 .86 α=.87

Subsample 2 (n=801) .93 .89 .84 .90 .85 α=.87
Test–retest reliability (n=208) .73* .73* .81* .71* rxx =.75

*p<.01
162 S.Y. Lim, E. Chapman

Table 9 Correlations between subscales of the ATMI and other constructs

ENJ MOT SC VAL FS-ANX FS-EASE Test score

ENJ – .84* .56* .60* −.59* .67* .41*

MOT – .54 *
.58 *
−.56 *
.70 *
.44*
SC – .33 *
−.90 *
.66 *
.52*
VAL – −.38 *
.38 *
.27*
FS-ANX – −.64* −.50*
FS-EASE – .49*
Test score –

*p<.01

sample of this study. Hence, further investigations using samples from other cultures
and schooling levels are necessary to generalize these relationships to other samples.
For the same reason, further studies are necessary before using this instrument with
students from other countries and schooling levels.
Nevertheless, this study is one of the first few that assessed the ATMI on an Asian sample and
the results of this study suggested the short ATMI is a promising instrument for measuring
mathematics attitudes. The short form developed in this study is particularly useful both for
practitioners, who want to monitor mathematics attitudes regularly, and for researchers, who are
often using batteries of instruments in their studies. A major impediment to routine monitoring
of mathematics attitudes is the time required for administration. Further, current concerns
regarding the “overtesting” of students (e.g., Porter, 2004) often leave researchers in the
position of having to “ration” their outcome measures to reduce the overall time commitment
required. This short form could be useful in such cases. The short form developed here took
students, on average, less than 10 min to complete, making it a more viable option for regular
use. This time would reduce further with the removal of the MOT subscale.

References

Aiken, L. R. (1974). Two scales of attitude toward mathematics. Journal for Research in Mathematics
Education, 5(2), 67–71.
Altonji, J. (1995). The effect of high school curriculum on education and labor market outcomes. The Journal
of Human Resources, 30(3), 409–438.
Antonnen, R. G. (1969). A longitudinal study in mathematics attitude. The Journal of Educational Research,
62, 467–471.
Atkinson, J. W., & Raynor, J. (1974). Motivation and achievement. Washington, DC: Winston.
Ball, S. (Ed.). (1977). Motivation in education. New York: Academic.
Barkoukis, V., Tsorbatzoudis, H., Grouios, G., & Sideridis, G. (2008). The assessment of intrinsic and
extrinsic motivation and amotivation: Validity and reliability of the Greek version of the academic
motivation scale. Assessment in Education: Principles, Policy & Practice, 15(1), 39–55.
Bell, J. (1987). Doing your research project. Philadelphia: Open University Press.
Bouchey, H. A., & Harter, S. (2005). Reflected appraisals, academic perceptions, and math/science perfor-
mance during early adolescence. Journal of Educational Psychology, 97(4), 673–686.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long
(Eds.), Testing structural equation models (pp. 136–161). Newbury Park: Stage Publications.
Byrne, B. M. (1989). A primer of LISREL: Basic applications and programming for confirmatory factor
analytic models. New York: Springer.
Chamberlin, S. A. (2010). A review of instruments created to assess affect in mathematics. Journal of
Mathematics Education, 3(1), 167–182.
Short form of the attitudes toward mathematics inventory 163

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale: Erlbaum.
Dossey, J. (1992). How school mathematics functions: Perspectives from the NAEP 1990 and 1992 assess-
ments. Princeton, NJ: National Assessment of Educational Progress. Retrieved from ERIC database.
(ED377057).
Fairbanks, P. J. (1992). Treating mathematics anxiety: The optional contract. Mathematics Teacher, 85(6),
428–430.
Fennema, E., & Sherman, J. A. (1976). Fennema-Sherman mathematics attitudes scales: Instruments designed
to measure attitudes toward the learning of mathematics by males and females. Journal for Research in
Mathematics Education, 7(5), 324–326.
Foire, G. (1999). Math-abused students: Are we prepared to teach them? Mathematics Teacher, 92, 403–407.
Gerardi, S. (1990). Academic self-concept as a predictor of academic success among minority and low-
socioeconomic status students. Journal of College Student Development, 31, 402–407.
Goolsby, C. B. (1988). Factors affecting mathematics achievement in high-risk college students. Research and
Teaching in Developmental Education, 4(2), 18–27.
Guay, F., Marsh, H. W., & Boivin, M. (2003). Academic self-concept and academic achievement: Develop-
mental perspectives on their causal ordering. Journal of Educational Psychology, 95(1), 124–136.
Hackett, G. (1985). Role of mathematics self-efficacy in the choice of math-related majors of college women
and men: A path analysis. Journal of Counseling Psychology, 32, 47–56.
Hauge, S. K. (1991). Mathematics anxiety: A study of minority students in an open admissions setting.
Unpublished manuscript. Washington, DC: University of the District of Columbia. Retrieved from ERIC
database. (ED335229).
Houser, R. (1998). Counseling and educational research: Evaluation and application. London: Sage.
Hu, L.-T., & Bentler, P. (1995). Evaluating model fit. In R. H. Hoyle (Ed.), Structural equation modeling.
Concepts, issues, and application (pp. 76–99). London: Sage.
Joensen, J. S., & Nielsen, H. S. (2009). Is there a causal effect of high school mathematics on labor market
outcomes? The Journal of Human Resources, 44(1), 171–198.
Kenschaft, P. (Ed.). (1991). Winning women into mathematics. Washington, DC: Mathematical Association of
America.
Lim, S. Y., & Chapman, E. (2011). A psychometric investigation of the Fennema-Sherman Mathematics
Anxiety Scale. Manuscript submitted for publication (Copy on file with authors).
Linn, M., & Hyde, J. (1989). Gender, mathematics, and science. Educational Researcher, 18(8), 17–27.
Longitudinal Study of American Youth (1990). The international center for the advancement of scientific
literacy. Retrieved from http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/30263?q0LSAY.
Ma, X., & Kishor, N. (1997). Assessing the relationship between attitude toward mathematics and achieve-
ment in mathematics: A meta-analysis. Journal for Research in Mathematics Education, 28, 26–47.
Minato, S. (1983). Some mathematical attitudinal data on eighth-grade students in Japan measured by a
semantic differential. Educational Studies in Mathematics, 14, 19–38.
Minato, S., & Yanase, S. (1984). On the relationship between students attitude towards school mathematics
and their levels of intelligence. Educational Studies in Mathematics, 15, 313–320.
Nunnally, J. C., & Bernstein, I. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill.
Porter, S. R. (Ed.). (2004). Overcoming survey research problems: New directions for institutional research
(Vol. 121). San Francisco: Jossey-Bass.
Randhawa, B. S., Beamer, J. E., & Lundberg, I. (1993). Role of the mathematics self-efficacy in the structural
model of mathematics achievement. Journal of Educational Psychology, 85, 41–48.
Rose, H., & Betts, J. (2004). The effect of high school courses on earnings. The Review of Economics and
Statistics, 86(2), 497–513.
Samuelsson, J., & Granstrom, K. (2007). Important prerequisite for students’ mathematical achievement.
Journal of Theory and Practice in Education, 3(2), 150–170.
Sherman, J., & Fennema, E. (1977). The study of mathematics by high school girls and boys: Related
variables. American Educational Research Journal, 14, 159–168.
Singh, K., Granville, M., & Dika, S. (2002). Mathematics and science achievement effects of motivation,
interest, and academic engagement. The Journal of Educational Research, 95(6), 323–332.
Tanaka, J. S. (1993). Multifaceted conceptions of fit in structural equation models. In K. A. Bollen & J. S.
Long (Eds.), Testing structural equation models. Newbury Park: Sage Publications.
Tapia, M. (1996, November). The attitudes toward mathematics instrument. Paper presented at the annual
meeting of the Mid-south Educational Research Association, Tuscaloosa, AL. Retrieved from ERIC
database. (ED404165).
Tapia, M., & II Marsh, G. E. (2000). Attitudes toward mathematics instrument: An investigation with middle
school students. Bowling Green: Paper presented at the Annual Meeting of the Mid-South Educational
Research Association.
164 S.Y. Lim, E. Chapman

Tapia, M., & II Marsh, G. E. (2002). Confirmatory factor analysis of the attitudes toward mathematics
inventory. Chattanooga: Paper presented at the Annual Meeting of the Mid-South Educational Research
Association.
Tapia, M., & II Marsh, G. E. (2004). An instrument to measure mathematics attitudes. Academic Exchange
Quarterly, 8(2), 16–21.
Terwilliger, J., & Titus, J. (1995). Gender differences in attitudes and attitude changes among mathematically
talented youth. Gifted Child Quarterly, 39(1), 29–35.
Thorndike-Christ, T. (1991). Attitudes toward mathematics: Relationships to mathematics achievement,
gender, mathematics course-taking plans, and career interests. Bellingham, WA: Western Washington
University. Retrieved from ERIC database. (ED347066).
Witherspoon, K. M., Speight, S. L., & Thomas, A. J. (1997). Racial identity attitudes, school achievement,
and academic self-efficacy among African American high school students. Journal of Black Psychology,
23, 344–357.
Zan, R., Brown, L., Evans, J., & Hannula, M. S. (2006). Affect in mathematics education: An introduction.
Educational Studies in Mathematics, 63(2), 113–121.