Content uploaded by Tim Woodman
Author content
All content in this area was uploaded by Tim Woodman
Content may be subject to copyright.
The relative impact of cognitive anxiety and
self-confidence upon sport performance: a meta-analysis
TIM WOODMAN* and LEW HARDY
School of Sport, Health and Exercise Sciences, University of Wales, Bangor, Gwynedd LL57 2DG, UK
Accepted 21 February 2003
This meta-analysis (k = 48) investigated two relationships in competitive sport: (1) state cognitive anxiety with
performance and (2) state self-confidence with performance. The cognitive anxiety mean effect size was
r=70.10 (P 50.05). The self-confidence mean effect size was r=0.24 (P 50.001). A paired-samples t-test
revealed that the magnitude of the self-confidence mean effect size was significantly greater than that of the
cognitive anxiety mean effect size. The moderator variables for the cognitive anxiety–performance relationship
were sex and standard of competition. The mean effect size for men (r = 70.22) was significantly greater than
the mean effect size for women (r = 70.03). The mean effect size for high-standard competition (r = 70.27) was
significantly greater than that for comparatively low-standard competition (r = 70.06). The significant
moderator variables for the self-confidence–performance relationship were sex, standard of competition and
measurement. The mean effect size for men (r = 0.29) was significantly greater than that for women (r = 0.04)
and the mean effect size for high-standard competition (r = 0.33) was significantly greater than that for low-
standard competition (r = 0.16). The mean effect size derived from studies employing the Competitive State
Anxiety Inventory-2 (r = 0.19) was significantly smaller than the mean effect size derived from studies using
other measures of self-confidence (r = 0.38). Measurement issues are discussed and future research directions
are offered in light of the results.
Keywords: cognitive anxiety, meta-analysis, self-confidence, sport performance.
Introduction
The relationship between anxiety and sport perfor-
mance has attracted much research attention over the
past 20 years, and researchers have tried to clarify this
relationship by advancing severa l models and theories.
These include multidimensional anxiety theory (Mar-
tens et al., 1990), catastrophe models (Hardy , 1990,
1996a), reversal theory (Apter, 1982; Kerr, 1990) and
zones of optimal functioning models (Hanin, 1980,
1986).
In multidimensional anxiety theory, Martens et al.
(1990) proposed a series of two-dimensional rel ation-
ships between cognitive anxiety, somatic anxiety, self-
confidence and performance. Cognitive anxiety was
defined as ‘negative expectations and cognitive con-
cerns about oneself, the situation at hand, and potential
consequences’ (Morris et al ., 1981, p. 541). Somatic
anxiety was conceptualized as the perception of one’s
physiological arousal. Self-confidence was conceptua-
lized as one’s belief in meeting the challenge of the task
to be performe d. In multidimensional anxiety theory
(Martens et al., 1990), cognitive anxiety is hypothesized
to have a negative linear relationship with performance;
somatic anxiety is hypothesized to have a quadratic
(inverted-U shaped) relationship with performance; and
self-confidence is hypothesized to have a positive linear
relationship with performance.
The hypothesized neg ative linear relationship be-
tween cognitive anxiety and performance was largely
based upon theories of attention (e.g. Wine, 1971,
1980), whereby cognitive resources are taken up by
worrying thoughts and so are not available for use on
the task at hand. As Martens et al. (1990) conceptua-
lized cognitive anxiety and self-confidence as lying at
opposite ends of a continuum, they hypothesized that
self-confidence and performance would be related in a
positive linear fashion. However, the rationale for the
hypothesized inverted-U relationship between somatic
anxiety and performance is much less clear. Martens et
al. (1990) cited Weinberg’s (1978) research, which
suggests that too much muscular tension will le ad to a
deterioration in performance. Howeve r, Martens et al.
* Author to whom all correspondence should be addressed.
e-mail: t.woodman@bangor.ac.uk
Journal of Sports Sciences, 2003, 21, 443–457
Journal of Sports Sciences ISSN 0264-0414 print/ISSN 1466-447X online
#
2003 Taylor & Francis Ltd
DOI: 10.1080/0264041031000101809
offered no theoretical explanation for the hypothesized
curvilinear relationship between the perception of one’s
physiological arousal (i.e. somatic anxiety) and perfor-
mance (see Woodman and Hardy, 2001a). Thus,
although somatic anxiety is a useful indirect measure
of the physiological indices of anxiety, it is of limited
theoretical value in explaining the relationship between
physiological arousal and performance. Cons equently,
we focus here only on the effects of cognitive anxiety
and self-confidence upon sport performance.
Several investigations have been cond ucted to test
the proposed relationships between cognitive anxiety
and performance and between self-confidence and
performance. For example, Burton (1988) found a
negative linear trend between cognitive anxiety and
swimming performance and a positive linear trend
between self-confidence and performa nce. In the two
samples investigated by Burton, cognitive anxiety
accounted for up to 46% of swimming performance
variance and self-confidence accounted for up to 21%.
Gould et al. (1984) also found a significant negative
linear relationship between cognitive anxiety and
performance, but no significant trend between self-
confidence and performance. Conversely, Mart in and
Gill (1991) found self-confidence to be significantly
and positively related to distance running perfor-
mance, but found no significant relationship between
cognitive anxiety and running performance. Similarly,
in their study of pistol shoo ters, Gould et al. (1987)
found no significant relationship between cognitive
anxiety and performance. However, in that study, a
significant negative relationship between self-confi-
dence and performance was revealed. Other studies
have revea led no significant relationships between
cognitive anxiety and performance (Maynard and
Cotton, 1993; Hammermeister and Burton, 1995 ;
Vadocz et al., 1997) or between self-confidence and
performance (Williams and Krane, 1992; Maynard
and Cotton, 1993). Thus, the relative impact of
cognitive anxiety and self-confidence upon competitive
sport performance remains unclear .
The inventory that was used to measure cognitive
anxiety and self-confidence in most of the above studies
was the Compe titive State Anxiety Inventory-2 (CSAI-
2; Martens et al., 1990). The CSAI-2 was origin ally
intended to be an anxiety scale comprising two
subscales: cognitive anxiety and somatic anxiety. How-
ever, in the exploratory factor analysis of the items
comprising the CSAI-2, Martens et al. (1990) found
that the cognitive anxiety items effectively separated
into two factors, one that included negatively phrased
items and one that included positively phrased items.
These factors were subsequently labelled cognitive
anxiety and self-confidence, respectively. Thus, a self-
confidence subscale was also included in the CSAI-2.
In the discussion of their factor analyses, Martens et al.
(1990) stated: ‘These findings sugg est that cognitive A-
state and state self-confidence represent opposite ends
of a cognitive evaluation continuum, state self-con-
fidence being viewed as the absence of cognitive A-
state, or conversely, cognitive A-state being the lack of
state self-confidence’ (p. 129).
Given that cognitive anxiety and self-confidence
emerged as orthogonal (i.e. independent) factors in
these factor analyses, it is surprising that Martens et al.
(1990) should view them as bipolar (i.e. interdepen-
dent). Furthermore, there appears to be sufficient
evidence to suggest that cognitive anxiety and self-
confidence are meaningfully distinct constructs (Bur-
rows et al., 1977; Thayer, 1978; Gould et al., 1984,
1987; Hardy and Whitehead, 1984; Jones and Cale,
1989; Hardy, 1996b; Parfitt and Pates, 1999). For
example, although Gould et al. (1984) found a
significant negative linear relationship between cogni-
tive anxiety and performance, they found no signifi-
cant trend between self-confidence and performance.
Also, in their work on the antecedents and temporal
patterning of cognitive anxiety and self-confidence,
Jones et al. (1990, 1991) provided more evidence for
the relative independence of cognitive anxiety and self-
confidence. Finally, both Hardy (1996b) and Parfitt
and Pates (1999) found that self-confidence accounted
for a significant proportion of performance variance
over and above that accounted for by cognitive
anxiety.
In light of the discrepant results revealed between
different studies that have reported cognitive anxie ty–
performance and self-confidence–performance relation-
ships, it is important to consider which variables might
be moderating these relationships. We consider three
major moderator variables: (a) measurement, (b) type
of sport and (c) individual differences.
Measurement
Intra-individual versus inter-individual measurement
Many researchers (e.g. Sonstroem and Bernardo,
1982; Burton, 1988) have contended that inter-
individual measurements are inappropriate when ex-
amining the relationships between anxiety and perfor-
mance, as such measurements are not sensitive to
individual differences in anxiety or performance. As
intra-individual measurements of anxiety, self-confi-
dence and performance control for such differen ces,
we hypothesized that the relationships between cogni-
tive anxiety and performance and between self-
confidence and performance would be stronger when
these constructs were measured intra-individually
rather than inter-individually.
444 Woodman and Hardy
Questionnaire
Several criticisms have been levelled at the Competitive
State Anxiety Inventory-2 as a measure of pre-
competition affect. These criticisms include the use of
the term ‘concern’ as a measure of cognitive anxiety
(Burton and Naylor, 1997; Woodman and Hardy,
2001a), its lack of specificity in relation to the task
(Moritz et al., 2000) and its poor overall fit (Lane et al.,
1999). However, researchers continue to use the CSAI-
2, probably because there is no obvious alternative
validated questionnaire that measures pre-competition
anxiety and self-confidence. In light of these issues, we
examined the use of the CSAI-2 (as opposed to other
measures of com petitive state cognitive anxiety and self-
confidence) as a possible moderator of the relationships
between cognitive anxiety and performance and be-
tween self-confidence and performance.
Type of sport
Sports can be broadly categorized into team sports and
individual sports. As there may be more pressure and
personal exposure associated with individual sports
than team sports, we hypothesized that cognitive
anxiety and self-confidence would be more strongly
associated with athletes’ perfo rmance in individual
sports.
Individual differences
Standard of competition
High-standard competition may be associated with
increased pressure. Cognitive anxiety probably reflects,
in part, athletes’ inability to deal with this pressure.
Conversely, self-confidenc e probably reflects, in part,
athletes’ ability to deal with this increased pressure.
Thus, cognitive anxiety and self-confidence are more
likely to affect subsequent perfo rmance in high-stan-
dard competitive settings. Thus, we hypothesized that
the relationships between cognitive anxiety and perfor-
mance and between self-confidence and performance
would be stronger for high-standard athletes than
relatively low-standard athletes.
Sex
Women typically report higher cognitive anxiety and
lower self-confidence than men (cf. Martens et al.,
1990; Jones et al., 1991). If it is accepted that there is a
cognitive anxiety threshold beyond which cognitive
anxiety will more likely affect performance, then
cognitive anxiety and performance should be more
strongly related for women if this threshold is high. This
is because women’s cognitive anxiety would be dis-
tributed below and above the threshold, whereas men’s
cognitive anxiety would largely be below the threshold.
Similarly, cognitive anxiety and performance may be
more strongly related for men if this threshold is low, as
men’s cognitive anxiety would be distributed below and
above this threshold, whereas women’s cognitive
anxiety would largely be above the threshold. The same
argument holds for self-confidence. That is, if there is a
threshold below which self-confidence will more likely
affect performance, then self-confidence may be more
strongly related to performance for men if this threshold
is high, as men’s self-confidence would be distributed
below and above this threshold, whereas most women’s
self-confidence would largely be below the threshold.
Similarly, self-confidence and performance should be
more strongly related for women if this threshold is low,
as women’s self-confidence would be distributed below
and above the threshold, whereas men’s self-confidence
would be largely above such a threshold. As the
existence of such thresholds is largely speculative, we
investigated sex as a moderator variable but did not
formulate any specific hypotheses regarding the differ-
ences between the sexes in the strength of relationships
between cognitive anxiety and performance or between
self-confidence and performance.
In summary, the aims of the present meta-analysis
were threefold: (1) To examine the fundamental
predictions of multidimensional anxiety theory; namely,
that cognitive anxiety has a negative relationship with
performance and that self-confidence has a positive
relationship with performance. (2) To examine the
relative magnitude of the cognitive anxiety and self-
confidence effect sizes to identify which (if either) is the
more important and whether it is empirically reasonable
to consider them as lying at opposite ends of the same
continuum. (3) To examine the moderating variables in
the relationships between cognitive anxiety and perfor-
mance and between self-confidence and perfo rmance.
Methods
Literature search
Computer-based literature searches were conducted to
locate published and unpublished research on cogni-
tive anxiety, self-confidence and performance. The
databases used for this search were: Applied Social
Sciences Index and Abstracts (ASSIA), Bath Informa-
tion and Data Services (BIDS), PsycINFO, PsycLIT,
Social Science Citation Index (SSCI) and Sport
Discus. The last search was conducted at the begin-
ning of January 2002. Keywords used for the searches
were: ‘cognitive anxiety’, ‘confidence’, ‘sport’ and
445Cognitive anxiety and self-confidence meta-analysis
‘performance’. Several ‘wild card’ searches were also
conducted to ensure that the search did not miss
studies containing related words such as ‘anxiety’,
‘worry’ and ‘competition’. The reference lists of the
located studies were examined for further possible
articles that might fulfil the criteria for inclusion.
Studies were included in the meta-analysis if they
fulfilled the following criteria:
1. A measure of state cognitive anxiety or state self-
confidence was taken before a sport competition.
2. Comp etitive sport performance was measured in a
field setting.
Statistical methods
The meta-analytic procedures used in the present study
are described in Rosenthal (1991). Effect sizes were
calculated for those studies that satisfied the criteria for
inclusion. The correlation coefficients (r) between
cognitive anxiety and performance and betwe en self-
confidence and performance were used to compute
effect sizes. As the population value of r gets further
from zero, the distribution of r’s becomes more and
more skewed (Rosenthal, 1991). Fisher’s (1928)
transformation converts r to z
r
, which results in a more
normal distribution. Hence, the present study used z
r
as
an estimate of effect size. The transformation from r to
z
r
is:
z
r
¼ 0:5 log
e
½ð1 þ rÞ=ð1 rÞ
To calculate the significance of the effect sizes, the
standard normal deviate Z was used. The transforma-
tion from r to Z is:
Z ¼ r
p
n
where n = sample size.
The cognitive anxiety Z’s were reversed to reflect
the expected (negative) direction of the effect. For
example, if r=70.20 and n = 100, then Z =2. If no
data were available to calculate the effect size (r)or
the level of significance (P, one-tailed), the primary
author of the study in question was contacted by
telephone or electronic mail. If clarification of the
data was not obtained from the primary author, P
was assumed to be 0.50 and r was assumed to be
0.00. This is because the omission of studies that
report non-significant results can artificially inflate
the effect size. However, this procedure is conserva-
tive and can result in effect size estimates that are
too low. Thus, following Rosenthal’s (1995) recom-
mendations, both procedures are presented in the
present study.
The following methods (Rosenthal, 1991) were used
for transforming a t statistic to r,oranF ratio to r,
respectively:
r ¼½t
2
=ðt
2
þ dfÞ
0:5
where df (the degrees of freedom) = n
1
+n
2
72, and
r ¼fF
1
;
-
=½F
1
;
-
þdf errorg
0:5
where F
1,-
represents any F with one degree of freedom
in the nume rator.
If more than one effect size estimate was availa ble
from one study, the method of mean result (Rosenthal,
1991) was employed. That is, each r from the study was
first converted to z
r
before calculating the mean of these
transformed effect sizes. To calculate the standard
normal deviate Z, the mean z
r
was converted back to r
using the following equation:
r ¼ðe
2zr
1Þ=ðe
2zr
þ 1Þ
where e is the base of the system of natural logarithms
(e&2.71828).
Study characteristics
Of the 48 studies retai ned for the meta-analysis, 46
contributed a cognitive anxiety effect size estimate and
43 contributed a self-confidence effect size estimate.
Forty-one of the 48 studies contributed both cognitive
anxiety and self-confidence effect size estimates to the
meta-analysis. Thirty-three studies were reported be-
tween 1991 and 2001, 14 studies were reported between
1981 and 1990, and one study was reported in 1979.
Forty-four studies were reported in journals and four
studies were reported in theses (three master’s theses
and one doctoral thesis). We decided to include the
results from theses in the meta-analyses to reduce the file
drawer threat (see File drawer analysis). However, given
that theses undergo a less stringent review process than
articles published in peer-review journals, we inclu ded
the source of the research (i.e. peer-review journal or
thesis) as a moderator variable in the analyses (see
Moderator variables). Finally, of the 14 authors con-
tacted for further information about the data, 10 (71%)
replied and 8 (57%) provided the necessary information
for the correlation coefficient not to be assumed as r=0.
Results
Outliers
Outliers were defined as values greater than 1.5 box-
lengths from the box, where the box represents the
446 Woodman and Hardy
range of scores from the 25th to the 75th percentile.
These outliers were removed from the data set. As a
result of this procedure, three cognitive anxiety effect
sizes and one self-confidence effect size were removed
from the data set. Consequently, the data set contained
47 studies, including 43 cognitive anxiety effect sizes
and 42 self-confidence effect sizes. A summary of all the
studies included in the meta-analysis is presented in
Table 1.
Descriptive statistics
Table 2 displays a stem-and-leaf plot of the cognitive
anxiety effect sizes included in the meta-analysis.
Table 3 displays a stem-and-leaf plot of the self-
confidence effect sizes included in the meta-analysis.
Table 4 contains information about central tendency,
variability, significance tests and confidence intervals
(using studies as the sampling unit) for the cognitive
anxiety data. This table presents two sets of results:
one with all cognitive anxiety studie s, the other
without those studies where r was assumed to be
zero. Table 5 contains this information for the self-
confidence data.
Effect sizes and significance testing
Cognitive anxiety
Of the 43 studies reporting a relationship between
cognitive anxiety and performance, 26 (60%) reported a
negative relationship, 7 (16%) reported non-significant
results (so r was assumed to be zero) and 10 (23%)
reported a positive relationship. The mean effe ct size
was 70.10. When studies were weighted for degrees of
freedom, the mean effect size was 70.11. When those
studies where the effect size was assumed to be 0 were
omitted from the analyses, the mean effect size was
70.12 and the weight ed mean effect size was 70.13 .
The Stouffer Z associated with the mean effect size was
statistically significant (Z = 4.73, P 50.001). The t-test
for the mean z
r
was also significant (t
42
= 2.73,
P 50.01).
Self-confidence
Of the 42 studies reporting a relationship between
self-confidence and performance, 32 (76%) reported a
positive relationship, 6 (14%) reported non-significant
results (so r was assumed to be zero) and 4 (10%)
reported a neg ative relationship. The mean effect size
was 0.24. When studies were weighted for deg rees of
freedom, the mean effect size was 0.23. When studies
where the effect size was assumed to be 0 were
omitted from the analyses, the mean effect size was
0.27 and the weighted mean effect size was 0.27. The
Stouffer Z associated with the mean effect size was
statistically significant (Z = 10.90, P 50.001). The t-
test for the mean z
r
was also significant (t
41
= 6.38,
P 50.001).
File drawer analysis
Non-significant results are less likely to be published
and more likely to remain in the file drawers of
researchers’ laboratories (Rosenthal, 1991). If adding
only a few such non-significant studies renders the
mean effect size non-significant, then the findings of a
meta-analysis are not robust to the threat posed by
studies hidden away in researchers’ file drawers.
Rosenthal (1991) suggested some simple calcul ations
for determining the extent to which a meta-analysis is
robust to this file drawer threat. The two questions that
are addressed here are: (1) How many non-significant
studies (where r=0, P=0.50) would have to be
unearthed to make the probability of the effect size
determined by the meta-analysis non-significant? (2)
What constitutes an unlikely number of unearthed non-
significant studies? If the number of non-significant
studies that would have to be unearthed is greater than
the ‘unlikely number of unearthed non-significant
studies’, then the meta-analysis is said to be robust to
the file drawer threat. The following figures for
cognitive anxiety and self-confidence are based on fairly
conservative calculations suggested by Rosenthal
(1991).
Cognitive anxiety
For the probability of the cognitive anxiety effect size to
become non-significant (P 40.05), 312 studies with a
mean probability of 0.50 would hav e to be stored away
in researchers’ file drawers. A figure of 225 would have
been considered robust to the file drawer threat. Thus,
the cognitive anxiety data are robust to the file drawer
threat.
Self-confidence
For the probability of the self-confidence effect size to
become non-significant, 1801 studies with a mean
probability of 0.50 would have to be stored away. A
figure of 220 would have been considered robust to the
file drawer threat. Thus, the self-confidence data are
highly robust to the file drawer threat.
Moderator variables
Heterogeneity tests revealed that the effect sizes were
heterogeneous for cognitive anxiety (w
2
42
¼ 146:73,
447Cognitive anxiety and self-confidence meta-analysis
Table 1. Summary of the studies (k = 47) included in the meta-analysis
Cognitive
anxiety
Self-
confidence
Authors Measures Sport nr Z r Z
Barnes et al. (1986) CSAI-2 Swimming
14 70.39 1.46 0.19 0.71
Bejek and Hagtvet (1996) CSAI-2 Artistic gymnastics
69 70.09 0.76 0.09 0.72
Bird and Horn (1990) CSAI-2 Softball
161 0.21 72.63 0.05 0.58
Burton (1988) CSAI-2 Swimming
98 70.39 3.85 0.30 2.97
Chapman et al. (1997) CSAI-2 Tae kwon-do
142 70.37 4.36 0.43 5.10
Cox et al. (2001) ARS-2 Basketball
248 70.13 2.06 0.15 2.33
Duesing (1984) CSAI-2 Middle-/long-distance
running
40 0.31 71.97
Edwards and Hardy (1996) CSAI-2 Netball
45 0.10 70.67 70.17 70.12
Gayton and Nickless (1987) SSCI Marathon
35 0.36 2.13
Gould et al. (1981) Wrestling questionnaire Wrestling
49 0.20 71.42 0.52 3.64
Gould et al. (1984) CSAI-2 Wrestling
37 70.29 1.74 0.02 0.09
Gould et al. (1987) CSAI-2 Pistol shooting
39 .0* 0.00 70.27 71.67
Gould et al. (1993) CSAI-2 Middle-/long-distance
running
11 70.07 0.23
Grasso (1999) CSAI-2 Basketball
42 70.12 0.76 0.18 1.17
Guest and Cox (1999) MRF-3 Golf
216 70.27 3.97 0.35 5.14
Hammermeister and Burton (1995) CSAI-2 Endurance sports
293 70.08 1.37
Hardy (1996a) CSAI-2 Golf
8 0.10 70.27 0.16 0.44
Highlen and Bennett (1979) Wrestling questionnaire Wrestling
39 0.56 3.47
Jerome and Williams (2000) CSAI-2 Bowling
143 .0* 0.00 .0* 0.00
Jones et al. (1993) CSAI-2 Artistic gymnastics
48 70.01 0.07 0.29 2.01
Krane and Williams (1987) CSAI-2 Golf and gymnastics
80 .0* 0.00 .0* 0.00
Krane et al. (1992) CSAI-2 Golf
100 0.04 70.40 0.07 0.70
Krane (1993) CSAI-2 Soccer
16 .0* 0.00 .0* 0.00
Martin and Gill (1991) CSAI-2 & SSCI Middle-/long-distance
running
86 70.10 0.86 0.57 4.83
Maynard and Howe (1987) CSAI-2 Rugby
22 70.20 0.93 70.01 70.05
Maynard and Cotton (1993) CSAI-2 Field hockey
20 .0* 0.00 .0* 0.00
Maynard et al. (1995) CSAI-2 Soccer
24 70.14 0.66 0.40 1.94
McAuley (1985) CSAI-2 Golf
7 70.11 0.28 0.01 0.02
McCann et al. (1992) CSAI-2 Road cycling
23 70.42 2.01 0.37 1.77
McKay et al. (1997) CSAI-2 Golf
15 0.07 0.27
Moraes (1987) CSAI-2 Judo
70 .0* 0.00 .0* 0.00
Parfitt and Pates (1999) CSAI-2 Basketball
12 70.07 0.26 0.49 1.69
Perreault and Marisi (1997) CSAI-2 Wheelchair basketball
37 70.02 0.23 70.02 70.15
Psychountaki and Zervas (2000) SCWI-C & SSCQ-C Swimming
143 70.12 1.44 0.22 2.58
Rodrigo et al. (1990) CSAI-2 Soccer
51 70.52 3.71 0.16 1.14
Smith et al. (2001) MRF-3 Volleyball
12 70.54 1.87 0.44 1.52
Swain and Jones (1996) CSAI-2 Basketball
10 70.18 0.57 0.34 1.07
Taylor (1987) CSAI-2 Mixture
84 0.35 72.09 0.34 2.05
Terry and Slade (1995) CSAI-2 Karate
208 70.46 6.49 0.42 5.92
Terry et al. (1996) CSAI-2 Tennis
100 70.12 1.15 0.42 4.20
Thelwell and Maynard (1998) CSAI-2 Cricket
20 70.32 1.43 0.64 2.86
Vadocz et al. (1997) CSAI-2 Roller skating
57 .0* 0.00 0.51 4.48
Wiggins and Henson (2000) CSAI-2 Tennis
7 0.05 70.13
Williams and Krane (1992) CSAI-2 Golf
83 70.22 2.00 .0* 0.00
Woodman et al. (1997) CSAI-2 Bowling
25 0.05 70.25
Yang (1994) CSAI-2 Mixture
56 0.49 3.67
Zhu and Fang (1998) CSAI-2 Distance running
88 0.39 73.69 0.26 2.39
Note: CSAI-2 = Competitive State Anxiety Inventory-2; ARS-2 = Anxiety Rating Scale-2; SSCI = State Sport-Confidence Inventory; MRF-
3 = Mental Readiness Form-3; SCWI-C = State Competitive Worries Inventory for Children; SSCQ-C = State Sport Confidence Questionnaire for
Children.
* Not significant, effect size assumed to be zero, P = 0.50, one-tailed.
448 Woodman and Hardy
P 50.001) and self-confidence (w
2
41
¼ 138:29,
P 50.001).
As the Rosenthal method can inflate effect size
estimates in the heterogeneous case (see Field, 2001),
we re-ran the meta-analysis using the Hunter-Schmidt
method (Hunter and Schmidt, 1990; Schmidt and
Hunter, 1999), which is more conservative. This
analysis revealed similar results for both cognitive
anxiety (mean r=70.11, Z = 3.32, P 50.001; mean r
excluding non-significant results = 70.12, Z = 3.36,
P 50.001) and self-confidence (mean r=0.22,
Z = 7.33, P 50.001; mean r excluding non-significant
results = 0.26, Z = 8.28, P 50.001). Consequently, we
proceeded with the Rosenthal method only.
The heterogeneity of the effect sizes suggests that
other factors were moderating the relationships between
Table 2. Cognitive anxiety stem-and-leaf plot
Stem
Leaf (with all studies
included), k=43 Stem
Leaf (excluding r=0
results), k=36
+0.4 +0.4
+0.3 1 4 9 +0.3 1 4 9
+0.2 0 0 +0.2 0 0
+0.1 0 +0.1 0
+0.0 00000003559 +0.0 3559
70.0 126789 70.0 126789
70.1 001223379 70.1 001223379
70.2 1 7 8 70.2 1 7 8
70.3 2689 70.3 2 6 8 9
70.4 2 6 70.4 2 6
70.5 2 4 70.5 2 4
70.6 70.6
Table 3. Self-confidence stem-and-leaf plot
Stem
Leaf (with all studies
included), k=42 Stem
Leaf (excluding
r=0 results), k=36
+0.7 +0.7
+0.6 4 +0.6 4
+0.5 1 2 6 7 +0.5 1 2 6 7
+0.4 0223499 +0.4 0223499
+0.3 044567 +0.3 044567
+0.2 2 6 9 +0.2 2 6 9
+0.1 56689 +0.1 56689
+0.0 000000125779 +0.0 125779
70.0 1 2 70.0 1 2
70.1 7 70.1 7
70.2 7 70.2 7
Table 4. Statistical summary of the cognitive anxiety studies (k = 43) included in the meta-analysis
Statistic
Value (including assumed r=0 results),
k=43
Value (excluding assumed r=0 results),
k=36
Central tendency (r)
Unweighted mean 70.10 70.12
Weighted mean 70.11 70.13
Significance tests
Combined Stouffer Z (SZ/Hk) 4.73, P 5 0.001 5.17, P 5 0.001
t-test for mean z
r
2.73, P 5 0.01 2.77, P 5 0.01
Variability (r)
Maximum 0.39 0.39
Quartile 3 ( Q
3
) 0.00 0.05
Median 70.08 70.11
Quartile 1 ( Q
1
) 70.22 70.28
Minimum 70.54 70.54
Q
3
7Q
1
0.22 0.33
Standard deviation (SD) 0.22 0.24
Standard error (SE; SD/Hk) 0.03 0.04
Confidence intervals (r)
90% (r + critical t
(df = k 71)
SE) 70.15 to 70.04 70.18 to 70.05
95% (r + critical t
(df = k 71)
SE) 70.16 to 70.03 70.19 to 70.04
99% (r + critical t
(df = k 71)
SE) 70.18 to 70.01 70.22 to 70.01
449Cognitive anxiety and self-confidence meta-analysis
Table 5. Statistical summary of the self-confidence studies (k = 42) included in the meta-analysis
Statistic
Value (including assumed r=0 results),
k=42
Value (excluding assumed r=0 results),
k=36
Central tendency (r)
Unweighted mean 0.24 0.27
Weighted mean 0.23 0.27
Significance tests
Combined Stouffer Z (SZ/Hk) 10.90, P 5 0.001 11.77, P 5 0.001
t-test for mean z
r
6.38, P 5 0.001 6.96, P 5 0.001
Variability (r)
Maximum 0.64 0.64
Quartile 3 ( Q
3
) 0.42 0.43
Median 0.20 0.30
Quartile 1 ( Q
1
) 0.01 0.08
Minimum 70.27 70.27
Q
3
7Q
1
0.41 0.35
Standard deviation (SD) 0.22 0.22
Standard error (SE; SD/Hk) 0.03 0.04
Confidence intervals (r)
90% (r + critical t
(df = k 71)
SE) 0.18 to 0.29 0.21 to 0.33
95% (r + critical t
(df = k 71)
SE) 0.17 to 0.30 0.20 to 0.34
99% (r + critical t
(df = k 71)
SE) 0.15 to 0.32 0.18 to 0.37
Table 6. Summary of the effect sizes (r) for the moderator variables
Cognitive anxiety mean effect size Self-confidence mean effect size
Measurement Inter-individual Intra-individual Inter-individual Intra-individual
70.08 70.11 0.24 0.20
(70.17 to 0.01) (70.21 to 70.01) (0.17 to 0.31) (0.04 to 0.36)
CSAI-2 Other CSAI-2 Other
70.09 70.18 0.19 0.38
a
(70.16 to 70.02) (70.54 to 0.18) (0.11 to 0.27) (0.22 to 0.54)
Sport type Individual Team Individual Team
70.09 70.14 0.25 0.19
(70.19 to 0.01) (70.26 to 70.02) (0.16 to 0.34) (0.05 to 0.34)
Individual differences High standard Low standard High standard Low standard
70.27** 70.06 0.33* 0.16
(70.43 to 70.11) (70.13 to 0.01) (0.19 to 0.47) (0.06 to 0.25)
Men Women Men Women
70.22*** 70.03 0.29*** 0.04
(70.34 to 70.10) (70.11 to 0.05) (0.18 to 0.40) (70.05 to 0.13)
a
Significantly higher than the CSAI-2 (P 5 0.05). * Significantly higher than low standard (P 5 0.05). ** Significantly higher (in absolute terms)
than low standard (P 5 0.01). *** Significantly higher (in absolute terms) than women (P 5 0.005). Confidence intervals (95%) are presented in
parentheses.
450 Woodman and Hardy
cognitive anxiety and performance and betwe en self-
confidence and performance. Measurement, sport type
and individual differences were considered to be
possible moderator variables. The results of these
analyses are given below and a summary is presented
in Table 6.
Measurement
Two measurement questions were addressed: (1) Did
the study employ an intra-individual or an inter-
individual design? (2) Was cognitive anxiety (or self-
confidence) measured using the CSAI-2 or using
another measure?
Intra-individual versus inter-individual measurement. We
coded studies based on whether the cognitive anxiety,
self-confidence and performance measurements were
either inter-individual or intra-individual. In light of the
increased sensitivity of intra-individual measurements,
we hypothesized that such measurements would yield
stronger effect sizes than inter-individual measure-
ments. However, independent means t-tests reveal ed
no significant difference between inter-individual and
intra-individual measurements for either cognitive
anxiety effect sizes (t
41
= 0.51, P=0.31) or self-con-
fidence effect sizes (t
40
= 0.54, P=0.30).
CSAI-2 versus other measures. Despite the criticisms
that have been levelled at the Competitive State
Anxiety Inventory-2, it remains the questionnaire of
choice for most researchers interested in the relation-
ships between cognitive anxiety, self-confide nce and
sport performance. To test whether the use of the
CSAI-2 moderates the relationships between cogni-
tive anxiety and performance and between self-
confidence and performance, we categorized studies
into those that used the CSAI-2 and those that used
other measures of cognitive anxiety or self-confi-
dence. For self-confidence, an independent means t-
test revealed that the mean effect size for studies
employing the CSAI-2 was significantly smaller than
the mean effect size for studies employing other
measures of self-confidence (t
39
= 2.14, P 50.05).
For cognitive anxiety, an independent means t-test
revealed no significant difference between the effect
sizes for studies employing the CSAI-2 and studies
employing other measures of cognitive anxiety
(t
41
= 0.91, P = 0.37).
Sport type
We hypothesized that the cognitive anxiety and self-
confidence effect sizes would be higher for individual
sports than team sports. However, independent means
t-tests revealed no significant differences between
individual and team sports for either the cognitive
anxiety effect sizes (t
36
= 0.64, P=0.26) or the self-
confidence effect sizes (t
36
= 0.85, P=0.20).
Individual differences
Two individual-difference comparisons were made:
standard of competition (high- and low-standard
athletes) and sex (men and women).
Standard of competition. It should be noted that
although the skill of the athlete and the standard of
competition are likely to be highly related, strictly
speaking the high- versus low-standard distinction
reflects the competitive setting rather than the skill of
the athlete. Studies were classified as ‘high standard’
if the sample studied was competing at national or
international standard. Studies were classified as ‘low
standard’ if the sample was competing at a compe-
titive standard below national standard (e.g. state,
regional, etc.). In line with the hypotheses, an
independent means t-test revealed that the mean
cognitive anxiety effect size of high-standard athletes
was significantly larger than the mean effect size of
low-standard athletes (t
39
= 2.93, P 50.01). Also, an
independent means t-test revealed that the mean self-
confidence effect size for high-standard athletes was
significantly larger than that for low-standard athletes
(t
37
= 2.23, P 50.05).
Sex. The mean cognitive anxiety effect size for men was
significantly larger than the mean effect size for women
(t
24
= 2.84, P 50.005). Also, the self-confidence effect
sizes for men were significantly larger than those for
women (t
25
= 3.19, P 50.005).
The possible confound of the standard of the
competitive setting and sex is addressed in the discus-
sion.
Publication status
As the meta-analysis included research results from
published and unpublished sources, we included the
publication status as a possi ble moderator variable.
Although it is possible that peer-review journals are
more likely to accept manuscripts that report sig-
nificant findings, we did not expect the source of the
research to be a significant moderator variable for
either relationship. Independent means t-tests con-
firmed that there were no significant differences
between peer-review journal articles and theses for
either the cognitive anxiety effect sizes (t
34
= 1.29,
P=0.21) or the self-confidence effect sizes (t
34
= 0.50,
P=0.62).
451Cognitive anxiety and self-confidence meta-analysis
The relative impact of cognit ive anxiety and
self-confidence
If cognitive anxiety and self-con fidence lie at opposite
ends of the same continuum, then they should have a
correlation of approximately r = 71 and their effects on
performance should mirror each other. That is, the
strength of the relationship between self-confidence and
performance should be similar to the strength of the
relationship between cognitive anxiety and perfor-
mance, only in the opposite direction. If cognitive
anxiety and self-confidence affect sport performance
independently, the strength of their relationships with
performance will probably be different. Thus, a pai red
samples t-test was run between the cognitive anxiety
and self-confidence effect sizes to determine whether
cognitive anxiety and self-confidence were indepen-
dently related to performance. To make meaningful
comparisons between cognitive anxiety and self-con-
fidence, cognitive anxiety effect sizes were first trans-
formed using y = 7x.
The paired samples t-test (with all effect sizes
included) revealed a significant difference between
cognitive anxiety and self-confidence effect sizes
(t
37
= 2.22, P 50.05). Wh en non-significant effect sizes
(i.e. those effect sizes where r = 0 was assumed) were
removed, a significant difference between cognitive
anxiety and self-confidence effect sizes remained
(t
29
= 2.42, P 50.05).
To test the degree of co-dependence between the
cognitive anxiety and self-confidence effects, correla-
tion coefficients were calculated between the effect sizes
for cognitive anxiety and self-confidence from those
studies that reported both effect sizes. The correlation
between the effect sizes for cognitive anxiety and self-
confidence was not significant (r = 70.25, P = 0.13).
Equally, when the non-si gnificant effect sizes (i.e. those
effect sizes where r = 0 was assumed) were removed
from the analyses, the correlation was not significant
(r = 70.22, P = 0.25).
Discussion
The focus of this meta-analysis was on two relation-
ships: (1) the relationship between cognitive anxiety and
competitive sport performance and (2) the relationship
between self-confidence and competitive sport perfor-
mance. The mean effect size for cognitive anxiety was
r = 70.10; the mean effect size for self-confidence was
r = 0.24. Both of these mean effect sizes were signifi-
cant, thus supporting two of the fundamental predic-
tions of multidimensional anxiety theory (Martens et
al., 1990). Sex and competitive standard were signifi-
cant moderating variables for the relationship between
cognitive anxiety and performance. Sex, competitive
standard and measurement were significant moderating
variables for the relationship between self-confidence
and performance. The results also revealed that self-
confidence was significantly more strongly related to
sport performance than was cognitive anxiety.
Both sets of effect sizes (cognitive anxiety and self-
confidence) were heterogeneous. Sex and competitive
standard were identified as significant moderating
variables, with the mean effect sizes being significantly
higher for men and high-standard athletes for both
cognitive anxiety and self-confidence. Also, measure-
ment was identified as a moderating variable for self-
confidence with the CSAI-2 measurements revealing
significantly lower effect sizes than other measurements
of self-confidence. The differences in mean effect sizes
between the sexes suggest that pre-competitive cogni-
tive anxiety and self-confidence have a greater impact
on the performance of men than that of women.
However, apart from the thresholds argument pre-
sented in the Introduction, there does not appear to be
any obvious reason why this should be. Indeed,
although previous research has shown that, compared
with men, women report higher cognitive anxiety
(Martens et al., 1990; Russell et al., 1998), lower self-
confidence (Martens et al., 1990; Jones et al., 1991;
Krane and Williams, 1994) and less stability before
competing (Jones and Cale, 1989; Jones et al., 1991),
such findings do not explain why cognitive anxiety and
self-confidence should be more related to performance
for either sex. The idea of a threshold is not new, as it is
central to catastrophe models of anxiety and perfor-
mance (Hardy and Fazey, 1987; Hardy, 1996a), where
performance suffers a catastrophic drop abov e a
physiological arousal threshold. In light of the present
results, further research on cognitive anxiety and self-
confidence thresholds appears worthwhile.
The cognitive anxiety and self-confidence mean
effect sizes were greater for high-standard athletes than
lower-standard athletes. One possible reason for these
differences is that high-standard performance is typi-
cally associated with increased pressure. If an athlete is
not able to deal with such pressure, then the effect upon
performance is likely to be fairly dramatic. Another
possible reason for these differences is that high-
standard performance is typically associated with fewer
‘random effects’. That is, high-standard athletes typi-
cally operate within a more controlled personal
environment than their comparatively low-standard
counterparts. In other words, athletes competing at a
higher standard are more likely to ‘control the
controllables’ (Hardy et al., 1996). As such, it is
reasonable to expect that the effect of self-confidence
(and cognitive anxiety) upon performance will be
clearer with elite athletes. In the present meta-analysis,
452 Woodman and Hardy
truly high-standard (international) performers were
investigated in one study only. The other studies
comprising the ‘high-standard’ group used national
standard athletes. This lack of studies involving truly
elite athletes poses a fairly serious problem in terms of
generalization of research findings to elite perform ers.
For example, the stress that elite athletes endure may be
rather different to that endured by relatively low-
standard athletes. Certainly, recent research (Wood-
man and Hardy, 1998, 2001b; Gould et al ., 1999) has
suggested that elite performers may be exposed to
various kinds of relational and organizational stress
before and during major international competitions.
Thus, generalizations of findings with lower-s tandard
sport performers to elite performers might be inap-
propriate (cf. Hardy et al., 1996; Balague, 1999).
Further research with truly high-standard performers
is needed to enhance our understanding of the effects of
stress, anxiety and self-confidence in an elite sport
environment.
Research investi gating female athletes in high-st an-
dard environments would be particularly helpful, as
most studies of high-standard athletes in this meta-
analysis were of men. More specifically, of the 12
studies conducted with high-standard athletes, seven
were with men and only one was with women (the
remaining fou r were with both men and women). Thus,
notwithstanding the threshold arguments presented
earlier, the most parsimonious explanation of sex as a
moderator variable is that it was confounded by the
standard of competition. Further research investigating
high-standard women athletes should help to clarify this
issue.
The vast majority of studies included in this meta-
analysis used the Competitive State Anxiety Inventory-
2 (Martens et al., 1990) as a measure of cognitive
anxiety and self-confidence. The moderator analyses
revealed the CSAI-2 to be a significant moderator of the
self-confidence–performance relationship, with the ef-
fect size being smaller for the CSAI-2 (r = 0.19) than for
the other measures of self-confidence (r = 0.38). This is
consistent with the results from a recent meta-analysis
of the self-efficacy–sport performance relationship
(Moritz et al., 2000), which revealed that task-specific
measures of self-efficacy correlated significantly more
strongly with performance (r = 0.38) than other meth-
ods of assessment such as the CSAI-2 (r = 0.24).
Certainly, studies that match the task with more specific
measures of self-confidence appear more likely to reveal
stronger effect sizes. The near sine qua non status that
the CSAI-2 seemingly holds for researchers interested
in pre-competition sport affect could be problematic for
at least two other reasons. First, a recent confirmatory
factor analysis (Lane et al., 1999) found the CSAI-2 to
have weak structural validity. However, as this factor
analysis tested only the structure of the three-factor
model (cognitive anxiety, somatic anxiety and self-
confidence) and not the structure of each factor
separately, it offers no direct evidence about the relative
structural integrity of the cognitive anxiety and self-
confidence subscales. Second, eight of the nine
cognitive anxiety items in the CSAI-2 use ‘concern’ as
an expression of cognitive anxiety (e.g. ‘I’m concerned
about reaching my goal’), and it has been argued that
the expression ‘I am concerned’ can be interpreted
positively or negatively (Barnes et al., 1986; Jones,
1991; Jones and Swain, 1992; Burton and Naylor,
1997; Woodman and Hardy, 2001a). These differences
in interpretation led Jones and his colleagues (Jones,
1991; Jones and Swain, 1992) to add an interpretation
scale to the CSAI-2, which measures the extent to
which performers interpret their anxiety symptoms as
either facilitative or debilitative. Research using this
modified scale suggests that interpretation may be an
important moderating variable in the relationship
between cognitive anxiety and performance. For
example, Jones et al . (1993) found that high- and low-
performance gymnasts did not differ in cognitive
anxiety intensity. However, the high- performance gym-
nasts reported their cognitive anxiety to be more
facilitative than the low-performance gymnasts. Similar
findings have been reported in other studies (e.g. Jones
et al., 1994; Swain and Jones, 1996; Perry and Williams,
1998). As the present analyses did not reveal the CSAI-
2 to be a significant moderator of the cognitive anxiety–
performance relationship, one cannot conclude that the
cognitive anxiety subscale of the CSAI-2 is problematic
in relation to other measures of cognitive anxiety.
However, too few studies employed other measures of
cognitive anxiety to make comparisons between the
CSAI-2 and any other single measure of cognitive
anxiety. If researc hers develop another measure of
cognitive anxiety, then one will be able to measure its
predictive validity in comparison to that of the CSAI-2.
This is a worthwhile avenue for future research.
The difference in magnitude between the cognitive
anxiety and self-confidence mean effect sizes is con-
sistent with past research (e.g. Gould et al., 1984; Jones
and Cale, 1989; Jones et al., 1990, 1991; Martens et al.,
1990; Hardy, 1996b) that has suggested that cognitive
anxiety and self-confidence are orthogonal constructs,
which do not lie at opposite ends of the same
continuum. T hus, future researchers would do well to
consider cogni tive anxiety and self-confidence either
independently or as an interactive dyad. It is the
interaction between cognitive anxiety and self-confi-
dence that is likely to yield the most fruitful findings
(Hardy, 1996b). Certainly, from an anecdotal perspec-
tive, it seems that many exceptionally fine performances
are achieved when athletes are both anx ious (‘I am so
453Cognitive anxiety and self-confidence meta-analysis
worried, this is the biggest competition of my life’) and
self-confident (‘I know I can do well, I have prepared so
well for this competition’). From a theoretical perspec-
tive, both processing efficiency theory (Eys enck and
Calvo, 1992; Smith et al., 2001) and higher-order
catastrophe models (Hardy, 1996b) would support this
view. More precisely, processing efficiency theory
predicts that cognitively anxious individuals will invest
more effort in the task at hand provided they perceive
themselves to have a reasonable chance of success.
Also, within a higher-order catastrophe model frame-
work, Hardy (1990, 1996b) has proposed that high self-
confidence might protect cognitively anxious perfor-
mers from catastrophic drops in performance. Thus,
both processing efficiency theory and catastrophe
models are worthy of further research with respect to
investigating interactions between cognitive anxiety and
self-confidence.
In conclusion, this meta-a nalysis has revealed that
both cognitive anxiety and self-confidence are signifi-
cantly related to competitive sport performance. The
mean effect sizes for cognitive anxiety and self-
confidence were significantly higher for men than for
women. They were also higher for high-standard
athletes than for low-standard athletes. Furthermore,
compared with other measures of self-confidence, the
CSAI-2 revealed a significantly smaller mean self-
confidence effect size. In view of the significant
difference in magnitude between the two mean effect
sizes, researchers should view cognitive anxiety and self-
confidence as distinct constructs, rather than two
extremes of a single construct. Finally, the interaction
between cognitive anxiety and self-confidence is likely
to be a fruitful avenue for future research, and the
current theoretical paradigms that are the m ost amen-
able to investigation of this intera ction are processing
efficiency theory and higher-order catastrophe models.
Acknowledgements
The authors would like to thank three anonymous reviewers
and David Markland for their helpful comments on earlier
drafts of the manuscript.
References
Studies preceded by an asterisk were included in the
meta-analysis
Apter, M.J. (1982). The Experience of Motivation: The Theory
of Psychological Reversal. London: Academic Press.
Balague, G. (1999). Understanding identity, value, and
meaning when working with elite athletes. The Sport
Psychologist, 13, 89–98.
*Barnes, M.W., Sime, W., Dienstbier, R. and Plake, B.
(1986). A test of construct validity of the CSAI-2
questionnaire on male elite college swimmers. International
Journal of Sport Psychology, 17, 364–374.
*Bejek, K. and Hagtvet, K.A. (1996). The content of pre-
competitive state anxiety in top and lower level of female
gymnasts. Anxiety, Stress and Coping: An International
Journal, 9, 19–31.
*Bird, A.M. and Horn, M.A. (1990). Cognitive anxiety and
mental errors in sport. Journal of Sport and Exercise
Psychology, 12, 217–222.
Burrows, G.C., Cox, T. and Simpson, G.C. (1977). The
measurement of stress in a sales training situation. Journal
of Occupational Psychology, 50, 45–51.
*Burton, D. (1988). Do anxious swimmers swim slower? Re-
examining the elusive anxiety-performance. Journal of
Sport Psychology, 10, 45–61.
Burton, D. and Naylor, S. (1997). Is anxiety really
facilitative? Reaction to the myth that cognitive anxiety
always impairs sport performance. Journal of Applied Sport
Psychology, 9, 295–302.
*Chapman, C., Lane, A.M., Brierley, J.H. and Terry, P.C.
(1997). Anxiety, self-confidence and performance in tae
kwon-do. Perceptual and Motor Skills, 85, 1275–1278.
*Cox, R.H., Robb, M. and Russell, W.D. (2001). Construct
validity of the revised Anxiety Rating Scale (ARS-2).
Journal of Sport Behavior, 24, 10–18.
*Duesing, L.J. (1984). Comparison of anxiety, confidence,
and performance levels of male and female distance
runners. Unpublished master’s thesis, Pennsylvania State
University.
*Edwards, T. and Hardy, L. (1996). The interactive effects
of intensity and direction of cognitive and somatic anxiety
and self-confidence upon performance. Journal of Sport
and Exercise Psychology, 18, 296–312.
Eysenck, M.W. and Calvo, M.G. (1992). Anxiety and
performance: the processing efficiency theory. Cognition
and Emotion, 6, 409–434.
Field, A.P. (2001). Meta-analysis of correlation coefficients:
a Monte Carlo comparison of fixed- and random-effects
methods. Psychological Methods, 6, 161–180.
Fisher, R.A. (1928). Statistical Methods for Research Workers,
2nd edn. London: Oliver & Boyd.
*Gayton, W.F. and Nickless, C.J. (1987). An investigation of
the validity of the trait and state sport-confidence
inventories in predicting marathon performance. Perceptual
and Motor Skills, 65, 481–482.
*Gould, D., Weiss, M. and Weinberg, R. (1981). Psycho-
logical characteristics of successful and nonsuccessful Big
Ten wrestlers. Journal of Sport Psychology, 3, 69–81.
*Gould, D., Petlichkoff, L. and Weinberg, R.S. (1984).
Antecedents of, temporal changes in, and relationships
between CSAI-2 subcomponents. Journal of Sport Psychol-
ogy, 6, 289–304.
*Gould, D., Petlichkoff, L., Simons, J. and Vevera, M.
(1987). Relationship between Competitive State Anxiety
Inventory-2 subscale scores and pistol shooting perfor-
mance. Journal of Sport Psychology, 9, 33–42.
454 Woodman and Hardy
*Gould, D., Tuffey, S., Hardy, L. and Lochbaum, M.
(1993). Multidimensional state anxiety and middle dis-
tance running performance: an exploratory examination of
Hanin’s (1980) Zone of Optimal Functioning Hypothesis.
Journal of Applied Sport Psychology, 5, 85–95.
Gould, D., Guinan, D., Greenleaf, C., Medbery, R. and
Peterson, K. (1999). Factors affecting Olympic perfor-
mance: perceptions of athletes and coaches from more and
less successful teams. The Sport Psychologist, 13, 371–394.
*Grasso, A.T. (1999). The relationship of competitive state
anxiety and athletic performance in high school basketball
players. Unpublished master’s thesis, University of Ore-
gon.
*Guest, J.Z. and Cox, R.H. (1999). Golf performance and
individual zone of optimal functioning. Applied Research in
Coaching and Athletics Annual, 14, 130–149.
*Hammermeister, J. and Burton, D. (1995). Anxiety and the
ironman: investigating the antecedents and consequences
of endurance athletes’ state anxiety. The Sport Psychologist,
9, 29–40.
Hanin, Y.L. (1980). A study of anxiety in sport. In Sport
Psychology: An Analysis of Athletic Behavior (edited by W.F.
Straub), pp. 236–249. Ithaca, NY: Mouvement Publica-
tions.
Hanin, Y.L. (1986). State trait anxiety research on sports in
the USSR. In Cross-cultural Anxiety, Vol. 3 (edited by C.D.
Spielberger and R. Diaz), pp. 45–64. Washington, DC:
Hemisphere.
*Hardy, L. (1990). A catastrophe model of anxiety and
performance. In Stress and Performance in Sport (edited by
J.G. Jones and L. Hardy), pp. 81–106. Chichester: Wiley.
Hardy, L. (1996a). Testing the predictions of the cusp
catastrophe model of anxiety and performance. The Sport
Psychologist, 10, 140–156.
Hardy, L. (1996b). A test of catastrophe models of anxiety
and sports performance against multidimensional theory
models using the method of dynamic differences. Anxiety,
Stress and Coping: An International Journal, 9, 69–86.
Hardy, L. and Fazey, J. (1987). The inverted-U hypothesis:
a catastrophe for sport psychology? Communication to the
Annual Conference of the North American Society for the
Psychology of Sport and Physical Activity, Vancouver, BC,
Canada, June.
Hardy, L. and Whitehead, R. (1984). Specific modes of
anxiety and arousal. Current Psychological Research and
Reviews, 3, 14–24.
Hardy, L., Jones, J.G. and Gould, D. (1996). Understanding
Psychological Preparation for Sport: Theory and Practice of
Elite Performers. Chichester: Wiley.
*Highlen, P.S. and Bennett, B.B. (1979). Psychological
characteristics of successful and nonsuccessful elite wres-
tlers: an exploratory study. Journal of Sport Psychology, 1,
123–137.
Hunter, J.E. and Schmidt, F.L. (1990). Methods of Meta-
analysis: Correcting Error and Bias in Research Findings.
Newbury Park, CA: Sage.
*Jerome, G.J. and Williams, J.M. (2000). Intensity and
interpretation of competitive state anxiety: relationship to
performance and repressive coping. Journal of Applied Sport
Psychology, 12, 236–250.
Jones, J.G. (1991). Recent developments and current issues
in competitive state anxiety research. The Psychologist, 4,
152–155.
Jones, J.G. and Cale, A. (1989). Relationship between
multidimensional competitive state anxiety and cognitive
and motor subcomponents of performance. Journal of
Sports Sciences, 7, 129–140.
Jones, G. and Swain, A.B.J. (1992). Intensity and direction
dimensions of competitive anxiety and relationships
with competitiveness. Perceptual and Motor Skills, 74,
467–472.
Jones, J.G., Swain, A.B.J. and Cale, A. (1990). Antecedents
of multidimensional competitive state anxiety and self-
confidence in elite intercollegiate middle distance runners.
The Sport Psychologist, 4, 107–118.
Jones, J.G., Swain, A.B.J. and Cale, A. (1991). Gender
differences in precompetition temporal patterning and
antecedents of anxiety and self-confidence. Journal of Sport
and Exercise Psychology, 13, 1–15.
*Jones, J.G., Swain, A.B.J. and Hardy, L. (1993). Intensity
and direction dimensions of competitive state anxiety and
relationships with performance. Journal of Sports Sciences,
11, 525–532.
Jones, J.G., Hanton, S. and Swain, A.B.J. (1994). Intensity
and interpretation of anxiety symptoms in elite and non-
elite sports performers. Personality and Individual Differ-
ences, 17, 657–663.
Kerr, J.H. (1990). Stress in sport: reversal theory. In Stress
and Performance in Sport (edited by J.G. Jones and L.
Hardy), pp. 107–131. Chichester: Wiley.
*Krane, V. (1993). A practical application of the anxiety–
performance relationship: the zone of optimal functioning
hypothesis. The Sport Psychologist, 7, 113–126.
*Krane, V. and Williams, J.M. (1987). Performance and
somatic anxiety, and confidence changes prior to competi-
tion. Journal of Sport Behavior, 10, 47–56.
Krane, V. and Williams, J.M. (1994). Cognitive anxiety,
somatic anxiety, and confidence in track and field athletes:
the impact of gender, competitive level and task char-
acteristics. International Journal of Sport Psychology, 25,
203–217.
*Krane, V., Williams, J. and Feltz, D. (1992). Path analysis
examining relationships among cognitive anxiety, somatic
anxiety, state confidence, performance expectations, and
golf performance. Journal of Sport Behavior, 15, 279–295.
Lane, A.M., Sewell, D.F., Terry, P.C., Bartram, D. and
Nesti, M.S. (1999). Confirmatory factor analysis of the
Competitive State Anxiety Inventory-2. Journal of Sports
Sciences, 17, 505–512.
Martens, R., Burton, D., Vealey, R.S., Bump, L.A. and
Smith, D.E. (1990). Development and validation of the
Competitive State Anxiety Inventory-2. In Competitive
Anxiety in Sport (edited by R. Martens, R.S. Vealey and
D. Burton), pp. 117–190. Champaign, IL: Human
Kinetics.
*Martin, J.J. and Gill, D.L. (1991). The relationships among
competitive orientation, sport-confidence, self-efficacy,
anxiety, and performance. Journal of Sport and Exercise
Psychology, 13, 149–159.
455Cognitive anxiety and self-confidence meta-analysis
*Maynard, I.W. and Cotton, P.C. (1993). An investigation
of two stress-management techniques in a field setting.
The Sport Psychologist, 7, 375–387.
*Maynard, I.W. and Howe, B.L. (1987). Interrelations of
trait and state anxiety with game performance of rugby
players. Perceptual and Motor Skills, 64, 599–602.
*Maynard, I.W., Hemmings, B. and Warwick-Evans, L.
(1995). The effects of a somatic intervention strategy on
competitive state anxiety and performance in semiprofes-
sional soccer players. The Sport Psychologist, 9, 51–64.
*McAuley, E. (1985). State anxiety: antecedent or result of
sport performance. Journal of Sport Behavior, 8, 71–77.
*McCann, S.C., Murphy, S.M. and Raedeke, T.D. (1992).
The effect of performance setting and the individual
differences on the anxiety–performance relationship for
elite cyclists. Anxiety, Stress and Coping: An International
Journal, 5, 177–187.
*McKay, J., Selig, S., Carlson, J. and Morris, T. (1997).
Psychophysiological stress in elite golfers during practice
and competition. Australian Journal of Science and Medicine
in Sport, 29, 55–61.
*Moraes, L.C. (1987). The relationship between anxiety and
performance of Brazilian judokas. Unpublished master’s
thesis, Michigan State University.
Moritz, S.E., Feltz, D.L., Fahrbach, K.R. and Mack, D.E.
(2000). The relation of self-efficacy measures to sport
performance: a meta-analytic review. Research Quarterly for
Exercise and Sport, 71, 280–294.
Morris, L., Davis, D. and Hutchings, C. (1981). Cognitive
and emotional components of anxiety: literature review
and revised worry-emotionality scale. Journal of Educa-
tional Psychology, 75, 541–555.
*Parfitt, G. and Pates, J. (1999). The effects of cognitive and
somatic anxiety and self-confidence on components of
performance during competition. Journal of Sports Sciences,
17, 351–356.
*Perreault, S. and Marisi, D.Q. (1997). A test of multi-
dimensional anxiety theory with male wheelchair basket-
ball players. Adapted Physical Activity Quarterly, 14, 108–
118.
Perry, J.D. and Williams, J.M. (1998). Relationship of
intensity and direction of competitive trait anxiety to skill
level and gender in tennis. The Sport Psychologist, 12, 169–
179.
*Psychountaki, M. and Zervas, Y. (2000). Competitive
worries, sport confidence, and performance ratings for
young swimmers. Perceptual and Motor Skills, 91, 87–94.
*Rodrigo, G., Lusiardo, M. and Pereira, G. (1990).
Relationship between anxiety and performance in soccer
players. International Journal of Sport Psychology, 21, 112–
120.
Rosenthal, R. (1991). Meta-analytic Procedures for Social
Research. Newbury Park, CA: Sage.
Rosenthal, R. (1995). Writing meta-analytic reviews. Psycho-
logical Bulletin, 118, 183–192.
*Russell, W.D., Robb, M. and Cox, R.H. (1998). Sex,
sport, situation, and competitive state anxiety. Perceptual
and Motor Skills, 86, 816–818.
Schmidt, F.L. and Hunter, J.E. (1999). Comparison of three
meta-analysis methods revisited: an analysis of Johnson,
Mullen, and Salas (1995). Journal of Applied Psychology,
84, 144–148.
*Smith, N.C., Bellamy, M., Collins, D.J. and Newell, D.
(2001). A test of processing efficiency theory in a team
sport context. Journal of Sports Sciences, 19, 321–332.
Sonstroem, R.J. and Bernardo, P. (1982). Intraindividual
pregame state anxiety and basketball performance: a re-
examination of the inverted-U curve. Journal of Sport
Psychology, 4, 235–245.
*Swain, A.B.J. and Jones, G. (1996). Explaining perfor-
mance variance: the relative contribution of intensity and
direction dimensions of competitive state anxiety. Anxiety,
Stress and Coping: An International Journal, 9, 1–18.
*Taylor, J. (1987). Predicting athletic performance with self-
confidence and somatic and cognitive anxiety as a function
of motor and physiological requirements in six sports.
Journal of Personality, 55, 139–153.
*Terry, P.C. and Slade, A. (1995). Discriminant effective-
ness of psychological state measures in predicting perfor-
mance outcome in karate competition. Perceptual and
Motor Skills, 81, 275–286.
*Terry, P.C., Cox, J.A., Lane, A.M. and Karageorghis, C.I.
(1996). Measures of anxiety among tennis players in
singles and doubles matches. Perceptual and Motor Skills,
83, 595–603.
Thayer, R.E. (1978). Toward a psychological theory of
multidimensional activation (arousal). Motivation and
Emotion, 2, 1–34.
*Thelwell, R.C. and Maynard, I.W. (1998). Anxiety–
performance relationships in cricketers: testing the zone
of optimal functioning hypothesis. Perceptual and Motor
Skills, 87, 675–689.
*Vadocz, E.A., Hall, C.R. and Moritz, S.E. (1997). The
relationship between competitive anxiety and imagery use.
Journal of Applied Sport Psychology, 9, 241–253.
Weinberg, R.S. (1978). The effects of success and failure on
the patterning of neuromuscular energy. Journal of Motor
Behavior, 10, 53–61.
*Wiggins, M.S. and Henson, W.D. (2000). An examination
of direction and intensity of competitive anxiety and
performance among collegiate tennis players. International
Sports Journal, 4, 54–59.
*Williams, J.M. and Krane, V. (1992). Coping styles and
self-reported measures on state anxiety and self-confi-
dence. Journal of Applied Sport Psychology, 4, 134–143.
Wine, J.D. (1971). Test anxiety and direction of attention.
Psychological Bulletin, 76, 92–104.
Wine, J.D. (1980). Cognitive-attentional theory of test
anxiety. In Test Anxiety: Theory, Research and Applications
(edited by I.G. Sarason), pp. 349–385. Hillsdale, NJ:
Lawrence Erlbaum.
Woodman, T. and Hardy, L. (1998). Le stress organisa-
tionnel: une e´tude de cas [Organisational stress: a case
study]. Communication to the Journe´es Nationales d’Etudes
de la Socie´te´ Franc¸aise de Psychologie du Sport, Poitiers,
France, March.
456 Woodman and Hardy
Woodman, T. and Hardy, L. (2001a). Stress and anxiety. In
Handbook of Sport Psychology (edited by R.N. Singer, H.A.
Hausenblas and C.M. Janelle), pp. 290–318. New York:
Wiley.
Woodman, T. and Hardy, L. (2001b). A case study of
organizational stress in elite sport. Journal of Applied Sport
Psychology, 13, 232–263.
*Woodman, T., Albinson, J.G. and Hardy, L. (1997). An
investigation of the zone of optimal functioning hypothesis
within a multidimensional framework. Journal of Sport and
Exercise Psychology, 19, 131–141.
*Yang, G. (1994). An investigation of relationships among
self-confidence, self-efficacy, competitive anxiety, and
sport performance. Unpublished doctoral dissertation,
Florida State University.
*Zhu, B. and Fang, Y. (1998). Study of relationship between
intensity, direction of competitive state anxiety and
performance in long-distance run. Journal of Tianjin
Institute of Physical Education, 13, 81–84.
457Cognitive anxiety and self-confidence meta-analysis
