Download Antecedents Of, Temporal Changes In, And

JOURNAL OF SPORT PSYCHOLOGY,1984, 6, 289-304
Antecedents Of, Temporal Changes In,
And Relationships
Between CSAI-2 Subcomponents
Daniel Gould and Linda Petlichkoff
University of Illinois
Robert S. Weinberg
North Texas State University
Two studies were conducted to examine antecedents of, relationships between, and
temporal changes in the cognitive anxiety, somatic anxiety, and the self-confidence
components of the Martens, Burton, Vealey, Bump, and Smith (1983) newly developed
Competitive State Anxiety Inventory-2 (CSAI-2). In addition, the prediction that
cognitive and somatic anxiety should differentially influence performance was examined.
In Study 1, 37 elite intercollegiatewrestlers were administered the CSAI-2 immediately
before two different competitions, whereas in Study 2,63 female high school volleyball
players completed the CSAI-2 on five different occasions (1 week, 48 hrs, 24 hrs,
2 hrs, and 20 min) prior to a major tournament. The results were analyzed using multiple
regression, multivariate multiple regression, univariate and multivariate analyses of
variance, and general linear model trend analysis techniques. The findigs supported
the scale development work of Martens and his colleagues by verifying that the CSAI-2
assesses three separate components of state anxiety. A number of other important find i g s also emerged. First, the prediction was confirmed that somatic anxiety increases
during the time leading to competition, while cognitive anxiety and confidence remain
constant. Second, CSAI-2 subscales were found to have different antscedents, although
the precise predictions of Martens and his colleagues were not supported. Third, the
prediction that cognitive anxiety would be a more powerful predictor of performance
than somatic anxiety was only partially supported. Fourth, the prediction that
precompetitive anxiety differences between experienced and inexperienced athletes initially found by Fenz (1975) result from somatic anxiety changes was not supported.
It was concluded that the CSAI-2 shows much promise as a multidimensional sportspecific state anxiety inventory, although more research is needed to determine how
and why specific antecedent factors influence various CSAI-2 components and to examine the predicted relationships between CSAI-2 components and performance.
Requests for reprints should be sent to Daniel Gould, Department of Physical Education,
University of Illinois, Urbana, IL 61801.
Much of this research was conducted while the first two authors were on the staff of the
Department of Physical Education, Dance and Leisure Studies at Kansas State University.
290
GOULD, PETLICHKOFF, AND WEINBERG
The examination of state anxiety, including factors associated with state anxiety
(Martens & Gill, 1976; Scanlan & Passer, 1979) and the state anxiety-motor performance
relationship (Carron & Morford, 1968; Martens & Landers, 1970) has been one of the
most extensively studied areas in sport psychology. Despite this long research history,
however, many relationships in the area remain only partially understood and theoretical
tests are incomplete. A primary factor limiting the development of anxiety theory and
research has been the lack of uniformly acceptable state anxiety assessment instruments
(Martens, 1977; Martens & Landers, 1970). Specifically, previous measures of state anxiety have been criticized for their failure to assess independent anxiety components and
for their lack of situational specific content.
Recently, state anxiety has been theoretically redefined as a multidimensional
construct by psychologists (Liebert & Morris, 1967; Schwartz, Davidson, & Goleman,
1978) and sport psychologists (Landers, 1980; Martens et al., 1983) alike, and new
multidimensional assessment instruments have been advocated. In response to this need,
Martens and his associates have developed a new sport-specific multidimensional state
anxiety inventory: The Competitive State Anxiety Inventory-2 (CSAI-2). The CSAI-2
is based on the theoretical developments of Spielberger, Gorsuch, and Lushene (1970)
and is derived in part from the original Competitive State Anxiety Inventory constructed
by Martens (1977). It is also highly congruent with contemporary anxiety research and
theory (Davidson & Schwartz, 1976; Liebert & Morris, 1967; Schwartz et al. 1978),
as it is a multidimensionalmeasure that separately assesses cognitive (worry) and somatic
(heightened physiological arousal) state anxiety components. Finally, the CSAI-2 is convenient because it is a paper and pencil test comprised of 27 4-point Liert-type scale items.
Extensive preliminary evidence conducted by Martens et al. (1983) revealed that
the CSAI-2 indeed did measure separate somatic and cognitive state anxiety components
which, as predicted, were only moderately correlated. ' In addition, factor analyses revealed
the emergence of a third component, self-confidence. Specifically, while self-confidence
was found to be negatively correlated with both the cognitive and somatic subcomponents,
these correlations were only moderate. Therefore, self-confidencewas viewed as a separate
CSAI-2 component and it was concluded that CSAI-2 assesses three independent (only
moderately correlated) subscales: cognitive anxiety, somatic anxiety, and self-confidence.
Not only has the CSAI-2 been shown to assess cognitive state anxiety, somatic
state anxiety, and self-confidence, but it has been found to have good internal consistency
(reliability) and construct validity (Martens et al., 1983). Specifically, Cronbach's alpha
coefficients ranged from .70 to .90, revealing that the internal consistency of the various
components was adequate, whereas concurrent validity was established by supporting
predicted relationships between the CSAI-2 subscales and a variety of trait (e.g., Martens'
Sport Competition Anxiety Test, 1977; Spielberger, Gorsuch, & Lushene's Trait Anxiety
Inventory, 1970) and state (e.g., Zuckerman's Affect Adjective Checklist, 1960; Schwartz
'Martens et al. (1983) argue that in a strict theoretical sense CSAI-2 subcomponents should be so
qualitatively different that they are orthogonal. However, most athletes would find it difficult to have
high cognitive anxiety without experiencing some somatic anxiety. Therefore, these investigatorsconcluded that it is unrealistic to view CSAI-2 subcomponents as totally independent, but subscale intercorrelations should be of low to moderate magnitude, sharing less than 50% common variance.
CSAI-2 SUBCOMPONENTS
29 1
et al.'s Cognitive-Somatic Anxiety Questionnaire, 1978) anxiety measures. CSAI-2
subscales were also found to be unbiased by the effects of social desirability.
Martens and his colleagues (1983) argue that the CSAI-2 subscales are not only
statistically independent but must be viewed separately because they are elicited by different antecedent conditions. They indicate, for example, that somatic anxiety is considered
to be a reflexive response to various environmental stimuli (e.g., playing field, game irnportance) associated with the onset of an evaluative event. In contrast, the degree of cognitive
state anxiety elicited by an athlete is dependent on his or her perceived ability, which is
primarily generated from previous competitive experiencealthough situational factors largely
independent of past experience (e.g., opponent ability) may influence it as well. Thus,
performance expectancies prior to competition should be more highly correlated with
cognitive than somatic anxiety.
Additional evidence supporting the independence of cognitive and somatic anxiety focuses on changes in temporal patterns in these states as the time of competition approaches. For example, evidence in both the sport psychology (Martens et al., 1983) and
the psychology (Morris & Fulmer, 1976; Smith & Moms, 1976) literature reveals that
somatic anxiety steadily increases prior to competition or evaluation, whereas cognitive
anxiety shows little temporal fluctuation, changing only with the expectation of success.
The separate assessment of cognitive and somatic anxiety components is especially important since they are thought to influence performance differently. For example, Martens and his colleagues (1983) indicate that "somatic anxiety is likely to reach
its peak at the onset of competition and dissipate once the contest begins." Thus, somatic
anxiety should influence performance less than cognitive anxiety, "unless the somatic anxiety becomes so great that attention is diverted from the task to these internal states" (Martens
et al., 1983). Similarly, Morris and Engle (1981) have concluded that cognitive anxiety,
as compared to somatic anxiety, is more strongly related to performance, while Wine (1971,
1980) has found that cognitive anxiety disrupts a performer's attentional mechanisms and,
in turn, performance. Based on previous theorizing and research, then, cognitive and somatic
anxiety should differentially influence performance.
In summary, the CSAI-2 appears to be a promising instrument for assessing
separate state anxiety components from both a theoretical and empirical perspective. Before
this instrument is widely used, however, further validation research is needed. For example, many of the initial results of Martens and his colleagues should be replicated. In
addition, predictions not tested by these investigators, such as the hypothesized differing
antecedent conditions for the various CSAI-2 subcomponents, must also be examined.
The present study was designed to replicate and test various predictions and
hypothesized relationships between CSAI-2 subscales. Specifically, two field studies were
conducted. The first involved 37 elite intercollegiate wrestlers who were administered the
CSAI-2 immediately before two different competitions. To test the independence of CSAI-2
subcomponents, intercorrelations between subscales were examined. In addition, the relationships between such antecedents as Martens' (1977) Sport Competition Anxiety Test
(SCAT), perceived ability, wrestling success/failure, past experience, and CSAI-2 subcomponents were investigated to determine if the various subcomponentshave differing
antecedents. Finally, the predicted relationships between CSAI-2 subcomponents and performance were examined. In Study 2, female high school varsity volleyball players were
administered the CSAI-2 on five different occasions prior to competition. As in Study
1, correlations between subcomponent scores were examined. Moreover, the prediction
292
GOULD, PETLICHKOFF, AND WEINBERG
was tested that somatic anxiety, as opposed to cognitive anxiety, would display greater
increases prior to competition.
STUDY 1
Method
Procedure and Sample
In October of 1982, 40 intercollegiate wrestlers, representing four of the top 10
ranked NCAA Division I teams in 1981, participated in the U.S.A. Wrestling Hall of
Fame Classic. 2 After receiving informed consents from the wrestlers and their coaches,
all participants were administered a demographic and background questionnaire the night
before this double dual meet competition. In addition, all competitors were asked to complete the CSAI-2 twice, 10 minutes prior to each individual's two wrestling matches on
the following day. Complete information was obtained for 37 of the 40 wrestlers.
Questionnaires
The demographic and background questionnaire was comprised of a series of
questions assessing past wrestling experience (e.g., age began wrestling, years of intercollegiate varsity experience), past success (e.g., intercollegiate record, awards won),
Martens' (1977) SCAT (a sport-specific measure of trait anxiety), a sources-of-stress questionnaire developed by Gould, Horn, and Spreeman (1983), Form E of the CSAI-2, and
selected subscales of Nideffer's (1976) Test of Attentional and Interpersonal Style. This
questionnaire was administered to the wrestlers in a group the night before competition.
For brevity, only the results of the demographic-background, CSAI-2, and SCAT sections
of the questionnaire will be discussed in this manuscript.
Form E of the CSAI-2 was again administered to each wrestler on the following
day, 10 minutes prior to each match wrestled in this double dual meet competition. On
all occasions the wrestler was taken to a quiet section of the warm-up area where he completed the inventory in the presence of the e~perimenter.~
ZTheauthors would like to thank Steve Combs and Tim Johnson of U.S.A. Wrestling for their s u p
port and assistance in conducting Study 1, and Gerald Fitzpatrick and Deb Gould for their assistance
in the data collection. Sincere appreciation is also extended to the coaches and athletes who volunteered
to participate in both studies.
3While all 40 wrestlers participating in the Hall of Fame Classic volunteered to take part in the investigation, complete data was obtained for only 37 of the 40 subjects. One subject missed the meeting
in which the background questionnaire was administered, one voluntarily withdrew from the experiment because he was too nervous to complete the first precompetitive CSAI-2, and a third wrestler
was eliminated when the experimenters saw that he was completing but not reading one of the
prematch CSAI-2 questionnaires.
4Martens et al. (1983) indicate that while Form E is less sensitive to social desirability than earlier
versions of the scale, antisocial desirability instructions should be given. These procedures were followed
in the present investigation.
CSAI-2 SUBCOMPONENTS
Results
Sample Characteristics
The 37 wrestlers ranged in age from 18 to 22, with the mean age being 20 years
(SD = 1.3). Ten of the wrestlers represented the University of North Carolina, nine
represented Oklahoma University, nine represented Oklahoma State University, and nine
represented San Jose State University. The sample comprised 10 freshman, 8 sophomores,
11juniors, and 6 seniors. As would be expected, most of the wrestlers had extensive wrestling experience (M years involved = 8.7, SD = 2.7), and on the average had begun wrestling competitively at age 11 (M = 11.1 years, SD = 2.6). Moreover, those wrestlers who
had wrestled intercollegiately the previous year (N = 27) were very successful (M wins
= 19.9, SD = 8.7; Mlosses = 6.5, SD = 3.5).
Correlations Between CSAI-2 Subcomponents
To test the prediction that the CSAI-2 measures two independent anxiety dimensions (cognitive and somatic anxiety), as well as self-confidence, correlations between
subscale scores were calculated. Table 1 contains the correlations between subscale scores
taken the night before competition, before Match 1, before Match 2, and averaged over
all three assessment intervals. In addition, this table contains the mean intercorrelations
of CSAI-2 subscales obtained by Martens et al. (1983).
Inspection of Table 1 reveals that these correlations were similar to or, in the
case of the somatic-to-confidence correlation, lower than those found by Martens and his
colleagues. That is, the predicted moderate correlation between cognitive and somatic anxiety (r = + .52) was obtained, as well as the predicted moderate negative correlations
between confidence and cognitive anxiety (r = - .48) and confidence and somatic anxiety
(r = -.40).
Antecedents of CSAI-2 Subcomponents
To examine the relationships between prematch CSAI-2 subscale scores and the
potential antecedent factors of trait anxiety (SCAT), perceived wrestling ability (an 11-point
Table 1
lntercorrelations Between CSAI-2 Subcomponent Scores
lntercorrelations
Time of Assessment
Cognitive
and
somatic
Night before competition
Prematch 1
Prematch 2
M
Martens et al. (1983) (3 samples)
All correlations significant @ < .05)
.44
.61
.50
.52
.50
Cognitive
and
confidence
Somatic
and
confidence
GOULD, PETLICHKOFF, AND WEINBERG
Table 2
Standardized Regression Coefficients Match 1 CSAI-2 Subscales
Criterion (CSAI-2 subscales)
Predictors
SCAT
Years experience
Perceived ability
Cognitive anxiety
Somatic anxiety
Confidence
.32*
-.51**
- .02
Likert scale rating of perceived wrestling ability), years wrestling experience, and previous
match outcome, a series of multivariate multiple regression, univariate multiple regression, and canonical correlation analyses were employed. In all cases the CSAI-2 subscale
scores served as the criterion variables and the various antecedents as predictors.
Match 1 Results. A significant multivariate relationship was found to exist between the predictor variables of SCAT, years experience and perceived ability, and the
three CSAI-2 subscale criterion measures, F (9, 75) = 4.98, p < .001. Moreover, an
inspection of the univariate multiple regression F tests for each criterion measure revealed that the predictor variables were significantly related to all three of the CSAI-2
subscale scores (cognitive anxiety, F (3, 33) = 8.87, p < .001; somatic anxiety, F (3,
33) = 4.94, p < .006; confidence, F (3, 33) = 6.38, p < .002). Finally, an examination
of the standardized regression coefficients contained in Table 2 reveals that SCAT was
highly related to both cognitive and somatic anxiety, years experience was highly related
to cognitive anxiety, and perceived ability was highly related to confidence.
Canonical correlation analysis was also used to examine the relationship between
linear combinations of the predictor and criterion variables. Two canonical variates were
found to be significant (Rcl = .6907, p < .001; Rc2 = .6072, p < .004),accounting for
48 and 37% of the variance, respectively. However, since the redundancy index has been
shown to provide a more accurate assessment of the strength of association between data
sets (Karpman, 1981; Weiss, 1972), a redundancy index was calculated for each canonical
variate. These indices showed that the first canonical variate accounted for 30% of the
variance, while the second variate accounted for 7% of the variance.
To further understand the contributions of the specific predictor and criterion
variables to the significant canonical correlations, the canonical loadings were examined. These loadings are contained in Table 3 and reveal that for the first canonical variate,
the criterion variables of cognitive and somatic anxiety and the predictor variable, years
experience, contributed most to the canonical correlation. An inspection of the loadings
51n the case of all regression analyses, correlations between predictor variables were examined to
ensure that multicollinearity could be ruled out as an explanation for nonsignificant findings.
6Standardizedregression coefficients were not used to assess the contribution of each predictor and
criterion variable to the canonical correlation because it has been shown (Weiss, 1972) that the canonical
loadings provide a more accurate assessment of these relationships.
CSAI-2 SUBCOMPONENTS
Table 3
Canonical Loadings: Match 1 CSAI-2 Subscales and Predictor Variables
Canonical correlation 1 loadings Canonical correlation 2 loadings
CSAI-2 subscales
Cognitive anxiety
Somatic anxiety
Confidence
Predictors
SCAT
Years experience
Perceived ability
.93
.79
- .65
.66
- .81
- .59
for the second variate reveals that the criterion variable, confidence, and the predictor
variable, perceived ability, contributed most to the canonical relationship. Thus, the
magnitude and signs of these loadings show that, as would be predicted, greater years
of experience were negatively related to heightened cognitive and somatic anxiety, while
perceived ability and confidence were positively related
Match 2 Results. The multivariate multiple regression for Match 2 revealed that
a significant relationship existed between the predictor variables of SCAT, years of experience, perceived ability, and Match 1 outcome (won-loss coded as a dummy variable)
and the CSAI-2 subscale score criterion measures, F (12,79) = 2.61, p < .006. However,
the univariate multiple regression F tests conducted on each criterion measure revealed
that only Match 2 cognitive anxiety was significantly related to the predictor variables,
F (4, 32) = 6.02, p < .001. Inspection of the standardized regression coefficients also
showed that years experience (-.43) and SCAT (- .31) contributed significantly more
( p < .05) to the regression equation than perceived ability (- .11) and Match 1 outcome
(.08).
The Match 2 canonical correlation results revealed that one significant canonical
variate emerged (Rc = .6784, p < .001, Rc2 = 46%), with the redundancy index indicating
that this relationship accounted for 20% of the variance. The canonical loadings for both
the predictor and criterion variables are contained in Table 4 and show that the criterion
variable of cognitive anxiety contributed most to the significant canonical correlation,
followed by somatic anxiety, with confidence contributing least. For the predictor variables,
years experience and SCAT were the greatest contributors to the relationship. Additionally, the signs of the loadings revealed that years experience was negatively related
to cognitive and somatic anxiety while SCAT was positively related to these CSAI-2
subscales.
In summary, when the results of the multivariate multiple regression, univariate
multiple regression, and canonical correlation analyses for both matches are considered,
several patterns of findings emerge. First, no single antecedent variable was strongly related
to all three CSAI-2 subcomponents. Second, years experience was the strongest predictor
of cognitive anxiety for both matches. Third, SCAT was moderately related to somatic
anxiety in Match 1 but was moderately related to cognitive anxiety for both Matches 1
and 2. Fourth, perceived ability was highly related to Match 1 confidence but showed
little relationship to cognitive and somatic anxiety. Fifth, little relationshipwas found between match outcome and any of the Match 2 CSAI-2 subscale scores. Finally, the various
GOULD,PETLICHKOFF, AND WEINBERG
Table 4
Canonical Loadings: Match 2 CSAI-2 Subscales and Predictor Variables
Canonical loadings
CSAI-2 subscales
Cognitive anxiety
Somatic anxiety
Confidence
Predictors
SCAT
Years experience
Perceived ability
Match 1 outcome
antecedents were more strongly related to Match 1 CSAI-2 subcomponents than Match
2 CSAI-2 subcomponents.
CSAI-2 Pe$omuznce Relationship
To test the prediction that cognitive anxiety rather than somatic anxiety would
show a stronger relationship to performance, a series of multivariate multiple regression,
univariate multiple regression, and canonical correlation analyses were conducted for both
Match 1 and 2. In all cases, CSAI-2 subscale scores were used as predictor variables
and performance measures were used as criterion variables. Performance measures included match outcome (won-loss) and the number of points scored in the first period of each
match. 'It should be noted that the number of points scored in the first period was selected
as a performance measure in an effort to provide a performance assessment as close to
the time of CSAI-2 assessment (5 to 10 rnin.prior to competition) as possible. This is important since precompetitive state anxiety may be more strongly related to immediate performance, as compared to later match performance.
Match 1Results. A multivariate multiple regression analysis was conducted using
match outcome and points scored in Period 1 as criterion variables and the three CSAI-2
prematch 1 subscale scores as predictor variables. The results revealed that no significant
multivariate relationship existed. Thus, no significant relationship was found between
cognitive anxiety, somatic anxiety, confidence, and performance.
Match 2 Results. The multivariate multiple regression analysis conducted on the
Match 2 performance measures was marginally significant, F (6, 64) = 2.01, p < .08.
T h e total points scored in period one of Match 1 ranged from 0 to 10, and in period one of Match
2 from 0 to 9. However, because f d or pin (the ultimate goal in wrestling) ends the match and eliminates
the further scoring of points, the system used for calculating the number of points scored in period
one dependent variable had to be devised to ensure that the wrestlers scoring pins were awarded the
highest point values. Therefore, 3 points were added to the highest number of points scored in the
period if a wrestler pinned his opponent in the first minute of the period, 2 points if he pinned his
opponent in the second minute of the period, and 1 point if he pinned his opponent in the third minute
of the period. Thus, a wrestler pinning his opponent in the third minute of the period in Match 1
was awarded 13 total points, a wrestler pinning his opponent in the second minute, 12 points, and
a wrestler pinning his opponent in the third minute, 11 points.
CSAI-2 SUBCOMPONENTS
297
Moreover, the follow-up univariate F tests conducted on the match outcome and points
scored in Period 1 criterion measures indicated that only the match outcome regression
equation was significant, F (3, 33) = 3.04, p < .04. Inspection of the Match 2 outcome
standardized regression coefficients also showed that cognitive anxiety (.53)was the only
significant @ < .01) predictor of match outcome while somatic anxiety (-. 12) and confidence (.03) contributed little to the relationship.
As was the case with the multivariate multiple regression analysis, the canonical
correlation analysis was only marginally significant, Rc = .4663, p < .08, R c ~= 22%.
Moreover, the redundancy index for the one marginally significant canonical correlation
showed that only 13 % of the variance was explained. Table 5 contains the canonical loadings
for this analysis, and an inspection of these loadings reveals that match outcome and
cognitive anxiety contributed most to the relationship.
In summary, no significant relationship existed between prematch CSAI-2 subscale scores and Match 1 performance. For Match 2, however, a marginally significant
relationship existed, with cognitive anxiety and match outcome contributing most to the
relationship.
STUDY 2
Method
Procedure and Sample
To test the hypothesized independence of CSAI-2 subcomponents and the prediction that precompetitive state anxiety increases prior to competition as a result of changes
in somatic, as opposed to cognitive, anxiety, 63 Michigan female high school varsity
volleyball players voluntarily completed the CSAI-2 on five different occasions prior to
a major tournament. The testing occasions occurred 1 week prior to competition, 48 hours
prior to competition, 24 hours prior to competition, 2 hours prior to competition, and
20 minutes prior to competition. In addition, all subjects completed a brief demographic
and background questionnaireat the first testing session. Finally, in all cases subjects completed the questionnaires in groups under standardized testing conditions.
Table 5
Canonical Loadings: Match 2 Performance
-
Canonical loadings
-
Performance
Match outcome
Points scored period 1
Predictors
Cognitive anxiety
Somatic anxiety
Confidence
298
GOULD, PETLICHKOFF, AND WEINBERG
Questionnaires
The demographic and background questionnaire consisted of a series of
questions focusing on the subject's age, year in school, previous volleyball experience,
and previous volleyball success (both team and individual). Moreover, as in the previous
experiment, Form E of Martens et al. (1983) CSAI-2 was utilized.
Results
Sample Characteristics
The 63 subjects ranged in age from 14 to 18, having a mean age of 16.4 years
(SD = .78). Most had begun playing volleyball between the ages of 12 and 13, although
some had begun as early as age 8 and others as late as age 16. In addition, the subjects
averaged 4.3 years (SD = 1.6) of competitive volleyball experience with the sample consisting of 31 seniors, 25 juniors, 6 sophomores, and 1 freshman.
Approximately 50% of the subjects indicated that they had started in more than
50% of the games of the last season, and also indicated that on the average they had participated in 77% of all games played. Finally, most of the subjects' teams had been successful, averaging 14 wins (SD = 6.9) and 5 losses (SD = 7.9) during the season.
Correlations Between CSAZ-2 Subcomponents
To test the prediction that the CSAI-2 measures two independent anxiety components, as well as self-confidence, intercorrelations between subscale scores were
calculated. Specifically, Table 6 contains the intercorrelations taken at the five different
precompetitive testing occasions, the correlations averaged over all time intervals, and
the Martens et al. (1983) intercorrelations. As can be seen in Table 6, these intercorrelations are similar to those of Martens and his colleagues, as well as those obtained in
Study 1. Thus, since the CSAI-2 subcomponents were only moderately correlated, support was again found for their independence.
Temporal Changes in CSAI-2 Subcomponents
In the initial validation studies of the CSAI-2, Martens and his colleagues
(1983) supported previous psychological research by finding that somatic anxiety increased as the time of competition approached, while cognitive anxiety and confidence remained stable. The generality of these findings was tested in this investigation by conducting
a one-way repeated measures MANOVA on three CSAI-2 subscale scores using the five
assessment times (1 wk, 48 hrs, 24 hrs, 2 hrs, and 20 min) as levels of the independent
variable. The results revealed that a significant multivariate effect emerged, F (12, 651)
= 4.56, p < .W1. Subsequent, univariate repeated measures ANOVAs conducted on the
five successive assessments for each CSAI-2 subscale score also showed that, as predicted,
only the somatic anxiety subcomponent significantly increased as the time of competition
approached, F (4, 248) = 12.69, p < .MI.
8The authors would like to thank Linda Bump of the University of Illinois for her assistance in this
phase of the data analysis.
CSAI-2 SUBCOMPONENTS
Table 6
lntercorrelations Between CSAI-2 Subcomponents Scores
lntercorrelations
Cognitive
and
somatic
Time of
assessment
1 week
48 hours
24 hours
2 hours
20 minutes
M
Martens et al. (1983)
findings
Cognitive
and
confidence
Somatic
and
confidence
- .52
- .58
- .35
.41
.58
.46
.47
.50
.48
.50
- .57
- .49
- .55
- .54
- .51
- .46
- .39
- .44
- .46
- .42
.52
All correlations significant (p < .05)
A foIlow-up general linear model trend analysis using the reduced mean square
error as the criterion for best fit revealed that this increase in somatic anxiety was best
explained by a cubic trend (M 1 wk = 15.09, M 48 hrs = 15.02, M 24 hrs = 14.49,
M 2 hrs = 16.24, M 20 min = 18.32). That is, somatic anxiety initially decreased, then
increased prior to competition (see Figure 1). Thus, the prediction that somatic anxiety
increases prior to competition, while cognitive anxiety and confidence remain stable, was
supported.
28
25
-A-
24
23
e~ 1
Y) 20
5
16
3:
"
-
r
- - -A
/
-. .........-.. ........-. .........-.. .......*
-
confidence
cognitive auciety
somatic anxiety
12
TIME BEFORE COMPETITION
Figure 1 - Precompetitive temporal changes in CSAI-2 subcomponent scores.
300
GOULD, PETLICHKOFF, AND WEINBERG
Temporal Changes in CSAI-2 Subcomponents
in Experienced Versus Inexperienced Players
Since Fenz (1975) initially found that experienced parachutists differed from inexperienced parachutists in preperformance state anxiety, a number of investigators have
attempted to replicate his findiingswith other samples of athletes. Unfortunately, the results
of these studies have been contradictory: Some investigators have found support for these
differences (Mahoney & Avener, 1977), while others have not (Gould et al., 1983; Highlen
& Bennett, 1979). It has recently been hypothesized (Gould et al. 1983) that the lack of
support for the initial Fenz (1975) finding may have resulted from the fact that nonvalidated
unidimensiond, as opposed to validated multidimensional, measures of anxiety have been
used and that the predicted differences may only occur with somatic anxiety.
Therefore, to test the prediction that inexperienced athletes will demonstrate a
linear increase in somatic state anxiety prior to competition, while experienced athletes
will demonstrate an inverted-V pattern (an increase followed by a decrease), a 2 x 5 (experience - inexperience by precompetitive time of assessment) repeated measure MANOVA
was utilized, with the CSAI-2 subcomponentscores serving as dependent variables. Inexperienced players were operationally defined as those with 1 to 3 years volleyball experience
(N = 21), while experienced players were defined as those with 5 to 9 years experience
(N = 27). The results revealed that no significant experience or experience by time main
or interaction effects occurred. Thus, no support was found for the predicted experience
versus inexperienced pattern of anxiety differences for either the cognitive or somatic
components.
DISCUSSION
The Multidimensional Nature of the CSAI-2
The primary reason for developing the CSAI-2 was to construct a sport-specific
multidimensional measure of state anxiety which assesses independent anxiety dimensions.
The CSAI-2 appears to have accomplished its stated purpose, as the intercorrelations found
between CSAI-2 subscales in both Study 1 and 2 of this investigation demonstrate strong
support for the recent scale development studies of Martens and his colleagues (1983).
Specifically, similar moderate positive correlations were obtained between cognitive and
somatic anxiety, as well as moderate negative correlations between confidenceand cognitive
anxiety and confidence and somatic anxiety. In fact, the only discrepancy in the findings
was the lower cognitive-to-somatic anxiety wrrelation (r = - .41)found in the present
investigation, as compared to that of Martens et al. (1983) (r = - .52). However, this
lower correlation only further supports the notion of independence since it reflects even
less common variance between subscales. In summary, the intercorrelational findings support the conception that the CSAI-2 is a multidimensional anxiety measure with independent subscales which only share between 16 and 29% common variance.
The temporal changes in CSAI-2 subcomponents found in Study 2 also support
the notion that independent anxiety subcomponents are being assessed. That is, if state
anxiety is unidimensional in nature, all three CSAI-2 subscales should show a similar
pattern of fluctuation prior to competition. As in the Martens et al. (1983) study, as well
as other studies using different multidimensional state anxiety measures (Morris & Engle,
CSAI-2 SUBCOMPONENTS
301
1981; Morris & Fulmer, 1976; Smith & Morris, 1976), the results of this investigation
revealed that only the somatic anxiety subcomponent showed a significant increase as the
time of competition approached.
A critical question not answered in this investigation is why somatic anxiety increases as the time of competition approaches, while cognitive anxiety and confidence
remain stable. Borkovec (1976) contends that these findings result from a conditioned
somatic anxiety response, stemming from somatic anxiety increases that have been conditioned over time to specific environmental stimuli such as the pregame locker room setting or the playing field itself. In contrast, cognitive anxiety and confidence are hypothesized
to be linked to changes in performance expectancies, which typically are not expected
to change until competitionactually begins. An important question for future investigators,
then, is to explain why CSAI-2 subscales show different patterns of temporal change prior
to competition.
While not fully explained, the temporal patterns of anxiety findings are important
in that they further reflect the utility of viewing anxiety as a multidimensional construct.
It is possible, for example, that one reason the anxiety-performancerelationship has been
so difficult to understand is because athletes who perform poorly may do so because of
changes in only one anxiety component (e.g., cognitive anxiety). However, when unidimensional measures are employed, any precompetitive cognitive anxiety or confidence changes
may be masked by the normal increase in somatic anxiety. Thus, specific changes in each
anxiety subcomponent must be examined in order to fully understand the anxietyperformance relationship.
Antecedents of the CSAI-2
While the intercorrelation and temporal change results strongly support the work
of Martens and his colleagues (1983), the findings testing the hypotheses that CSAI-2
subcomponents are elicited by different antecedents were not clearly supported. Specifically,
no support was found for the prediction that SCAT, because it has more items judged
to be somatic in nature, would be a better predictor of somatic anxiety than cognitive anxiety or confidence. In contrast, SCAT was found to be a significant predictor of cognitive
anxiety for both Match 1 and 2, while it was only a significant predictor of Match 1 somatic
anxiety and was not found to be related to confidence in either match.
Performance expectancies were also hypothesized to be more highly predictive
of cognitive, as compared to somatic, anxiety, but both the perceived ability rating and
Match 1 outcome antecedents were not found to be significant predictors of either anxiety
component. However, it can only be assumed that previous experience and Match 1 outcome are true measures of performance expectancies. That is, since the quality of past
experience as well as the subject's perception of Match 1 outcome were not directly assessed,
it is possible that these were inaccurate measures of wrestler expectancies. For example,
losing to a top ranked wrestler by only a point or two might be perceived as a successful
performance by a wrestler of lesser ability. Future investigators should therefore reexamine these relationships using direct measures of expectancies.
In a more general sense, while the specific hypotheses made by Martens et al.
(1983) regarding various antecedents of CSAI-2 subcomponentswere not supported, the
present results did show that the various antecedent factors do influence CSAI-2 dimensions in different ways. That is, while significantrelationshipswere found between various
302
GOULD, PETLICHKOFF, AND WEINBERG
antecedents and CSAI-2 subscale scores, no single antecedent was strongly related to all
three subcomponents. For example, the years experience and SCAT predictors showed
little relationship to the confidence subcomponent, but were related to both the cognitive
and somatic anxiety subcomponents. The perceived ability antecedent, on the other hand,
was strongly related to the CSAI-2 confidence subcomponent but showed little relationship to the cognitive and somatic anxiety subcomponents. Thus, support was found for
the general prediction that the various CSAI-2 subdimensionshave different antecedents,
although the precise predictions of Martens and his colleagues were not supported.
Clearly, more research pertaining to the antecedents of CSAI-2 dimensions is needed.
Moreover, future studies should examine antecedent conditions which are specifically
predicted to affect each CSAI-2 subscale differently.
Performance and the CSAI-2
The results of those analyses examining the predicted relationships between
CSAI-2 levels and performance were contradictory. On one hand, no support was found
for the predictions when Match 1 performance was examined. On the other hand, however,
the prediction that cognitive anxiety would be a more powerful predictor of performance
than somatic anxiety received some support in that match outcome and cognitive anxiety
were found to contribute most to this marginally (p < .O8) significant relationship.
The lack of consistent support for the predicted CSAI-2 performance relationships may have resulted for several reasons. First, the CSAI-2 may be similar to previous
state anxiety inventories in that it is not consistently a powerful predictor of athlete performance. Or second, it is possible that the predicted CSAI-2 performance relationships
were not demonstrated because of the way in which the CSAI-2 was used and performance assessed. Sonstroem and Bernardo (1982), for example, found that absolute levels
of state anxiety have little relationship to performance. Consistent relationships were found
only when variations around the subjects' own optimal levels of state anxiety were examined. Thus, based on the Sonstroem and Bernardo (1982) findings, it appears more
appropriate to examine intraindividual relationships between CSAI-2 levels and performance. It was not possible to examine this in the present study since a minimum of three
state anxiety and performance measures is needed for such an analysis. Consequently,
it is recommended that when the relationships between CSAI-2 components and performance are examined in future investigations, intraindividual tests be designed and
implemented.
It is also important to recognize that while the CSAI-2 performance findings
did not support the hypothesized relationships, they are in many ways similar to those
found by Martens and his colleagues (1983). Specifically, these investigators found no
relationship between precompetitive CSAI-2 subscale scores and initial performance (score
on the first nine holes) in junior golfers. However, precompetitive cognitive and somatic
subscale scores were found to be significantly related to later match performance (score
on the last nine holes). There is no clear explanation for these findings, although the consistency of these delayed relationships warrants further examination.
Athlete Experience and the CSAI-2
Gould, Horn, and Spreeman (1983) have suggested that anxiety differences between experienced and inexperienced athletes found by previous investigators (e.g., Fenz,
CSAI-2 SUBCOMPONENTS
303
1975) result from changes in somatic anxiety. Although it was not a primary purpose of
this investigation, this relationship was examined. No significant between-group anxiety
differences were found when experienced and inexperienced high school volleyball players
were examined. It should be recognized, however, that the lack of between-group differences in Study 2 may have resulted from the fact that in previous studies, the athletes
had a wider range of experience than those in the present investigation. This explanation
seems especially credible in light of the strong experience related differences found in
CSAI-2 subscale scores by Martens and his colleagues (1983) when athletes with a wider
range of experience were used. Moreover, Fenz (1975) utilized sport parachutists as subjects in his investigation and suggested that the anxiety differences he obtained may be
specific to sports in which the concern over life itself is of particular importance.
Summary
In conclusion, the results of this investigation support the scale development work
of Martens and his colleagues by verifying that the CSAI-2 assesses three separate components of state anxiety. Second, the results confirmed the prediction that somatic anxiety
increases during the time leading to competition, while cognitive anxiety and confidence
remain constant. Third, only marginal support was found for the prediction that cognitive
anxiety should be a more powerful predictor of performance than somatic anxiety. Finally, while it was shown that CSAI-2 subcomponents are not equally influenced by the same
antecedent factors, the precise prediction Martens and his colleagues (1983) made about
differing relationships between CSAI-2 subscales and various antecedents were not supported. Consequently, more research is needed to determine how and why specific antecedent factors influence various CSAI-2 components, and to test the hypothesized relationships between CSAI-2 components and performance.
References
Borkovec, T.D. (1976). Physiological and cognitive processes in the regulation of anxiety. In
G. Schwartz & D. Shapiro (Eds.), Consciousness and self regulation:Advances in research
(Vol. 1). New York: Plenum.
Carron, A.V., & Morford, W.R. (1968). Anxiety, stress, and motor learning. Perceptual and Motor
Skills, 27, 507-5 11.
Davidson, R.J., & Schwartz, G.E. (1976). The psychobiology of relaxation and related states:
A multiprocess theory. In D.I. Mostofsky (Ed.) Behavior control and rnod$cation of
physiological activity. Englewood Cliffs, NJ: Prentice-Hall.
Fenz, W.D. (1975). Coping mechanisms and performance under stress. In D.M. Landers (Ed.),
Psychology of sport and motor behavior IZ. University Park, PA: Penn State HPER Series
No. 10.
Gould, D., Horn, T., & Spreemann, J. (1983). Competitive anxiety in junior elite wrestlers.
Journal of Sport Psychology, 5 , 58-71.
Highlen, P.S., & Bennett, B.B. (1979). Psychologicd characteristics of successful and nonsuccessful elite wrestlers: An exploratory study. Journal of Sport Psychology, 1 , 123-137.
Karpman, M.B. (1981). Redundancy in canonical analysis. Research Quarterly for Exercise and
Sport, 52, 291-292.
Landers, D.M. (1980). The arousd-performance relationship revisited. Research Quarterly for
Exercise and Sport, 51, 77-90.
304
GOULD, PETLICHKOFF, AND WEINBERG
Liebert, R.M., & Morris, L.W. (1967). Cognitive and emotional components of test anxiety:
A distinction and some initial data. Psychological Reports, 20, 975-978.
Mahoney, M.S., & Avener, M. (1977). Psychology of the elite athlete: An exploratory study. Cognitive lhzrapy and Research, 1, 135-141.
Martens, R. (1977). Sport competition anxiety test. Champaign, Il: Human Kinetics.
Martens, R., Burton, D., Vealey, R., Smith, D., & Bump, L. (1983). i%e development of the
Competitive State Anxiety Inventory-2 (CSAI-2). Unpublished manuscript.
Martens, R., & Gill, D.L. (1976). State anxiety among successful and unsuccessful competitors
who differ in competitive trait anxiety. Research Quarterly, 47, 698-700.
Martens, R., & Landers, D.M. (1970). Motor performance under stress: A test of the invertedU hypothesis. Journal of Personality and Social Psychology, 16, 29-37,
Moms, L.W., & Engle, W.G. (1981). Assessing various coping strategies and their effects on
test performance and anxiety. Journal of Clinical Psychology, 37, 165-171.
Moms, L.W., & Fulmer, R.S. (1976). Test anxiety (worry and emotionality) changes during
academic testing as a function of feedback and test importance. Journal of Educational
Psychology, 68, 817-824.
Nideffer, R.M. (1976). Test of attentional and interpersonal style. Journal of Personality and
Social Psychology, 34, 394-404.
Scanlan, T.K., & Passer, M.W. (1979). Sources of competitive stress in young female athletes.
Journal of Sport Psychology, 1, 151-159.
Schwartz, G.E., Davidson, R.J., & Goleman, D. (1978). Patterning of cognitive and somatic
processes in the self-regulationof anxiety: Effects of meditation versus exercise. Psychosomatic
Medicine, 40, 321-328.
Smith, C.A., & Moms, L.W. (1976). Effects of stimulative and sedative music on two components of test anxiety. Psychological Reports, 38, 1187-1193.
Sonstroem, R.J., & Bemardo, P. (1982). Intraindividualpregame state anxiety and basketball performance: A re-examination of the inverted-U curve. Journal of Sport Psychology, 4, 235-245.
Spielberger, C.D., Gorsuch, R.L., & Lushene, R.E. (1970). Manual for the State-Trait Anxiety
Inventory. Palo Alto, CA: Consulting Psychologists Press.
Weiss, D.J. (1972). Canonical correlation analysis in counseling psychology research. Journal
of Counseling Psychology, 19, 241-252.
Wine, J.D. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104.
Wine, J.D. (1980). Cognitive-attentional therapy of test anxiety. In I.G. Sarason (Ed.), Test
anxiety: meory, research, and applications. Hillsdale, NJ: Erlbaum.
Zuckerman, M. (1960). The development of an affect adjective check list for the measurement
of anxiety. Journal of Consulting Psychology, 24, 457-462.
Manuscript submitted: November 10, 1983
Revision received: April 10, 1984