the effects of anxiety on intellectual performance: when and why are they found?

17
JOURNAL OF RESEARCH IN PERSONALITY 20, 385-401 (1986) The Effects of Anxiety on Intellectual Performance: When and Why Are They Found? GERALD MATTHEWS University of Cambridge An experiment tested the effects on intelligence and creativity test performance of two 16PF primary anxiety traits, and general mood and activation components of state anxiety. Two attentional theories of the deleterious effects of anxiety on performance were also tested. Subjects were 80 male students. Trait and state components of anxiety appeared to affect creativity test performance independently. Intelligence test performance was insensitive to anxiety variables. The 0 anxiety primary factor was significantly negatively correlated with creativity test per- formance, but the unique variance of the other anxiety primary (Q4) was associated with higher levels of performance. Neither specific attentional theory was supported, but data were generally compatible with “dual-mechanism” theories of anxiety, which posit separate detrimental and sometimes facilitative effects of anxiety on performance. 0 1986 Academic Press, 113~. ANXIETY AND INTELLECTUAL PERFORMANCE The effects of anxiety on complex cognitive or intellectual tasks are notoriously inconsistent. Adverse effects of anxiety on intelligence test performance have only modest replicability (see review by Matarazzo, 1972). Three studies have shown (deleterious) main effects of trait anxiety or neuroticism on creativity test or verbal fluency performance (Tapasak, Roodin, & Vaught, 1978; Upmanyu, Gill, & Singh, 1982, White, 1968); two others have not (Di Scipio, 1971; Leith, 1972). Two approaches to resolving these inconsistencies have been followed. One line of research, which often involves correlational methods, is directed toward identifying different types or components of anxiety and determining which affect performance. The other approach is to investigate the dependency of anxiety effects on other factors, such as extraversion (Gupta, 1977), time This research was carried out in partial fulfillment of the requirements for the University of Cambridge Ph.D. degree, while the author was supported by an SSRC grant. Contributions to research expenses were made by the Perrott-Warwick Studentship and the 23 Club. 1 am grateful to Carl Sargent for supervising this research. I also thank an anonymous referee for comments made about a previous draft of this paper. Requests for reprints should be sent to Dr. Gerald Matthews, Applied Psychology Division, Aston University, Aston Triangle, Birmingham BA 7ET, England. 385 OO92-6566/86 $3.00 Copyright 0 1986 by Academic Press, Inc. All rights of reproduction in any form reserved.

Upload: gerald-matthews

Post on 30-Aug-2016

218 views

Category:

Documents


0 download

TRANSCRIPT

JOURNAL OF RESEARCH IN PERSONALITY 20, 385-401 (1986)

The Effects of Anxiety on Intellectual Performance: When and Why Are They Found?

GERALD MATTHEWS

University of Cambridge

An experiment tested the effects on intelligence and creativity test performance of two 16PF primary anxiety traits, and general mood and activation components of state anxiety. Two attentional theories of the deleterious effects of anxiety on performance were also tested. Subjects were 80 male students. Trait and state components of anxiety appeared to affect creativity test performance independently. Intelligence test performance was insensitive to anxiety variables. The 0 anxiety primary factor was significantly negatively correlated with creativity test per- formance, but the unique variance of the other anxiety primary (Q4) was associated with higher levels of performance. Neither specific attentional theory was supported, but data were generally compatible with “dual-mechanism” theories of anxiety, which posit separate detrimental and sometimes facilitative effects of anxiety on performance. 0 1986 Academic Press, 113~.

ANXIETY AND INTELLECTUAL PERFORMANCE

The effects of anxiety on complex cognitive or intellectual tasks are notoriously inconsistent. Adverse effects of anxiety on intelligence test performance have only modest replicability (see review by Matarazzo, 1972). Three studies have shown (deleterious) main effects of trait anxiety or neuroticism on creativity test or verbal fluency performance (Tapasak, Roodin, & Vaught, 1978; Upmanyu, Gill, & Singh, 1982, White, 1968); two others have not (Di Scipio, 1971; Leith, 1972). Two approaches to resolving these inconsistencies have been followed. One line of research, which often involves correlational methods, is directed toward identifying different types or components of anxiety and determining which affect performance. The other approach is to investigate the dependency of anxiety effects on other factors, such as extraversion (Gupta, 1977), time

This research was carried out in partial fulfillment of the requirements for the University of Cambridge Ph.D. degree, while the author was supported by an SSRC grant. Contributions to research expenses were made by the Perrott-Warwick Studentship and the 23 Club. 1 am grateful to Carl Sargent for supervising this research. I also thank an anonymous referee for comments made about a previous draft of this paper. Requests for reprints should be sent to Dr. Gerald Matthews, Applied Psychology Division, Aston University, Aston Triangle, Birmingham BA 7ET, England.

385 OO92-6566/86 $3.00

Copyright 0 1986 by Academic Press, Inc. All rights of reproduction in any form reserved.

386 GERALD MATTHEWS

pressure (Siegman, 1956), and the nature of the task (see M. W. Eysenck, 1982). This approach suggests an experimental approach to test for in- teractive effects of anxiety and other variables on performance.

The research reported on here was part of a wider investigation of the effects of personality traits and states on intellectual performance (Mat- thews, 1983). The experimental procedure was designed to provide both correlational and experimental data on effects of anxiety-related variables on performance. The remainder of this section discusses the dimensions and modifying variables investigated, and identifies unresolved issues to which the present research was directed.

Two separate distinctions between different types of anxiety are of interest here, that between trait and state anxiety, and that between worry and emotionality. According to Spielberger (1966, 1972), state anxiety may be characterized as a transitory emotional state, while trait anxiety reflects a stable predisposition to respond to stimuli threatening to self-esteem with state anxiety. It is clear that state but not trait anxiety consistently impairs working memory (M. W. Eysenck, 1981), and the- oretically effects of trait anxiety are often said to be wholly dependent on state anxiety. Since working memory may be an element of intelligence test performance (see Humphreys & Revelle, 1984). it is plausible that intelligence test performance should be more susceptible to effects of state than trait anxiety. There are few empirical tests of this hypothesis, though the insensitivity of IQ to trait anxiety provides circumstantial evidence for it. Leon and Revelle (1985) obtained significant deleterious effects of state but not trait anxiety on analogical reasoning. Conversely, Samuel (1980) found a small but significant negative correlation (r = - .27) between trait anxiety and IQ, but no effect of state anxiety on IQ (r = - . IO). Unfortunately this study measured state variables post- task rather than pretask, which raises interpretational difficulties. Also, effects of trait anxiety might be expected to the extent that ability testing acts as an ego threat (see Sarason, 1975), and so induces higher state anxiety in trait anxious subjects (see Hodges, 1968).

The research on creativity and anxiety referred to above was confined to trait measures of anxiety or neuroticism. Idzikowski and Baddeley (1983) found a small but significant deleterious effect of situational anxiety prior to public speaking on verbal fluency (one component of “creativity”: Cattell, 1971), but found no significant effect on logical reasoning. This study suggests that state anxiety might be more consistently related to creativity test performance than trait anxiety. Further research comparing effects of trait and state anxiety on intellectual task performance is clearly desirable.

The most important component of self-report anxiety is displeasure, with high arousal or activation making a lesser contribution (Russell & Mehrabian, 1977). If anxiety effects are considered to be dependent on

ANXIETY AND INTELLECTUAL PERFORMANCE 387

state anxiety, it becomes important to determine whether such effects are determined by the pleasure-displeasure and/or activation components of the state.

Morris and Liebert (1969) have shown that (time-pressure dependent) adverse effects of general anxiety on intelligence depend on “worry” rather than “emotionality” items. It would be desirable to extend this distinction to psychometrically more rigorous traits. H. J. Eysenck’s (1967) neuroticsm (N) variable is indivisible, but Cattell (e.g., 1973) has claimed that several primary traits contribute to the anxiety factor ap- proximately equivalent to N and to other trait anxiety measures (see Cattell & Kline, 1977, chap. 5). Two Cattell anxiety primaries appear to correspond approximately to worry and emotionality-0 and Q4, re- spectively. The high-0 subject is said to be apprehensive, self-reproaching, insecure, worrying, and troubled. The high-Q4 subject is tense, frustrated, driven, and overwrought (Cattell & Kline, 1977). A preliminary experiment (Matthews, 1983), suggested that only 0 seemed to be consistently neg- atively associated with performance. Table 1 shows correlations between 16PF anxiety primaries in groups run in quiet and noise. The experiment used a repeated-measures Latin square design, with 18 subjects run in noise followed by quiet, and 18 run in quiet followed by noise. The two testing sessions were a week apart. The Table 1 quiet and noise data were pooled across testing session.

MECHANISMS FOR EFFECTS OF ANXIETY ON PERFORMANCE

Theoretically anxiety is said to induce a generalized attentional deficit (Sarason, 1972; Wine, 1971) due to the tendency of anxious subjects to divert attention toward self-deprecatory cogitation. Attempts to account for effects of anxiety on performance in more detail have led to several “dual-mechanism” theories of anxiety. Such theories posit one mechanism

TABLE 1 PRELIMINARY EXPERIMENT (MATTHEWS, 1983): CORRELATIONS BETWEEN 16PF TRAIT

ANXIETY VARIABLES AND PERFORMANCE, IN QUIET AND NOISE (N = 36)

Trait anxiety variable

Group 0

Intelligence test Quiet -33* Noise -22

Creativity test Quiet -29 Noise -35*

Note. All correlations are Pearson r. * P < .OS. two-tailed test.

Q4 Anxiety

-06 - 18 09 05

-04 - 15 -20 - 33*

388 GERALD MATTHEWS

deleterious to performance and one which may sometimes be facilitatory . The second mechanism accounts for improvements with anxiety sometimes found with easy tasks (see M. W. Eysenck, 1981). The mechanisms of three such theories are shown in Table 2. The difference in the effects of anxiety on effort according to the M. W. Eysenck (1981) and Humphreys and Revelle (1984) theories may be more apparent than real. Eysenck sees effort as related to the intensity of motivation, while Humphreys and Revelle associate “on-task effort” with the direction of motivation. Two of these theories were tested here, those of Hamilton (1975) and M. W. Eysenck (1981). Because effects of anxiety on intellectual per- formance, where found, tend to be detrimental, empirical tests were directed toward those components of anxiety said to impair performance.

Hamilton (1975) sees anxiety as increasing task-irrelevant information load in a limited capacity processing system. If information load exceeds a fixed capacity, then performance decrements will result. Different sources of information load are additive, so tasks with a high rate of information processing would be most sensitive to deleterious effects of anxiety. Facilitative effects of anxiety are accounted for in terms of increased drive. Similarly, M. W. Eysenck (1981) considers that (state) anxiety has both cognitive and motivational/physiological effects. Cognitively, anxiety reduces working memory capacity by generating task-irrelevant cognitions. Motivationally, anxiety is considered to increase effort, such that anxious subjects are often able to maintain performance efficiency at the cost of increased effort. On simple tasks this expenditure of effort may allow anxious subjects to outperform those low in anxiety. There have been few direct tests of these models (though see M. W. Eysenck, 1982). According to Hamilton, though, anxiety decrements should be enhanced primarily by manipulations increasing total information input, regardless of requirements to retain information, while according to M. W. Eysenck, such decrements should be enhanced by conditions reducing working memory capacity.

TABLE 2

“DUAL-MECHANISM” THEORIES OF ANXIETY: DELETERIOUS AND POTENTIALLY FACILITATIVE MECHANISMS FOR THE EFFECTS OF ANXIETY ON PERFORMANCE

Theory Deleterious Potentially facilitative Mechanism mechanism

Hamilton (1975)

Eysenck (1981) Humphreys and Revelle

(1984)

Information overload

Working memory overload

Reduced “on-task

effort”

Drive

Effort

Arousal

ANXIETY AND INTELLECTUAL PERFORMANCE 389

Aims Two approaches to anxiety have been outlined above: the isolation of

the component(s) of anxiety responsible for performance decrements, and the construction of precise attentional theories of anxiety able to predict how anxiety effects on performance will interact with task and situational variables. The first aim of the study was to compare the following components of anxiety as predictors of intellectual performance: activation and general mood (pleasure-displeasure) components of state anxiety, and 16PF 0 and Q4 anxiety primary trait factors. The second aim was to test Hamilton’s (1975) and M. W. Eysenck’s (1981) theories of anxiety, by determining how other theoretically relevant variables (noise and time of day) modified effects of anxiety on performance.

Hypotheses Existing research suggests the following hypotheses concerning rela-

tionships between anxiety-related trait and state variables and performance. First, state variables (mood and activation) should be stronger predictors than trait variables (0 and 44). However, of the two trait variables 0 should be a stronger predictor of performance deficits than Q4. These hypotheses were tested by correlational and multiple regression analyses.

To derive hypotheses from the Hamilton (1975) and M. W. Eysenck (1981) theories it is necessary to find manipulations which have differing effects on external information load and on working memory capacity. Noise vs quiet and late vs early time of day were used for this purpose. Clearly noise but not time of day affects information load. There is strong evidence, from a variety of paradigms, that working memory efficiency peaks in the late morning, and declines during the rest of the day (see Folkard, 1983). White noise tends to reduce working memory capacity (see M. W. Eysenck, 1976, 1982, chap. 8). though the effects may be more complex than a simple capacity loss. In particular, effects of noise on memory are modified by strategy use and the extent to which per- formance depends on order information (see M. W. Eysenck, 1982, chap. 8).

In general, then, the Hamilton (1975) model predicts that noise but not later time of day should enhance anxiety decrements. while on the M. W. Eysenck (1981) model both time of day and noise should enhance anxiety decrements, with these two variables interacting to produce poorest performance in noise in the evening. Both tasks required speeded responses, the intelligence items because time pressure was imposed, the creativity tests because scores depend on the number of responses produced in a given time interval. Thus, on Hamilton’s (1975) model, both tasks should require high rates of information processing, and so would be affected by extraneous information load. The probable role of working memory in intelligence test performance has been referred to above (see Humphreys

390 GERALD MATTHEWS

& Revelle, 1984). M. W. Eysenck (1974a, 1974b) argues that one process on which creativity test performance depends is retrieval from semantic memory. Since items retrieved from long-term memory presumably have to be retained in working memory before being output, it is plausible that creativity test performance will be affected by working memory capacity. Baddeley, Lewis, Eldridge, and Thomson (1985) report that speed of retrieval from semantic memory is sensitive to attentional re- sources, though it is unclear whether this effect involves working memory.

Interactive effects between anxiety and experimental manipulations could also reflect an activation-dependent mechanism obeying the Yerkes- Dodson Law (Broadhurst, 1959). Here the relationship between trait anxiety and activation could be checked. Activation is a component of state anxiety (Russell & Mehrabian, 1977), but analyses of interactive effects of mood and activation on performance were not performed, to prevent overanalysis of the data.

To summarize, the correlational analyses indicated which components of anxiety were linearly related to intellectual task performance. The interactive effects of variables related to anxiety, and noise and time of day distinguished between the cognitive anxiety theories of M. W. Eysenck (1981) and Hamilton (1975).

METHOD

Subjects The subjects were 80 male Cambridge University students and postgraduate students

or recent former students.

Design The experiment used a 2 x 2 design with time of day (12 PM vs 6 PM) and noise vs quiet

as between-subjects factors. Subjects were tested on one occasion only.

Materials Independent variables were chosen as follows: Cattell, Eber, and Tatsuoka’s (1970)

Sixteen Personality Factor Questionnaire (16PF) measures 0 (Apprehensive vs Self-assured) and Q4 (Tense vs Relaxed) and the anxiety secondary factor. 0 and Q4, for which Cattell, Eber, and Tatsuoka (1970) give 2&month test-retest reliabilities of .70 and .66, respectively, are sufficiently stable over time to be considered trait measures. 16PF trait anxiety has been shown to be psychometrically indistinguishable from Eysenck Personality Inventory (EPI: H. J. Eysenck & S. B. G. Eysenck, 1964) neuroticism on a large (N = 1185) sample of British students (Saville & Blinkhom, 1976). A problem here is the questionable psy- chometric discriminability of 0 and Q4 factors (H. J. Eysenck, 1972; Saville & Blinkhom, 1976). However, the preliminary study of Matthews (1983) showed a Pearson correlation between 0 and Q4 of .54 (p < .OOl, N = 36). Although this r is quite high, it is substantially smaller than the immediate test-retest correlations for 0 (.81) and Q4 (.79) reported by Cattell, Eber, and Tatsuoka (1970), which suggests that in the population used here the two factors may be measuring something more than just general anxiety.

The Eight State Questionnaire (SSQ; Curran & Cattell, 1974) gives bipolar state factors clearly distinguished from trait and trait-change factors (Cattell, 1973). including state

ANXIETY AND INTELLECTUAL PERFORMANCE 391

anxiety. Alternate form reliabilities given by Curran and Cattell (1974) are in the range .74-.89. The preliminary experiment showed a general factor, on which all eight states loaded according to the association of their scale items with felt pleasure or displeasure. An index of this general factor labeled “mood” was therefore constructed by subtracting the sum of the 8SQ factors associated with displeasure (stress, depression, fatigue, regression, guilt, anxiety) from the sum of those factors associated with pleasure (arousal, extraversion). Both the SSQ anxiety and arousal factors are positively correlated with physiological indications of arousal or activation such as high pulse rate and skin conductance (see Cattell. 1972; Mefferd. 1966). However, these two factors loaded in opposite directions on the general factor, which had no obvious association with nonspecific activation. To obtain an activation index which was not contaminated with mood, as anxiety and arousal appeared to be, the anxiety and arousal scales were summated to give an activation index. This index thus has no factor analytic basis, and, to the extent that activation is multi- dimensional, may be influenced by several independent activation systems.

Other measures used, but not reported on here, were the Test Anxiety Scale (Sarason, 1972). the Motivation Analysis Test (Cattell. Horn. & Sweney, 1970), the Telic Dominance Scale (Apter. 1982). and post-task introspection questionnaires.

Dependent variables were selected to measure two intellectual abilities: verbal intelligence and creativity. Both were measured using standard test items (see below), but the intelligence items were administered individually to subjects to distinguish effects on speed and accuracy. The anxiety data in fact showed no systematic differences between these dependent variables, which were in any case positively correlated. Thus results here are reported in terms of overall intelligence score. The discriminability of creativity from intelligence has been questioned (e.g.. Cronbach. 1968). but in groups of high ability creativity appears to be both independent of intelligence and psychometrically well defined (Cattell. 1971; Yamamoto, 1965).

The intelligence task used a set of 32 verbal and 4 arithmetical items, taken from the AH5 Group Test of High-Grade Intelligence, Part 1 (Heim. 1968), and the AH6 Group Test of High-Level Intelligence (Heim, Watts, & Simmonds. 1970). Items were mounted on cards for individual presentation. The items were divided into 27 items with a time limit of 2.5 s and 9 with a time limit of 40 s to impose time pressure while allowing for differences in item length. Four creativity tests were selected from the Comprehensive Ability Battery (CAB: Hakstian & Cattell, 1976)-Spontaneous Flexibility (Fs). Ideational Flexibility (Fi), Word Fluency (W). and Originality (Or). These are similar in content to tests of verbal fluency and flexibility conventionally described as creativity tests, and are considered by Hakstian and Cattell (1976) to reflect a global “originality-creativity con- stellation,” and to be moderately positively intercorrelated. though Hakstian and Cattell do not report the magnitude of these correlations. All these creativity tests require the production of as many verbal responses as possible within a given time limit.

Procedure Each subject first completed all the questionnaires used in their own time, except for

the SSQ. Subjects began the experiment at 12 PM or 6 PM. The order of task administration was SSQ (full version), first performance task, introspection questionnaire. SSQ (short version), second performance task, introspection questionnaire. The short version of the SSQ consisted of the arousal and anxiety scales only. Task order and parallel form of the SSQ were counterbalanced across the cells defined by time of day and noise. Testing took approximately 90 min per session. A complex noise stimulus which would impose a high irrelevant stimulus load on processing was used. Subjects allocated to the noise condition were supplied with noise during performance tasks only, using a reel-to-reel tape recorder played through a stereo system and headphones. The tape carried BBC sound effects of nonverbal material such as the sounds of warfare, machinery, and animals on one channel.

392 GERALD MATTHEWS

and nonmelodic electronic music on the other. Using a Brtiel and Kjaer Artificial Ear Type 4153, the audio equipment was calibrated to give a mean intensity of approximately 90 dB SPL, though intensity was highly variable within a range of about 80-105 dB. In quiet, subjects wore headphones with sound supplied. The room used was sound attenuated and generally quiet. Subjects were randomly allocated to the experimental conditions.

RESULTS

Data Analysis

Two sets of analyses were performed, in addition to preliminary analyses. First, simple correlations and multiple regressions were performed to determine linear effects of the Cattell anxiety primaries 0 and Q4 and activation and mood on intelligence and creativity test performance. Second, ANOVAs were performed to test the effects on performance of (1) manipulated variables and trait anxiety and (2) manipulated variables and state anxiety.

All data analysis was carried out by means of the SPSS computer pakage (Nie, Hull, Jenkins, Steinbrenner, & Bent, 1975; Hull & Nie, 1981). In the ANOVAs presented here random assignation of subjects to experimental conditions made unequal cell numbers likely. Hence all ANOVAs used a regression model such that the sum of squares for each term in the model was corrected for all other terms.

Preliminary Analyses

The mean levels of trait anxiety, 0, and Q4 were 5.57, 5.67, and 5.98 respectively, and hence were similiar to normative means of 5.5. The mean level of state anxiety was 5.4 (normative mean 5.5). The positive correlation of .54 (p < .OOl, N = 80) between 0 and Q4 was again substantially smaller than the reliabilities of these variables. The activation and mood indices were independent (r = .15). Correlations between trait and state variables related to anxiety are shown in Table 3. The trait variables appeared to be related to the mood but not the activation component of state anxiety.

TABLE 3 CORRELATIONS BETWEEN TRAIT AND STATE ANXIETY VARIABLES

Trait anxiety variable

0 Q4 Anxiety

State anxiety variable Anxiety Mood Activation

52+ 48* .5x* - 4J* -45* -I%*

20 -02 05

*P < .OOl, two-tailed test.

ANXIETY AND INTELLECTUAL PERFORMANCE 393

Correlations between dependent variables are shown in Table 4, and are similar to those found in the preliminary experiment referred to above. As in that experiment, a creativity index was constructed by summating scores on Fs, Fi, and 0. W was not added to the index because this subtest was more strongly correlated with intelligence than with the other creativity subtests. The creativity index was independent of intelligence score (r = .04, NS, N = 80).

Correlational and Multiple Regression Analyses

The correlational analyses involved the following stages. First, simple correlations between all predictor variables used (plus trait and state anxiety) and the two dependent variables were computed. Second, scat- terplots of dependent variables plotted against predictors were examined for curvilinearity. Third, multiple regressions were computed for the two dependent variables using 0, Q4, and the two pretest state measures as predictors. Fourth, the multiple regressions were repeated using “midtest” state measures derived from the short version of the SSQ given between the two performance tests. Activation could be calculated as before. Mood was estimated by subtracting anxiety from arousal. This measure is a good index of mood calculated from the full version of the 8SQ; the correlation between mood and (arousal-anxiety) in the pretest SSQ data was .91. These last regressions were computed because pretest state measures may be poor predictors of the subject’s state during the per- formance of a task which is itself stressful. However, correlations between midtest state measures and performance might reflect effects of the subject’s appraisal of their performance on their state, rather than the causal effects of state on performance posited by the theories of anxiety discussed above.

Pearson correlations between performance test scores and the trait and state variables related to anxiety are given in Table 5. No anxiety variables were significantly associated with intelligence. Creativity, how- ever, was significantly associated with low 0 and pleasurable pretest and

TABLE 4 INTERCORRELATIONS OF TEST SCORES

12 3 4 5

1. Intelligence - 02 10 36* -04 2. Fs - 51* 16 52* 3. Fi - 10 38* 4. w 14 5. 0 -

*P < ,001, two-tailed test.

394 GERALD MATTHEWS

TABLE 5 CORRELATIONS OF PERFORMANCE TESTS WITH

ANXIETY-RELATED VARIABLES

Performance tests

Anxiety variables

Q4O

Anxiety trait 04 -21 Pretest anxiety state -05 - 10 Pretest mood 11 28** Pretest activation 05 06 Midtest anxiety state -06 -32** Midtest mood 16 40*** Midtest activation 16 12

Note. Int: intelligence. Cr: creativity; all tests two tailed.

*P < .05. **P < .Ol.

***p < .OOl.

midtest mood. Inspection of scatterplots corresponding to these T’S did not show any marked curvilinear relationships.

A multiple regression with all predictors entered into the regression equation indicates whether the unique variance of any variable makes a significant contribution to the overall multiple correlation, provided that the overall equation reaches significance. The magnitude of these unique contributions of variables can be expressed as a partial correlation. Trait- state theory (Spielberger, 1966) predicts that either or both of the com- ponents of state anxiety should make such a significant contribution. Since the correlations between trait and state predictor variables (see Table 3) were substantially than their reliabilities, the multiple regressions were capable of discriminating effects of the two types of variable.

Table 6 gives summary statistics for the regression of intelligence score

TABLE 6 MULTIPLE REGRESSION SUMMARY STATISTICS: INTELLIGENCE SCORE REGRESSED ON ANXIETY

TRAIT VARIABLES AND PRETEST ANXIETY STATE VARIABLES

Variable Beta

:4 -0.14 0.29 Mood 0.16 Activation 0.06

Partial r

-0.11 0.23 0.14 0.06

F Sig. of F

4.3 0.9 NS .04 1.5 NS 0.2 NS

ANXIETY AND INTELLECTUAL PERFORMANCE 395

on 0, Q4, and pretest mood and activation. The overall multiple R for the regression equation was .26, which was not significant (F(4, 75) = 1.33).

Table 7 gives equivalent multiple regression summary statistics for the creativity score as the dependent variable. Here the overall equation was significant (F(4, 75) = 2.94, p < .05): the multiple R value was .37. Of the individual variables, the unique variance of 0 made a significant contribution to predicting creativity variance, and the unique variances of Q4 and pretest mood made marginally significant contributions.

Table 8 gives summary statistics for the multiple regression of intelligence score on 0, Q4, and midtest mood and activation. The overall equation here, with an R of .33, was marginally significant (F(4, 75) = 2.36, p < .lO), with the strongest individual predictor being Q4.

Table 9 gives summary statistics for 0, Q4, and midtest state measures and creativity. Here the overall R was .51, which was highly significant (F(4,75) = 6.64, p < .OOl). Midtest mood, 0, and Q4 all made significant individual contributions to the equation.

Effects of Time of Day, Noise, and Anxiety on Performance

Cell means and N’s for the effects of time of day, noise, and trait anxiety on intelligence and creativity scores are given in Tables 10 and Il. Subjects were divided into high- and low-anxiety groups on the basis of a median split. No significant effects on intelligence involving anxiety were found; the only significant effect in the entire analysis was a main effect of time of day (F(1, 72) = 6.62, p < .05), with mean intelligence score lower in the evening. As for effects of these independent variables on creativity, a marginally significant main effect of trait anxiety was found (F(1, 72) = 2.96, p < . lo), and a significant interactive effect of trait anxiety and time of day (Fl, 72) = 3.85, p = .05). The interaction was due to adverse effects of trait anxiety on creativity being confined to the morning. No significant main or interactive effects of state anxiety were found from the ANOVAs for effects of time of day, noise, and state anxiety on performance. The only significant effect in these analyses was a main effect of time of day on intelligence score (F( 1, 72) = 6.84, p -=I .05).

TABLE 7 MULTIPLE REGRESSION SUMMARY STATISTICS: CREATIVITY SCORE REGRESSED ON ANXIETY

TRAIT VARIABLES AND PRETEST ANXIETY STATE VARIABLES

Variable Beta Partial r F Sig. of F

-0.28 -0.22 3.9 .05 0.24 0.20 3.3 .07

Mood 0.24 0.21 3.5 .06 Activation 0.09 0.09 0.6 NS

396 GERALD MATTHEWS

TABLE 8 MULTIPLE REGRESSION SUMMARY STATISTICS: INTELLIGENCE SCORE REGRESSED ON ANXIETY

TRAIT VARIABLES AND MIDTEST ANXIETY STATE VARIABLES

Variable

:4 Mood Activation

Beta Partial r

-0.19 0.34 -0.16 0.27 0.23 0.19 0.13 0.13

F Sig. of F

2.0 5.8 NS .02 3.5 .08 1.2 NS

TABLE 9 MULTIPLE REGRESSION SUMMARY STATISTICS: CREATIVITY SCORE REGRESSED ON ANXIETY

TRAIT VARIABLES AND MIDTEST ANXIETY STATE VARIABLES

Variable

0 Q4 Mood Activation

Beta Partial Y

-0.31 - 0.28 0.36 0.31 0.46 0.41 0.08 0.08

F Sig. of F

6.4 .Ol 8.0 .Ol

15.0 .OOl 0.5 NS

TABLE 10 INTELLIGENCE SCORE AS A FUNCTION OF TIME OF

DAY, QUIET vs NOISE, AND TRAIT ANXIETY

Group N Intelligence

12 PM Quiet

High anxiety Low anxiety

Noise High anxiety Low anxiety

6 PM Quiet

High anxiety Low anxiety

Noise High anxiety Low anxiety

9 24.2 11 25.6

11 24.7 9 22.6

11 21.6 9 21.9

9 20.4 11 21.8

Note. M = 22.89, SD = 4.86, N = 80.

ANXIETY AND INTELLECTUAL PERFORMANCE 397

TABLE 11 CREATIVITY SCORE AS A FUNCTION OF TIME OF

DAY, QUIET vs NOISE, AND TRAIT ANXIETY

Group

12 PM

Quiet High anxiety Low anxiety

Noise High anxiety Low anxiety

6 PM

Quiet High anxiety Low anxiety

Noise High anxiety Low anxiety

N Creativity

9 - 1.16 11 1.61

11 -0.59 9 0.56

11 -0.61 9 - 0.69

9 0.42 11 0.25

Note. M = 0.00, SD = 2.42, N = 80.

DISCUSSION

The data here provide evidence on (1) the components or types of anxiety to which intellectual performance is sensitive and (2) theoretical accounts of effects of anxiety on performance. A general observation is that the small magnitude of the significant effects here implies that the data are of more theoretical than practical importance.

Empirical Effects of Anxiety-Related Variables on Performance

No evidence was found here to suggest that effects of anxiety on performance are wholly dependent on state anxiety. Intelligence test performance was generally insensitive to anxiety, but both creativity multiple regressions showed significant effects of predictor variables on performance. These regressions showed deleterious effects of 0 on crea- tivity, which were not mediated by mood or activation. The data were also suggestive of an independent facilitative effect of the unique variance of Q4. Mood data were somewhat equivocal. While midtest mood was relatively strongly related to creativity performance, pretest mood showed only a marginally significant relationship in the multiple regression. These data could reflect either midtest mood being a better measure than pretest mood of subjects’ mood during task performance, or causal effects of performance on mood, mediated by subjects’ appraisals. In general, though, the data suggest trait and state anxiety variables independently predict creativity test performance.

398 GERALD MATTHEWS

The association between 0 and poor creativity performance supports the idea that detrimental effects of anxiety are dependent on worry (Morris & Liebert, 1969), given Cattell, Eber, and Tatsuoka’s (1970a) description of 0. Thus it appears that worry can be related to a factor analytically derived personality trait. However, it is likely that the dis- crimination of the specific effects of 0 and Q4 found here were contingent upon the unusually low correlation found between 0 and Q4. Other psychometric data (Saville & Blinkhom, 1976) suggest that the psychometric discriminability of 0 and Q4 could be much improved.

Theoretical Implications of the Data Theoretically, data on the interaction of anxiety with noise and time

of day failed to support the theories of either Hamilton (1975) or M. W. Eysenck (1981). No significant main or interactive effects of state anxiety on performance were found. The significant interactive effect of time of day and trait anxiety on performance cannot be predicted from the Hamilton (1975) theory, and is exactly opposite to the prediction from M. W. Eysenck (1981). There is no obvious explanation for this interaction. It cannot be attributed to arousal, since here neither trait anxiety nor time of day affected activation.

However, the data here do support some “dual-mechanism” account of the effects of anxiety on performance, in that while 0 affected creativity adversely, the unique variance of Q4 was associated with significant or nearly significant facilitatory effects on this task. Interactive effects of 0 and time of day and noise were not tested in order to limit the number of analyses performed. Thus no strong statements about the nature of the mechanism for the effects of 0 can be made. However, given the lack of evidence for the information load and working memory hypotheses, the data suggest that 0 is associated with a generalized attentional deficit of the kind postulated by Wine (1971) and Sarason (1972). Leon and Revelle (1985) also obtained evidence supporting a simple attentional theory over a working memory theory (for state anxiety under nonstressful conditions). One possible mechanism is suggested by Humphreys and Revelle (1984) who distinguish processing resources for “sustained in- formation transfer” (processing not requiring retention of information) from those for short-term memory. 0, and perhaps mood also, may therefore reduce sustained information transfer resources. This suggestions implies that performance of creativity tests is not in fact strongly limited by demands on working memory.

In general, then, the data support the idea that there are two mechanisms by which anxiety affects intellectual task performance. However, it may be that trait worry and state anxiety (or its mood component) independently affect attention adversely. The present data suggest that this mechanism should be distinguished from effects of anxiety on short-term memory,

ANXIETY AND INTELLECTUAL PERFORMANCE 399

which do seem to be entirely dependent on state rather than trait anxiety (M. W. Eysenck, 1981). Thus the Humphreys and Revelle (1984) theory provides the best account of detrimental effects of anxiety on creativity test performance. Facilitatory effects of Q4 on performance were less reliable than effects of 0, but, if real, do not seem to reflect an arousal mechanism, given the independence of creativity and activation. Such effects would be consistent with the effort mechanism posited by M. W. Eysenck (1981). Thus future research could most usefully be directed toward developing dual-mechanism theories of anxiety.

REFERENCES Apter, M. J. (1982). The experience of motivation: The theory of psychological rersersals.

Orlando/London: Academic Press. Baddeley, A., Lewis. V.. Eldridge, M., & Thomson, N. (1985). Attention and retrieval

from long-term memory. Journal qf Experimental Psychology: General, 113, 518-540. Broadhurst, P. L. (1959). The interaction of task difficulty and motivation: The Yerkes-

Dodson law revisited. Acta Psychologica, 16, 321-338. Cattell, R. B. (1971). Abilities: Their structure, growth and uction. New York: Houghton

Mifflin. Cattell, R. B. (1972). The nature and genesis of mood states: A theoretical model with

experimental measurements concerning anxiety. depression. arousal, and other mood states. In C. D. Spielberger (Ed.), Anxiety: Current trends in theor?, and research (Vol. 1). New York: Academic Press, 1972.

Cattell. R. B. (1973). Personality and mood by questionnaire. San Francisco: Jossey- Bass..

Cattell, R. B.. Eber, H. W., & Tatsuoka, M. M. (1970). Handbook for the Sixteen Personulity Factor Questionnaire. Champaign, IL: IPAT.

Cattell, R. B.. Horn, J., & Sweney. A. B. (1970). Motivation Analysis Test. Champaign, IL: IPAT.

Cattell, R. B., & Kline, P. (1977). The scientiJic ana/ysis of personality and motirution. New York: Academic Press.

Cronbach, L. J. (1968). Intelligence? Creativity? A parsimonious reinterpretation of the Wallach-Kogan data. American Educational Research Journal, 5, 491-511.

Curran. J. P., & Cattell, R. B. (1974). The Eight State Questionnaire. Champaign, IL: IPAT.

Di Scipio, W. J. (1971). Divergent thinking: A complex function of interacting, dimensions of extraversion-introversion and neuroticism-stability. British Journal ofPsychology, 62, 545-550.

Eysenck, H. J. (1967). The biological basis of personality. Springfield, IL: Thomas. Eysenck, H. J. (1972). Primaries or second-order factors: A critical consideration of Cattell’s

16PF Battery. British Joarnul of Social and Clinical Psychology, 11, 265-269. Eysenck, H. J., & Eysenck, S. B. G. (1964). The Eysenck Personality Inventory. London:

Univ. of London Press. Eysenck, M. W. (1974a). Individual differences in speed of retrieval from semantic memory.

Journal of Research in Personality, 8, 307-323. Eysenck, M. W. (1974b). Extraversion, arousal, and retrieval from semantic memory.

Journal of Personality, 42, 3 19-33 1. Eysenck, M. W. (1976). Arousal, learning and memory. Psychological Bulletin, 83, 7_5-

90. Eysenck. M. W. (1981). Learning, memory and personality. In H. J. Eysenck (Ed.), A

model for personality. Berlin: Springer-Verlag.

400 GERALD MATTHEWS

Eysenck, M. W. (1982). Arousal and attention: Cognition and performance. Berlin: Springer- Verlag.

Folkard, S. (1983). Diurnal variation. In G. R. J. Hockey (Ed.), Stress and fatigue in human performance. London: Wiley.

Gupta, B. S. (1977). Dextroamphetamine and measures of intelligence. Intelligence, 1, 274-280.

Hakstian, A. R., & Cattell, R. B. (1976). Manual for the Comprehensive Ability Battery (CAB). Champaign, IL: IPAT.

Hamilton, V. (1975). Socialization anxiety and information processing: A capacity model of anxiety-induced performance deficits. In I. G. Sarason & C. D. Spielberger (Eds.). Stress and anxiety (Vol. 2). Orlando/London: Academic Press.

Heim, A. W. (1968). AH5 Group Test of High-grade Intelligence manual. Windsor: NFER. Heim, A. W., Watts, K. P., & Simmonds. V. (1970). Manual for the AH6 Group Tests

of High-level Intelligence. Windsor: NFER. Hodges, W. F. (1968). Effects of ego threat and pain on state anxiety. Journal of Personality

and Social Psychology, 8, 364-372. Hull, C. H.. & Nie, H. H. (1981). SPSS update 7-9: New procedures for Releases 7-9.

New York: McGraw-Hill. Humphreys, M. S., & Revelle, W. (1984). Personality, motivation and performance: A

theory of the relationship between individual differences and information processing. Psychological Review. 91, 153-184.

Idzikowski, C. J. F., & Baddeley, A. D. (1983). Waiting in the wings: Apprehension, public speaking and performance. Ergonomics, 26, 575-584.

Leith, G. (1972). The relationships between intelligence, personality and creativity under two conditions of stress. British Journal of Educational Psychology, 42, 240-247.

Leon, M. R., & Revelle. W. (1985). The effects of anxiety on analogical reasoning: A test of three theoretical models. Journal of Personality and Social Psychology, 49, 1302- 1315.

Matarazzo, J. (1972). Wechsler’s measurement and appraisal of adult intelligence (5th ed.). Baltimore: Williams & Wilkins.

Matthews, G. (1983). Personality, arousal states and intellectual performance. Unpublished doctoral dissertation, University of Cambridge.

Mefferd, R. B., Jr. (1966). Structuring physiological correlates of mental processes and states: The study of biological correlates of mental processes. In R. B. Cattell (Ed.), Handbook of multivariate experimental psychology. Chicago: Rand McNally.

Morris, L. W., & Liebert, R. M. (1969). Effects of anxiety on timed and untimed intelligence tests: Another look. Journal of Consulting and Clinical Psychology, 33, 240-244.

Nie, N. H., Hull, C. H., Jenkins, J. G., Steinbrenner, K., & Bent, D. H. (1975). Statistical package for the social sciences (2nd ed.). New York: McGraw-Hill.

Russell, J. A., & Mehrabian, A. (1977). Evidence for a three-factor theory of emotion. Journal of Research in Personality, 11, 273-294.

Samuel, W. (1980). Mood and personality correlates of IQ by race and sex of subject. Journal of Personality and Social Psychology, 38, 993-1004.

Sarason, 1. G. (1972). Experimental approaches to test anxiety: Attention and the use of information. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2). Orlando/London: Academic Press.

Sarason, I. G. (1975). Anxiety and self-preoccupation. In I. G. Sarason & C. D. Spielberger (Eds.), Stress and anxiety (Vol. 2). Washington, DC: Hemisphere.

Saville, D.. & Blinkhorn, S. (1976). Undergraduate personality by factored scales: A large scale study on Cattell’s 16PF and Eysenck Personality Inventory. Windsor: NFER.

Siegman, A. W. (1956). The effect of manifest anxiety on a concept formation task, a non- directed learning task. and on timed and untimed intelligence tests. Journal of Consulting Psychotogy. 20. 176-178.

ANXIETY AND INTELLECTUAL PERFORMANCE 401

Spielberger, C. D. (1966). Theory and research on anxiety. In C. D. Spielberger (Ed.), Anxiety and behavior. New York: Academic Press.

Spielberger, C. D. (1972). Anxiety as an emotional state. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 1). Orlando/London: Academic Press.

Tapasak, R. C., Roodin, P. A., & Vaught, G. M. (1978). Effects of extraversion, anxiety, and sex on children’s verbal fluency and coding task performance. The JoumuI QJ Psychology, 100, 49-55.

Upmanyu, V. V., Gill, P. S., & Singh, S. (1982). Nature of unusual responses in Kent- Rosanoff Word Association Test and Torrance Test of Creativity. Personality Study and Group Behu\*iour, 2, 44-53.

White, K. (1968). Anxiety, extraversion-introversion and divergent thinking ability. Journal of Creative Behaviour,2, 119-127.

Wine, J. (1971). Test anxiety and the direction of attention. Psychological Bu//etin. 76, 92-104.

Yamamoto, K. (1965). Effects of restriction of range and test unreliability on correlation between measures of intelligence and creative thinking. British Journal of Edwxtional Psychology, 30, 300-305.