Journal of Abnormal Child Psychology, VoL 7, No. 1, 1979, pp. 45-59
Test Anxiety and Evaluation Threat: Children's Behavior in the Classroom 1 Jeri Dawn Wine The Ontario Institute for Studies in Education
Children differing in test anxiety level were observed in two art classes, one immediately preceding a classroom examination, the other when no examination was expected. 'The observational analyses indicated the following: (a) When an examination was expected there were general increases both in children's task orientation and in their concern with other's evaluations, and a decrease in general activity level. It was suggested that future research examine the effects o f additional situational variables on children's classroom behaviors. (b ) Highly test-anxious children were generally hardworking and inactive. They reacted to examination expectancy with a decrease in task orientation, reduction in communications, and attentional constriction. Less anxious children reacted to examination expectancy by working harder, eliminating task-irrelevant behaviors, and seeking feedback. These results were interpreted as supporting a cognitive theory o f test anxiety. The effects of test anxiety on task performance have been well documented. The majority of studies indicate an interaction between evaluative stress and test anxiety in their effects on task performance (see Phillips, Martin, & Meyers, 1972; Sarason, 1972; Wine, 1971, for reviews). The highly test-anxious person performs most poorly in highly evaluative situations, and best under nonevaluatire conditions, while the reverse is true for the low test-anxious persons. Research with children suggests that this effect is obtained as tasks are varied along a dimension from "garnelike" to "testlike" (Lighthall, Ruebush, Sarason, & Zweibelson, 1959; Sarason, Davidson, Lighthall, Waite, & Ruebush, 1960; Zweilbelson, 1956) and with audience absence versus presence (Cox, 1966, 1968). Manuscript received in final form April 27, 1978. 1This research was supported by Canada Council Grant No. S71-0272. 45 0091-0627/79/0300-0045503.00/0 C) 1979 Plenum Publishing Corporation
46
Wine
Though there is a good deal of evidence regarding systematic variations in task performance as a function of test anxiety and stress, there is little evidence regarding variations in other behaviors. The classroom seems an ideal setting within which to research naturalistic variations in stress and their effects on children's behaviors. However, the effects of situational variables are so strong that it is quite difficult to make comparisons of children's behaviors across classroom situations. For example, it is probably not meaningful to compare a child's behavior while taking a classroom test with that child's behavior during a class in oral reading, because the appropriate behaviors in the two situations are different. These difficulties were noted in classroom observational studies of high and low test-anxious children reported by Sarason, Davidson, Lighthall, and Waite (1958) and Davidson and Sarason (1961). Consequently, these researchers settled for global clinical ratings rather than moment-to-moment systematic behavioral observations. An alternative to examining children's behavior across all classroom situations is to manipulate evaluative stress within a particular kind of class that makes reasonably consistent demands on children's behaviors, preferably one that allows for a fairly wide range of behaviors and for interactions among the children and between the children and the teacher. Art class is one such situation; it is a regularly scheduled event in most elementary school classrooms; it often takes place in the normal classroom with the regular teacher; and typically, children are given a measure of freedom to move around and interact. It would be difficult, and perhaps inappropriate, to introduce evaluative stress within this situation, but it is possible to introduce the expectancy of an impending examination. Research has indicated that "worry" regarding a particular evaluating situation tends to remain constant across time and does not fluctuate with contiguity to the evaluating situation (Liebert & Morris, 1967). Barnard, Zimbardo, and Sarason (1961) interviewed high and low test-anxious children, half of them in a permissive atmosphere and half with the instruction that there would be a test following the interview. In the evaluative interview condition the high anxious children used more negative expressions than the low anxious children, while there were no differences in negative expressions in the permissive interview. These results indicate that the expectation of evaluation interacts with anxiety level in affecting children's overt behaviors. In the present study, the classroom behaviors of children varying in test anxiety level were systematically observed in an art class immediately followed by a difficult classroom examination, and during an art class when no tests were scheduled for the entire day. The following questions were addressed: How are differences in test anxiety level manifested in children's overt classroom behaviors? Do these behaviors change systematically with changes in evaluative stress? The earlier classroom observation studies (Davidson & Sarason, 1961; Sarason et al., 1958) reported that test anxiety related more clearly to the behaviors of boys than to those of girls. Sex differences were investigated further in the present study.
Anxiety, Evaluation Threat, and Classroom Behavior
47
In a broader sense, the study examined the issue of the nature of test anxiety. Sarason (1966) noted in a review of test anxiety research with children that high test-anxious children show none of the overt behavioral signs of unambiguous anxiety, such as agitation, flight, extreme physiological reactivity, high distractibility. On the contrary, test anxiety is associated with caution, constriction, and a passive orientation to authority. He suggests that "the cognitive consequences of anxiety affect personality to a greater degree than does the anxiety to which they were a response" (p. 70). Other test anxiety reviewers (Sarason, 1972, 1975; Wine, 1971) have supported cognitive interpretations of the nature of test anxiety. The present study sought to provide additional information on the issue of a cognitive versus an emotional reactivity interpretation of test anxiety.
PROCEDURE
Design The design of the present study is a quasi-experimental one, i.e., data were systematically collected on naturally occurring behaviors in a naturalistic setting, with experimental control exercised over the variable of Examination Expectancy. Sixty-six children in three Grade 4 classrooms were observed during an art class in which the children expected a difficult classroom examination to be immediately following, and during an art class when no tests were scheduled for the entire day. The children were divided into subgroups on the basis of sex, and high, middle, and low test-anxiety scores. The subgroups were equated for IQ and socioeconomic status. The art classes were videotaped and the children's behaviors coded from the tapes on the basis of 22 observational categories. The overall design of the study was a 2 (Sex) • 3 (Test Anxiety) • 2 (Examination Expectancy) analysis of variance design, with repeated measures on the third variable.
Subjects and Setting The subjects were children in three Grade four classrooms in three Waterloo County, Ontario, schools. All of the classes were traditional, one-teacher, oneroom, single-grade, nonstreamed classes, two taught by females, the third by a male. In all three classes, art lessons were given for 1 hour each week by the regular teacher. And in all three the children occupied assigned desks. There were 87 children in total in the three classrooms. The parents of 9 children refused permission for their children to participate in the study; data on 4 children were not used because of absence on 1 of the 2 observation days.
48
Wine
Five subjects were dropped from the final analyses in order to achieve comparability in IQ and socioeconomic status across subgroups. Three additional subjects were dropped randomly in order to achieve equal Ns in the subgroups for the analyses of variance. The final sample consisted of 66 children, 11 in each of the six subgroups formed by the Sex by High, Middle, and Low testanxiety combinations.
Individual Difference Measures
The Test Anxiety Scale for Children (Sarason et al., 1960) was groupadministered to each class approximately 1 month prior to the 1st day of observation. The distributions of TASC scores were split into thirds separately for boys and girls forming high (HA), middle (MA), and low (LA) test-anxious groups. Information was collected on IQ and socioeconomic status since these variables have been shown to interact with test anxiety in effects on behaviors. The children in all three classrooms had been administered the Otis-Lennon Mental Ability Test at the end of Grade 3. Paternal occupations served as the basis for socioeconomic status classification (Blishen, 1958). Table I presents the test anxiety, IQ, and SES data for the six subgroups. The groups are roughly comparable in IQ and SES. Observational Situations
Each classroom was observed on two occasions during their weekly art classes, which involved, in each case, individual projects completed at the child's assigned desk. In the Examination Expected condition, the children were informed of the examination by their teacher several days prior to the observed art class; they were reminded at the beginning of the art class that they would be taking the test in the hour following the class. For two of the classrooms, the No ExTable I. Means and Standard Deviations on IQ, SES, and TASC Boys LA IQ SES TASC
)~ SD X SD X" SD
MA
Girls HA
102.00 98.00 101.00 11.09 12.74 10.88 4.27 4.55 4.55 1.56 1.57 .93 8.46 14.55 19.09 1.37 2.98 2.88
LA
MA
HA
101.91 99.18 98.91 10.08 11.45 14.40 4.46 4.91 4.55 1.21 1.45 1.29 9.82 16.27 21.82 2.49 1.68 1.78
Anxiety, Evaluation Threat, and Classroom Behavior
49
amination art class was scheduled first; for the third class the Examination Expected class was scheduled first.
Observational Method Video recordings were used to obtain behavior records. Each class was observed for a half day prior to the classes of interest in order to accustom the children to the presence of the observer. On the actual observation days, the video equipment was set up several hours prior to the actual observational situation, the equipment was run continuously, and the children were not made aware of the situation which was of interest to the observer. The equipment was set up in the rear of the classroom so that it would not be in the children's line of vision. In order to obtain codable records, each classroom was divided into three sections, each containing approximately one-third of the children, and the recordings were made on a time-sampling basis of 5 minutes per section. The art classes lasted an hour, with the first 5 to 10 minutes occupied by the teacher's instructions. There were 15 minutes in total of videotape per section of the classroom. Thus each child was videotaped for 15 minutes in both the Examination Expected and the No Examination classes. Three observers, trained on a classroom videotape not used in this study, coded the children's behaviors. The observers' task was to observe and code each child's behavior individually; during coding the videotapes were replayed as many times as there were children in the section of the classroom on each videotape, 7-10 children. Checks were made on agreement among observers following their training and on two occasions during the actual coding. Percentages of agreement between pairs of observers on total day records for individual children were uniformly high, ranging from 92% to 98% agreement, with total number of coded behaviors agreed upon divided by the total number of coded behaviors. The observers coded blind, without knowledge of individual children's test anxiety scores, or of Examination Expectancy conditions.
Observational Categories Unfortunately, in a busy classroom it was not possible to record the verbal content of the children's communications. The set of 22 observational categories was designed, through several months of pretesting, to be descriptive of other observable task, activity, and interactional behaviors. The observational unit was the discrete observable behavior. An arbitrary time unit of 30 seconds was also used, i.e., at least one behavior was coded in each 30-second interval, though more behaviors were recorded if more occurred. Behaviors were coded in the following categories:
50
Wine
Attending Behaviors. These were coded when children oriented head and eyes in the direction of speaker(s): (1) Attends to teacher's communication to the class, (2) attends to teacher's communication to another child, (3) attends to other children's conversations. Task-Related Behaviors. These were coded if the child engaged in no other codable behavior for a 30-second interval: (1) Works quietly on art tasks, (2) sits idle. Activity. (1) Takes more art materials, (2) stands, (3) leaves desk, (4) looks at camera. Communication. (1) Initiates communication with teacher, (2) initiates communication with another child, (3) receives communication from teacher, (4) receives communication from another child. Interactional Behaviors. (1) With teacher: (a) raises hand, (b) displays work to the teacher, (c) receives help or material from the teacher. (2) With another child: (a) displays work to child, (b) prosocial behavior (gives help, shares, passes materials, etc.), (c) jokes or laughs, (d) task-distruptive or aggressive behaviors, (e) helped by child. In addition to the observations of the children, several observations were made of the teachers to ensure there were no significant variations in their behavior across the two observation occasions. As the videotapes were intended to yield codable records of the children's behaviors and not of the teachers', the teachers were frequently out of view. While the children's verbal communications were rarely audible, the reverse was true for the teachers; most of their communications were loud and clear. The teachers' audible verbal behaviors were coded in several broad categories: (a) total amount of time spent talking, (b) number of communications, (c) communication directed to the class or to an individual child, and (d) communication content: task instruction, evaluation of work or behavior, or question. These codings of the teachers' verbal behaviors revealed no significant differences between the Examination Expected and No Examination classes.
RESULTS AND DISCUSSION
Data on 5 of the observational categories are not reported here either because too few behaviors were observed, i.e., the means for none of the subgroups exceeded 1.00, or no significant Fs emerged from the analyses of variance. They were: attends to other children's conversations, receives communication from the teacher, task disruptive or aggressive behaviors, prosocial behaviors, and jokes or laughs. Data on the remaining 17 observational indices are reported in terms of the Sex by Anxiety by Examination Expectancy (2 • 3 • 2) repeatedmeasures analyses of variance. The large number of observational categories, coupled with the rather brief length of time each child was observed in each art
Anxiety, Evaluation Threat, and Classroom Behavior
51
class, resulted in small numbers of behaviors coded in some categories, tn view of this, results are reported on F ratios that approached significance, between the .05 and. 10 levels.
Examination Expectancy Conditions Perhaps the most surprising aspect of the results was the pervasive effects that the anticipation of an examination had upon the behaviors of the children in general, disregarding anxiety level. In the Examination Expected (E) class as compared to the No Examination (NE) class, the children spent more time working quietly on their art tasks, F(1, 60) = 4.09, p < .05 (E M = 12.70, NE M = 11.74), and less time sitting idle, F(1, 60) = 6.38, p < .025 (E M = 1.18, NE M = 2.21). They attended more to the teacher's communications to the class, F(1, 60) = 18.93, p < .001 (E M = 4.80, NE M = 2.53), displayed their work to the teacher, F(1, 6 0 ) = 27.38, p < .001 ( E M = .95, N E M = 15)and to other children more often, F(1,60) = 15.51, p < .001 (E M = 1.29, NE M = .47). They took materials less often, F(1, 60) = 25.14, p < .001 (E M = .53, NEM = 3.36), stood at their desks less,/7(t, 60) = 10.76, p < .005 (E M = 3.03, NE M = 4.97), and looked at the camera less often, F(1, 60) = 9.65, p < .005 (E M = 2.35, NE M = 3.96). They also received fewer communications from other children, F(1, 60) = 4.16, p < .05 (E M = 3.80, NE M = 5.02). The overall picture is one of change from the No Examination class to the Examination Expected class toward more intense task orientation, greater concern with others' evaluations, and behavioral constriction.
Test Anxiety and Examination Expectancy The pattern of the results for test-anxiety level and its interaction with Examination Expectancy is, overall, consistent and meaningful within a f}amework of a cognitive interpretation of test anxiety. Attending Behaviors. As noted above, all of the children listened to the teacher's communications to the class more in the Examination Expected than the No Examination class. However, an Anxiety by Examination Expectancy interaction, which approached significance, F(2, 60) = 2.37, .10 > p > .05, suggested that LA and MA children listened to the teacher's communications with other children more in the Examination Expected than the No Examination class while HA children listened to this kind of communication slightly tess in the Examination Expected class (see Figure 1). Though these results are of borderline significance, they are provocative. With the threat of evaluation all of the children became more attentive to the evaluating authority figure in the situation. However, under evaluative stress, the HA child tended to listen only to what she/he was "supposed" to listen to, i.e., the teacher's communications
Wine
52 3.0-
2.5-
2.0-
o
4-"
~3 1,5
/ /
/
/
/
/
a MA
/
//
E Z
1.0~ //
""-.b~
LA
.5-
Conditions
Fig. 1. Attending to teacher's communication to another child: anxiety X conditions interaction.
to the entire class but not to other children, while the MA and LA child became more alert to both kinds of teacher communications. The result is suggestive of the proposition that attentional constriction accompanies increases in anxiety (Easterbrook, 1959; Wachtel, 1968; Wine, 1971), if one interprets teacher's communications to other children as a kind of peripheral task cue. Task-Related Behaviors. Sarason et al. (1958) and Davidson and Sarason (1961) reported results regarding test anxiety and sex effects on task orientation. The classroom behaviors of LA boys were rated as more task-oriented than HA boys, while HA girls were more task-oriented than LA girls. In the present study, Sex and Anxiety level did not interact in their effects on task orientation.
53
Anxiety, Evaluation Threat, and Classroom Behavior 15-
14-
'""--i
13-
HA
,,e LA
r O
sr
/
t, o
/
sS
.B
MA
.1
12-
e~
/
E
gS s/S
z
sS
11-
sS g gS
Ir
10-
Conditions
Fig. 2. Working quietly: anxiety • conditions interaction.
There was a significant main effect for test anxiety on the amount of time spent working on the art tasks, F(2, 60) = 4.05, p < .05, with the HA children spending more time overall working quietly (M = 13.18) than the MA (M = 11.80) or LA children (M -- 11.68). However, a near-significant interaction between Anxiety and Examination Expectancy, F(2, 60) = 2.79,. 10 > p > .05, indicated that LA and MA children spent more time working quietly in the Examination Expected than in the No Examination class, while HA children spent slightly less time working in the Examination Expected class. This Anxiety by Expectancy effect is mirrored by the amount oft/me spent sitting idle, F(2, 60) = 2.96, .I0 > p > .05. The LA and MA children spent less time sitting idle in the Examination Expected class
54
Wine 5-
4-
O
I~%x% %x
3-
"~xx
.B
o ~6 E z
x
2-
~
HA
~-
1-
~LA ~r~MA
0
9
N'E
6 Conditions
Fig. 3. Sitting idle: anxiety X conditions interaction.
than in the No Examination class, while HA children were idle slightly more in the Examination Expected class. These effects are presented in Figures 2 and 3. The pattern, though again only approaching significance, is quite consistent with laboratory findings that with increases in evaluation emphasis, the low testanxious person becomes more task-oriented and works harder, while the high test-anxious person becomes self-preoccupied and less task-oriented (e.g., Sarason, 1975; Wine, 1971). More information regarding task-orientation differences between anxiety levels was provided by the finding of a main effect, F(2, 60) = 3.58, p < .05, for anxiety level on looking at other children's work - in the rcontext of an art class, an entirely appropriate behavior and a source of task-relevant cues. The HA children looked at the work of others less (.41/= 7.80) than did the MA (M = 9.21) and LA children ( 3 / = 11.07). These results suggest that the high testanxious child may be less alert than the less anxious child to peripheral task cues.
Anxiety, Evaluation Threat, and Classroom Behavior
55
Child-Teacher Interactions. Highly test-anxious children, especially boys, have been described as dependent upon the approval of authority figures, direction seeking, and occasionally unresponsive (Barnard et al., 1961 ; Sarason, 1966). The single result bearing upon this issue was a significant interaction between Sex and Anxiety level on the number of times the teacher helped a child, F(2, 60) = 6.84, p < .025, with the HA boys and the MA girls receiving the most help of the Sex X Anxiety subgroups. Child-Child Interactions. The literature regarding interactions among children as a function of test anxiety is sparse. The results of two studies of children's sociometric choices, reported by Hill (1963) and Phillips, Adams, Gotts, and McNeil (cited in Phillips et al., 1972), suggest that the HA child may be something of a social isolate in the classroom. Several analyses in the present study shed some light on how this isolation is manifested in the HA child's classroom interactions. The HA children did not initiate or receive fewer communications overall than did the other anxiety groups. However, there was a significant interaction between Anxiety and Examination Expectancy on the number of communications initiated, F(2, 60) = 6.62, p < .01. The LA and MA children initiated slightly more communications with other children in the Examination Expected than in the No Examination class, while the HA children initiated markedly fewer communications with other children in the Examination Expected than in the No Examination class. The HA children were also the only test-anxiety subgroup to receive significantly fewer communications in the Examination than in the No Examination situation, individual means comparison, F(1,20) = 5.91 ,p < .05. These effects are presented in Figures 4 and 5. The HA child was not an isolate under normal, nonevaluative classroom conditions, but under stress she/he did become isolated. Another result has implications for the nature of the relationship between the HA child and other children. A significant main effect for anxiety level, F(2, 60) = 8.27, p < .01, revealed that the HA child was more likely to be helped by other children (3/= 1.41) than were MA (M = .64) and LA children (3/ = ~ The test-anxiety literature has indicated that HA children are dependent in their relationships with authority figures; this result suggests they are dependent in their relationships with other children as well. Activity Indices. Analyses of the activity indices suggested that the HA children were less active, in general, than the LA children, standing at their desks less often, individual means comparison, F(1,42) = 4.10, p < .05 (HAM = 3.27, LA M = 4.59), and leaving their desks fewer times, individual means comparison, F(1, 42) = 6.90, p < .025 (HA M = 1.46, LA M = 2.4 t) than LA children. Two test anxiety by Examination Expectancy interactions and appropriate individual means comparisons suggested that both MA and LA children engaged in fewer task-irrelevant behaviors in the Examination Expected class than in the No Examination class. LA children looked at the camera markedly fewer times in the Examination Expected (M = 2.14) than in the No Examination class (3'/=
56
Wine 7-
,,m LA 6-
s
~ 5-
s
ssSSSSssS ssSIsS jjSS //BMA,///,//,
O
> o
4rq/
..O
E Z 3HA
2-
N'E
I~ Conditions
Fig. 4. Communications initiated to other children: anxiety • conditions interaction.
5.36), F(1, 20) = 13.01, p < . 0 0 5 , while MA children stood at their desks significantly less often in the Examination Expected (M = 2.09) as compared to the No Examination class (M = 6.19), F(1,20) = 16.17, p < .005. Sex Differences. There were surprisingly few sex differences, or interactions between Sex and the Test Anxiety and Examination Expectancy variables. Boys were significantly more likely (M = 2.56) than girls (M--- 1.18) to spend time sitting idle, F(1, 60) = 8.17, p < .01, and to look at other children's work (Boys M = 10.61, Girls M = 8.10), F(1, 60) = 5.89, p < .05. A significant interaction between Sex and Examination Expectancy, F(1, 60) = 4.30, p < .05, indicated that though both boys and girls displayed their work to the teacher more
Anxiety, Evaluation Threat, and Classroom Behavior
57
7.
6.
.......
o
-o LA
o
:IVIA
E Z
3-
2-
N'E Conditions
Fig. 5. Communicationsreceived from other children: anxiety X conditions interaction.
often in the Examination Expected than in the No Examination class, this effect was much more marked for boys. A significant interaction between Sex and Anxiety, F(2, 60) = 5.42, p < .01, revealed that among the Sex • Anxiety subgroups LA boys and MA girls were most likely to look at other children's work.
CONCLUSIONS AND IMPLICATIONS The strongest set of findings in the present study bear upon the pervasive effects that the expectation of an impending examination had upon the behaviors of children, regardless of their test-anxiety level or sex. When expecting an ex-
58
Wine
amination, as compared to when no examination was expected, children worked harder, sought evaluations of their work from both the teacher and other children, and restricted their general activity level. Observational researchers have often noted difficulties in obtaining consistent observations of behaviors across different situations. As Mischel (1973) has amply documented, the nature of situational variables must be taken into account in predicting the behaviors of individuals. The present study suggests that examination of the effects of situational variables in their impact on individual children's behaviors is a potentially rich area for classroom observational researchers. Evaluative stress is only one of a host of variables that might be manipulated in the naturalistic setting of the classroom. There were also interesting findings bearing upon the effects of test-anxiety level, as well as upon the interaction between test anxiety and examination expectancy on children's classroom behaviors. In general, these results form an interpretable pattern within a cognitive framework, as opposed to an emotional reactivity one. The results suggest that test-anxious children tend, in general, to be "good," hardworking, inactive children. When evaluative stress is introduced, their task-oriented activities decrease. They spend more time sitting idle, apparently preoccupied, and attention to peripheral task cues becomes constricted. Though the amount of test-anxious children's interaction with other children does not differ from that of less anxious children under normal classroom conditions, they tend to become isolates under evaluative stress, initiating and receiving fewer communications. Interactions with other children, as well as with the teacher, may be characterized by passivity and dependency. Behavioral changes in low and middle anxious children that occurred with evaluative stress indicate that while the high test-anxious child became more preoccupied, less effective, the less anxious child became more attentive to the task and to task cues, worked harder, sought feedback from the teacher and other children, and eliminated task-irrelevant behaviors. Consistent with a cognitive interpretation of the present test-anxietyrelated results, recent treatment research in test anxiety (Allen, 1971; Holroyd, 1976; Little & Jackson, 1974; Meichenbaum, 1972; and Wine, 1971) strongly supports the use of cognitively based treatment approaches with test-anxious adolescents and adults. The results of the present study indicate that similar cognitive approaches may be highly appropriate for the treatment of test anxiety in children.
REFERENCES Allen, G. J. Effectiveness of study counseling and desensitization in alleviating test anxiety in college students. Journal of Abnormal Psychology, 1971, 77, 282-289. Barnard, J. W., Zimbardo, P. G., & Sarason, S. B. Anxiety and verbal behavior in children. Child Development, 1961, 32, 379-392.
Anxiety, Evaluation Threat, and Classroom Behavior
59
Barnard, J. W., Zimbardo, P. G., & Sarason, S. B. Teachers' ratings of student personality traits as they relate to LQ. and social desirability. Journal o f Educational Psychology, 1968, 59, 128-132. Blishen, B. R. The construction and use of an occupational class scale. Canadian Journal o f Economics and Political Science, 1958, 24, 519-531. Cox, F. N. Some effects of test anxiety and presence or absence of other persons on boys' performance of a repetitive motor task. Journal o f Experimental Child Psychology, 1966,3, 100-112. Cox, F. N. Some relationships between test anxiety, presence or absence of male persons, and boys' performance of a repetitive motor task. Journal of Experimental ChildPsychology, i968, 6, 1-12. Davidson, K. S., & Sarason, S. B. Test anxiety and classroom observations. Child Development, 1961,32, 199-210. Easterbrook, J. A. The effect of emotion on cue utilization and the organization of behavior. Psychological Review, 1959, 66, 183-201. Hill, L. Relation of test anxiety, defensiveness, and intelligence to sociometric status. Child Development, 1963, 31, 767-776. Holroyd, K. A. Cognition and desensitization in the group treatment of test anxiety. Journal o f Consulting and Clinical Psychology , 1976, 44, 991-t001. Liebert, R. M., & Morris, L. W. Cognitive and emotional components of test anxiety: A distinction and some initial data. Psychological Reports, 1967, 20, 975-978. Lighthall, F., Ruebush, B., Sarason, S. B., & Zweibelson, I. Change in mental ability as a function of test anxiety and type of mental test. Journal of Consulting Psychology, 1959, 23, 34-38. Little, S., & Jackson, B. The treatment of test anxiety through attentional and relaxation training. Psychotherapy: Theory, Research and Practice, 1974, 11, 175-178. Meichenbaum, D. Cognitive modification of test anxious college students.JournalofConsulting and Clinical Psychology, 1972, 39, 370-380. Mischel, W. Toward a cognitive social learning reconceptualization of personality. Psychological Review, 1973, 80, 252-283. Morris, L. W., & Liebert, R. M. Effects of anxiety on timed and untimed intelligence tests. Journal o f Consulting and Clinical Psychology, 1969, 33, 240-244. Phillips, B. N., Martin, R. P., & Meyers, J. Interventions in relation to anxiety in school. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2). New York: Academic Press, 1972. Sarason, I. G. Experimental approaches to test anxiety: Attention and the uses of information. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2). New York: Academic Press, 1972. Sarason, I. G. Test anxiety, attention, and the general problem of anxiety. In C. D. Spielberger & L G. Sarason (Eds.), Stress and anxiety (Vol. 1). Washington, D.C.: Hemisphere Publishing, 1975. Sarason, S. B. The measurement of anxiety in children: Some questions and problems. In C. D. Spielberger (Ed.), Anxiety and behavior. New York: Academic Press, 1966. Sarason, S. B., Davidson, K., Lighthall, F., &Waite, R. Classroom observation of high and low anxious children. Child Development, 1958, 29, 287-295. Sarason, S. B., Davidson, K. S., Lighthall, F., Write, R., & Ruebush, B. K. Anxiety in elementary school children. New York: Wiley, 1960. Wachtel, P. L. Anxiety, attention and coping with threat. Journal ofAbnormalPsychology, 1968, 73, 137-143. Wine, J. Test anxiety and direction of attention. Psychological Bulletin, 1971, 76, 92-104. Winer, B. J. Statistical principles in experimental design (2nd ed.). New York: McGrawHi/l, 1971. Zweibelson, I. Test anxiety and intelligence test performance. Journal o f Consulting Psychology, 1956, 20, 479-481.
Test Anxiety and Evaluation Threat: Children's Behavior in the Classroom 1 Jeri Dawn Wine The Ontario Institute for Studies in Education
Children differing in test anxiety level were observed in two art classes, one immediately preceding a classroom examination, the other when no examination was expected. 'The observational analyses indicated the following: (a) When an examination was expected there were general increases both in children's task orientation and in their concern with other's evaluations, and a decrease in general activity level. It was suggested that future research examine the effects o f additional situational variables on children's classroom behaviors. (b ) Highly test-anxious children were generally hardworking and inactive. They reacted to examination expectancy with a decrease in task orientation, reduction in communications, and attentional constriction. Less anxious children reacted to examination expectancy by working harder, eliminating task-irrelevant behaviors, and seeking feedback. These results were interpreted as supporting a cognitive theory o f test anxiety. The effects of test anxiety on task performance have been well documented. The majority of studies indicate an interaction between evaluative stress and test anxiety in their effects on task performance (see Phillips, Martin, & Meyers, 1972; Sarason, 1972; Wine, 1971, for reviews). The highly test-anxious person performs most poorly in highly evaluative situations, and best under nonevaluatire conditions, while the reverse is true for the low test-anxious persons. Research with children suggests that this effect is obtained as tasks are varied along a dimension from "garnelike" to "testlike" (Lighthall, Ruebush, Sarason, & Zweibelson, 1959; Sarason, Davidson, Lighthall, Waite, & Ruebush, 1960; Zweilbelson, 1956) and with audience absence versus presence (Cox, 1966, 1968). Manuscript received in final form April 27, 1978. 1This research was supported by Canada Council Grant No. S71-0272. 45 0091-0627/79/0300-0045503.00/0 C) 1979 Plenum Publishing Corporation
46
Wine
Though there is a good deal of evidence regarding systematic variations in task performance as a function of test anxiety and stress, there is little evidence regarding variations in other behaviors. The classroom seems an ideal setting within which to research naturalistic variations in stress and their effects on children's behaviors. However, the effects of situational variables are so strong that it is quite difficult to make comparisons of children's behaviors across classroom situations. For example, it is probably not meaningful to compare a child's behavior while taking a classroom test with that child's behavior during a class in oral reading, because the appropriate behaviors in the two situations are different. These difficulties were noted in classroom observational studies of high and low test-anxious children reported by Sarason, Davidson, Lighthall, and Waite (1958) and Davidson and Sarason (1961). Consequently, these researchers settled for global clinical ratings rather than moment-to-moment systematic behavioral observations. An alternative to examining children's behavior across all classroom situations is to manipulate evaluative stress within a particular kind of class that makes reasonably consistent demands on children's behaviors, preferably one that allows for a fairly wide range of behaviors and for interactions among the children and between the children and the teacher. Art class is one such situation; it is a regularly scheduled event in most elementary school classrooms; it often takes place in the normal classroom with the regular teacher; and typically, children are given a measure of freedom to move around and interact. It would be difficult, and perhaps inappropriate, to introduce evaluative stress within this situation, but it is possible to introduce the expectancy of an impending examination. Research has indicated that "worry" regarding a particular evaluating situation tends to remain constant across time and does not fluctuate with contiguity to the evaluating situation (Liebert & Morris, 1967). Barnard, Zimbardo, and Sarason (1961) interviewed high and low test-anxious children, half of them in a permissive atmosphere and half with the instruction that there would be a test following the interview. In the evaluative interview condition the high anxious children used more negative expressions than the low anxious children, while there were no differences in negative expressions in the permissive interview. These results indicate that the expectation of evaluation interacts with anxiety level in affecting children's overt behaviors. In the present study, the classroom behaviors of children varying in test anxiety level were systematically observed in an art class immediately followed by a difficult classroom examination, and during an art class when no tests were scheduled for the entire day. The following questions were addressed: How are differences in test anxiety level manifested in children's overt classroom behaviors? Do these behaviors change systematically with changes in evaluative stress? The earlier classroom observation studies (Davidson & Sarason, 1961; Sarason et al., 1958) reported that test anxiety related more clearly to the behaviors of boys than to those of girls. Sex differences were investigated further in the present study.
Anxiety, Evaluation Threat, and Classroom Behavior
47
In a broader sense, the study examined the issue of the nature of test anxiety. Sarason (1966) noted in a review of test anxiety research with children that high test-anxious children show none of the overt behavioral signs of unambiguous anxiety, such as agitation, flight, extreme physiological reactivity, high distractibility. On the contrary, test anxiety is associated with caution, constriction, and a passive orientation to authority. He suggests that "the cognitive consequences of anxiety affect personality to a greater degree than does the anxiety to which they were a response" (p. 70). Other test anxiety reviewers (Sarason, 1972, 1975; Wine, 1971) have supported cognitive interpretations of the nature of test anxiety. The present study sought to provide additional information on the issue of a cognitive versus an emotional reactivity interpretation of test anxiety.
PROCEDURE
Design The design of the present study is a quasi-experimental one, i.e., data were systematically collected on naturally occurring behaviors in a naturalistic setting, with experimental control exercised over the variable of Examination Expectancy. Sixty-six children in three Grade 4 classrooms were observed during an art class in which the children expected a difficult classroom examination to be immediately following, and during an art class when no tests were scheduled for the entire day. The children were divided into subgroups on the basis of sex, and high, middle, and low test-anxiety scores. The subgroups were equated for IQ and socioeconomic status. The art classes were videotaped and the children's behaviors coded from the tapes on the basis of 22 observational categories. The overall design of the study was a 2 (Sex) • 3 (Test Anxiety) • 2 (Examination Expectancy) analysis of variance design, with repeated measures on the third variable.
Subjects and Setting The subjects were children in three Grade four classrooms in three Waterloo County, Ontario, schools. All of the classes were traditional, one-teacher, oneroom, single-grade, nonstreamed classes, two taught by females, the third by a male. In all three classes, art lessons were given for 1 hour each week by the regular teacher. And in all three the children occupied assigned desks. There were 87 children in total in the three classrooms. The parents of 9 children refused permission for their children to participate in the study; data on 4 children were not used because of absence on 1 of the 2 observation days.
48
Wine
Five subjects were dropped from the final analyses in order to achieve comparability in IQ and socioeconomic status across subgroups. Three additional subjects were dropped randomly in order to achieve equal Ns in the subgroups for the analyses of variance. The final sample consisted of 66 children, 11 in each of the six subgroups formed by the Sex by High, Middle, and Low testanxiety combinations.
Individual Difference Measures
The Test Anxiety Scale for Children (Sarason et al., 1960) was groupadministered to each class approximately 1 month prior to the 1st day of observation. The distributions of TASC scores were split into thirds separately for boys and girls forming high (HA), middle (MA), and low (LA) test-anxious groups. Information was collected on IQ and socioeconomic status since these variables have been shown to interact with test anxiety in effects on behaviors. The children in all three classrooms had been administered the Otis-Lennon Mental Ability Test at the end of Grade 3. Paternal occupations served as the basis for socioeconomic status classification (Blishen, 1958). Table I presents the test anxiety, IQ, and SES data for the six subgroups. The groups are roughly comparable in IQ and SES. Observational Situations
Each classroom was observed on two occasions during their weekly art classes, which involved, in each case, individual projects completed at the child's assigned desk. In the Examination Expected condition, the children were informed of the examination by their teacher several days prior to the observed art class; they were reminded at the beginning of the art class that they would be taking the test in the hour following the class. For two of the classrooms, the No ExTable I. Means and Standard Deviations on IQ, SES, and TASC Boys LA IQ SES TASC
)~ SD X SD X" SD
MA
Girls HA
102.00 98.00 101.00 11.09 12.74 10.88 4.27 4.55 4.55 1.56 1.57 .93 8.46 14.55 19.09 1.37 2.98 2.88
LA
MA
HA
101.91 99.18 98.91 10.08 11.45 14.40 4.46 4.91 4.55 1.21 1.45 1.29 9.82 16.27 21.82 2.49 1.68 1.78
Anxiety, Evaluation Threat, and Classroom Behavior
49
amination art class was scheduled first; for the third class the Examination Expected class was scheduled first.
Observational Method Video recordings were used to obtain behavior records. Each class was observed for a half day prior to the classes of interest in order to accustom the children to the presence of the observer. On the actual observation days, the video equipment was set up several hours prior to the actual observational situation, the equipment was run continuously, and the children were not made aware of the situation which was of interest to the observer. The equipment was set up in the rear of the classroom so that it would not be in the children's line of vision. In order to obtain codable records, each classroom was divided into three sections, each containing approximately one-third of the children, and the recordings were made on a time-sampling basis of 5 minutes per section. The art classes lasted an hour, with the first 5 to 10 minutes occupied by the teacher's instructions. There were 15 minutes in total of videotape per section of the classroom. Thus each child was videotaped for 15 minutes in both the Examination Expected and the No Examination classes. Three observers, trained on a classroom videotape not used in this study, coded the children's behaviors. The observers' task was to observe and code each child's behavior individually; during coding the videotapes were replayed as many times as there were children in the section of the classroom on each videotape, 7-10 children. Checks were made on agreement among observers following their training and on two occasions during the actual coding. Percentages of agreement between pairs of observers on total day records for individual children were uniformly high, ranging from 92% to 98% agreement, with total number of coded behaviors agreed upon divided by the total number of coded behaviors. The observers coded blind, without knowledge of individual children's test anxiety scores, or of Examination Expectancy conditions.
Observational Categories Unfortunately, in a busy classroom it was not possible to record the verbal content of the children's communications. The set of 22 observational categories was designed, through several months of pretesting, to be descriptive of other observable task, activity, and interactional behaviors. The observational unit was the discrete observable behavior. An arbitrary time unit of 30 seconds was also used, i.e., at least one behavior was coded in each 30-second interval, though more behaviors were recorded if more occurred. Behaviors were coded in the following categories:
50
Wine
Attending Behaviors. These were coded when children oriented head and eyes in the direction of speaker(s): (1) Attends to teacher's communication to the class, (2) attends to teacher's communication to another child, (3) attends to other children's conversations. Task-Related Behaviors. These were coded if the child engaged in no other codable behavior for a 30-second interval: (1) Works quietly on art tasks, (2) sits idle. Activity. (1) Takes more art materials, (2) stands, (3) leaves desk, (4) looks at camera. Communication. (1) Initiates communication with teacher, (2) initiates communication with another child, (3) receives communication from teacher, (4) receives communication from another child. Interactional Behaviors. (1) With teacher: (a) raises hand, (b) displays work to the teacher, (c) receives help or material from the teacher. (2) With another child: (a) displays work to child, (b) prosocial behavior (gives help, shares, passes materials, etc.), (c) jokes or laughs, (d) task-distruptive or aggressive behaviors, (e) helped by child. In addition to the observations of the children, several observations were made of the teachers to ensure there were no significant variations in their behavior across the two observation occasions. As the videotapes were intended to yield codable records of the children's behaviors and not of the teachers', the teachers were frequently out of view. While the children's verbal communications were rarely audible, the reverse was true for the teachers; most of their communications were loud and clear. The teachers' audible verbal behaviors were coded in several broad categories: (a) total amount of time spent talking, (b) number of communications, (c) communication directed to the class or to an individual child, and (d) communication content: task instruction, evaluation of work or behavior, or question. These codings of the teachers' verbal behaviors revealed no significant differences between the Examination Expected and No Examination classes.
RESULTS AND DISCUSSION
Data on 5 of the observational categories are not reported here either because too few behaviors were observed, i.e., the means for none of the subgroups exceeded 1.00, or no significant Fs emerged from the analyses of variance. They were: attends to other children's conversations, receives communication from the teacher, task disruptive or aggressive behaviors, prosocial behaviors, and jokes or laughs. Data on the remaining 17 observational indices are reported in terms of the Sex by Anxiety by Examination Expectancy (2 • 3 • 2) repeatedmeasures analyses of variance. The large number of observational categories, coupled with the rather brief length of time each child was observed in each art
Anxiety, Evaluation Threat, and Classroom Behavior
51
class, resulted in small numbers of behaviors coded in some categories, tn view of this, results are reported on F ratios that approached significance, between the .05 and. 10 levels.
Examination Expectancy Conditions Perhaps the most surprising aspect of the results was the pervasive effects that the anticipation of an examination had upon the behaviors of the children in general, disregarding anxiety level. In the Examination Expected (E) class as compared to the No Examination (NE) class, the children spent more time working quietly on their art tasks, F(1, 60) = 4.09, p < .05 (E M = 12.70, NE M = 11.74), and less time sitting idle, F(1, 60) = 6.38, p < .025 (E M = 1.18, NE M = 2.21). They attended more to the teacher's communications to the class, F(1, 60) = 18.93, p < .001 (E M = 4.80, NE M = 2.53), displayed their work to the teacher, F(1, 6 0 ) = 27.38, p < .001 ( E M = .95, N E M = 15)and to other children more often, F(1,60) = 15.51, p < .001 (E M = 1.29, NE M = .47). They took materials less often, F(1, 60) = 25.14, p < .001 (E M = .53, NEM = 3.36), stood at their desks less,/7(t, 60) = 10.76, p < .005 (E M = 3.03, NE M = 4.97), and looked at the camera less often, F(1, 60) = 9.65, p < .005 (E M = 2.35, NE M = 3.96). They also received fewer communications from other children, F(1, 60) = 4.16, p < .05 (E M = 3.80, NE M = 5.02). The overall picture is one of change from the No Examination class to the Examination Expected class toward more intense task orientation, greater concern with others' evaluations, and behavioral constriction.
Test Anxiety and Examination Expectancy The pattern of the results for test-anxiety level and its interaction with Examination Expectancy is, overall, consistent and meaningful within a f}amework of a cognitive interpretation of test anxiety. Attending Behaviors. As noted above, all of the children listened to the teacher's communications to the class more in the Examination Expected than the No Examination class. However, an Anxiety by Examination Expectancy interaction, which approached significance, F(2, 60) = 2.37, .10 > p > .05, suggested that LA and MA children listened to the teacher's communications with other children more in the Examination Expected than the No Examination class while HA children listened to this kind of communication slightly tess in the Examination Expected class (see Figure 1). Though these results are of borderline significance, they are provocative. With the threat of evaluation all of the children became more attentive to the evaluating authority figure in the situation. However, under evaluative stress, the HA child tended to listen only to what she/he was "supposed" to listen to, i.e., the teacher's communications
Wine
52 3.0-
2.5-
2.0-
o
4-"
~3 1,5
/ /
/
/
/
/
a MA
/
//
E Z
1.0~ //
""-.b~
LA
.5-
Conditions
Fig. 1. Attending to teacher's communication to another child: anxiety X conditions interaction.
to the entire class but not to other children, while the MA and LA child became more alert to both kinds of teacher communications. The result is suggestive of the proposition that attentional constriction accompanies increases in anxiety (Easterbrook, 1959; Wachtel, 1968; Wine, 1971), if one interprets teacher's communications to other children as a kind of peripheral task cue. Task-Related Behaviors. Sarason et al. (1958) and Davidson and Sarason (1961) reported results regarding test anxiety and sex effects on task orientation. The classroom behaviors of LA boys were rated as more task-oriented than HA boys, while HA girls were more task-oriented than LA girls. In the present study, Sex and Anxiety level did not interact in their effects on task orientation.
53
Anxiety, Evaluation Threat, and Classroom Behavior 15-
14-
'""--i
13-
HA
,,e LA
r O
sr
/
t, o
/
sS
.B
MA
.1
12-
e~
/
E
gS s/S
z
sS
11-
sS g gS
Ir
10-
Conditions
Fig. 2. Working quietly: anxiety • conditions interaction.
There was a significant main effect for test anxiety on the amount of time spent working on the art tasks, F(2, 60) = 4.05, p < .05, with the HA children spending more time overall working quietly (M = 13.18) than the MA (M = 11.80) or LA children (M -- 11.68). However, a near-significant interaction between Anxiety and Examination Expectancy, F(2, 60) = 2.79,. 10 > p > .05, indicated that LA and MA children spent more time working quietly in the Examination Expected than in the No Examination class, while HA children spent slightly less time working in the Examination Expected class. This Anxiety by Expectancy effect is mirrored by the amount oft/me spent sitting idle, F(2, 60) = 2.96, .I0 > p > .05. The LA and MA children spent less time sitting idle in the Examination Expected class
54
Wine 5-
4-
O
I~%x% %x
3-
"~xx
.B
o ~6 E z
x
2-
~
HA
~-
1-
~LA ~r~MA
0
9
N'E
6 Conditions
Fig. 3. Sitting idle: anxiety X conditions interaction.
than in the No Examination class, while HA children were idle slightly more in the Examination Expected class. These effects are presented in Figures 2 and 3. The pattern, though again only approaching significance, is quite consistent with laboratory findings that with increases in evaluation emphasis, the low testanxious person becomes more task-oriented and works harder, while the high test-anxious person becomes self-preoccupied and less task-oriented (e.g., Sarason, 1975; Wine, 1971). More information regarding task-orientation differences between anxiety levels was provided by the finding of a main effect, F(2, 60) = 3.58, p < .05, for anxiety level on looking at other children's work - in the rcontext of an art class, an entirely appropriate behavior and a source of task-relevant cues. The HA children looked at the work of others less (.41/= 7.80) than did the MA (M = 9.21) and LA children ( 3 / = 11.07). These results suggest that the high testanxious child may be less alert than the less anxious child to peripheral task cues.
Anxiety, Evaluation Threat, and Classroom Behavior
55
Child-Teacher Interactions. Highly test-anxious children, especially boys, have been described as dependent upon the approval of authority figures, direction seeking, and occasionally unresponsive (Barnard et al., 1961 ; Sarason, 1966). The single result bearing upon this issue was a significant interaction between Sex and Anxiety level on the number of times the teacher helped a child, F(2, 60) = 6.84, p < .025, with the HA boys and the MA girls receiving the most help of the Sex X Anxiety subgroups. Child-Child Interactions. The literature regarding interactions among children as a function of test anxiety is sparse. The results of two studies of children's sociometric choices, reported by Hill (1963) and Phillips, Adams, Gotts, and McNeil (cited in Phillips et al., 1972), suggest that the HA child may be something of a social isolate in the classroom. Several analyses in the present study shed some light on how this isolation is manifested in the HA child's classroom interactions. The HA children did not initiate or receive fewer communications overall than did the other anxiety groups. However, there was a significant interaction between Anxiety and Examination Expectancy on the number of communications initiated, F(2, 60) = 6.62, p < .01. The LA and MA children initiated slightly more communications with other children in the Examination Expected than in the No Examination class, while the HA children initiated markedly fewer communications with other children in the Examination Expected than in the No Examination class. The HA children were also the only test-anxiety subgroup to receive significantly fewer communications in the Examination than in the No Examination situation, individual means comparison, F(1,20) = 5.91 ,p < .05. These effects are presented in Figures 4 and 5. The HA child was not an isolate under normal, nonevaluative classroom conditions, but under stress she/he did become isolated. Another result has implications for the nature of the relationship between the HA child and other children. A significant main effect for anxiety level, F(2, 60) = 8.27, p < .01, revealed that the HA child was more likely to be helped by other children (3/= 1.41) than were MA (M = .64) and LA children (3/ = ~ The test-anxiety literature has indicated that HA children are dependent in their relationships with authority figures; this result suggests they are dependent in their relationships with other children as well. Activity Indices. Analyses of the activity indices suggested that the HA children were less active, in general, than the LA children, standing at their desks less often, individual means comparison, F(1,42) = 4.10, p < .05 (HAM = 3.27, LA M = 4.59), and leaving their desks fewer times, individual means comparison, F(1, 42) = 6.90, p < .025 (HA M = 1.46, LA M = 2.4 t) than LA children. Two test anxiety by Examination Expectancy interactions and appropriate individual means comparisons suggested that both MA and LA children engaged in fewer task-irrelevant behaviors in the Examination Expected class than in the No Examination class. LA children looked at the camera markedly fewer times in the Examination Expected (M = 2.14) than in the No Examination class (3'/=
56
Wine 7-
,,m LA 6-
s
~ 5-
s
ssSSSSssS ssSIsS jjSS //BMA,///,//,
O
> o
4rq/
..O
E Z 3HA
2-
N'E
I~ Conditions
Fig. 4. Communications initiated to other children: anxiety • conditions interaction.
5.36), F(1, 20) = 13.01, p < . 0 0 5 , while MA children stood at their desks significantly less often in the Examination Expected (M = 2.09) as compared to the No Examination class (M = 6.19), F(1,20) = 16.17, p < .005. Sex Differences. There were surprisingly few sex differences, or interactions between Sex and the Test Anxiety and Examination Expectancy variables. Boys were significantly more likely (M = 2.56) than girls (M--- 1.18) to spend time sitting idle, F(1, 60) = 8.17, p < .01, and to look at other children's work (Boys M = 10.61, Girls M = 8.10), F(1, 60) = 5.89, p < .05. A significant interaction between Sex and Examination Expectancy, F(1, 60) = 4.30, p < .05, indicated that though both boys and girls displayed their work to the teacher more
Anxiety, Evaluation Threat, and Classroom Behavior
57
7.
6.
.......
o
-o LA
o
:IVIA
E Z
3-
2-
N'E Conditions
Fig. 5. Communicationsreceived from other children: anxiety X conditions interaction.
often in the Examination Expected than in the No Examination class, this effect was much more marked for boys. A significant interaction between Sex and Anxiety, F(2, 60) = 5.42, p < .01, revealed that among the Sex • Anxiety subgroups LA boys and MA girls were most likely to look at other children's work.
CONCLUSIONS AND IMPLICATIONS The strongest set of findings in the present study bear upon the pervasive effects that the expectation of an impending examination had upon the behaviors of children, regardless of their test-anxiety level or sex. When expecting an ex-
58
Wine
amination, as compared to when no examination was expected, children worked harder, sought evaluations of their work from both the teacher and other children, and restricted their general activity level. Observational researchers have often noted difficulties in obtaining consistent observations of behaviors across different situations. As Mischel (1973) has amply documented, the nature of situational variables must be taken into account in predicting the behaviors of individuals. The present study suggests that examination of the effects of situational variables in their impact on individual children's behaviors is a potentially rich area for classroom observational researchers. Evaluative stress is only one of a host of variables that might be manipulated in the naturalistic setting of the classroom. There were also interesting findings bearing upon the effects of test-anxiety level, as well as upon the interaction between test anxiety and examination expectancy on children's classroom behaviors. In general, these results form an interpretable pattern within a cognitive framework, as opposed to an emotional reactivity one. The results suggest that test-anxious children tend, in general, to be "good," hardworking, inactive children. When evaluative stress is introduced, their task-oriented activities decrease. They spend more time sitting idle, apparently preoccupied, and attention to peripheral task cues becomes constricted. Though the amount of test-anxious children's interaction with other children does not differ from that of less anxious children under normal classroom conditions, they tend to become isolates under evaluative stress, initiating and receiving fewer communications. Interactions with other children, as well as with the teacher, may be characterized by passivity and dependency. Behavioral changes in low and middle anxious children that occurred with evaluative stress indicate that while the high test-anxious child became more preoccupied, less effective, the less anxious child became more attentive to the task and to task cues, worked harder, sought feedback from the teacher and other children, and eliminated task-irrelevant behaviors. Consistent with a cognitive interpretation of the present test-anxietyrelated results, recent treatment research in test anxiety (Allen, 1971; Holroyd, 1976; Little & Jackson, 1974; Meichenbaum, 1972; and Wine, 1971) strongly supports the use of cognitively based treatment approaches with test-anxious adolescents and adults. The results of the present study indicate that similar cognitive approaches may be highly appropriate for the treatment of test anxiety in children.
REFERENCES Allen, G. J. Effectiveness of study counseling and desensitization in alleviating test anxiety in college students. Journal of Abnormal Psychology, 1971, 77, 282-289. Barnard, J. W., Zimbardo, P. G., & Sarason, S. B. Anxiety and verbal behavior in children. Child Development, 1961, 32, 379-392.
Anxiety, Evaluation Threat, and Classroom Behavior
59
Barnard, J. W., Zimbardo, P. G., & Sarason, S. B. Teachers' ratings of student personality traits as they relate to LQ. and social desirability. Journal o f Educational Psychology, 1968, 59, 128-132. Blishen, B. R. The construction and use of an occupational class scale. Canadian Journal o f Economics and Political Science, 1958, 24, 519-531. Cox, F. N. Some effects of test anxiety and presence or absence of other persons on boys' performance of a repetitive motor task. Journal o f Experimental Child Psychology, 1966,3, 100-112. Cox, F. N. Some relationships between test anxiety, presence or absence of male persons, and boys' performance of a repetitive motor task. Journal of Experimental ChildPsychology, i968, 6, 1-12. Davidson, K. S., & Sarason, S. B. Test anxiety and classroom observations. Child Development, 1961,32, 199-210. Easterbrook, J. A. The effect of emotion on cue utilization and the organization of behavior. Psychological Review, 1959, 66, 183-201. Hill, L. Relation of test anxiety, defensiveness, and intelligence to sociometric status. Child Development, 1963, 31, 767-776. Holroyd, K. A. Cognition and desensitization in the group treatment of test anxiety. Journal o f Consulting and Clinical Psychology , 1976, 44, 991-t001. Liebert, R. M., & Morris, L. W. Cognitive and emotional components of test anxiety: A distinction and some initial data. Psychological Reports, 1967, 20, 975-978. Lighthall, F., Ruebush, B., Sarason, S. B., & Zweibelson, I. Change in mental ability as a function of test anxiety and type of mental test. Journal of Consulting Psychology, 1959, 23, 34-38. Little, S., & Jackson, B. The treatment of test anxiety through attentional and relaxation training. Psychotherapy: Theory, Research and Practice, 1974, 11, 175-178. Meichenbaum, D. Cognitive modification of test anxious college students.JournalofConsulting and Clinical Psychology, 1972, 39, 370-380. Mischel, W. Toward a cognitive social learning reconceptualization of personality. Psychological Review, 1973, 80, 252-283. Morris, L. W., & Liebert, R. M. Effects of anxiety on timed and untimed intelligence tests. Journal o f Consulting and Clinical Psychology, 1969, 33, 240-244. Phillips, B. N., Martin, R. P., & Meyers, J. Interventions in relation to anxiety in school. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2). New York: Academic Press, 1972. Sarason, I. G. Experimental approaches to test anxiety: Attention and the uses of information. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2). New York: Academic Press, 1972. Sarason, I. G. Test anxiety, attention, and the general problem of anxiety. In C. D. Spielberger & L G. Sarason (Eds.), Stress and anxiety (Vol. 1). Washington, D.C.: Hemisphere Publishing, 1975. Sarason, S. B. The measurement of anxiety in children: Some questions and problems. In C. D. Spielberger (Ed.), Anxiety and behavior. New York: Academic Press, 1966. Sarason, S. B., Davidson, K., Lighthall, F., &Waite, R. Classroom observation of high and low anxious children. Child Development, 1958, 29, 287-295. Sarason, S. B., Davidson, K. S., Lighthall, F., Write, R., & Ruebush, B. K. Anxiety in elementary school children. New York: Wiley, 1960. Wachtel, P. L. Anxiety, attention and coping with threat. Journal ofAbnormalPsychology, 1968, 73, 137-143. Wine, J. Test anxiety and direction of attention. Psychological Bulletin, 1971, 76, 92-104. Winer, B. J. Statistical principles in experimental design (2nd ed.). New York: McGrawHi/l, 1971. Zweibelson, I. Test anxiety and intelligence test performance. Journal o f Consulting Psychology, 1956, 20, 479-481.
