1 The Role of Habits in Test Anxiety
Koen van Dalfsen
University of Amsterdam Student number: 10535616 Teacher: Poppy Watson Word count: 3694
2 Abstract
The balance between habitual behavior versus goal-directed behavior in relationship with Test Anxiety was investigated with a group of 69 Dutch participants with an age of 18 to 40. A slips-of-action test was used to asses habitual versus goal-directed behavior. To measure test anxiety,
participants filled in the test anxiety inventory (Ploeg, 1988; Szafranski, Barrera, & Norton, 2012). Furthermore, training length and response time were manipulated. No relationship between test anxiety and the balance between habitual versus goal-directed behavior was found. Furthermore there was no effect of response time on the slips-of-action test. However, there was an effect of training length. Contrary to expectations, the more training participants had the less slips-of-actions they made. Possibly because more training led to more familiarization of the test.
These findings indicate that there is no relationship between test anxiety and a bias towards habitual decision-making.
3 The Role of Habits in Test Anxiety
The nervous feeling before a test is a familiar feeling to most of us. Take for example a driving exam. Before the exam one might feel more sweaty than usual, maybe even a bit shaky and one worries that one might fail. People react differently to such circumstances. One person might experience a slight excitement another might blank out, forget everything. Excessive nervousness or anxiety can stand in the way of performing well. In the case of a driving exam, it can lead to failing the test, resulting in loss of money and time. Anxiety for test situations is a common story in high schools and even universities. According to research by Putwain and Daly (2014) 16.4% of students experience this sort of anxiety, which is called Test Anxiety (TA). TA is excessive worry and fear of performing in test situations and can be measured with the Test Anxiety Inventory (Ploeg, 1988; Szafranski, Barrera, & Norton, 2012). Typical
symptoms of TA include sweating, increased heart rate, excessive worry and difficulty organizing thoughts (Salend, 2011; de Lara-Kroon, & van Efferen-Wiersma, 2009). These symptoms can be disabling and lead to a vicious cycle where poor school performance in turn leads to more anxiety concerning tests (Peleg, 2009) and thus worse performance. According to Salend (2011), efficient studying, effective test-taking skills and use of anxiety reduction strategies are useful tools in dealing with TA. Nonetheless, many students still experience TA. Since stress is known to increase habitual behavior (Schwabe & Wolf, 2009; Dias-Ferreira et al., 2009) a possible explanation of the persistence of TA and poor performance on test taking is that test-anxious people might fall back on habitual behavior during test taking instead of being more goal-orientated and using the efficient test-taking skills that they have been rehearsing. For example, a student, when receiving a test, may fall into old habits and start answering without properly reading the question, thus making more mistakes. In the current study we investigated
4
the relationship between TA and the balance of habitual versus goal-directed decision-making (as measured with a slips-of-action task). Investigating this relationship is important in
understanding TA, because if test anxious people make decisions in a different manner this can be taken into account when designing efficient strategies to overcome TA.
Associative theories of decision-making differentiate between goal-directed and habitual actions (Wit, & Dickinson, 2009). Goal-directed actions are those where one knows the preferred outcome and chooses the best response leading to that outcome because it is currently desired (Adams, 1981; Adams, 1982). In contrast, habitual actions are based on stimulus-response associations and carried out even when the outcome is devalued (Adams, 1981; Adams, 1982). For example, if I have to drive to work and know that there is a roadblock on the way, I make the goal-directed intention to take another route to work. However, on “automatic pilot”, it is
possible to step in the car and drive the usual route to work, thus encountering the roadblock. This is considered a habitual slip-of-action where the action is triggered by the stimuli instead of the preferred outcome. In other words, the habit of stepping in the car and driving the usual road is stronger than the thought of the roadblock as an outcome, resulting in driving the usual road and being late for work.
To measure the balance between goal-directed and habitual actions in the current study, participants performed a slips-of-action test (De Wit, Niry, Wariyar, Aitken, & Dickinson, 2007; Worbe, Savulich, De Wit, Fernandez-Egea & Robbins, 2015). The slips-of-action test is a computerized test in which participants first learn to press keys in order to win certain outcomes (and financial rewards) in the presence of certain stimuli. In the test phase some outcomes are now devalued which means that to avoid a financial penalty, participants should no longer press the key for that outcome, when the stimulus is shown. When a participant presses a key although
5
the outcome is now devalued, it is called a slip-of-action and indicates habitual behavior (driven by the stimulus rather than the outcome) instead of goal-directed behavior.
Previous research, using a variety of different tasks, has highlighted that anxiety and stress can affect decision-making. Stracke and Brand (2012) conclude in their selective review on decision making under stress, that stress can change decision-making by increasing the reliance on automatic response tendencies instead of rational strategies, especially in the case of an uncertain decision. In line with the previous findings Alvares, Balleine and Guastella (2014) found that people with social anxiety were less sensitive to devaluation (thus showing less goal-directed behavior) on an instrumental learning task compared to healthy individuals. This was in turn linked to a lower response rate to CBT. It has been suggested that anxiety leads to a loss of goal-directed control due to the effects on inhibition and attentional processes (Eysenck,
Derakshan, Santos & Calvo, 2007). While the slips of action task has not been specifically investigated with TA, it has been investigated with Obsessive Compulsive Disorder (OCD). Gillan et al. (2011) found that patients with OCD made more action during a slips-of-action test compared to healthy individuals, thus indicating an overreliance on habits triggered by stimuli.
To investigate the relationship between TA and slips of action, stress and amount of training were manipulated in the current study. Stress was manipulated to investigate the
previous findings that stress can increase habitual behavior. Stress was manipulated by reducing the time participants had available to respond on each trial during the test phase. Furthermore, the effect of training was investigated to see if amount of training has influence on habitual behavior. To investigate this, the number of blocks of training was varied in the training phase. Based on previous research it was expected that more stress would lead to an increase in habitual
6
behavior and more training leads to stronger habits, which in turn leads to more habit behavior (Adams, 1981; Adams, 1982).
In summary the current study used a slips-of-action task to measure habitual versus goal-directed decision making and relate this to test anxiety. Various parameters of the slips of action task were also manipulated. Respectively the time window participants had to respond and the training length. These parameters were manipulated to further investigate the relationship
between TA and the balance between habitual- versus goal-directed behavior. Based on previous research it was hypothesized that those individuals reporting higher TA would show more habitual behavior because they have more difficulty inhibiting functions, leading to more slips-of-action especially when the pressure of a test is increased.
Method
Participants
Sixty-nine people with an age between 18 and 40 participated in this study for research credit or 12.50 euros which could add up to 17.50 euros, depending on their final score during the task. Participants were randomly assigned to different conditions leading to a 2 x 2 design. There were two conditions. Firstly the training condition, consisting of long training condition and short training condition. Secondly the stress condition consisting of fast response condition (high stress) and slow response condition (low stress). These conditions will be further explained in the materials section. Participant demographics can be seen in Table 1.
7
Test Anxiety Inventory: To measure Test Anxiety participants were asked to fill in the
Test Anxiety Inventory (TAI). The TAI contains 20 items with a 4 point Likert scale that can be scored with “almost never”, “sometimes”, “often” and “almost always”, with a minimum of 20 points and a maximum of 80. An example of a question: “I start feeling very uneasy just before getting a test paper back.” (Ploeg, 1988; Szafranski, Barrera, & Norton, 2012). This test is reported to have a Cronbach alpha of .93 and validity of r = .49 (Taylor & Deane, 2002). Slips-of-Action Task: To measure decision-making participants played the computerized
“Fabulous Fruit Game” (De Wit, et al., 2007; Worbe et al., 2015). This slips-of-action test consists of three phases and made use of twelve different pictures of fruits (300 x 300 pixels). During the instrumental training phase a box was shown with a particular fruit on the outside, the stimulus, when either the left or right key was pressed, the response, the box opened. When the correct key was pressed, points were gained and a piece of fruit was shown inside the box, the outcome. The more points were gathered during the game, the higher the money reward was in the end. When the response was incorrect no points were gained and an empty box was shown. By means of trial and error the correct response to a certain fruit on the outside of a box was learned together with the relationship with the paired fruit inside the box, for example banana on the outside of the box led to strawberry inside the box when the left key was pressed (see figure 1A1) and lemon led to pear when the right key was pressed (see figure 1A2). There were six of these stimulus-outcome pairs. The instrumental training phase consisted of either ten blocks (short training condition) or 16 blocks (long training condition), each block consisted of the six pairs showed twice in a random order, meaning either 120 trials (short training condition) or 192 trials (long training condition) in total. The stimuli were shown for 2 seconds while participants made their response. During training the amount of points gained was shown. The
8
faster the response the higher the score and the higher the money reward was. When a response was faster than 500ms 4 points were given, between 500ms and 1000ms 3 points were given, between 1000ms and 1500ms 2 points were given and 1 point was given when responded in between 1500ms or 2000ms. The Inter Trial Interval (ITI) was 2500ms to 4500ms in a random order.
During the slips-of-action test phase certain outcome fruits were devalued, meaning that response was no longer needed when the stimulus previously associated with the devalued outcome was shown. Points were deducted for a response to a stimulus paired with a devalued outcome. During this phase participants did not receive any feedback and were not shown the points they gained or the outcome fruit inside the box. At the start of each block the six outcome pictures were shown and two of these were devalued as indicated by a red cross through them (See figure 1B). This devaluation screen was shown for 5 seconds (see figure 1B). Across nine blocks, every right-response paired outcome was devalued with every left-response paired outcome. After the devaluation screen had been shown participants received 18 trials in which the six stimuli were shown three times in a random order (162 trials in total). The response window was varied. The participants either had 1500 ms (long condition) or 1000 ms (short condition) to respond. ITI was the same as in the training-phase.
The final phase of the task was the baseline test, designed to control for differences in working memory and general inhibitory control. This is the same as the slips-of-action test phase, the only difference being the devaluation of stimuli instead of outcomes (see figure 1C). Due to an error in the task program the data of the baseline test was not used in the analysis.
9
Figure 1. During the instrumental training phase (A) participants were shown a box with a particular fruit on the outside and had to respond by either pressing a left or a right key. In this example when shown banana on the outside of the box a left key press led to strawberry inside
10
the box and points (A1), when shown lemon a right key press led to a pear inside the box and points (A2). When participants did not press the correct key, no points were earned and they were shown an empty box. During the slips-of-action test phase (B) certain outcome fruits were devalued (indicated by the red cross), meaning that participants no longer had to respond when the stimulus fruit connected with the devalued outcome fruit was shown (No Go trial). For the fruits that were not devalued participants still had to respond (Go trial). A wrong response led to point deduction. At the start of each block the participants had 5 seconds to memorize the two stimulus-outcome pairs that were devalued. The final phase of the task was the baseline test (C), this was the same as the slips-of-action test phase, the only difference being the devaluation of stimuli instead of outcomes.
Procedure
First participants were asked to turn off their phone, read the information and sign the informed consent. Next the participants were seated in front of the computer. The experimenter started a demo of the game. This demo was a short version of the game and still included the three phases. Instead of pictures of fruits, pictures of four drinks were used as stimuli and another four drink pictures functioned as outcomes. The experimenter remained in the room during the demo phase to answer possible questions. After the demo phase finished the experimenter asked again if everything was clear, if so the experimenter started the real task and left the room. To increase motivation of the participants they were told that the more points they earned the bigger their money reward would be. After the task was finished participants filled in the TAI and a questionnaire asking the connection of the stimuli, response and outcome to test their S-R, R-O
11
and S-O knowledge. Participants also filled in the OCI and performed a test of working memory (data not presented here). When done the participants received either money or research credit.
Data analysis
To analyze the data from both the test phase and the stimulus phase of the slips-of-action test the Devaluation Sensitivity Index (DSI) as described by Snorrason, Lee, de Wit and Woods (2016) was used. The percentage of responses to devalued stimuli was subtracted from the percentage of responses to valued stimuli. A higher score indicates more sensitivity to
devaluation, indicating goal-directed behavior. To further analyze the data an ANCOVA was conducted. For both the test phase and the baseline phase DSI was the dependent variable and TA the independent together with manipulated stress and amount of training.
Results
Of the 69 people participating in this study 14 participants were excluded. Three participants were excluded for reporting use of the medicine Methylphenidate, known to influence attention (Sykes, Douglas, & Morgenstern, 1972). Furthermore, four participants indicated that they had a DSM axis one disorder in the last two years, respectively depression, PTSD, ADD and social anxiety disorder. The data of these four participants was not used in the analysis to make sure the study concerned a healthy population only. Furthermore, seven
participants had a DSI score of 0 or less, indicating that they did not understand the test and were also excluded.
12
Of the 55 remaining participants, 12 were male and 43 were female, and the age of participants varied between 18 and 40 with a mean age of 23.49 (SD = 4,13). Participants were randomly assigned to the four experimental conditions. Firstly the long-fast group, where participants received the long training condition with the short response window at the test phase; secondly long-slow group where participants received the long training condition with the long response window at the test phase; thirdly short-fast group where participants received the short training condition with the fast response window at the test phase and lastly short-slow group where participants received the short training condition with the slow response window at the test phase. See table 1 for mean age and number of participants across the four experimental conditions.
Table 1. Mean age with Sd in brackets and number of participants in each group.
Group Mean age(Sd) N
Long-Fast Long-Slow Short-Fast Short-Slow 21 (1.07) 26.63 (5.01) 21 (2.12) 22.58 (2.55) 8 19 9 19
To check whether the groups were matched on age a one-way ANOVA was conducted. A main effect of group (F(3,51) = 8,56, p < .001) was found. Using independent samples t-tests it appeared that the difference between the groups was driven by the long-slow condition where age was significantly higher than the short-fast group (t(25.9) = 4,17, p < .001) but no significant difference between the short-fast group and the short-slow group was found (p = 0,12). Given
13
that the groups differed in age we also added age as a covariate to the following ANCOVA analysis.
As shown in table 2 and figure 2, participants in all groups were highly accurate on the final training block, indicating that they understood the task. An ANCOVA, with age as covariate, showed no significant main effect of age (F(1, 50) = .72, p = .40), no significant difference between participants in either the long or short training condition (F(1, 50)= 1.7 , p = .20) and no sig difference between participants in the high or low stress condition on accuracy in the final training block (F(1, 50)= .47 , p = .50).
Table 2. Accuracy final training block for each group and standard deviation.
Group Accuracy SD Long-Fast Long-Slow Short-Fast Short-Slow 1.0 0.95 0.95 0.94 0.0 8.01 11.11 10.70
Figure 2. Proportion of correct responses of the different groups over the blocks.
0,50 0,55 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 P ropor ti on Blocks LongFast LongSlow ShortFast ShortSlow
14
To see if there was an effect of training length, response window and TAI score an
ANCOVA was conducted with independent factors of training condition and stress condition and covariates of age and test anxiety score. There was a significant effect of age F(1, 49) = 9.23, p = .004 where participants who were older did not perform as well on the task. Importantly there was no effect of TAI score, F(1, 49) = .36, p = .55. After controlling for these two variables there was a significant effect of training length, F(1, 49) = 8.72, p = .01, as seen in figure 3 long training led to a higher DSI score indicating that participants who had more training were better able to inhibit their responses to devalued outcomes. However there was no effect of response window, F(1, 49) = 2.41, p = .13 and no interaction between training length and response window F(1,49) = .01, p = .94. See figure 3 for a visualization of the DSI scores between the different groups after controlling for age.
Figure 3. DSI score of the different conditions.
*with a significant difference between training-condition 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
LongFast LongSlow ShortFast ShortSlow
DS
I
Condition *
15
Discussion
This study investigated the relationship between TA and the balance of habitual versus goal-directed decision-making. However, no significant relationship between TA and the balance of habitual versus goal-directed decision-making was found. Furthermore, the stress
manipulation (time participants had to respond) did not lead to an increase in slips of action. However, we did observe a relationship between the amount of training and slips of action with our results suggesting that more training led to more goal-directed behavior and less habitual behavior. These results will be discussed in more detail below.
Surprisingly, no relationship between test anxiety and decision-making was found in the current study. There are a number of possible reasons for this. For example the sample of participants was extremely homogenous. The majority of participants were Dutch, female and student at the University of Amsterdam. Furthermore, another possible explanation that there was no relationship, is that participants did not perceive the task as a real test. Participants did not feel the same pressure they normally feel in test situations. A possible way to increase this pressure could be to tell participants that the test measures IQ score and that they will be compared to other participants. Furthermore, implementing the tasks in a battery of tests might increase pressure. For example, before participants do the slips-of-action task they make a math test. Future studies could take these suggestions in consideration.
The effect of longer training might be against expectations, however it is not a total surprise. It was assumed that the more training one has, the stronger the connection between the stimulus and response. This connection would in turn lead to more difficulty inhibiting the response when needed (e.g. making more slips of action). On the other hand, good performance
16
on the task relies on learning the relationships between the stimuli, the responses and the
outcomes during training. So in this case the short training group did not appear to have enough training to be accurate at the task. Furthermore, when training takes longer, for example hours instead of minutes, participants might be more likely to show habitual behavior. In addition when one is familiar with a task one might get more confident which can lead to better results. The exact parameters required for participants to learn the task associations well and begin to make automatic slips of action needs to be further investigated.
No effect of manipulated stress was found. It was thought that by decreasing the response window stress would increase, leading to more slips of action. Since no effect was found, there is a possibility that participants did not perceive the decreased response window as extra stressful, thus explaining that there was no effect. Maybe the difference in response time of half a second between the two groups was not enough, this needs further investigation. Future research could consider using different ways to manipulate stress. For example stress can be increased by putting a camera in the room.
Lastly inhibitory control was not checked for. An error in the task program made it impossible to use the information of the baseline test. When participants cannot inhibit their responses they might press a key although they do not intend to. This can cause more slips of action during the task, thus making inhibitory control a possible cofounding variable. Meaning that the differences found between people might partly be explained by a difference in inhibitory control.
For now the conclusion can be that there is no relationship between test anxiety and the balance between goal-directed- versus habitual behavior. Thus there are no implications arising from this study for current treatment. So when we have a (driving) exam, we do not have to
17
worry about relying too much on habitual behavior during the test, for now we can just worry about the test itself.
References
Adams, C. D. (1982). Variations in the sensitivity of instrumental responding to reinforcer devaluation. The Quarterly Journal of Experimental Psychology, 34(2), 77-98.
Adams, C. D., & Dickinson, A. (1981). Instrumental responding following reinforcer devaluation. The Quarterly journal of experimental psychology, 33(2), 109-121.
Alvares, G. A., Balleine, B. W., & Guastella, A. J. (2014). Impairments in goal-directed actions predict treatment response to cognitive-behavioral therapy in social anxiety disorder. PloS one, 9(4), 1-7. de Lara-Kroon, N. C., & van Efferen-Wiersma, E. S. (2009). Faalangst. Psychosociale problemen, 53-69. De Wit, S., Niry, D., Wariyar, R., Aitken, M. R. F., & Dickinson, A. (2007). Stimulus-outcome
interactions during instrumental discrimination learning by rats and humans. Journal of
Experimental Psychology: Animal Behavior Processes, 33(1), 1-11.
Dias-Ferreira, E., Sousa, J. C., Melo, I., Morgado, P., Mesquita, A. R., Cerqueira, J. J., ... & Sousa, N. (2009). Chronic stress causes frontostriatal reorganization and affects
decision-making. Science, 325(5940), 621-625.
Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive performance: attentional control theory. Emotion, 7(2), 336-353.
Gillan, C. M., Papmeyer, M., Morein-Zamir, S., Sahakian, B. J., Fineberg, N. A., Robbins, T. W., & de Wit, S. (2011). Disruption in the balance between goal-directed behavior and habit learning in
18
obsessive-compulsive disorder. American Journal of Psychiatry, 168(7), 718-726. Hartley, C. A., & Phelps, E. A. (2012). Anxiety and decision-making. Biological
psychiatry, 72(2), 113-118.
Peleg, O. (2009). Test anxiety, academic achievement, and self-esteem among Arab adolescents with and without learning disabilities. Learning Disability
Quarterly, 32(1), 11-20.
Ploeg, H.M. van der. (1988). Handleiding bij de Examen/Toets Attitude Vragenlijst ETAV. Een vragenlijst voor het meten van examenangst. Een Nederlandstalige bewerking van de Spielberger Test Anxiety Inventory. Lisse, Nederland: Swets & Zeitlinger.
Putwain, D., & Daly, A. L. (2014). Test anxiety prevalence and gender differences in a sample of English secondary school students. Educational Studies, 40(5), 554-570.
Salend, S. J. (2011). Addressing test anxiety. Teaching exceptional children,44(2), 58-68. Schwabe, L., & Wolf, O. T. (2009). Stress prompts habit behavior in humans. The Journal of
Neuroscience, 29(22), 7191-7198.
Snorrason, I., Lee, H. J., de Wit, S., & Woods, D. W. (2016). Are nonclinical
obsessive-compulsive symptoms associated with bias toward habits?. Psychiatry research, 241, 221-223.
Starcke, K., & Brand, M. (2012). Decision making under stress: a selective review. Neuroscience
& Biobehavioral Reviews, 36(4), 1228-1248.
Sykes, D. H., Douglas, V. I., & Morgenstern, G. (1972). The effect of methylphenidate (Ritalin) on sustained attention in hyperactive children. Psychopharmacologia, 25(3), 262-274.
doi:10.1007/bf00422506
19 Stress & Coping, 25(6), 667-677.
Taylor, J., & Deane, F. P. (2002). Development of a Short Form of the Test Anxiety Inventory (TAI). The Journal of General Psychology, 129(2), 127-136. doi:10.1080/00221300209603133
Wit, S., & Dickinson, A. (2009). Associative Theories of Goal-Directed Behaviour: a Case for Animal–
hhhhhhHuman Translational Models. Psychological Research Psychologische Forschung, 73(4),
463-hhhhhh476. doi:10.1007/s00426-009-0230-6
Worbe, Y., Savulich, G., De Wit, S., Fernandez-Egea, E., & Robbins, T. W. (2015). Tryptophan depletion promotes habitual over goal-directed control of appetitive responding in
