Introduction
Given that developmental pathways are triggered or become rooted during adolescence, mental health problems in adolescence may have long-term consequences (Ormel et al.
2015). The prevalence of psychiatric illnesses rises from childhood to young adulthood (Copeland et al.
2011; Newman et al.
1996). Adolescence and young adulthood especially are periods with increasing demands for coping with stress resulting from the multiple transitions in these periods (Leadbeater et al.
2012). This underscores the importance of understanding mechanisms involved in the development of mental disorders during adolescence and young adulthood. Two prevalent classes of disorders with age-of onset in childhood and adolescence are anxiety disorders and behavioral disorders (Ormel et al.
2015). Two important traits that have often been linked to symptoms of anxiety and behavioral disorders are sensitivity to punishment (PS) and sensitivity to reward (RS).
Punishment and reward sensitivity stem from the reinforcement sensitivity theory (Gray
1970,
1982,
1987; Gray and McNaughton
2000). According to this theory, people who are sensitive to punishment will have a more negative response to punishment, more attention to punishment-relevant cues and a stronger tendency to avoid punishment. People who are sensitive to reward will have a more positive response to reward, more attention to reward-relevant cues and a stronger tendency to approach reward (Gray
1970; Gray and McNaughton
2000; Davis and Fox
2008).
The attentional system provides the mechanism for detecting and monitoring the environment for stimuli that are relevant to the motivational state of the organism (Mogg and Bradley
1998). People who are heightened punishment sensitive are motivated to avoid punishment and are therefore expected to be more prone to detect punishing signals in the environment; people who are heightened reward sensitive are motivated to obtain reward and are therefore expected to be more prone to detect rewarding signals in the environment (Gray
1970; Gray and McNaughton
2000).
A heightened proneness to detect punishing signals in the environment may result in prolonged anxious states, limited attention for fear-disconfirming information, and feelings of uncontrollability, making people more vulnerable for the development of anxiety disorders (Harvey et al.
2004). A heightened proneness to detect signals of reward may result in positive affect in rewarding situations; however, if the person does not succeed in getting the preferred outcome, this might result in non-reward elicited anger and behavioral problems (Corr
2013) as was found in multiple studies conducted in non-clinical samples of adolescents (Carver
2004; Hundt et al.
2013; Harmon-Jones
2003). Reward and punishment sensitivity are presumed to represent orthogonal dimensions that can vary independently in strength, indicating that all combinations of (relatively) high and (relatively) low PS and RS may be evident in a particular population (Carver and White
1994). Individuals at the far poles of the punishment sensitivity and/or the reward sensitivity dimensions are expected to have an increased risk for developing mental health problems (Pickering and Gray
1999), which might especially become evident during periods with increasing demands, such as adolescence and young adulthood (Leadbeater et al.
2012).
Multiple studies have investigated the associations between self-reported punishment and reward sensitivity and anxiety and behavioral problems, respectively. These self-report measures are well suited to assess the affective component of punishment and reward sensitivity. A review conducted by Bijttebier et al. (
2009) indicated that on global measures of mental disorder symptoms, internalizing problems were associated with higher PS, whereas behavioral problems were associated with higher RS. When looking more specifically at anxiety disorders within the internalizing domain, there is ample evidence linking PS to anxiety symptoms in non-clinical child and adult samples (Takahashi et al.
2015; Bijttebier et al.
2009) and linking PS to anxiety disorders in both child and adult clinical samples (Vervoort et al.
2010; Bijttebier et al.
2009). In line with the view that high PS may be a risk factor for the development of anxiety disorders, a longitudinal study showed that (high) self-reported PS in adolescence had predictive value for the level of anxiety symptoms in adulthood, even when controlling for anxiety in adolescence (Izadpanah et al.
2016).
With regard to behavioral disorders, multiple studies indicated an association between RS and behavioral problems. More specifically, RS has been associated with self-reported conduct problems in clinical adolescents (Morgan et al.
2014), trait anger in non-clinical students (Smits and Kuppens
2005; Harmon-Jones
2003), self-reported verbal and physical aggression in non-clinical students (Smits and Kuppens
2005), and self-reported hostility in non-clinical students (Harmon-Jones
2003). Heightened reward sensitivity may result in a higher proneness to detect signals of reward in the environment, higher motivation to approach reward, and may result in positive affect in rewarding situations. However, when more reward sensitive persons experience failures to obtain anticipated reward, this is expected to result in non-reward elicited anger and behavioral problems (Corr
2013; Carver
2004; Hundt et al.
2013; Harmon-Jones
2003).
The evidence mainly stems from cross-sectional studies using self-report measures of punishment and reward sensitivity. However, it is doubtful whether the attention component of punishment and reward sensitivity, namely the proneness to detect cues of punishment and reward respectively, can be adequately assessed by means of self-reports. Performance-based measures seem required to assess this component of RS/PS. Attentional processes help in selecting specific stimuli for further processing and prevent us from being overwhelmed by all information that surrounds us. In this way, it involves the initial filtering of the environment and if there is a bias in this first filtering of information, it might likely contribute to further processing biases that might result in clinical problems (Derryberry and Reed
2002).
In the current study, we therefore decided to use a performance measure (spatial orientation task; Derryberry and Reed
2002) to examine the relation between sensitivity of the punishment system and reward system with anxiety and behavioral problems. The spatial orientation task (SOT) was developed to explore to what extent people direct and hold their attention to places of potential reward and punishment, and was in previous studies successfully used in the context of substance use and addiction in non-clinical adolescent and young adult samples (Colder and O’Connor
2002; van Hemel-Ruiter et al.
2013; van Hemel-Ruiter et al.
2015) eating disorders in clinical and non-clinical adolescents (Jonker et al.
2016; Matton et al.
2017), and depression in clinical adolescents and young adults (Vrijen et al.
2018).
The SOT is a reaction time task which consists of games in which participants can gain points (winning games), and games where points can be lost (losing games). Before each target appears, a cue is presented that either signals a high chance of reward (in winning games)/non-punishment (in losing games) or a high chance of punishment (in losing games)/non-reward (in winning games). The target can occur either in the cued or uncued location. The difference in reaction time between the cued and uncued location represents the cue validity effect. This cue validity effect indicates the attentional bias of individuals to cues predicting punishment or reward. Separate cue validity scores were calculated for short (250 ms) and long (500 ms) delays between cues and targets, which provides the opportunity to examine the relative importance of early (short delay) attentional processes and attentional processes that allow for some regulatory control (long delay) (Derryberry and Reed
2002).
Previous cross-sectional research indicated that anxious students showed an enhanced cue validity effect for cues signaling punishment (Derryberry and Reed
2002). This effect was found with short cue delays, but not with longer cue delays (Derryberry and Reed
1994,
2002) and is suggested to be largely automatic (McNally
1995; Mogg et al.
1995). Allocating attention to objectively threatening stimuli can be regarded as an adaptive mechanism that serves rapid detection and avoidance of danger (Mogg and Bradley
1998). However, an attentional bias to subjective or ambiguous threat may contribute to the development and maintenance of anxiety problems.
An enhanced cue validity effect for non-punishing cues may reflect a tendency to seek safety. Within a threatening situation, attention to safety may help the person attenuate their anxiety, enable to remain in, and learn from, the environment. This may generally be adaptive, however, an enhanced cue validity effect for non-punishing cues may prevent the habituation and reappraisal of stimuli perceived to be threatening, and thereby maintain anxiety (Harvey et al.
2004; MacLeod and Grafton
2016). Previous research indeed found evidence indicating that high anxious individuals showed an heightened cue validity effect for cues signaling non-punishment (regarded as safety cues (Derryberry and Reed
2002). This effect was evident for cues with longer cue delay, suggesting that this process may be less automatic and more voluntary.
Increased reward sensitivity leading to enhanced responses to reward is assumed in young children with clinical behavioral problems (Quay
1993). Previous research showed that enhanced attentional engagement to cues signaling reward and difficulty disengaging from cues signaling reward were related to adolescent substance use (van Hemel-Ruiter et al.
2013; Colder and O’Connor
2002), and that this bias measured during adolescence was predictive for substance use in young adulthood (van Hemel-Ruiter et al.
2015). Furthermore, it was found that an attentional bias to reward, measured with an adapted version of the Posner spatial attention-cueing task, was associated with behavioral problems in 5 year old children (He et al.
2016). It is however untested whether an attentional bias to reward as indexed by a spatial orientation task is related to and has prognostic value for the development of behavioral problems in adolescence and young adulthood.
The current study was designed to investigate how individual differences in attentional bias for cues predicting punishment and reward are associated with symptoms of anxiety and behavioral disorders in adolescence and young adulthood.
We will try to replicate the findings from Derryberry and Reed (
2002) in an adolescent sample to see whether (i) having a stronger cue validity effect for cues signaling punishment with short cue delay is associated with higher anxiety symptoms. We will extend previous research by also looking at (ii) the prognostic value of this cue validity effect for cues signaling punishment with short cue delay for anxiety symptoms at six years follow-up. We (iii) will also try to replicate the findings from Derryberry and Reed (
2002) in an adolescent sample to see whether a stronger cue validity effect for cues signaling non-punishment with long cue delay is associated with having higher anxiety symptoms and extend this line of research by (iv) also looking at the prognostic value of this cue validity effect at six years follow-up. Our study is the first study to investigate an attentional bias for reward on behavioral problems in adolescence and young adulthood. Based on studies investigating the role of an attentional bias for reward on substance use in adolescents and young adults (van Hemel-Ruiter et al.
2013; van Hemel-Ruiter et al.
2015; Colder and O’Connor
2002) and a reward bias on behavioral problems in young children (He et al.
2016) we will (v) test whether a stronger cue validity effect for cues signaling reward with both short and long cue delay is associated with behavioral problems in adolescents and has (vi) prognostic value for behavioral problems at 6 years follow-up. We expect (vii) this association between a stronger cue validity effect for cues signaling reward with behavioral problems to be most pronounced on trials with long cue delay since than both automatic and more voluntary processes are expected to play a role and are expected to have an added effect.
Results
We have missing data in the current study, which can produce biased estimates due to differences between missing and included participants and can reduce the statistical power of a study, leading to invalid conclusions (Kang
2013). Therefore, we checked whether the missing data in our study would pose a threat to the validity of our conclusions due to having a biased sample or a too large reduction in power.
For the cross-sectional sample, we could include 696 of the 715 participants (97%). This means that for the cross-sectional analyses only 3% is missing, indicating that bias and loss of power are both likely to be inconsequential (Graham
2009). Furthermore, for the prospective sample, including participants with complete data only, did not seem to lead to a biased sample, given that we did not find any significant differences between the individuals with missing prospective data (
n = 117, 16%) and individuals with complete prospective data (
n = 598) on anxiety and behavioral problems at t3 as well as with regard to the cue validity effects.
1 With regard to power, a sample size of 544 participants is needed to be able to find an effect with a small effect size, α of 0.0125 and power of 0.80. Therefore, given our sample size of 598 participants, power should also not be a problem for the prospective analyses.
Descriptives
In line with Jonker et al. (
2016), trials during which participants did not respond to the target were deleted, which resulted in deletion of 3.3% of the trials. Also trials on which participants responded before the target appeared were removed, resulting in the deletion of 8.3% of the trials. Furthermore, reaction times below 125 ms, which are expected to be anticipation errors, were deleted, resulting in the deletion of 8.5% of the remaining trials. The mean reaction times for each game type (winning and losing) and trial type (easy cue/hard cue and cued/uncued) were calculated after these deletions and are presented in Tables
3 and
4.
Table 3
Mean reaction times and standard deviations of the Spatial Orientation Task in the cross-sectional sample
Losing game |
Short cue delay time (250 ms) | 330 (47) | 358 (53) | 456 (89) | 458 (92) |
Long cue delay time (500 ms) | 332 (59) | 366 (68) | 380 (83) | 374 (79) |
Winning game |
Short cue delay time (250 ms) | 336 (43) | 366 (48) | 468 (90) | 471 (90) |
Long cue delay time (500 ms) | 342 (58) | 379 (67) | 384 (79) | 377 (74) |
Table 4
Mean reaction times and standard deviations of the Spatial Oriental Task in the prospective sample
Losing game |
Short cue delay time (250 ms) | 327 (43) | 356 (49) | 454 (86) | 458 (92) |
Long cue delay time (500 ms) | 330 (56) | 363 (66) | 378 (80) | 373 (75) |
Winning game |
Short cue delay time (250 ms) | 335 (41) | 364 (46) | 466 (89) | 468 (88) |
Long cue delay time (500 ms) | 341 (57) | 377 (67) | 380 (77) | 375 (72) |
Task Design Check
In line with previous studies (Jonker et al.
2016; van Hemel-Ruiter et al.
2015) a series of paired sample
t-tests were carried out to test the expectation that people in general respond faster to cued blue trials compared to cued red trials, and have faster responses on uncued red compared to uncued blue trials (see Table
5). Participants were faster on the cued blue than cued red trials for both winning and losing games, irrespective of the cue delay time, indicating a general preference to direct attention to cues that predict reward or non-punishment compared to cues that predict punishment and non-reward. Thus, in line with the task design, participants showed a generally enhanced attentional engagement to stimuli signaling reward and non-punishment. Furthermore, participants were slower on uncued blue trials than uncued red trials on long cue delay time trials in winning games, indicating a difficulty to disengage from reward with longer cue delay.
Table 5
Task design check; differences between red and blue targets
Short cue delay time (250 ms) |
Wining game | Cued red – cued blue | 28.07 | 32.85 | <0.001* |
Uncued red- uncued blue | −2.24 | 7.03 | 0.311 |
Losing game | Cued red- cued blue | 25.34 | 30.62 | <0.001* |
Uncued red – uncued blue | −2.42 | 7.23 | 0.328 |
Long cue delay time (500 ms) |
Winning game | Cued red-cued blue | 32.63 | 39.94 | <0.001* |
Uncued red- uncued blue | −11.35 | −3.00 | 0.001* |
Losing game | Cued red – cued blue | 29.87 | 37.67 | <0.001* |
Uncued red – uncued blue | −10.35 | −1.33 | 0.011* |
Step 1
Bivariate correlations were calculated between the cue validity effects, anxiety symptoms, and behavioral problems, see Table
6.
Table 6
Bivariate correlations of cue validity effects with internalizing and behavioral problems at T3 and T5
1 Anxiety 3 (RCADS) | – | | | | | | | | | | | | |
2 Anxiety t3 (YSR) | 0.713* | – |
3 Behavioral problems t3 | 0.288* | 0.271* | – |
4 Anxiety t5 | 0.467* | 0.473* | 0.224* | – |
5 Behavioral problems t5 | 0.385* | 0.337* | 0.388* | 0.641* | – |
6 CV-reward short | −0.061 | −0.065 | −0.065 | −0.056 | 0.058 | – |
7 CV-reward long | −0.067 | −0.081 | −0.046 | −0.057 | −0.046 | 0.702* | – |
8 CV-nonreward short | −0.044 | −0.037 | −0.048 | −0.043 | −0.040 | 0.897* | 0.706* | – |
9 CV-nonreward long | −0.072 | −0.091 | −0.056 | −0.070 | −0.065 | 0.654* | 0.748* | 0.689* | – |
10 CV-punishment short | −0.018 | 0.030 | −0.055 | −0.009 | −0.007 | 0.803* | 0.710* | 0.807* | 0.672* | – |
11 CV-punishment long | −0.094 | −0.087 | −0.022 | −0.087 | −0.062 | 0.646* | 0.717* | 0.671* | 0.741* | 0.692* | – |
12 CV-nonpunishment short | −0.067 | −0.065 | −0.034 | 0.067 | −0.068 | 0.795* | 0.676* | 0.769* | 0.665* | 0.795* | 0.669* | – |
13 CV-nonpunishment long | −0.070 | −0.066 | −0.034 | −0.095 | −0.084 | 0.698* | 0.741* | 0.681* | 0.723* | 0.697* | 0.746* | 0.697* | – |
Significant correlations were found between anxiety at T3 and T5 and behavioral problems at T3 and T5. Furthermore, significant correlations were found between anxiety and behavioral problems. However, no significant bivariate relationships were found between the cue validity effects and anxiety or behavioral problems.
Step 2 Main Analyses
Unexpectedly, no significant bivariate relationships were found between the cue validity effects and anxiety or behavioral problems. The planned main analyses were still conducted to test whether the whole model including multiple cue validity effects was significant, which would be indicative of their joint effects.
Cross-Sectional Analyses
Anxiety symptoms (T3): No significant associations between the cue validity effects for cues signaling punishment or non-punishment with anxiety were found (see Table
7). The full model was also not significant.
Table 7
regression model with anxiety (T3) and cue validity effects for punishment and non-punishment
Constant b0 | .414 | 0.019 | | 21.50 | <0.001 |
CV-punishment-short | 0.000 | 0.000 | 0.158 | 2.34 | 0.019 |
CV-punishment-long | 0.000 | 0.000 | −0.124 | −2.02 | 0.044 |
CV-non-punishment-short | 0.000 | 0.000 | −0.094 | −1.41 | 0.159 |
CV-non-punishment-long | 0.000 | 0.000 | −0.023 | −0.373 | 0.709 |
R2change. = 0.017 | F = 2.93 | p = 0.020 | | | |
Behavioral problems (T3): No significant associations between the cue validity effects for cues signaling reward or non-reward with behavioral problems were found (see Table
8).
Table 8
Regression model with behavioral problems (T3) and cue validity effects for reward and non-reward
Constant b0 | 0.318 | 0.017 | | 18.51 | <0.001 |
CV-reward-short | 0.000 | 0.000 | −0.070 | −1.05 | 0.295 |
CV-reward-long | 0.000 | 0.000 | 0.014 | 0.23 | 0.822 |
CV-non-reward-short | 0.000 | 0.000 | 0.024 | 0.35 | 0.729 |
CV-non-reward-long | 0.000 | 0.000 | −0.038 | −0.62 | 0.537 |
R2change. = 0.005 | F = 0.85 | p = 0.494 | | | |
Prospective Analyses
Anxiety symptoms (T5): It was found that having a larger attentional bias to cues signaling punishment with short cue delay predicted higher anxiety symptoms. The full model was significant (see Table
9).
Table 9
Regression model with anxiety (T5) and cue validity effects for punishment and non-punishment
Constant b0 | 0.380 | 0.026 | | 14.74 | <0.001 |
CV-punishment-short | 0.001 | 0.000 | 0.192 | 2.64 | 0.008* |
CV-punishment-long | 0.000 | 0.000 | −0.080 | −1.22 | 0.222 |
CV-non-punishment-short | 0.000 | 0.000 | −0.090 | −1.24 | 0.216 |
CV-non-punishment-long | 0.000 | 0.000 | −0.107 | −1.56 | 0.120 |
R2change. = 0.021 | F = 3.22 | p = 0.012* | | | |
Behavioral problems: The cue validity effects for cues signaling reward or non-reward did not predict behavioral problems (see Table
10).
Table 10
Regression model with behavioral problems (T5) and cue validity effects for reward and non-reward
Constant b0 | 0.216 | 0.018 | | 11.88 | <0.001 |
CV-reward-short | 0.000 | 0.000 | −0.062 | −0.86 | 0.393 |
CV-reward-long | 0.000 | 0.000 | 0.020 | 0.29 | 0.776 |
CV-non-reward-short | 0.000 | 0.000 | 0.044 | 0.59 | 0.555 |
CV-non-reward-long | 0.000 | 0.000 | −0.069 | −1.04 | 0.298 |
R2change. = 0.005 | F = 0.82 | p = 0.514 | | | |
Step 3 Exploratory Analyses
In order to see whether effects are game specific, we conducted the same analyses also with the cue validity effects with the other game type. We did not find significant associations with these analyses. Furthermore, we investigated whether the cue validity effects predicted change in anxiety and behavioral disorder symptoms from T3 to T5. We did not find significant effects. The findings of these exploratory analyses can be found in
Appendix A.
Step 4 (Only when Necessary Based on Previous Results)
We did not find effects in both the losing and winning games on either behavioral problems or anxiety symptoms. Therefore, this step was not conducted.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.