Response inhibition (RI) refers to the ability to suppress habitual behaviour that is inappropriate given current task demands. It is suggested to reflect a basic ability that is central to executive functions [1], higher-order cognitive processes that enable goal-directed behaviour[2]. Moreover, impaired RI, as commonly measured in the lab using go/no-go and Stop signal tasks [3], has been linked to multiple psychopathologies [4], and other negative outcomes, such as poor academic and career success [5]. Given the central role of RI, it is important to identify its correlates. Such identification might inform theoretical considerations concerning causes and consequences of variations in RI ability, and ultimately interventions to improve this ability and/or have positive effects on its correlates. In the present study, we focussed on two such correlates, including their own association, that have been examined before in the literature with partly mixed results: childhood socioeconomic status (SES) and anxiety levels.
SES and RI
Many studies have provided evidence that a low SES of the parents or caregivers of an individual is associated with poor executive functioning [6–9], including brain and/or behavioural measures of RI specifically [10–13]. The general notion is that SES is positively associated with the development of brain areas that are importantly involved in RI and other components of executive functioning, such as the prefrontal cortex [14]. Various mechanisms have been proposed to underly this association. These mechanisms are related to the concept of early-life adversities [15], which may include deprivation of cognitive stimulation from the home environment [16], and family stress, with low SES being conducive to psychological distress and poor parenting behavior that is characterized by harsh discipline and detachment [17].
The mentioned studies examining the relation between SES and RI ability as indexed by go/no-go task performance, are characterized by: 1) a focus on children with a low SES background [12, 13], 2) an examination of young childhood samples [12], and/or 3) using a (very) restricted number of SES indicators [10, 13]. The present study aimed to focus on 14–18-year adolescents because previous studies suggest relatively large SES effects on the development of brain structure and functioning related to EFs [18], and on RI in particular [11, 19] in this age group. Moreover, we aimed to include adolescents with a wide range of SES level, which we measured using a variety of indicators.
SES and anxiety
Anxiety is commonly subdivided into a more transient mental state elicited by stressful situations, also known as state anxiety, and a more stable way of experiencing environmental stimuli, to the extent that it can be conceived of as a personality trait, known as trait anxiety. Although state and trait anxiety are often correlated [20], they appear to have partially different neural correlates [21]. Low family SES has been found to be associated with high trait anxiety, even to the level of developing anxiety disorders [22, 23]. A mechanism suggested to underly this association concerns negative effects of a lack of basic resources, such as food, clothing, and medical care, on the development of brain structures (e.g., hippocampus and prefrontal cortex) and their functioning and connectivity. These brain-related effects are in turn thought to be critically involved in developing mental health issues, including depression and anxiety. An additional potential mechanism is that disadvantageous living conditions associated with low SES are directly conducive to the development of depression and anxiety.
Low SES has also been linked to high state anxiety [24], including test anxiety [25, 26]. These associations may be mediated by experienced punitive measures, the frequency of which is assumed to be higher for children from a low SES background. These experiences could subsequently enhance anxiety when the individual has to perform daily tasks (including school tasks) that involve a risk of failure. Another proposed mechanism is that a high SES is associated with more cognitive stimulation and availability of learning and financial resources compared to a low SES. These in turn enhance the chance of (academic) success, thereby lowering (test) anxiety [27].
Anxiety and RI
Regarding associations between state and trait anxiety on one hand and inhibition on the other, results of previous studies are mixed. High trait anxiety has been found to be negatively associated with certain types of inhibition, specifically inhibition of task-irrelevant or distracting stimuli, also termed interference control, especially when the distractors concern threat stimuli [28, 29]. A negative association of high trait anxiety has also been found in a recent study in a sample of Chinese students, measuring performance in a go/no-go response inhibition task with multiple go and no-go stimuli [30]. Specifically, the number of correct responses to no-go stimuli (inverse of number of incorrect responses or false alarms) was smaller in high trait anxiety participants than their low anxiety counterparts. However, some studies also specifically looking at response inhibition as assessed with classic go/no-go tasks found trait or “clinical” anxiety (i.e., those with a confirmed anxiety disorder) in adolescence and adulthood to have no, or even a negative association with the number of false alarms. For example, one study examining anxiety-disordered and healthy adults, found no group difference in the number of false alarms [31]. Another study examining a large sample of young adolescents, also failed to find a significant association between caregiver-reported trait anxiety and false alarms [32]. Grillon et al. [33] studied adults with and without anxiety disorder and found that the two participant groups did not differ on a measure of inhibition based on the difference between number of responses to go stimuli (also known as “hits”) and false alarms, namely d-prime (see also below under data analysis). However, when looking at the number of hits separately, these authors found less hits in the anxious participants relative to the controls. This reduced number of hits suggests that the anxious participants had a relatively strong tendency or bias to not respond, possibly reflecting a more cautious response style [see also 34]. Finally, Sehlmeyer et al. [35] examined undergraduate students and found higher self-report trait anxiety to be negatively related to number of false alarms and positively to the amplitude of evoked brain responses that are indicative of exerting cognitive control. The authors used an explanation put forward by Eysenck and Calvo [36] as a possible underlying mechanism, namely that anxious participants tend to exert elevated cognitive effort to avoid aversive states.
Studies on the effect of state anxiety on response inhibition also report contrasting findings. For example, Grillon et al. [33, 37] and Robinson et al. [38] manipulated state anxiety using a threat-of-shock paradigm and found higher state anxiety to cause less false alarms, while the number of hits was unaffected. These results imply a positive effect of state anxiety on response inhibition. Grillon and colleagues suggested that a heightened state anxiety might facilitate early perceptual and sensory processing. Instead, Wilson and colleagues [39] manipulated state anxiety by presenting negatively valenced emotional or neutral pictures in undergraduate students. They found that the negative pictures both increased the number of false alarms and reduced the number of hits, implying a negative effect of state anxiety on response inhibition. The authors suggested that this negative effect might reflect that the negative pictures induced cognitive control processes related to the suppression of negative thoughts, which compete with the control processes necessary for response inhibition.
This overview of previous studies suggests the necessity of a detailed analysis of different outcome measures from RI tasks in general, and from go/no-go tasks in particular. Associations between anxiety and RI, and between SES and RI for that matter, may differ depending on the exact measure that is used to operationalize RI ability.
Present study
The purpose of the current study was to perform a detailed analysis of response inhibition in a relatively large sample of 14–18-year-old Chinese high school students in relation to their family SES level and state and trait anxiety. We used a classic go/no-go task to measure response inhibition ability, from which we derived five outcome measures (see under Data analysis for details): 1) percentage of hits (responses to go stimuli), 2) percentage of false alarms (erroneous response to no-go stimuli), 3) sensitivity to distinguish between go and no-go stimuli, expressed as d-prime (d'), 4) general response bias, expressed as response criterion (c), and 5) inhibition efficiency. Measures 3 and 4 were based on signal detection theory [40], whereas the first two measures allowed us to specify the reason for specific scores on Measures 3 and 4, namely whether they were primarily due to hits and/or false alarms. Response inhibition efficiency was computed by taking into account both the number of false alarms and response speed on go trials. A strong efficiency means that the individual is well capable of suppressing responding to no-go stimuli while not trading response speed for accuracy [13] (see below for further details). For determining SES, we adopted both objective and subjective indices. These indices have been shown to be correlated, but not perfectly, implying that they can provide non-redundant and necessary information [41, 42]. Finally, we assessed both state and trait anxiety via a commonly used self-report questionnaire.
Based on previous studies, we expected that SES indices are associated with most of the response inhibition outcome measures, specifically negatively with false alarms, and positively with inhibition sensitivity and efficiency. However, given the mixed findings in the literature, we had no clear a priori hypotheses about the association between state and/or trait anxiety and the diverse response inhibition measures. After presenting the empirical results, our discussion focussed on theoretical considerations about possible mediation and moderation models underlying the results.