Notes. RMSE = Root mean squared error. N2 = maximal amplitude for Nogo N2 signal. P3a = mean amplitude of Nogo P3a signal. P3b = mean amplitude of Nogo P3b signal. IES = inverse efficiency score. Nested cv = nested cross validation. p value = proportion of RMSE from permutation test smaller than RMSE from nested cv.
Discussion
The present study investigated the relationship between self-reported impulsivity and compulsivity with response inhibition in a large non-clinical sample. In order to do this, we obtained behavioral (RT, Nogo accuracy and inverse efficiency score) and electrophysiological (N2, P3a, P3b) data from a standard Go/Nogo task. Data were analyzed with robust linear regression as well as regression trees, which allow non-linear exploration via a machine-learning algorithm. The linear analyses yielded no significant associations of impulsivity or compulsivity with task performance or EEG measures except for the effect of the UPPS subscale premeditation on Nogo accuracy and IES. We did not observe any non-linear relationship either, as the regression trees did not appear to outperform the linear regression or permutated models as measured by RMSE . Possible explanations for this will be discussed below.
Task and sample characteristics
Firstly, as pointed out before, literature on the relationship between impulsivity and compulsivity and response inhibition is ambiguous, as said effect is often not found at either the behavioral or psychophysiological level. This could be due to the fact that the Go/Nogo tasks employed are not always entirely comparable. One feature of the tasks seems to be emotion: Accordingly, adapted Go/Nogo tasks show an incline in activation for emotional vs. neutral stimuli, which in turn is modulated by higher trait impulsivity. This could be shown for Nogo P3 amplitude as well as activation in the middle temporal gyrus , an area linked to disorders of cognitive control . Behaviorally, emotional stimuli seem to impair response inhibition , especially when the demand on executive functions is high . Since emotion and physiological arousal are intricately linked, arousal is thought to be related to impulsivity as well . For example, have shown scores on the BIS-11 to be positively correlated to corticolimbic structures such as the ventral amygdala, caudate and dorsal anterior cingulate gyrus, while negatively correlated with control circuits, e.g., the ventral prefrontal cortex. Thus, trait impulsivity might be associated with an imbalance of physiological arousal and inhibitory control. This would likely also reflect on self-report measures, as these items refer to behavior in the individual’s everyday life, which is not taking place in an experimental setting, but influenced by states of affect and arousal. Furthermore, both impulsivity and compulsivity are defined by affective elements and the respective behavior is heavily influenced by emotion. However, the Go/Nogo paradigm in the present study was designed to examine response inhibition as a cognitive process, hence emotion or arousal processes were not involved. This may have attenuated the relationship between response inhibition and its neurophysiological correlates and dispositional impulsivity and compulsivity as indexed by the BIS-11, UPPS and OCI-R. Similarly, in studies investigating the role of cognitive load, the negative effect of trait impulsivity on response inhibition became even more pronounced when cognitive demands were high or executive functioning itself was worse . Again, research on compulsivity mainly focuses on individuals with OCD, but there decreases in stop-signal task performance as well as activity in the supplementary motor area and inferior parietal lobule have been found with higher cognitive demand as well . As our Go/Nogo paradigm was intentionally low on cognitive demand, possible relationships to impulsive or compulsive behavior in the participant’s everyday life, where they are confronted with a myriad of different stimuli and simultaneous processes, may have been dampened.
One could also raise the question how our participants’ questionnaire scores relate to other studies, as this might affect comparability. Regarding the literature mentioned above that examined healthy control groups, our scores for impulsivity and compulsivity (see table 2) are quite similar to their distributions or lie between their groups of high- and low-scoring participants . Notably, our impulsivity scores strongly resemble those of the respective general population samples used for validation of the BIS-11 and UPPS , while participants in the original publications of the OCI-R achieved higher scores . In some cases, our participants achieved slightly higher impulsivity scores than in other studies using healthy control participants . In sum, our sample characteristics resemble those of earlier studies.
Yet, our analyses also differ from those with clinical populations, where results appear more stable. As OCD is defined by obsessive and compulsive symptomatology, individuals with OCD score generally score higher on the OCI-R . Further deficits in behavioral response inhibition have been found , as well as alterations in Nogo ERP signals and higher BIS-11 scores predicted by symptom severity . Similar changes have been found for other disorders often associated with impulsivity and/or compulsivity such as substance use , behavioral addictions , borderline personality disorder , and schizophrenia . Again, other studies have found elevated scores on varying impulsivity or compulsivity measures in these clinical groups compared to our sample . A possible explanation for the difference is that high manifestations of trait impulsivity and compulsivity, as often seen in clinical groups, are needed for its effects to unfold. Present psychopathology may also exacerbate the effect of trait impulsivity and compulsivity when typically compensating processes are impaired. This could be due to either to adverse effects of the mental disorder itself or a common predisposition that facilitated its onset.
It is thus not surprising that many of the studies that did find significant differences in behavioral or electrocortical response inhibition did so via group comparisons based on diagnostic status or high vs. low scores on impulsivity or compulsivity measures. A clear effect is probably found more easily in these distinct groups than in our rather homogenous sample. Furthermore, using patient groups as proxies for high impulsivity and compulsivity complicates interpretability, as potential confounding factors increase. For example, impairments in neurocognitive functions such as attention and processing speed that would in turn affect performance in a Go/Nogo task are reported among others for OCD , alcohol use disorder and borderline personality disorder .
Methodological considerations
When analyzed directly, impulsivity as indexed by self-report measures and cognitive tasks are associated with everyday behavior but show little or no correlation with each other . Presumably these measures do not simply reflect one overarching characteristic of impulsivity, but instead tap into different aspects of it. Sharma et al. (2014) proposed self-report scales to depict more long-term, emotionally laden response-patterns, which would also complement the findings on emotional response inhibition tasks described earlier. Cognitive tasks, on the other hand, may represent purer and “microlevel” processes, which would then also apply to the event-related potentials we analyzed. Thus, this problem of construct validity could explain the lacking association between self-report scales and behavioral and psychophysiological variables, as they measure rather independent dimensions.
Limitations and future directions
It should also be noted that the effect of impulsivity or compulsivity could affect the time course of the ERP instead of their amplitude. Later onsets have been found for the N2 with higher impulsivity and for individuals with OCD , while P3 latency was positively associated with impulsivity on a trend level . Thus, alterations in inhibitory control might not be represented in the strength of the conflict or inhibitory processing, but in the time and resources needed. In an exploratory analysis we did not find the N2 latency to be explained by impulsivity or compulsivity. Unfortunately, as the P3a and P3b were computed through the mean in a specific time window, latencies could not be reported. Future studies investigating the effect on the time course of response inhibition would be desirable. Another problem arises from the use of the gradient-boosting regression trees for the analysis of possible non-linear effects. This method is rather susceptible to the influence of outliers , as it explicitly focuses on the data points that could not be explained in the previous iterations. While this might have a slight effect on the reliability of our analyses, we chose not to exclude outliers to ensure a better depiction of the true population in our sample.
Conclusions
In sum, our non-clinical sample did not show any clear linear or non-linear relationship between self-reported impulsivity and compulsivity on the one hand and response inhibition as indexed by behavioral performance and N2, P3a and P3b signals in a Go/Nogo paradigm on the other hand. Possibly the effect of said personality traits on response inhibition does not hold true in a general population sample, but needs a clinical sample to unfold, or else to be exacerbated by adaptations to the Go/Nogo task. As research on the effect of impulsivity and compulsivity on cognitive control is vital, but as of yet discordant, a discussion on the best ways to uncover possible associations and interactions is highly warranted. With our comparatively large sample and the different statistical methods we applied, we hope this is a first step in elucidating the question at hand.