Introduction
Adaptive human behavior is commonly taken to reflect the operation of cognitive-control processes that organize lower-level information-processing streams according to the current task and intention. Increasing evidence suggests that individuals exhibit systematic differences in the way they perform cognitive-control processes, and it was these differences that the present study was aimed at. In particular, people have shown systematic interindividual and intraindividual differences regarding the degree to which their performance reflects cognitive persistence and flexibility (Hommel,
2015; Hommel & Colzato,
2017a,
b). Persistence and flexibility have been considered two antagonistic metacontrol strategies (i.e., strategies that control cognitive control; Goschke,
2003; Cools & d’Esposito,
2011) that can be considered as the extreme poles of a common metacontrol dimension (Hommel,
2015). Changing tasks and environmental conditions require continuous readjustments of the balance between persistence and flexibility, which induces intraindividual variability (Akbari Chermahini & Hommel,
2010; Dreisbach & Goschke,
2004; Herd et al.,
2014; Müller et al.,
2007), and people differ systematically with respect to the efficiency of the degree to which this balance can be achieved (Arbula, Capizzi, Lombardo, & Vallesi,
2016; Babcock & Vallesi,
2017; for a review, see Hommel & Colzato,
2017b).
The goal of this study was to investigate a possibly potent personality characteristic that might be associated with a systematic metacontrol bias, at least with respect to some control-relevant tasks: rumination. According to Whitmer and Gotlib (
2013, p. 1036), “rumination is generally defined as repetitive thinking about negative personal concerns and/or about the implications, causes, and meanings of a negative mood”. In terms of metacontrol theory, this amounts to a strong bias toward persistence, at the expense of flexibility. Rumination and cognitive reactivity to sad mood in general has important clinical implications, and the degree and amount of rumination are informative regarding the probability of becoming chronically depressed (Beck,
1967; Nolen-Hoeksema, Morrow, & Fredrickson,
1993; Kuehner & Weber,
1999; Nolen-Hoeksema,
2000; Spasojevic & Alloy,
2001; Moulds et al.,
2008). However, the relevance of rumination is not restricted to clinical cases, more so as many healthy individuals tend to ruminate at least occasionally. From a metacontrol perspective, it is interesting to consider that individuals who tend to ruminate show difficulties in disengaging from or forgetting information that is no longer relevant (Whitmer & Gotli,
2013), which has been shown in studies with healthy individuals. For instance, the personal tendency to ruminate was positively correlated with bigger struggle in disengaging from no-longer-relevant information, but not in ignoring external distracters, in a modified Sternberg task (Joormann & Gotlib,
2010), and with greater difficulty to ignore irrelevant information in a Stroop task (Philippot and Brutoux,
2008). Unfortunately, however, the overall picture is disturbed by at least partial non-replication. For instance, other studies found Stroop performance to be better (Altamirano, Miyake & Whitmer,
2010) or unrelated to rumination in samples of dysphoric and nondysphoric participants (Krompinger and Simons,
2011 and Meiran, Diamond, Toder, and Nemets,
2011). On the one hand, this may raise serious questions regarding replicability but, on the other, differences in task and design cannot be ruled out as possible factors.
Given the present state of affairs, we were interested in testing the impact of rumination in healthy participants on a broader set of control-related tasks. In addition to various demographic indicators, we assessed rumination and cognitive reactivity to transient changes in sad mood by means of the revised Leiden Index of Depression Sensitivity (LEIDS-r; van der Does & Williams,
2003). We studied the impact of LEIDS-r scores on four tasks or task-performance indicators. The first choice was obviously the Stroop task, as this has been used in various rumination studies previously. Given previous observations, the question was whether we could replicate the observation of a positive correlation between rumination tendency and the inability to suppress irrelevant information (as indicated by the size of the Stroop effect).
The second was the event-file task developed by Hommel (
1998). The task assesses the degree to which specific combinations (bindings) of stimulus and response features are automatically retrieved in the next trial. Previous studies have shown that individuals with suboptimal top-down cognitive-control abilities, such as children and elderly (Hommel, Kray & Lindenberger,
2011), people low in fluid intelligence (Colzato, van Wouwe, Lavender & Hommel,
2006), or individuals suffering from autistic spectrum disorders (Zmigrod et al.,
2013), are more likely to retrieve (or less likely to suppress) bindings between irrelevant stimulus features and the response. At the same time, neurofeedback-based cognitive-control training was able to eliminate the retrieval of these bindings, in addition to increasing the intelligence score (Keizer, Verment & Hommel,
2010; Keizer, Verschoor, Verment & Hommel,
2010). This suggests that the degree to which bindings between irrelevant stimulus features and the response formed in the previous trial affects performance in the present trial represents the degree to which people can control binding retrieval. If so, more rumination should be associated with more pronounced effects of bindings between irrelevant stimulus features and the response.
The Stroop and the event-file task tap into the efficiency of handling of memory information that is activated by the current stimulus. A more pronounced effect of task-irrelevant information would thus indicate a lack of selectivity with respect to the information that the present stimulus ought to (re-)activate, with the result that falsely (re-)activated memory traces compete with task-relevant information for selection. A related, but somewhat different aspect of information-processing efficiency is assessed by an attentional set-shifting task developed by Dreisbach and Goschke (
2004). Participants usually perform a letter or digit classification task and the performance is measured before and after the switch to a different version of the task, so that the difference represents set-switching costs. In a “perseveration” condition, participants switch to target stimuli in a novel color and try to ignore the previous target, which continues to be present. Switching costs in this condition are compared with switching costs in a “learned irrelevance” condition, where participants switch to target stimuli in a previously ignored color, while ignoring the previously relevant color. Individual differences related to persistence and flexibility should affect performance in the two conditions differently. A lack of persistence or a bias toward flexibility should support switching in the perseveration condition, but impair switching in the learned irrelevance condition, as flexibility should bias attention toward novel stimuli. Accordingly, one would expect the opposite pattern in individuals with a strong rumination tendency: they would have a hard time turning to something novel and thus show relatively poor performance in the perseveration condition, but relatively good performance in the learned irrelevance condition. Given that the attentional set-shifting task shares the characteristic of the Stroop in the event-file task of tapping into memory control, but focuses more on the impact of past attentional settings on present attentional control (a factor that can be suspected to affect Stroop and switching performance differently: Herd et al.,
2014), we decided to also include it as the third task in the present study.
Note that both the Stroop task and the event-file task assess selectivity in handling stimulus-induced activations of internal information. A Stroop effect can only be obtained if the meaning of the color word is retrieved to a degree that it interferes with the determination of the stimulus color. Hence, the size of the Stroop effect should depend on retrieval control, just as assessed by the event-file task. However, the recent emphasis on attentional processes in the context of rumination (Whitmer & Gotlib,
2013) suggests that more direct assessments of attention to external information might also be of interest. The most comprehensive assessment in this respect is provided by the Attentional Network Task (ANT) developed by Fan et al. (
2002). This task is a hybrid that combines Posner’s (
1980) cued reaction time (RT) task and Eriksen and Eriksen’s (
1974) flanker task. It provides three indicators that are assumed to reflect the efficiency of the alerting network (i.e., preparation for a stimulus event), the orienting network [i.e., (re-)allocating visual attention to a new stimulus event], and the executive-control network (i.e., resolving stimulus-induced response-selection conflict). Note that none of these measures tap into retrieval control, which is an issue that will be important for interpreting our findings.
As the ANT has not yet been used to investigate the impact of rumination or to study individual differences in metacontrol, different predictions are possible. Given the emphasis of recent rumination research on attentional control, one would expect a systematic association between the personal tendency to ruminate and performance on all three indicators of the ANT task. Hence, alertness, orientation ability, and response-conflict resolution should be impaired more the more people tend to ruminate. Another possibility is that rumination is more selectively related to more central processes, with response selection being an obvious choice (Johnston, McCann & Remington,
1995). If so, it is possible that rumination only affects tasks or task indicators that induce response conflict, such as the Stroop, the event-file task, the attentional set-shifting task, and the executive-control indicator of the ANT. Finally, it is interesting to note that the previous metacontrol studies did not show any systematic impact of individual metacontrol policies on the Simon effect (Colzato, Sellaro, Samara, & Hommel,
2015), even though this effect is commonly taken to reflect the same kind of response conflict that is induced by the flanker task. Accordingly, it might also be possible that rumination is even more selectively related to online retrieval control, in which case the correlation might be restricted to the Stroop and the event-file task.
Discussion
The aim of the study was to throw more light on the relationship between rumination and cognitive-control processes. For that purpose, we tried to replicate the previously observed positive relationship between rumination tendencies in the Stroop effect and added eight further indicators of performance in theoretically relevant tasks or task aspects. With regard to the Stroop effect, we made two interesting observations. For one, our standard analysis replicated the previous findings of Philippot and Brutoux (
2008) that more rumination goes with a more pronounced Stroop effect. At the same time, however, we also found that controlling for depressive tendencies eliminated this effect and that analysis within the Bayesian framework found no evidence in supporting H
0 or H
1. This means that with respect to the Stroop effect, our data are inconclusive in providing an account for the previous failures to replicate the relationship between rumination and Stroop (Philippot & Brutoux,
2008; Krompinger & Simons,
2011; Meiran et al.,
2011).
Less sensitive to self-reported depression symptoms was our second measure, the degree to which previous bindings between irrelevant stimulus information and the response was retrieved in the next trial. The corresponding color-by-response effect strongly correlated with all three scales irrespective of depressive tendencies, but we found moderate evidence for H
1 only for the rumination score. This provides reasonable evidence that rumination is related to the control of retrieving internal information and, in particular, to preventing irrelevant information from being retrieved. Nevertheless, given that the correlation of rumination scores with the effect of the binding between the relevant stimulus feature and the response and of the correlation between rumination and the three ANT scores received no evidence or only anecdotal evidence for H0 or H1, it is difficult to say how specific the impact of rumination is to the control of retrieving internal information. Supported by Bayesian inference, future studies need to search for converging evidence that rumination is associated with impairments of the handling of distractor information (Whitmer & Gotlib,
2013) and for the idea that it is not the mere presence of an external distractor that ruminating individuals find difficult to deal with, but the handling of information that such distractors may activate in one’s memory.
Our conclusion is that rumination reliably affects the efficiency of memory-retrieval control in the face of events that activate irrelevant memory information. This allows the prediction that rumination might be beneficial in tasks that require or that benefit from less selective memory retrieval. In other words, the fact that a greater tendency to ruminate was associated with less efficient processing in the present study might be taken as a reflection of the choice of tasks rather than a demonstration of a generalized deficit associated with rumination.
As standard procedure in our laboratory, at the start of each session, participants completed a visual analog scale (range of scores from 0 to 100) that measured the subjective self-reported current level of anxiety, nervousness, insecurity, and stress. Next, heart rate (HR) data were measured for 5 min using a Polar H7 heart rate monitoring system (Polar Electro, Kempele, Finland), which wirelessly receives HR data from a chest strap worn by the participants. As a result of technical problems, several data were lost across all five sessions. Further, in the first session participants were weighed and their BMI was measured using an OMRON Body Composition Scale Karada Scan. Moreover, their daily level of physical activity and smoking behavior were recorded.