fMRI noise and cognitive control 1 The effect of fMRI (noise) on cognitive control

Stressful situations, the aversiveness of events, or increases in task difficulty (e.g., conflict) have repeatedly been shown to be capable of triggering attentional control adjustments. In the present study we tested whether the particularity of an fMRI testing environment (i.e., EPI noise) might result in such increases of the cognitive control exerted. We found that participants were more effective in controlling episodic retrieval of previous stimulus-response bindings (Experiment 1), in switching to a new task (Experiment 2), and shielding a current goal from distracting response tendencies (Experiment 3) if they were exposed to challenging task situations, such as 70 dB echo planar imaging noise sampled from an fMRI scanner. These findings have considerable theoretical implications, in questioning the widespread assumption that people are equally devoted to easy and more challenging tasks, and methodological implications, in raising the possibility that experiments carried out in fMRI scanners or under otherwise challenging conditions systematically overestimate contributions from cognitive control processes. fMRI noise and cognitive control 3 For about two decades, increasingly many neuroscientific studies have been using fMRI to investigate cognitive processes, a method that provides restricted temporal but highly resolved spatial information about the brain areas involved in a given mental operation or task. It seems obvious to most researchers that the findings obtained in an fMRI scanner can be straightforwardly related to findings obtained outside the scanner; indeed, some authors have even refrained from collecting behavioral measures inside the scanner and directly related performance measured outside the scanner to fMRI results (e.g., Marois, Chun, & Gore, 2000). However, it is obvious that testing situations differ considerably between an fMRI environment and a behavioral laboratory setting. Moreover, recently the scanning environment has been shown to increase endocrinological stress responses (e.g., heightened cortisol levels) especially for scanner-naive (inexperienced) participants (e.g., Tessner, Walker, Hochman, & Hamann, 2006) and in adolescent participants (Eatough, Shirtcliff, Hanson, & Pollak, 2009). Apart from many setting-related differences, one particularly salient characteristic of an fMRI environment to often sound attenuated behavioral testing environments is the high level of noise created by the fast echo planar imaging (EPI) pulse sequences used in magnetic resonance imaging (MRI/fMRI; Okada & Nakai, 2003). Most people find this noise challenging, disturbing and/or annoying. Therefore, one may wonder in as much extensive EPI noise levels might represent an additional challenge or difficulty that potentially biases the cognitive performance itself. That this is a realistic possibility was suggested to us by two observations. First, high (loud) EPI noise has been shown to be a particularly effective distractor even in a visual task and to change brain activity substantially: Higher noise produces increased change blood oxygenation level dependent (BOLD) responses fMRI noise and cognitive control 4 bilaterally in temporal, occipital, and prefrontal cortices, and the cerebellum, and decreased BOLD responses (i.e., smaller signal changes) bilaterally in the anterior cingulated cortex (ACC) and the putamen. This has been claimed to support the idea that attentional networks are more strongly recruited to compensate for interference due to increased scanner noise (Tomasi, Caparelli, Chang & Ernst, 2005). If the tonic baseline activation of control areas is elevated, so the consideration, less extra activation can be observed if control is temporarily recruited on more difficult trials. Consistent with that, a PET study has shown that scanner noise increases the regional cerebral blood flow in the ACC (Mazard et al., 2002). Second, we recently developed a (rather complicated) version of Kahneman, Treisman, and Gibbs’ (1992) preview task with face and house stimuli for use in an fMRI scanner (cf., Keizer, Colzato & Hommel, 2008). This task allows for the study of involuntary episodic retrieval of feature bindings upon the repetition of one or multiple visual features (Hommel, 2004). Comparing the behavioral findings obtained in pilot studies carried out inside and outside the scanner suggested that some of the retrievalrelated effects were smaller inside the scanner. Given that retrieval in this task is unnecessary and thus involuntary, this observation may suggest that the higher noise level inside the scanner led to a stronger engagement of control processes, which may have worked against retrieval. Indeed, involuntary retrieval is more pronounced in people low in fluid intelligence (Colzato, van Wouwe, Lavender, & Hommel, 2006a), in young children and elderly individuals (Hommel, Kray & Lindenberger, 2010), and in cannabis users (Colzato & Hommel, 2008), which are all groups that are impaired with respect to cognitive control in general and the ACCdorsolateral prefrontal cortex (DLPFC) circuit assumed to resolve cognitive conflict in particular. fMRI noise and cognitive control 5 Given these observations, it is possible that fMRI studies provide a distorted picture of cognitive processing with respect to both behavioral and imaging results. In particular, one might wonder whether high noise levels, likely representing a challenging, stressful, and/or annoying situation, might trigger compensatory responses in terms of increased attentional control thus, leading to performance improvements (e.g. Plessow, Fischer, Kirschbaum, & Goschke, 2010; Kofman, Meiran, Greenberg, Balas, & Cohen, 2006). If so, this would have important consequences as fMRI studies would systematically overestimate the amount of cognitive control exerted in a given task and they would show more activation of control-related brain areas than normal. But how and why would the scanner environment and scanner noise in particular result in increased amounts of attentional control? First of all, noise is known to generally increase arousal and stress levels. Unfortunately, despite several decades of research dedicated to the influence of noiseinduced stress and drive levels (arousal) on cognitive performance, to date results remain rather inconclusive (Loeb, 1986; Smith & Broadbent, 1985). This might be mainly due to substantial differences in procedures, inconsistent and often relatively vague definitions of noise and stress (e.g. continuous vs. phasic noise), and the observation that the effects of noise also critically depend on the intensity of the applied noise level (Broadbent, 1971, p.416). Using predominantly traditional Stroop tasks (Stroop, 1935) some authors, for example, found speeded responses and reduced interference when being stressed by loud bursts of white noise compared to unstressed controls without noise (O’Malley & Gallas, 1977; O’Malley & Poplawsky, 1971; see also Booth and Sharma, 2009 for a similar approach). Such findings have typically been interpreted in terms of an increased selectivity under stress. (Noise-induced) stress or high drive levels in general overload the cognitive system. As a consequence, the fMRI noise and cognitive control 6 attentional focus is adjusted to task-relevant processing thus, reducing interference by less relevant information (e.g., Callaway, 1959; Easterbrook, 1959; see also Chajut & Algom, 2003 and Wells & Matthews, 1994 for recent discussions). Although plausible, other authors demonstrated increased interference under noise-induced stress (e.g., Hartley & Adams, 1974) or increased anxious states (e.g., Eysenck, Derakshan, Santos, & Calvo, 2007), so that straight-forward conclusion from the noise literatures have to be rather handled with care. A second possibility of how scanner noise might lead to increases in attentional control can be traced back to early ideas of Hillgruber (1912). His ―difficulty law of motivation‖ says that the difficulty of an action is the motive for investing more effort and devoting more cognitive control to reach the task goal. He assumed that increasing task difficulty automatically (―drive-like‖) increases will power without conscious deliberation (cf., Ach, 1935), an idea that has lived on in several disguises (e.g., Kahneman, 1973; Kukla, 1972; Sanders, 1983). One particularly influential disguise is currently under lively debate, which relates to the question of how people learn from or even avoid stimulus-induced action errors. Based on modeling work and neuroscientific observations, Botvinick, Braver, Barch, Carter, and Cohen (2001) suggested that registering a conflict or registering its aversiveness (Botvinick, 2007) leads to a stronger focus of attention on task-relevant stimuli or stimulus dimensions, so that distractorinduced response conflict (such as in a Stroop or flanker task) can be minimized or avoided on the next occasion. However, Botvinick et al. were mainly interested in explaining trial-to-trial effects—i.e., in the question of how we learn from previous conflict—but recent observations by Egner and Hirsch (2005) suggest that registering conflict may even have immediate attentional consequences. In particular, these authors obtained evidence that ACC-mediated conflict detection leads to a stronger activation of fMRI noise and cognitive control 7 task-relevant cortical representations induced by signals from the DLPFC, which is assumed to translate the action goal into top-down support for goal-related processes (e.g., Desimone & Duncan, 1995; Miller & Cohen, 2001). In a sense, then, there is both behavioral and neuroscientific evidence for Hillgruber’s (1912) claim that increasing the challenge of the task spontaneously increases one's effort to compensate for and to overcome that challenge. In such a conception, however, noise would represent or be perceived as a challenge and/or aversive signal which, following Botvinick (2007), is capable of triggering attentional adjustments (e.g., increased attentional control). However, up to now only few studies have looked into the particular impact of scanner noise on brain activation (e.g., Elliott, Bowtell & Morris, 1999; Tomasi et al., 2005) and, to our knowledge, no evidence has been reported that scanner noise would affect behavioral measures. Accordingly, we set out to provide an empirical test that scanner noise affects behavior in general and cognitive-control processes in particular, and that it does so in a way that fits with Hillgruber's law, namely, that it increases the exertion of cognitive control. Given that fMRI environments differ from behavioral laboratories in more ways than just noise, which may create confounds, we carried out all experiments outside the scanner in the same lab and just manipulated the presence or absence of previously recorded EPI noise. We used three tasks to look into the impact of this noise. First, we attempted to formally verify our informal observation that involuntary episodic retrieval is less pronounced in the presence of noise (Experiment 1). Second, given that our preview task cannot be considered to provide a process-pure measure of cognitive control, we used two further paradigms that are commonly assumed to tap into cognitive control processes. In particular, we implemented a more standard taskswitching design to investigate the flexibility of task set shifting (Experiment 2) and a fMRI noise and cognitive control 8 Simon task to investigate the shielding of the current goal from competing response tendencies (Experiment 3).