|Home | About | Journals | Submit | Contact Us | Français|
Behavioral regulation is a hallmark feature of executive functioning (EF). The present study investigated whether commonly used neuropsychological test measures of EF (i.e., working memory, Stroop, trail making, and verbal fluency) were related to ability to down-regulate emotion: (a) spontaneously, and (b) when instructed to suppress emotional expressions. To ensure a wide range of EF, 24 frontotemporal lobar degeneration patients, 7 Alzheimer’s patients, and 17 neurologically normal controls participated. Participants were exposed to an acoustic startle stimulus (single aversive noise burst) under three conditions: unwarned, warned with no instructions (to measure spontaneous emotion down-regulation), and warned with instructions to suppress (to measure instructed emotion down-regulation). Results indicated that higher verbal fluency scores were related to greater emotion regulation (operationalized as reduction in body movement and emotional facial behavior when warned of the impending startle) in both regulation conditions. No relationships were found between emotion regulation in these conditions and the other EF measures. We conclude that of four commonly used measures of EF, verbal fluency best indexes the complex processes of monitoring, evaluation, and control necessary for successful emotion regulation - both spontaneously and following instructions to suppress.
The ability to monitor and modify ongoing behavior is essential to many cognitive and emotional functions. A set of psychological processes collectively called executive functioning (EF) have been associated with coordinating perceptual and motor processes in the service of behavioral goals (Lezak, Howieson, Loring, Hannay, & Fischer, 2004; Norman & Shallice, 1986; Smith & Jonides, 1999). These processes allow us to do such things as participating in a conversation while we have something else on our minds, modifying routine responses, learning new tasks, noticing and correcting inappropriate behavior, and regulating our emotional responses (Posner & Rothbart, 2000; Stuss & Alexander, 2000; Zelazo & Cunningham, 2007).
It is well documented that deficits in EF can result in behaviors becoming repetitive and stimulus bound, and in difficulties successfully integrating perceptual information in cognitive tasks (Baddeley, 1986; Stuss & Levine, 2002). However, the role of EF deficits in emotion regulation is less well-understood. Emotion regulatory processes are goal directed behaviors, functioning to modify dynamic features of emotion, such as the magnitude and duration of behavioral (i.e., expressive), experiential, and physiological responses (Thompson & Gross, 2007). Successful emotion regulation draws heavily on EF in realms such as anticipating outcomes, planning, and executing responses (e.g., Banfield, Wyland, Macrae, Münte & Heatherton, 2004; Denckla, 1996). For example, in order to down-regulate manifestations of fear in response to a threatening stimulus, one has to integrate perceptual cues, anticipate one’s responses to these cues, devise an action plan (e.g., keep one’s breathing steady and facial muscles immobile), and continuously monitor and adjust ongoing behavior.
Indeed, studies using neurologically intact populations show that EF is related to the ability to modulate emotionally laden responses, such as reducing prejudiced behaviors (von Hippel, Silver, & Lynch, 2000), reducing biased opinions (Payne, 2005), refraining from expressing disgust in a socially unacceptable setting (von Hippel & Gonsalkorale, 2005), and delaying gratification (Eigsti et al., 2006). Because EF are thought to be related to the integrity of the frontal lobes (e.g., Royall, et al., 2002) lesion studies that report diminished emotion regulatory functioning among patients with circumscribed focal damage also support the idea that EF is related to emotion regulation. Specifically, researchers suggest that frontal lobe damage and resultant EF deficits compromise abilities to integrate emotional cues into decision making (e.g., Eslinger & Damasio, 1985) and to evaluate the relevance of emotional cues to the task at hand (Rule, Shimamura & Knight, 2002).
Recent fMRI research also implicates frontal brain structures, linked to EF, in the conscious control and cognitive reappraisal of emotional stimuli (e.g., Beauregard, Levesque, & Bourgouin, 2001; Ochsner, Bunge, Gross, & Gabrieli, 2002, Ochsner et al., 2004). Importantly, two putatively different emotion regulation strategies - suppression of emotional expression and reappraisal of emotional eliciting stimuli - have been found to recruit largely overlapping frontal lobe structures in medial and bilateral dorsolateral PFC and lateral OFC (Goldin, McRae, Ramel, & Gross, 2008).
Taken together, the research reviewed above suggests a tight link between EF and emotion regulatory ability. However, no direct evidence connecting EF and emotion regulation is available to date. Therefore, the purpose of the current research was to address this gap by investigating whether EF is related to performance on two established laboratory tests emotion regulation (a) the ability to down-regulate spontaneously, and (b) the ability to down-regulate when instructed to suppress emotional expressions. Next, we turn to an overview of EF measurement.
The need for reliable and valid measures of EF has led to the development of a number of neuropsychological tests that putatively assess dissociable facets of EF (Miyake et al., 2000; Royall et al., 2002). Research shows that although EF tests tend to be inter-correlated (Miyake et al., 2000), they measure different processes. In the present study, we selected tests commonly used to measure four different facets of EF: (1.) working memory capacity, (2.) inhibition, (3.) task switching, and (4.) cognitive flexibility—to determine their relationship to a laboratory-based battery of emotion regulation ability (Levenson, 2007).
Working-memory refers to a set of processes that enables maintenance and manipulation of information in short-term memory in the service of a particular goal (Baddeley, 1986; Norman & Shallice, 1986). Working memory is measured by having individuals memorize, and in their head manipulate groups of items (e.g., letters or numbers). Working memory might aid emotion regulation by enabling storage and manipulation of perceptual and contextual information. Inhibition refers to the ability to suppress one behavioral response in favor of another (Cohen, Dunbar, & McClelland, 1990). It is typically assessed by an interference procedure in which subjects are instructed to respond to one aspect of a stimulus (e.g., ink color) and not to another more perceptually salient aspect (e.g., color word) and therefore inhibit a pre-potent response from occurring. Inhibition plays a role in emotion regulation by enabling suppression of the more automatic aspects of emotional responding. Task switching refers to the ability to redirect attention quickly between tasks. It is typically measured by having individuals engage in activities with alternating instruction sets (e.g., tracing mazes by alternating between letters and numbers). Task switching is important in emotion regulation because of the need to switch attention between various competing stimuli (e.g., the emotion-eliciting cue versus what kind of emotional expression is appropriate). Finally, cognitive flexibility (Lezak et al., 2004; Ruff, Light, Parker, & Levin, 1997) is the ability to generate responses quickly in keeping with predetermined criteria. Cognitive flexibility is typically measured with fluency tasks (Lezak et al., 2004) that require the subject to continually (e.g., for 1 minute) generate lists of words or objects that fit some criteria (e.g., words starting with the letter A). Cognitive flexibility is important for emotion regulation because of the need to remember rules governing appropriate emotional expression, devise regulatory plans, maintain and monitor emotional responding.
The present study was designed to determine whether the four aspects of EF reviewed above (working memory, inhibition, task switching, and cognitive flexibility) are related to the ability to down-regulate emotional responding either spontaneously or following instructions to suppress emotional responses. EF was measured using standard neuropsychological testing procedures. Emotion regulation was measured in the laboratory using procedures that generate strong emotions and create conditions designed to assess ability to down-regulate emotional response (Levenson, 2007). Although prior studies have found that certain EF tests predict emotion and self-regulatory success in socially complex situations (Payne, 2005; von Hippel & Gonsalkorale, 2005; von Hippel et al, 2000), there have been no previous examinations of the relationship between EF tests and the ability to down-regulate emotional responses.
To ensure that participants had a wide range of EF, we included patients with neurodegenerative brain diseases - frontotemporal lobar degeneration (FTLD) or Alzheimer’s disease (AD) - as well as neurologically normal controls. Deficits in EF have been documented in dementing disorders that impact the frontal lobes (Baddeley, 1986; Duke & Kaszniak, 2000). FTLD is characterized by neuronal loss in the frontal and anterior temporal lobes (Kertesz, Davidson, & Munoz, 1999; Neary et al., 1998). Although degeneration in AD typically starts in medial temporal areas, it often spreads to include the frontal lobes (Tikofsky, Hellman, & Parks, 1993). Thus, patients with these dementias were expected to manifest varying degrees of EF deficits as compared to controls.
The emotion regulation battery used a very simple emotion elicitor, an aversive acoustic startle stimulus, so that the dementia patients would not have difficulties with interpretation and processing (Levenson, 2007). This stimulus is known to produce a strong, defensive response consisting of: (a) reflexive body movements that serve to protect the torso and head, and (b) emotional facial behaviors including surprise and fear (Ekman, Friesen, & Simons, 1985; Hagemann, Levenson, & Gross, 2006; Keltner & Ekman, 1996; Sturm, Rosen, Allison, Miller, & Levenson, 2006). We and others have used similar startle stimuli in studies of emotional reactivity and regulation (e.g., Ekman et al., 1985; Hagemann et al., 2006; Roberts et al, 2004; Soto, Levenson, & Ebling, 2005).
In the present study we presented the acoustic startle stimulus under three different instructional conditions designed to create different levels of demand for down-regulating emotional behavior and thus test different emotion regulatory strategies. In an unwarned condition, the startle appeared without warning and participants were not given any instructions about regulating their emotional response. This condition did not impose regulatory demands and was included to measure differences in baseline reactivity to the stimuli. In a warned only condition, the startle appeared at the end of a 20-sec countdown period and, again, participants were not given any instructions about regulation their emotional response. This condition was included to tap into spontaneous attempts to regulate emotions (Ekman, et al., 1985) and was designed to mobilize the participants own choice of regulatory strategy. In a warned with instructions to suppress condition, the startle appeared at the end of a 20-second countdown period and participants were explicitly instructed to hide their emotional responses.
Based on the role that EF plays in emotion regulation, we hypothesized that deficits in EF would be associated with poor ability to down-regulate emotional responding in all groups (patients and controls). We did not expect to find this relationship in the unwarned condition, but did expect to find it in the two warned conditions (the warning period provides ample opportunity for formulating and activating regulation strategies of one’s choice and to formulate plans to suppress emotional behavior). By including the two different warned conditions, we were able to examine the relationship between EF and two kinds of emotion regulation. In the warned only condition, we were able to assess spontaneous emotion regulation in that participants were not explicitly told to down-regulate. Further, if they chose to down-regulate they were free to do so in any way they wanted. In the warned with instructions to suppress condition, we were able to assess the ability to comply with instructions to down-regulate in a particular way (using expressive suppression; Gross & Levenson, 1993). Existing research provided a basis for predicting that EF measures would be related to the ability to down-regulate both spontaneously (e.g., von Hippel & Gonsalkorale, 2005) and the when instructed to suppress (e.g., Goldin et al., 2008).
Given the lack of research on the connection between laboratory tests of emotion regulation and EF, we did not make any strong predictions about which particular EF tests would be most strongly related to down-regulatory ability. However, given the complexities involved with down-regulating emotion in this kind of semi-naturalistic situation, we speculated that tests of EF that reflected more complex processing (e.g., verbal fluency) would show the strongest relationships with down-regulatory ability.
FLTD (N = 24) and AD (N = 7) patients were recruited through the Memory and Aging Center at the University of California San Francisco (UCSF). Neurologically normal control participants (N = 17) were selected through newspaper ads. Participants were evaluated using clinical interviews, questionnaires, MRI structural brain scans, and neuropsychological measures (including EF tests). FTLD diagnoses were determined using the Neary clinical criteria (Neary et al., 1998). AD diagnoses were determined using the National Institute of Neurological and Communication Diseases and Stroke/Alzheimer’s Disease and Related Disorders Association criteria (McKhann, et al., 1984). Control participants were free of neurological and psychiatric symptoms. Demographics and neurological status of the participants are described in Table 1. Despite attempts to match the three groups on age, control participants in the sample were significantly older than FTLD patients; therefore, age was used as a covariate in all analyses.
A trained staff at UCSF administered the EF tests. Descriptive statistics and correlations among these measures are presented in Table 2.
To assess working memory, the Digit Span and its visual analogue, the Spatial Span subscales (forward and backward) of the Wechsler Adult Intelligence Scale (Version III, Wechsler, 1997) were administered. Scores were summed across the correctly completed trials on the digit and spatial span trials to create a single composite score, with higher scores indicating greater working memory capacity. Reliabilities of both the digit and spatial span subscales are high, having test-retest correlations of 83 and .71 respectively (Wechsler, 1997).
To assess inhibition, the Stroop task from the Delis-Kaplan Executive Function System (D-KEFS; Delis, Kaplan, & Kramer, 2001) was used. In this task, there are two separate trials; color words are either printed in incongruent ink colors (e.g., the word “red” printed in green ink) or in congruent ink colors (e.g., the word “red” printed in red ink) and the participant is asked to name the ink color. To control for individual differences in color naming speed, an overall inhibition score was created by predicting incongruent color word reading time from congruent color word reading time and saving the residuals.1 These residual scores were then used as a measure of inhibition with longer times, and higher scores indicating poorer inhibition. Reliability of the Stroop is reasonable, with test-retest correlations of .70-.79 (Delis et al., 2001).
To assess task switching, the D-KEFS Trail Making Test was used (Delis et al., 2001). In this task, participants are instructed to alternate between connecting letters and connecting numbers printed in scrambled order on a card (connect from “A” to “1” to “B” to “2”). If the participant made an error, the examiner pointed it out and instructed the participant to return to the correct location while the clock kept running. To hold individual differences in number and letter processing speed constant, participants also completed two trials where the task was to connect letters-to-letters and numbers-to-numbers. To control for individual differences on these latter trials, an overall task switching score was created by predicting time to completion on the numbers-to-numbers trials from the letters-to-letters and numbers-to-numbers trials and saving the resulting residuals. These residual scores were used a measure of task switching where shorter times and lower scores indicate better task-switching. Reliability of the Trail Making test is reasonable, with test-retest correlations of .70-.79 (Delis et al., 2001).
To assess cognitive flexibility, the Verbal fluency test (Delis et al., 2001) was used. On separate one-minute trials, participants were asked to generate words that began with the letters F, A, and S (excluding proper nouns and not repeating the same word with different suffixes). Verbal fluency was calculated as the total number of correct words produced across the three trials. More words produced indicate better cognitive flexibility. Reliability of the verbal fluency test is high, with a test-retest correlation of .90 (Delis et al., 2001).
Means on the EF variables2 are presented in Table 3 for the FTLD, AD, and normal controls. As expected, these show lower levels of EF for patients compared to controls. This was particularly the case for verbal fluency. Additionally, Stoop scores were lower among FTLD patients than AD patients and controls.
Two aspects of the motor response to the startle were measured: (a) somatic activity, and (b) facial expressive behavior.
General body movement associated with the startle reflex was continuously monitored with an electromagnetic transducer attached to a platform under the participant’s chair. The transducer generates an electrical signal proportional to the amount of movement in any direction and is very sensitive to the kinds of head and trunk movements that occur in response to the startle stimulus. The signal from the transducer was amplified using a Grass Model 7 polygraph, processed continuously (at 300 Hz sampling rate) and was averaged for the 5 seconds following the startle stimulus using software written by one of the authors (RWL). A number of autonomic nervous system responses were also monitored as part of the experimental procedure, but they are not included here because of our primary interest was in measuring regulation of visible motoric components of the startle response.
A front-view of the participant’s face was videotaped using a partially concealed, remotely-controlled camera. A team of research assistants coded facial expressive behavior from the videotapes using the Emotional Expressive Behavior Coding System (Gross & Levenson, 1993). Two coders who were blind to diagnostic status and experimental instructions viewed each tape without sound. Conditions were coded in randomized order. Coders rated the intensity of emotional expressions of anger, contempt, confusion, disgust, fear, happiness, embarrassment, interest, sadness and surprise on a second-by-second basis during the 5 seconds following the startle stimulus using a 4-point intensity scale (0: absent, 3: strong). Inter-coder reliability for the full set of codes was high (α = .79). To capture the wide variety of emotional expressions typically produced by the startle stimulus (Ekman et al., 1985; Keltner & Ekman, 1996), a measure of overall facial expressive behavior was computed by averaging across all categories of emotional expressions.
To measure emotion regulatory strategies, participants were asked an open-ended question at the end of the regulation trials. Specifically, in the warned only trial participants were asked whether they “did anything different knowing this time that the noise was coming?” After the warned with instructions to suppress trial, participants were asked to describe how they tried to hide their reactions. Participants’ responses were transcribed and categorized by one of the authors (AG) into five different categories: (a) no regulation; (b) suppression of physiological and behavioral response (e.g., “held breath, held facial muscles steady”); (c) modified sensory features of the noise either internally (“paid attention to the quality of the noise”) or externally (e.g., “covered head to buffer noise”); (d) cognitive reinterpretation (e.g., “thought of the noise as a rifle starting a race”); (e) no specific regulatory strategy (e.g., “just prepared”)].
Participants came to our laboratory at the University of California, Berkeley for a single experimental session scheduled within 3 months (M = 15.44 days, SD = 7.5) after their EF was assessed at UCSF3. On arrival, they signed the consent forms, were seated in a chair located 1.25 m from a 48-cm color television monitor, and had sensors attached for physiological monitoring. The full experimental protocol consisted of a series of tasks designed to assess emotional functioning (e.g., emotion eliciting films, recall of emotional memories; (Levenson, 2007). The present article focuses on the three startle trials where participants were presented with a loud noise (115-dB 100 ms long burst of white noise) administered through two loudspeakers located behind the subject’s head4. The sound can be likened to a “gunshot” and, as noted earlier, it produces a large reflexive defensive startle reaction (Hagemann et al., 2006).
Each of the three startle trials began with a 1-minute baseline period, during which the participant was instructed to view an “X” on the TV screen. On the first trial (unwarned), the startle occurred without warning at the end of the 1-minute baseline period. On the second trial (warned only), participants were told by the experimenter and on the television screen that the startle stimulus would occur at the end of a 20-second countdown presented on the television screen. Under these conditions, most people spontaneously down-regulate their emotional response, bracing for the upcoming aversive stimulus (Ekman et al., 1985). On the third trial (warned with instructions to suppress), participants were told that the startle stimulus would occur at the end of the countdown and that: “This time if you have any reaction as you hear the noise, try your best not to let your feelings show. Pretend that someone is watching you and you do not want them to be able to tell that you are reacting to the noise.”
We have utilized these three instructional conditions with the acoustic startle in a number of previous studies with patients and normals (Hagemann et al., 2006; Roberts et al., 2004; Soto et al., 2005; Sturm et al., 2006) and have generally used them in the same fixed order (unwarned, warned only, warned with instructions to suppress). To get a relatively pure measure of reactivity to the startle when it occurs without prior warning, the unwarned condition has to be administered first. For the two warned conditions, our desire to get an estimate of spontaneous regulation ability requires that the warned only condition come next (because the warned with instruction to down-regulate condition would cue the subject to use a suppression strategy on subsequent trials).
To assess the reliability of the emotion regulation battery, we examined data from a small sample (N = 9) of dementia patients and normal controls who returned to our laboratory and repeated the battery approximately one year after their initial visit. Reliabilities for the overall emotional responses (aggregated body movement and emotional facial response averaged over the three startle conditions) were quite high (the test-retest correlation was .74). Examining the three startle conditions separately, reliabilities were high for the unwarned and warned only conditions (test-retest correlations of .71 and .70 respectively), and moderate for the warned with instructions to suppress condition (the test-retest correlation was .55). Within the limits of our small reliability sample, we believe that these data suggest that, even over this relatively long time interval, the emotion regulation battery has quite reasonable levels of reliability.
For each startle trial, the somatic activity and facial expressive averages for the 5 seconds following the startle burst were normalized and an emotional responding composite score was calculated by averaging the two normalized values. This emotional responding composite was used as the criterion variable in all analyses. Statistical analyses were performed using the SAS statistical package (Version 9; SAS Institute Inc, Cary, NC). To examine the relationship between EF and emotional responding, data from all participants were pooled (this maximized variability in EF). Where significant relationships were found between EF and emotional responding, we followed up with an analysis that tested the interaction of diagnostic group and EF scores to determine whether diagnostic group moderated the relationship between EF and emotional responding.
To determine whether the experimental manipulation used to produce emotional responding was effective, we tested reduction in somatic activity and facial expressive behavior separately for the two warned trials compared to the unwarned trial using paired sample t-tests. Results for somatic activity showed reduction from the unwarned to the warned only condition (unwarned M = 3.37, SD = 2.33; warned only M = 2.07, SD = 1.97; t(47) = 5.16, p < .05) and from the unwarned to the warned with instructions to suppress trial (warned with instructions to suppress M = 1.49, SD = 1.24; t(47) = 5.76, p < .05). Results for facial expressive behavior showed reduction from the unwarned to the warned only trial (unwarned M = .67, SD = .57; warned only M = .43, SD = .54; t(47) = 2.80, p < .05) and from the unwarned to the warned with instructions to suppress trial (warned with instructions to suppress M = .27, SD = .50; t(47) = 3.85, p < .05). These results suggest that the unwarned condition elicited more emotional behaviors both on the face and in somatic activity than either the warned only or warned with instructions to suppress trials.
Open-ended responses were analyzed for both regulation trials to elucidate the type of emotion regulatory strategies used in each condition. In the warned only trial, 49 % of participants reported not using a strategy. Among those who reported using a strategy, 12.5 % reported using suppression, 16.67 % reported using attention distraction and 4.17% reported using cognitive reinterpretation and 66.67% did not provide detailed enough information to reliably categorize their emotion regulatory strategy. Thus, reflecting the lack of specific instructions this condition produced a wide range of reported strategies (and lack of). In contrast, in the warned with instructions to suppress trial 97% reported using a regulatory strategy. Of these, 56.82 % reporting using suppression, 20.45 % reported using attention distraction and 22.73% of the participants did not provide enough detail to categorize the emotional regulatory strategy reliably. Thus, in this condition most subjects appeared to follow the instructions and used the prescribed suppression strategy.
We hypothesized that EF would be related to emotional responding on the two warned trials, which created a demand for and provided time for down-regulation, but not to responses on the unwarned trial. Consistent with this prediction, we found no significant correlations between the four EF functions and emotional responding on the unwarned trial; partial correlations (controlling for age) were: Working memory r(46) = .09, ns; Stroop r(46) = .07, ns; Trail making r(46) = .00, ns; and Verbal fluency r(46) = -.01, ns.
We hypothesized that deficits in EF would be associated with deficits in emotion regulation on the warned trials. Because the warned only and warned with instructions to suppress trials differed in how explicit the demand for regulation was, we considered each separately5. Specifically, for each trial, we conducted a series of four separate regression analyses, each with one EF test as the predictor variable, our composite measure of emotional responding as the criterion, and age and emotional responding on the unwarned trial as covariates. Controlling for emotional responding on the unwarned trial enabled us to examine emotional behavior in the two regulation trials free of any individual differences in the magnitude of emotional responding (Ekman et al., 1985). To control for multiple statistical tests, we applied a Bonferroni adjustment and adopted a more conservative significance level of .025 for a priori one-tailed tests.
For the warned only trial, the only EF measure that predicted emotional responding with verbal fluency, such that higher verbal fluency scores were related to more reduction in emotional responding, F(1,45) = 7.77, p < .05, β = -.31). Results of these analyses for all four EF measures are presented in Table 4. No significant interactions were found between verbal fluency and diagnostic group membership (F < 1)6. Furthermore, we also examined the unique predictive ability of verbal fluency above and beyond the other three measures of EF, using a simultaneous regression analysis. This revealed that the relationship between verbal fluency and emotional responding in the warned only trial was attenuated some, but remained significant (F(1,41) = 4.92, p < .05, β = -.28) after the effects of the other three EF tests were held constant.
For the warned with instructions to suppress trial, again, the only EF measure that predicted emotional responding was verbal fluency such that higher verbal fluency scores were related to more reduction in emotional responding (F(1,45) = 7.97, p < .05, β = -.39). Results of these analyses for all four EF measures are presented in Table 4. No significant interactions were found between verbal fluency and diagnostic group membership (F < 1)7. Finally, we also examined the unique predictive ability of verbal fluency above and beyond the other three measures of EF using a simultaneous regression analysis. This analysis revealed that the relationship between verbal fluency and emotional responding on the warned with instructions to suppress remained significant when controlling for the other three measures of EF (F(1,41) = 8.48, p < .05, β = -.44).
Because outliers can unduly influence findings with small sample sizes, we examined regression scatter plots (verbal fluency versus residualized emotional responding after controlling for age and for emotional responding on the unwarned trial for both the warned only (Figure 1) and warned with instructions to suppress (Figure 2) trials. Both visual examination and formal outlier analyses (i.e., re-running the analyses after excluding any observation that had a Cook’s D greater than 4/N) indicated that outliers were not responsible for the findings.
This study tested the hypothesis that deficits in EF, which are believed to underlie goal directed behavior and regulation, would be associated with deficits in emotion regulation. Using a sample of patients (FTLD and AD) and neurologically healthy controls with a wide range of EF and an aversive acoustic startle stimulus known to produce sizeable emotional responding, we found support for this hypothesis both when the demand for regulation was implicit and participant used spontaneously recruited emotion regulatory strategies (our warned only condition) and when it was explicit and participants were instructed to suppress their responses (our warned with instructions to suppress condition). Importantly, there was considerable specificity to this finding in that: (a) EF only predicted emotional responding on the regulation trials, and not when the startle stimulus occurred unexpectedly without a warning; and (b) of four commonly used measures of EF -working-memory, Stroop, trail making, and verbal fluency - only deficits in verbal fluency (a measure of cognitive flexibility) predicted deficits in emotion regulation.
We had predicted that EF deficits would not be related to emotional responding to the unwarned startle. This prediction was based on the view that emotional reactivity to simple stimuli such as the acoustic startle are largely subserved by brain stem and limbic circuits (Davis, Gendelman, Tischler, & Gendelman, 1982), which are not assessed by EF measures. Additionally, we had predicted that EF deficits would be related to emotional regulation in the two regulation conditions, both when participants were free to use a regulatory strategy of their choice spontaneously, and in the suppression condition where participants are given explicit instructions to use suppression as emotion regulatory strategy. This prediction was based on the view that emotion regulation is subserved by frontal-subcortical circuits (Roberts, et al., 2004; Rolls, 2000; Stuss & Levine, 2002), which are assessed by EF measures. Together these findings are also supportive of a differentiated view of emotional circuitry in the brain, namely that different circuits are involved for having emotional responses than for regulating these responses.
The finding that among EF measures, only verbal fluency predicted emotion regulation may seem surprising, but it underscores the fact that putative measures of EF are not interchangeable. Rather, they capture clinically, functionally and anatomically different aspects of EF (Miyake et al., 2000; Royall et al., 2002). For example the Stroop test, which assesses inhibition of a dominant response, might have been viewed as the most likely candidate for predicting emotion regulation. The result that Stroop performance did not correlate with emotion regulation underscores that the ability to mount a successful emotion regulatory response to an aversive auditory stimulus, both spontaneously and to suppress when instructed, involves more than just response inhibition. In our view, successful emotion regulation draws on a complex set of skills, including strategy formulation, behavioral monitoring, evaluation, and adjustment. Viewed from this perspective, there are similarities between the process of EF as indexed by generating words of a certain kind on the verbal fluency task, and down-regulating an emotion, insofar as both tests require the individual to be able to devise an advantageous strategy, to monitor ongoing performance, and to inhibit erroneous responses (Lezak et al., 2004; Shimamura, 2000; Troyer, Moscovitch, & Winokur, 1997).
Although emotion down-regulation to acute aversive stimulus is arguably highly prototypical, there clearly are other kinds of emotion regulation and kinds of stimuli that occasion regulation. Additional research will be necessary to determine whether EF as measured by tests of verbal fluency is associated with the ability to engage in other kinds of emotion regulation (e.g., amplification, cognitive reappraisal and attentional control) and other kinds of stimuli (e.g., those that are more chronic, those that evoke different kinds of emotion).
In summary, were able to present evidence that the relationship between verbal fluency and EF was consistent across the three groups and across two different emotion regulatory conditions. Thus, we have confidence that these findings establish a true link between measures of verbal fluency and these two forms of emotion regulation. Furthermore, we demonstrated that the findings do not merely result from the measures associated with diagnosis, and we found that findings hold robust across two different emotion regulatory conditions.
This study bridges the domains of cognitive and emotional functioning using a novel methodology that links neuropsychological tests of EF with emotional regulation in both neurological patients and normal controls. Because EF was measured using performance-based neuropsychological tests and emotion regulation was assessed using laboratory measures of emotional facial behavior and somatic muscle activity, concerns about shared method variance were minimized.
The study has several limitations that may have affected our findings. We tested only a subset of the available tests of EF and only a single emotion elicitor. Thus, our findings might have differed with other tests (e.g., N-back, Wisconsin Card Sort) and other elicitors (e.g., films, pictures). To provide a wide range of EF, we included patients with two kinds of dementia and neurologically-normal participants. Thus, our findings may not generalize to other neurological populations. Additionally, because the three startle conditions occurred in a fixed order, we cannot rule out order or habituation effects. Furthermore, individual patterns of neurodegeneration have not been characterized for the patients in this sample, thus, we cannot directly determine relationships between regional neural loss and deficits in our EF and emotion regulation measures. The fact that no diagnostic group differences emerged in our analyses could be due to small sample size, especially in the AD group. Finally, as noted above, our results establish a link between verbal fluency and the ability to down-regulate the emotional response to an acute aversive stimulus spontaneously and when instructed to suppress expressive responses. Thus, we cannot know the extent to which these findings generalize to other kinds of emotion regulation and to other emotional stimuli.
This study provided the first evidence that an aspect of EF (cognitive flexibility as measured by a test of verbal fluency) predicts the ability to down-regulate emotional responses both when the demand was implicit and when it was explicit to suppress. This findings provide a behavioral counterpoint to imaging studies that show important inhibitory relationship between frontal circuits and key emotion centers in the brain (e.g., medial prefrontal and amygdala circuits, Ochsner et al., 2002). The striking specificity of these findings, insofar as: (a) EF predicted both spontaneous emotion regulation and instructed suppression when participants knew when the stimulus would occur, but did not predict emotional responding when they did not; and (b) only a test of verbal fluency, and not tests of other aspects of EF including working memory, inhibition, and task-switching, predicted emotion regulation underscores the precise nature of these relationships and the importance of using differentiated measures of EF and emotional functioning.
This research was supported by the National Institute on Aging Grants AG17766 and P01-AG19724 (subcontract) and National Institute of Mental Health Grant T32 MH20006 awarded to Robert W. Levenson, National Institute on Aging Grants 1K08AG02076001, AG10129, P50-AG05142, and AG16570 awarded to Bruce L. Miller, and the Alzheimer’s Disease Research Center of California.
We would like to thank the members of the Berkeley Psychophysiology Laboratory and the Memory and Aging Center for their help in conducting this research.
1There are a variety of ways used to calculate interference (Jensen & Rohwer, 1966). We chose residualized scores over other methods because this resulted in a distribution that best approximated normality.
2Examination of the distribution (SD and variance) and skewness, a measure of asymmetry, suggest that all four measures of EF used in the current analyses approximate normality and have sufficient spread.
3Repeating our major analyses using the days between the two assessments as a covariate did not change any of the reported findings.
4The paradigm employed here is different from the emotion modulated startle paradigm, which uses repeated presentations of a lower amplitude noise (typically 95-105 dB).
5We considered the possibility that forewarning Ps about the impeding startle noise might have interacted with EF and diagnosis. Thus, we created facial behavior and body movement composites scores for the 20 second-countdown leading up to the startle in the warned only and warned with instructions to suppress trials. We then conducted a set of analyses similar to those reported in the main analyses with these new composite scores as the dependent measures. These analyses confirmed the general findings reported here, namely that higher verbal fluency was related to lower levels of facial behavior and body movement during the anticipatory period. No relationship was found for the other three measures of EF and there were no interactions with patient status or diagnosis.
6Controlling for diagnostic group membership attenuated the relationship between verbal fluency and emotional responding on the warned only trial (F(5,42) = 3.35, p = .07).
7Controlling for diagnostic group membership reduced the relationship between verbal fluency and emotional responding on the warned with instructions to suppress trial to non-significant levels (F(1,42) = 2.73, ns).