Power and Executive Functions 1
Running head: POWER AND EXECUTIVE FUNCTIONS
Lacking Power Impairs Executive Functions
Pamela K. Smith
Radboud University Nijmegen
Nils B. Jostmann
VU University Amsterdam
Adam D. Galinsky
Wilco W. van Dijk
VU University Amsterdam
IN PRESS, PSYCHOLOGICAL SCIENCE
Power and Executive Functions 2
The current research explores whether lacking power impairs executive functioning. The authors hypothesized that the cognitive presses of lacking power make individuals more vulnerable to performance decrements during complex executive tasks. In three experiments, low power impaired performance on executive function tasks, demonstrating that the powerless are less effective at updating (Experiment 1), inhibiting (Experiment 2), and planning (Experiment 3). Existing power research suggests that the powerless have difficulty distinguishing between what is goal-relevant and -irrelevant in the environment. A fourth experiment establishes that executive function impairment by low power is driven by goal neglect. The authors suggest that the cognitive alterations of lacking power may help foster stable social hierarchies and discuss how empowering employees may reduce costly organizational errors.
Keywords: social power, executive functions, goal neglect
Power and Executive Functions 3
Lacking Power Impairs Executive Functions
Societies are structured around social hierarchies, with some individuals and groups achieving positions of power and dominance over others (cf. Pratto, Sidanius, & Levin, 2006). These social orders are often rooted in immutable characteristics such as race and sex, which is unfair and ineffective because talented members of disadvantaged groups are prevented from moving into positions of power. Many contemporary societies, in response to this injustice, have shifted from hierarchies based on aristocracy to ones based on meritocracy, with high achievers filling more powerful positions than low achievers.
An implication of meritocracies is that those who lack power are low achievers because they are less capable or less motivated than those who acquire power. In the present research, we challenge this assumption. We propose that powerless people often achieve less because lacking power itself fundamentally alters cognitive functioning, and makes individuals more vulnerable to performance decrements during complex, executive tasks.
Power and Executive Functions
The powerless face a world of threats and uncertainty (Keltner, Gruenfeld, & Anderson, 2003). They must wait for instructions before they can act (Galinsky, Gruenfeld, & Magee, 2003) and also attempt to discern the goals of the powerful. Even when the powerless can act, they often cannot fully commit to action, but must be prepared to change course if their superiors’ goals change. As a result, the powerless must constantly engage in perspective-taking (Galinsky, Magee, Inesi, & Gruenfeld, 2006) and be vigilant of their environment.
Existing power research provides tentative evidence that low power fundamentally alters an individual’s mental world. Low-power individuals focus on the details at the
expense of the “bigger picture” (Smith & Trope, 2006). They are less cognitively flexible (Guinote, 2007a), attending to both peripheral and central attributes in the environment, and
Power and Executive Functions 4
fail to distinguish between what is goal-relevant versus -irrelevant about a stimulus (Overbeck & Park, 2001, 2006). In addition, low-power individuals from both human (Keltner et al., 2003) and animal populations (Shepherd, Deaner, & Platt, 2006) tend to be more vigilant than high-power individuals. Such heightened self- and other-monitoring impairs executive functions, as demonstrated in research on the cognitive stress of interracial interactions (Richeson & Shelton, 2003).
Because of these cognitive changes, the powerless may be less successful on difficult tasks, consistent with research on stereotype threat (Steele, Spencer, & Aronson, 2002). Members of stigmatized groups whose low status is made salient display worse self-control (Inzlicht, McKay, & Aronson, 2006) and decreased performance, partially by impairing working memory (Beilock, Rydell, & McConnell, 2007; Schmader & Johns, 2003). Indeed, neurophysiological correlates of low power (i.e., low levels of serotonin; Moskowitz, Pinard, Zuroff, Annable, & Young, 2001; Raleigh, McGuire, Brammer, & Yuwiler, 1984) also correlate with worse performance during complex tasks (Park et al., 1994).
We suggest that low power causes performance deficits because being powerless impairs executive functions. Executive functions reflect an attentional control mechanism that coordinates various cognitive subprocesses such as the updating of goal-relevant information and the inhibition of goal-irrelevant information (cf. Engle, 2002; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000). Executive functions are necessary for goal-directed behavior, allowing individuals to remain goal-directed despite interference and distraction (cf. Shah, Friedman, & Kruglanski, 2002). Thus, losing goal focus often reflects an insufficiency of executive functions, a situation referred to as goal neglect (Duncan, Emslie, Williams,
Johnson, & Freer, 1996; cf. Jostmann & Koole, in press; Kane & Engle, 2003).
The current research sought to establish that lacking power impairs executive functions. Although executive functions are considered to reflect a general attentional control
Power and Executive Functions 5
mechanism (Engle, 2002), the quality of executive functions can have a variety of manifestations (Miyake et al., 2000). Executive functions are reflected in cognitive subprocesses like updating and inhibiting, as well as in performance on more complex executive tasks like planning, which itself relies on updating and inhibiting. Thus, Experiments 1-2 explored whether the powerless are less effective at updating by using a 2-back task (Experiment 1), and inhibiting by using a Stroop task (Experiment 2). Experiment 3 tested whether the powerless are less effective at planning by using a Tower-of-Hanoi task. Finally, Experiment 4 examined general attentional control deficits among the powerless. Using variations of an inhibition task (e.g., Stroop), which has previously been employed to demonstrate goal neglect (Jostmann & Koole, in press; Kane & Engle, 2003), we tested whether lacking power leads individuals to have difficulty maintaining goal focus.
Experiment 1 examined the effect of power on the executive function of updating. Updating involves monitoring whether information is relevant for a present goal: new information is monitored for relevance, and relevant information replaces old, irrelevant information in working memory. We used a 2-back task (Braver et al., 1997) because it requires participants to update working memory constantly to respond accurately. We predicted that low-power participants would make more errors than high-power participants. Method
Participants were 102 students from a Dutch university. They received ?3 for
participating. Six participants were dropped from analyses: four for suspicions and two for extreme 2-back performance (more than 3 SD from mean). Overall, 95 participants (65
1females) were analyzed.
Using a procedure adapted from Richeson and Ambady (2003), participants were assigned to be either a superior or a subordinate in a computer-based task. They were told
Power and Executive Functions 6
that the superior would direct, evaluate, and monetarily reward the task performance of the subordinate.
The computer-based task was the 2-back task. Participants were told they would first complete the task separately to obtain an accurate baseline measure of team performance before working on the task interactively with their partner. In reality, they only completed the 2-back task once, which served as our dependent measure.
In the 2-back task, participants viewed a series of letters and were instructed to indicate, as quickly and accurately as possible, whether the current letter matched the letter shown two trials previously. In each trial, a black letter was presented in the center of the white screen for 500 ms, followed by a blank screen for 2000 ms. Participants were told to indicate during this 2500 ms interval whether the letter matched the one shown two trials previously (target trial), or not (nontarget trial).
Participants completed 20 practice trials (7 targets, 13 nontargets) with accuracy feedback before the actual task. The task consisted of 120 trials without feedback, divided into 4 blocks of 10 target and 20 nontarget trials.
Finally, participants completed manipulation checks of power and how much effort they put into the 2-back task and perceptions of their performance. At the end of this and all subsequent experiments, participants were probed for suspicion and debriefed. Results
Low-power participants (M = -1.02, SD = 1.98) indicated they had less relative power
22than high-power participants (M = 2.30, SD = 1.49), F(1, 93) = 84.48, p > .99, η = .48. repp
Power conditions did not differ in effort or perceived performance on the 2-back task, Fs <
4Accuracy in the 2-back task was assessed with error rate (e.g., Friedman & Förster, 2005) and d’ (e.g., Gray & Braver, 2002). d’ was calculated using the loglinear approach
Power and Executive Functions 7
(Stanislaw & Todorov, 1999) to include participants with hit or false-alarm rates of 0 or 1. Analyses were only based on trials in which participants responded (Wacker, Chavanon, & Stemmler, 2006). Low-power participants (M = 0.09, SD = 0.05) had a higher error rate than
2high-power participants (M = 0.07, SD = 0.04), F(1, 93) = 4.90, p = .91, η = .05. Low-repp
power participants (M = 2.68, SD = 0.59) were also less sensitive in terms of d’ scores (M =
23.02, SD = 0.71), F(1, 93) = 6.50, p = .945, η = .07. repp
Participants in a low-power role performed worse on a 2-back task, a standard executive function measure of updating, than participants in a high-power role. Although these results support our hypothesis, the power manipulation allows for an alternative explanation: Low-power participants may have been preoccupied with their impending evaluation and this evaluation concern might have driven our results. To address such potential confounds, in the remainder of our experiments we manipulated power via priming. Priming power has been shown to manipulate a sense of power and has produced similar results as actual role assignments (Galinsky et al., 2003).
Additionally, the high-power role may have improved participants’ executive function (Smith & Trope, 2006), rather than a low-power role impairing it. Because Experiment 1 only used low- and high-power conditions, we cannot be certain of the direction of the effects. The remaining experiments include a control condition to resolve this ambiguity.
Experiment 2 examined the effect of power on the executive function of inhibition. Inhibition involves the suppression of unwanted and/or irrelevant responses that may interfere with a present goal. We used a Stroop (1935) task as our dependent measure because it requires maintaining the goal of naming the color of words and inhibiting the prepotent tendency to read them (MacLeod, 1991). We predicted that low-power-primed (LPP)
Power and Executive Functions 8
participants would show more Stroop interference than high-power-primed (HPP) and control participants.
Participants were 77 students from a Dutch university who received course credit or ?3 for participating. Five participants were dropped from analyses: four for extreme Stroop performance (more than 3 SD from mean) and one for not following directions. Overall, 72 participants (65 females) were analyzed.
Participants first completed a 17-item scrambled sentences priming task (Smith & Trope, 2006). For each item, participants had to use four out of the five listed words to make a grammatically correct sentence. For LPP participants, 9 items contained a word related to lacking power (e.g., subordinate, obey). For HPP participants, those same 9 items contained a word related to having power (e.g., authority, dominate). For the control prime, all 17 items contained only power-irrelevant words.
In the Stroop task that followed, participants were instructed to indicate, as quickly and accurately as possible, whether each of a series of letter strings was written in red or in blue ink. Participants were instructed to ignore the meaning of the words and to focus on the ink colors only. Each trial started with a 1-s fixation asterisk in the center of the screen, immediately followed by a colored letter string. Participants responded to the string by indicating if it was in blue ink or in red ink. A 2-s blank screen appeared in between trials.
Participants first completed 10 practice trials with accuracy feedback after each trial. The actual task followed, consisting of 120 trials without feedback. There were 40 congruent trials (i.e., RED in red or BLUE in blue), 40 neutral trials (i.e., XXXX in red or blue), and 40
incongruent trials (i.e., RED in blue or BLUE in red), presented in random order.
Power and Executive Functions 9
Stroop interference is typically assessed by contrasting performance on incongruent trials with performance on neutral trials. Error rates were entered into a 3 (Power: low power, control, high power) x 2 (Trial Type: incongruent, neutral) mixed-model ANOVA, with the second factor within subjects (see Table 1). Participants made more errors on incongruent
2trials than on neutral trials, indicating a robust Stroop effect, F(1, 69) = 20.82, p > .99, η repp
2= .23. This was moderated by a significant 2-way interaction, F(2, 69) = 3.63, p = .91, η repp
= .10. Power did not affect performance on neutral trials, F < 1, but did affect performance on
2incongruent trials, F(2, 69) = 4.01, p = .91, η = .10. LPP participants made more errors on repp
incongruent trials than either control participants or HPP participants, ps > .90, with the rep
latter groups not differing, p = .43. Participants primed with low power showed more rep
difficulty with inhibition than both participants primed with high power and control participants.
Experiment 3 extends the results of the previous two experiments by testing the more complex executive ability to plan. Planning involves continuous switching between the main goal and subgoals and thus requires people to regularly update their current goal focus and to inhibit currently irrelevant (sub-)goals (cf. Miyake et al., 2000). We used the Tower-of-Hanoi (TOH) task, which involves moving an arrangement of disks from a start position to a goal position in as few moves as possible (Goel & Grafman, 1995). TOH trials vary in whether it is functional to move disks temporarily away from their final peg position. As a result, optimal performance on the TOH sometimes requires noticing and then resolving conflict between the goal (i.e., to move disks toward their final position) and the subgoal (i.e., to move disks temporarily away from it). Our version of the TOH involves trials varying in whether goal-subgoal conflict resolution is required (Morris, Miotto, Feigenbaum, Bullock, & Polkey, 1997). We predicted that LPP participants would have more difficulty in resolving
Power and Executive Functions 10
goal-subgoal conflict on the TOH than HPP and control participants. That is, LPP participants should make more errors, requiring more moves to solve conflict trials, relative to no-conflict trials.
Participants were 85 students (47 females) from a Dutch university, who received ?5
Participants started with a practice TOH. They subsequently engaged in a writing task used to prime the experience of power (Galinsky et al., 2003). LPP participants wrote about a time when someone had control over them, HPP participants about a time when they had control over others, and control participants about what they did yesterday. Afterwards they
5completed the actual TOH, followed by manipulation checks of power.
TOH task. We used a computerized TOH (Morris et al., 1997). In each trial,
participants saw two disk-rod sets, each consisting of three vertical rods and three different-sized disks placed on the rods. Participants had to rearrange the bottom set (the “start position”) so it looked like the top set (the “goal position”). They could only move one disk at a time and could not place a larger disk on top of a smaller disk. Moving a disk required two clicks: one to select a disk and one to indicate to which rod it should be moved. Participants worked on each trial until the start position matched the goal position.
Participants started with a warm-up trial and then continued with four experimental trials. For each trial, the computer counted the number of meaningful clicks (i.e., clicks leading to the selection or movement of a disk) and measured the time that passed before each click.
Each trial required a minimum of four moves to be solved but varied in complexity. The first two trials were no-conflict trials. Here a simple, effective strategy was to move the
first disk immediately into the direction of its final goal position. Thus, the subgoal (i.e., the