Can ‘risky’ impulsivity explain sex differences in aggression?

5
Can ‘risky’ impulsivity explain sex differences in aggression? Anne Campbell * , Steven Muncer Durham University, Psychology, South Road, Durham DH1 3LE, United Kingdom article info Article history: Received 9 January 2009 Received in revised form 24 March 2009 Accepted 6 April 2009 Available online 7 May 2009 Keywords: Aggression Impulsivity Sex differences abstract Impulsivity is a heterogeneous concept and sex differences are most apparent on those inventories that sample involvement in risky behaviours. However these instruments often fail to emphasise the ‘impul- sive’ component of risky action. Exploratory and confirmatory factor analyses were used to develop a 12- item scale of risky impulsivity. To examine its construct validity, the Zuckerman–Kuhlman impulsive sen- sation seeking scale was used. Self-reported aggression (physical and verbal) and angry behaviours (explosive and defusing) were also measured. Although risky impulsivity correlated with both of the two Z–K facets (Impulsivity and sensation seeking), it showed a stronger correlation than either of them with physical and verbal aggression. Sex differences in physical and verbal aggression were completely eliminated (and the sex difference in Explosive Anger was significantly decreased) when risky impulsivity was controlled. The risky impulsivity instrument appears very suitable for examining that form of impul- sivity most relevant to aggression. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction Impulsivity has been strongly implicated in aggression and externalising disorders (Fox, Henderson, Marshall, Nichols, & Ghera, 2005) and the relationship is as strong for females as for males (Cale, 2006). Sex differences are reliably found in aggressive behaviour, with the magnitude of the effect size increasing in line with the dan- gerousness of the form of aggression (Archer, 2004). In light of the conspicuous absence of sex differences in anger (Archer, 2004), it seems reasonable to propose that males’ greater aggression may re- sult from their higher levels of ‘risky’ impulsivity (Campbell, 2006). Although ‘impulsivity’ is broadly understood to refer to ‘a ten- dency to act spontaneously and without deliberation’ (Carver, 2005, p. 313), the existence and magnitude of sex differences in impulsivity seems to depend upon the measurement instrument used. Many impulsivity inventories pose general statements about impulsive behaviour where no explicit danger is implied. Examples include Are you an impulsive person? (I 7 Impulsiveness, Eysenck, Pearson, Easting, & Allsopp, 1985), I have trouble resisting my crav- ings (NEO personality inventory-revised: impulsiveness, Costa & McCrae, 1992) and I act on the spur of the moment (Barratt impul- siveness scale-11, Barratt, 1994). These inventories tend to produce weak, inconsistent or null sex differences (e.g. Costa, Terracciano, & McCrae, 2003; Feingold, 1994). Other impulsivity inventories incorporate items which carry an element of risk or danger, such as Eysenck I 7 Venturesomeness (Eysenck et al., 1985), NEO PI-R: Excitement Seeking (Costa et al., 2003), sensation seeking scale (Zuckerman, 1994), and MPQ harm avoidance reversed (Moffitt, Caspi, Rutter, & Silva, 2001). Here sex differences are apparent. These sex differences may derive from the well-established sex difference in fear of harm or injury (Campbell, 1999, 2006). In childhood, girls are more fearful than boys (Else-Quest, Hyde, Goldsmith, & Van Hulle, 2006). In adults, physiological studies (McManis, Bradley, Berg, Cuthbert, & Lang, 2001) and international surveys (Brebner, 2003) find that women experience more fre- quent and intense fear than men. Women make less risky decisions than men, especially when the risks are physical or life-threatening (Byrnes, Miller, & Schafer, 1999). Most relevant to the present argument, fear moderates the sex difference in aggression. Sex dif- ferences in aggression are larger to the extent that women rate the situation as more dangerous than men (Bettencourt & Miller, 1996; Eagly & Steffen, 1986). Fear, by triggering freezing or flight in threatening situations, restrains impulsive action (Fox et al., 2005). Sensation seeking scales explicitly address risky situations and sex differences are reliably found. But is sensation-seeking really a form of impulsivity or does it reflect a distinct trait? At an empir- ical level, Whiteside and Lynam’s factor analysis of 21 impulsivity scales found sensation seeking to be an orthogonal factor but Ger- bing, Ahadi, and Patton’s (1987) factor analysis of 373 impulsivity items reported a correlation of .52 between thrill seeking and impulsivity factors. Zuckerman (1994) argued that sensation seek- ing implicitly incorporates a ‘‘willingness to take physical and so- cial risks”. However parachute jumpers do not jump from planes on impulse; they plan carefully and check their equipment, drop site, parachute and timings. Although the Zuckerman–Kuhlman impulsive sensation seeking scale (Zuckerman, Kuhlman, Joireman, Teta, & Kraft, 1993; Zuckerman & Kuhlman, n.d.) attempts to blend the constructs of impulsivity and sensation seeking, it nonetheless 0191-8869/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.paid.2009.04.006 * Corresponding author. Tel.: +44 0191 334 3235. E-mail address: [email protected] (A. Campbell). Personality and Individual Differences 47 (2009) 402–406 Contents lists available at ScienceDirect Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid

Transcript of Can ‘risky’ impulsivity explain sex differences in aggression?

Page 1: Can ‘risky’ impulsivity explain sex differences in aggression?

Personality and Individual Differences 47 (2009) 402–406

Contents lists available at ScienceDirect

Personality and Individual Differences

journal homepage: www.elsevier .com/locate /paid

Can ‘risky’ impulsivity explain sex differences in aggression?

Anne Campbell *, Steven MuncerDurham University, Psychology, South Road, Durham DH1 3LE, United Kingdom

a r t i c l e i n f o

Article history:Received 9 January 2009Received in revised form 24 March 2009Accepted 6 April 2009Available online 7 May 2009

Keywords:AggressionImpulsivitySex differences

0191-8869/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.paid.2009.04.006

* Corresponding author. Tel.: +44 0191 334 3235.E-mail address: [email protected] (A. Ca

a b s t r a c t

Impulsivity is a heterogeneous concept and sex differences are most apparent on those inventories thatsample involvement in risky behaviours. However these instruments often fail to emphasise the ‘impul-sive’ component of risky action. Exploratory and confirmatory factor analyses were used to develop a 12-item scale of risky impulsivity. To examine its construct validity, the Zuckerman–Kuhlman impulsive sen-sation seeking scale was used. Self-reported aggression (physical and verbal) and angry behaviours(explosive and defusing) were also measured. Although risky impulsivity correlated with both of thetwo Z–K facets (Impulsivity and sensation seeking), it showed a stronger correlation than either of themwith physical and verbal aggression. Sex differences in physical and verbal aggression were completelyeliminated (and the sex difference in Explosive Anger was significantly decreased) when risky impulsivitywas controlled. The risky impulsivity instrument appears very suitable for examining that form of impul-sivity most relevant to aggression.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction (Zuckerman, 1994), and MPQ harm avoidance reversed (Moffitt,

Impulsivity has been strongly implicated in aggression andexternalising disorders (Fox, Henderson, Marshall, Nichols, & Ghera,2005) and the relationship is as strong for females as for males (Cale,2006). Sex differences are reliably found in aggressive behaviour,with the magnitude of the effect size increasing in line with the dan-gerousness of the form of aggression (Archer, 2004). In light of theconspicuous absence of sex differences in anger (Archer, 2004), itseems reasonable to propose that males’ greater aggression may re-sult from their higher levels of ‘risky’ impulsivity (Campbell, 2006).

Although ‘impulsivity’ is broadly understood to refer to ‘a ten-dency to act spontaneously and without deliberation’ (Carver,2005, p. 313), the existence and magnitude of sex differences inimpulsivity seems to depend upon the measurement instrumentused. Many impulsivity inventories pose general statements aboutimpulsive behaviour where no explicit danger is implied. Examplesinclude Are you an impulsive person? (I7 Impulsiveness, Eysenck,Pearson, Easting, & Allsopp, 1985), I have trouble resisting my crav-ings (NEO personality inventory-revised: impulsiveness, Costa &McCrae, 1992) and I act on the spur of the moment (Barratt impul-siveness scale-11, Barratt, 1994). These inventories tend to produceweak, inconsistent or null sex differences (e.g. Costa, Terracciano, &McCrae, 2003; Feingold, 1994). Other impulsivity inventoriesincorporate items which carry an element of risk or danger, suchas Eysenck I7 Venturesomeness (Eysenck et al., 1985), NEO PI-R:Excitement Seeking (Costa et al., 2003), sensation seeking scale

ll rights reserved.

mpbell).

Caspi, Rutter, & Silva, 2001). Here sex differences are apparent.These sex differences may derive from the well-established sex

difference in fear of harm or injury (Campbell, 1999, 2006). Inchildhood, girls are more fearful than boys (Else-Quest, Hyde,Goldsmith, & Van Hulle, 2006). In adults, physiological studies(McManis, Bradley, Berg, Cuthbert, & Lang, 2001) and internationalsurveys (Brebner, 2003) find that women experience more fre-quent and intense fear than men. Women make less risky decisionsthan men, especially when the risks are physical or life-threatening(Byrnes, Miller, & Schafer, 1999). Most relevant to the presentargument, fear moderates the sex difference in aggression. Sex dif-ferences in aggression are larger to the extent that women rate thesituation as more dangerous than men (Bettencourt & Miller, 1996;Eagly & Steffen, 1986). Fear, by triggering freezing or flight inthreatening situations, restrains impulsive action (Fox et al., 2005).

Sensation seeking scales explicitly address risky situations andsex differences are reliably found. But is sensation-seeking reallya form of impulsivity or does it reflect a distinct trait? At an empir-ical level, Whiteside and Lynam’s factor analysis of 21 impulsivityscales found sensation seeking to be an orthogonal factor but Ger-bing, Ahadi, and Patton’s (1987) factor analysis of 373 impulsivityitems reported a correlation of .52 between thrill seeking andimpulsivity factors. Zuckerman (1994) argued that sensation seek-ing implicitly incorporates a ‘‘willingness to take physical and so-cial risks”. However parachute jumpers do not jump from planeson impulse; they plan carefully and check their equipment, dropsite, parachute and timings. Although the Zuckerman–Kuhlmanimpulsive sensation seeking scale (Zuckerman, Kuhlman, Joireman,Teta, & Kraft, 1993; Zuckerman & Kuhlman, n.d.) attempts to blendthe constructs of impulsivity and sensation seeking, it nonetheless

Page 2: Can ‘risky’ impulsivity explain sex differences in aggression?

Table 1Item loadings Factors 1 and 2.

Item (‘‘On impulse I would . . . ”) Factor1

Factor2

9. Have a one night stand with an attractive stranger .65 �.3026. Tear up a parking ticket .65 �.0522. Have unprotected sex .62 �.144. Smoke cannabis if someone offered it to me .61 �.057. Make a gesture at an inconsiderate driver .61 �.1114. Gamble more money than I actually have .60 �.0823. Drive too fast when I am feeling upset .59 .0625. Skip a lecture because I am not in the mood .58 .0929. Turn right across oncoming traffic with only just enough

time to make it.58 �.09

8. Have another drink even when I am already drunk .55 .0515. ‘Streak’ at a public event, just for a laugh .54 �.215. Make the first move to kiss someone I find attractive .53 �.2124. Put purchases on a credit card without having enough

money to pay it off.53 .28

2. Drive through an amber traffic light .50 .0117. Run across a road to beat the oncoming traffic if I am in a

hurry.47 �.08

27. Walk out of a restaurant because the service is too slow .46 .126. Steal something from a shop .43 �.0916. Go up to a stranger and begin a conversation if I find them

attractive.45 �.07

30. Walk away from someone who is annoying me .47 �.0418. Book a ‘last-minute’ foreign holiday .47 .1628. Buy an item of clothing that I like even if I don’t need it .18 .7011. Go into an expensive shop just because I am walking past .21 .681. Buy a ‘treat’ to cheer myself up .10 .6613. ‘Binge’ eat my favourite food .20 .5621. Hug someone out of happiness �.10 .5312. Blurt something out without thinking .22 .4410. Hang up the phone on an unsolicited sales call .39 .043. Cancel my plans and go for a drink if I unexpectedly met an old

friend.30 .14

19. Get on a really scary ride at a funfair .36 �.1820. Say something to someone that I later regret .30 .2031. ‘‘In general, I avoid impulsive actions if I might get injured”

(diagnostic item)�.32 .32

Variance explained % 21.92 8.89

A. Campbell, S. Muncer / Personality and Individual Differences 47 (2009) 402–406 403

contains two distinct item populations: Sensation Seeking itemsemphasising risk but not spontaneity (I like to have new and excitingexperiences and sensations even if they are a little frightening) andImpulsivity items emphasising spontaneity but not riskiness (Iam an impulsive person). When factor analysed, the scale items splitinto these two components. Thus it is unclear whether the overallsex difference on this scale (Aluja, Garcia, & Garcia, 2003; Hojat &Zuckerman, 2008; Zuckerman & Kuhlman, 2000) is attributable tothe impulsivity, sensation seeking or both.

Women’s desistance from aggression may reflect behaviouralinhibition under risky conditions, rather than a lower level of gen-eral impulsivity. To examine this, we aim to construct an inventoryof risky acts with the items worded to incorporate a clear elementof impulsivity. We can then examine sex differences and their po-tential to mediate sex differences in aggression.

Meta-analyses have confirmed robust sex differences in directphysical and verbal aggression in laboratory (Bettencourt & Miller,1996) and real world studies (Archer, 2004). However, anger is fre-quently discharged without recourse to aggression (defined as ‘anintent to harm or injure’, Baron, 1977). Such non-injurious angryactions have been measured and factorially confirmed as explosiveand defusing acts (Campbell & Muncer, 2008). ‘Explosive’ actionsare those in which there is an acute, high-energy behavioural dis-charge of anger in the absence of the provoker. The four-item mea-sure of Explosive acts includes (when alone) hitting walls,throwing inanimate objects, destroying property, and screamingabuse. ‘Defusing’ actions involve attempts to reduce the intensityof the aggression-precipitating angry emotion. Defusing items in-clude retreating from the scene to calm down, discussing the inci-dent with a third party, giving the offender the silent treatmentand crying. Explosive and defusing acts are a safer strategic optionthan direct aggression, being unlikely to provoke physical retalia-tion. We therefore predict that risky impulsivity will be positivelyrelated to verbal and physical aggression (but not to explosive ordefusing acts) and that sex differences in physical and verbalaggression will be mediated by risky impulsivity.

Note. Items in bold were retained on the final scales.

2. Method

2.1. Sample 1

2.1.1. ParticipantsParticipants were 329 undergraduate students aged 18–24 at a

northern British university. This opportunity sample was com-posed of 165 men and 164 women who were studying subjectsother than psychology.

2.1.2. InstrumentsCandidate items for the impulsivity inventory were initially

developed from focus group discussions (5 groups of 6 student par-ticipants) who described occasions, during the prior week, wherethey had acted impulsively. Impulsivity was not formally defined,as the aim was to access lay people’s own perceptions and experi-ences. Following this, synonymous items were merged, contextualdetails were deleted to achieve greater generality and items refer-ring to aggressive acts were removed to eliminate content overlapwith the aggression questionnaire. This resulted in 30 behaviouralitems (see Table 1), each prefaced by the lead statement ‘‘On im-pulse I would . . .”. Respondents rated the probability of impul-sively engaging in each act on a Likert scale from Very unlikely(1) to Very likely (5). One global self-descriptive item was added:‘‘In general, I avoid impulsive actions if I might get injured” to as-sist in interpreting the factors resulting from the factor analysis.

The angry behaviour questionnaire (Campbell & Muncer, 2008)was also distributed. The 16-item questionnaire taps four expres-

sions of anger; physical aggression, verbal aggression, explosiveacts and defusing acts. Each scale has four items. Respondentsrated their probability of engaging in each behaviour when angryon a Likert scale from Very unlikely (1) to Very likely (5).

2.2. Sample 2

2.2.1. ParticipantsParticipants were 356 undergraduate students aged 18–24 at a

northern British university. This opportunity sample was com-posed of 174 men and 182 women, studying subjects other thanpsychology.

2.2.2. InstrumentsParticipants completed the 30-item impulsivity scale and the

angry behaviour questionnaire as described for Sample 1. In addi-tion, they completed the 19-item impulsive sensation seeking scalefrom the Zuckerman–Kuhlman personality inventory (Zuckerman& Kuhlman, n.d.; Zuckerman et al., 1993). Eight of the items assessimpulsivity and 11 assess sensation seeking. The questionnaireswere distributed and completed as for Sample 1.

3. Results

3.1. Exploratory factor analysis

Exploratory factor analysis of the 31 candidate items for the Ris-ky impulsivity scale was performed on the Sample 1 data. Bartlett’s

Page 3: Can ‘risky’ impulsivity explain sex differences in aggression?

Table 2Pattern of item loadings from confirmatory factor analysis (Sample 1, N = 329).

Item Loadingontoone factor

Loading ontothree factors

2. Drive through an amber traffic light .444 Physical injury.622

17. Run across the road to beat the traffic if I amin a hurry

.426 Physical injury.516

23. Drive too fast when I am feeling upset .591 Physical injury.677

29. Turn right across oncoming traffic with onlyjust enough time to make it

.578 Physical injury728

4. Smoke cannabis if someone offered it to me .567 Health risk .6048. Have another drink when I am already drunk .501 Health risk .5439. Have a one night stand with an attractive

stranger.622 Health risk .721

22. Have unprotected sex .616 Health risk .6526. Steal things from a shop .417 Criminal risk .46614. Gamble more money than I actually have .605 Criminal risk .75424. Put purchases on credit card without having

enough money to pay it off.523 Criminal risk .677

26. Tear up a parking ticket .622 Criminal risk .709

404 A. Campbell, S. Muncer / Personality and Individual Differences 47 (2009) 402–406

test, X2 = 3000, df = 465, p < .001, and KMO = 0.84, indicated thatthe impulsivity data were suitable for factor analysis. Nine factorshad eigenvalues greater than 1 but the Scree test suggested thatthe first two factors were important, cumulatively explaining30.0% of the variance (see Table 1). Following the second factor, ex-plained variance dropped to 5.95% and the slope levelled off. Ourgeneral diagnostic item (‘‘In general I avoid impulsive actions if Imight get injured”) loaded negatively on Factor I and positivelyon Factor II suggesting these two factors might be characterisedas risky and safe impulsivity, respectively. Factor I had high load-ings from four behaviours that contained a risk of physical injury(turning across oncoming traffic, driving through an amber light,running across a busy road, driving fast when upset) and a furtherfour other items that carried clear health risks (unprotected sex,one night stand, cannabis and alcohol use). Four items entailedfinancial or criminal risk (gambling, credit card overuse, destroyingparking ticket, stealing). However the remaining items were char-acterised by a general disregard for social convention (makingoffensive gestures, abandoning ongoing interactions, initiatingromantic encounters, streaking) and personal obligations (foreignholiday, skipping lectures). Because these six items did not containa clear element of risk to health or physical integrity, they wereeliminated from further analysis.

Factor II had loadings from four items that were clearly con-sumerist (buy clothes, enter expensive shop, buy treat, binge eatfood). Nonetheless, they were factorially distinct from financial riskitems (gambling, credit card abuse) and from items addressing theretail context more generally (behaviour in shops and restaurants).The items reflected indulgence or self reward. A further two itemswere about spontaneity in interpersonal relationships (conversa-tional ‘blurting’ and hugging). Because these acts did not carryany physical, financial or social risks, this was tentatively called aSafe Impulsivity factor.

3.2. Confirmatory factor analysis (Sample 1, N = 329)

The exploratory factor analysis suggested that there were twelveitems loading onto a Risky Impulsivity factor which could begrouped in terms of risk of physical injury, risk to health and crim-inal/financial risk. We investigated the structure of this twelve itemscale by using AMOS 7 (Arbuckle & Wothke, 1999). In all cases,maximum likelihood estimation was employed and missing caseswere excluded from the analysis. Two models were compared;the first suggesting that all twelve items loaded onto one factorcalled risky impulsivity and the second suggesting three relatedfactors of physical injury risk, criminal risk and health risk. The pat-tern of item loadings for both models is presented in Table 2.

The one factor model showed a poor fit to the data; {2/df = 5.65,comparative fit index (CFI) = .76, Tucker Lewis Index (TLI) = .711and root mean square error of approximation (RMSEA) = .119 (90% confidence interval .106 to .132). The three factor model wasmore acceptable with a {2/df = 2.44, an RMSEA = .066 (90%CI = .052�.081), TLI = .91 and CFI = .93. The three factor modelwas a significantly better fit than the one factor model({2 = 180.8, df = 3, p < .0001), and fitted the data reasonably well.The health risk factor correlated with the physical injury factor atr = .68 and with the criminal risk factor at r = .67, and the physicalinjury factor correlated with the criminal risk factor at r = .50. Thesix-item Safe Impulsivity scale had acceptable fit statistics as aone-factor construct; {2/df = 2.016, RMSEA = .056 (90%CI .015–.090), CFI = .97 and TLI = .95.

3.3. Confirmatory factor analysis (Sample 2, N = 356)

The three factor model of risky impulsivity fitted the datafrom the second sample reasonably well with a {2/df = 2.07,

RMSEA = .055 (90% CI .04–.07), TLI = .91, CFI = .93. It was a signifi-cantly better fit to the data than a one factor model ({2 = 155.99,df = 3, p < .0001). The health risk factor correlated with the physicalinjury factor at r = .62 and with the criminal risk factor at r = .60and the physical injury factor correlated with the criminal risk fac-tor at r = .48.

Our attempt to confirm the 6-item Safe Impulsivity scale wasnot successful. The fit statistics were {2/df = 6.435, RMSEA = .124,CFI = .85 and TLI = .76. Given this result and our chief focus ondeveloping a measure of risky impulsivity, we dropped this vari-able from further analysis.

3.4. Scale construction and psychometrics

To examine the psychometric properties of the risky impulsivityand angry behaviour scales, samples 1 and 2 were combined(N = 685). The risky impulsivity scale (12 items) had a Cronbach’salpha of 0.81 which could not be improved by the removal ofany item. The four angry behaviour scales had the following Cron-bach’s alphas; physical aggression 0.84, verbal aggression 0.79,defusing anger 0.59, explosive anger 0.72. The Z–K impulsive sen-sation seeking scale (N = 356) had a Cronbach’s alpha of 0.84 (sen-sation seeking alpha = 0.75, impulsivity alpha = 0.78.

3.5. Correlations

Correlations between all variables are presented in Table 3. TheZ–K impulsivity and sensation seeking subscales were correlated atr = .51 suggesting about 26% shared variance. As we hoped, our ris-ky impulsivity measure correlated moderately with both Z–KImpulsivity r = .40 and Z–K sensation seeking r = .46.

Physical aggression, verbal aggression and explosive anger allinter-correlated above r = .46. All three were highly correlated withrisky impulsivity, but showed smaller associations with the Z–Kmeasures, ranging from r = .10 (verbal aggression with Z–K sensa-tion seeking) to r = .21 (verbal aggression with Z–K impulsivity). Itappears that risky impulsivity may be more useful than these alter-native measures in the context of explaining aggressive behaviour.

Defusing anger was clearly distinct from the other triad of angrybehaviours. This scale measured behaviours that reduce the inten-sity of the angry emotion (e.g. avoiding contact with the provoker,leaving the scene). The correlations of defusing anger with the Z–Ksubscales and with risky impulsivity were all significantly negativebut modest in magnitude.

Page 4: Can ‘risky’ impulsivity explain sex differences in aggression?

Table 3Intercorrelations of all variables.

Variable 1 2 3 4 5 6 7 8

1. Risky impulsivity – .53*** .50*** .37*** �.14*** .49*** .40*** .46***

2. Physical aggression – .64*** .50*** �.09* .17** .17** .13*

3. Verbal aggression – .46*** .02 .17** .21*** .104. Explosive anger – .02 .16** .15** .13*

5. Defusing anger – �.23*** �.18*** �.21***

6. Z–K impulsive sensation seeking – .85*** .89***

7. Z–K Impulsivity – .51***

8. Z–K sensation seeking –

* p < .05.** p < .01.

*** p < .001.

Table 4Means and standard deviations for males and females on all variables.

Variable Males Females t da

M SD M SD

Risky impulsivity 21.01 8.32 17.68 7.11 5.63*** 0.42Z–K Impulsive sensation seeking 10.55 4.28 8.66 4.74 3.92*** 0.41Z–K Impulsivity 3.56 2.35 2.91 2.51 2.52* 0.26Z–K Sensation seeking 6.99 2.65 5.76 2.92 4.16*** 0.43Physical aggression 3.77 3.38 3.03 3.27 2.92** 0.22Verbal aggression 8.95 3.69 8.23 3.55 2.62** 0.20Explosive anger 6.28 3.64 5.19 3.26 4.13*** 0.31Defusing anger 8.50 2.60 11.85 2.54 �17.09*** �1.09

a d is an effect size measure computed as the difference between male and female means divided by the pooled standard deviation.* p < .05.

** p < .01.*** p < .001.

A. Campbell, S. Muncer / Personality and Individual Differences 47 (2009) 402–406 405

3.6. Sex differences

A two-way analysis of variance examined sex differences acrossthe eight scales. The interaction term was significant, F (7,2478) = 18.96, p < .001. Means and standard deviations are shownin Table 4. Men scored significantly higher than women on bothrisky impulsivity, d = .42 and the full Z–K impulsive sensationseeking scale, d = .41. The effect size was similar for the sensationseeking facet, d = .43, but smaller for the impulsivity facet,d = .26. Men were higher on three of the four angry behaviourscales: physical aggression, d = .22; verbal aggression, d = .20, andexplosive anger, d = .31. Women were very much higher on defus-ing anger, d = �1.09. Lest this was attributable to women’s greaterlikelihood of crying (or admitting to it), the scale was reanalysedwith the crying item omitted. Women still scored significantlyhigher, t = �9.12, p < .001, d = �.66.

3.7. Does Risky Impulsivity explain sex differences in aggression andanger?

To examine this question, we ran mediation analyses with thedifferent forms of anger expression as the dependent variables,sex as the independent variable and risky impulsivity as the medi-ator (Table 5). The amount of mediation is the indirect effectthrough the mediating variable (ab) which is equivalent to the to-tal effect of the independent variable on the dependent variable (c)minus its direct effect (c’). As recommended by MacKinnon, Lock-wood, and Lang (2004), estimates of indirect effects (ab) weremade using a bias corrected bootstrap technique with 1000 sam-ples and confidence intervals set at 95% (Preacher & Hayes,2004). Risky impulsivity completely explained the sex differencesin physical and verbal aggression, reducing the respective correla-tions from r = �.11 to r = .00, and from r = �.10 to r = .00. For explo-

sive anger, the sex difference was significantly mediated by riskyimpulsivity but the effect was partial because the correlation be-tween sex and explosive anger remained significant (although re-duced in magnitude from r = �.16, p < .001 to r = �.09, p < .05).The absence of a significant unique effect of risky impulsivity ondefusing anger meant that this variable did not meet the criterionfor mediation analysis (Baron & Kenny, 1986).

4. Discussion

The literature on impulsivity is in a deeply confused conceptualstate (Carver, 2005; Evendon, 1999). This is also true at the level ofmeasurement. For example, the multidimensional personalityquestionnaire describes the impulsive person as ‘‘spontaneous,can be reckless and careless”, while the NEO–PI–R cautions thatimpulsiveness ‘‘should not be confused with spontaneity, risk-tak-ing or rapid decision time” (NEO–PI–R).

We suspected that this confusion might have obscured sex differ-ences. Although sex differences are found on scales that incorporatean element of risk, the item wording often fails to make reference toacting on impulse. The risky impulsivity scale developed in the pres-ent study explicitly fills this gap. The new scale captures the middlegrounds between Zuckerman–Kuhlman’s impulsivity and sensationseeking facets. The sex difference we found is of similar magnitudeto the Z–K sensation seeking facet and larger than the impulsivity fa-cet. Especially noteworthy for aggression researchers is the fact thatthe risky impulsivity scale shows the strongest correlation with self-reported direct aggres- sion.

By measuring four forms of aggression and angry behaviour, wewere able to examine the discriminative validity of risky impulsiv-ity. As expected, risky impulsivity indeed correlated strongly andpositively with the direct (physical and verbal) forms of aggression.(We note that this correlation may be specific to impulsive or reac-

Page 5: Can ‘risky’ impulsivity explain sex differences in aggression?

Table 5Summary of mediation results (unstandardised coefficients).

Independentvariable

Mediatingvariable

Dependentvariable

Effect of IVon mediator (a)

Unique effectof mediator (b)

Indirect effect(ab) (and CIa)

Totaleffect (c)

Directeffect (c’)

Degree ofmediation

Sex Risky Physical �3.33*** 0.22*** �0.75 �0.77** �0.01 CompleteImpulsivity Aggression (�1.06, �0.47)

Sex Risky Verbal �3.33*** 0.23*** �0.76 �0.75** 0.02 CompleteImpulsivity Aggression (�1.04, �0.47)

Sex Risky Explosive �3.33*** 0.15*** �0.51 �1.11*** �0.60* PartialImpulsivity Anger (�0.73, �0.30)

a CI = Confidence interval.* p < .05.

** p < .01.*** p < .001.

406 A. Campbell, S. Muncer / Personality and Individual Differences 47 (2009) 402–406

tive forms of aggression). However risky impulsivity also showed asignificant positive correlation with explosive acts (though weakerthan those found for direct aggression). In light of the correlationbetween explosive acts and direct forms of aggression, this is notsurprising. Yet it appears paradoxical that impulsive individualsstill retain sufficient self control to ‘leave the room’ before dis-charging their anger. Perhaps riskily impulsive individuals aremore willing not only to risk injury from others (in the form ofretaliation to direct aggression), but also to risk self injury (fromexplosive acts such as hitting walls).

As predicted, defusing acts were significantly, though modestly,negatively correlated with risky impulsivity suggesting that indi-viduals who avoid dangerous impulsive actions also avoid directconfrontation. Defusing acts were higher among women (evencontrolling for women’s greater admission of crying) but this wasnot attributable to the sex difference in risky impulsivity. Women’sgreater likelihood of leaving the scene, avoiding the target, cryingand seeking third party consolation may result from their prefer-ence for social affiliation when stressed (Taylor et al., 2000) andgreater willingness to express powerless emotions (Timmers,Fischer, & Manstead, 1998).

A key finding was the elimination of sex differences in directaggression when the effect of risky impulsivity was controlled.The next research step is to see if risky impulsivity can explainbehavioural as well as self-reported measures of aggression. In thisway, common method variance can be eliminated as potentiallycontributing to the current findings. Given the absence of sex dif-ferences in the impelling motivation of anger, a clear alternativeexplanation for sex differences in direct aggression must be interms of inhibitory control and its converse, impulsivity. Howeverthe present data suggest that it is specifically where impulsive ac-tions carry potential risks, that men exceed women. This findinglinks well-established sex differences in fear, risk-aversion andharm-avoidance with impulsivity, and can potentially explain theequally well-established sex difference in direct aggression.

References

Aluja, A., Garcia, O., & Garcia, L. F. (2003). Psychometric properties of theZuckerman–Kuhlman personality questionnaire (ZKPQ-III-R): A study of ashortened form. Personality and Individual Differences, 34, 1083–1097.

Arbuckle, J. L., & Wothke, W. (1999). Amos 4.0 user’s guide. Chicago: SmallwatersCorporation.

Archer, J. (2004). Sex differences in aggression in real-world settings: A meta-analytic review. Review of General Psychology, 8, 291–322.

Baron, R. A. (1977). Human aggression. New York: Plenum Press.Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in

social psychological research: Conceptual, strategic and statisticalconsiderations. Journal of Personality and Social Psychology, 51, 1173–1182.

Barratt, E. S. (1994). Impulsiveness and aggression. In J. Monahan & H. J. Steadman(Eds.), Violence and mental disorder (pp. 61–79). Chicago: University of ChicagoPress.

Bettencourt, B. A., & Miller, N. (1996). Gender differences in aggression as a functionof provocation: A meta-analysis. Psychological Bulletin, 119, 422–447.

Brebner, J. (2003). Gender and emotions. Personality and Individual Differences, 34,387–394.

Byrnes, J. P., Miller, D. C., & Schafer, W. D. (1999). Gender differences in risk taking:A meta-analysis. Psychological Bulletin, 125, 367–383.

Cale, E. M. (2006). A quantitative review of the relationship between the ‘‘Big 3”higher order personality dimensions and antisocial behaviour. Journal ofResearch in Personality, 40, 250–284.

Campbell, A. (1999). Staying alive: Evolution, culture and women’s intra-sexualaggression. Behavioral and Brain Sciences, 22, 203–252.

Campbell, A. (2006). Sex differences in direct aggression: What are thepsychological mediators? Aggression and Violent Behavior, 11, 237–264.

Campbell, A., & Muncer, S. (2008). An intent to harm or injure? Gender and theexpression of anger. Aggressive Behavior, 34, 282–293.

Carver, C. S. (2005). Impulse and constraint: Perspectives from personalitypsychology, convergence with theory in other areas and potential forintegration. Personality and Social Psychology Review, 9, 312–333.

Costa, P. T., & McCrae, R. R. (1992). NEO PI-R professional manual. Odessa, FL:Psychological Assessment Resources.

Costa, P. T., Terracciano, A., & McCrae, R. R. (2003). Gender differences in personalitytraits across cultures: Robust and surprising findings. Journal of Personality andSocial Psychology, 81, 322–331.

Eagly, A. H., & Steffen, V. (1986). Gender and aggressive behavior: A meta-analyticreview of the social psychological literature. Psychological Bulletin, 100, 3–22.

Else-Quest, N. M., Hyde, J. S., Goldsmith, H. H., & Van Hulle, C. A. (2006). Genderdifferences in temperament: A meta-analysis. Psychological Bulletin, 132, 33–72.

Evendon, J. L. (1999). Varieties of impulsivity. Psychopharmacology, 146, 348–361.Eysenck, S. B. G., Pearson, P. R., Easting, G., & Allsopp, J. F. (1985). Age norms for

impulsiveness, venturesomeness and empathy in adults. Personality andIndividual Differences, 6, 613–619.

Feingold, A. (1994). Gender differences in personality: A meta-analysis.Psychological Bulletin, 116, 429–456.

Fox, N. A., Henderson, H. A., Marshall, P. J., Nichols, K. E., & Ghera, M. M. (2005).Behavioral inhibition: Linking biology and behavior within a developmentalframework. Annual Review of Psychology, 56, 235–262.

Gerbing, D. W., Ahadi, S. A., & Patton, J. (1987). Toward a conceptualization ofimpulsivity: Components across the behavioural and self-report domains.Multivariate Behavioral Research, 22, 357–379.

Hojat, M., & Zuckerman, M. (2008). Personality and speciality interest in medicalstudents. Medical Teacher, 30, 400–406.

MacKinnon, D. P., Lockwood, C. M., & Lang, P. J. (2004). Confidence limits for theindirect effect: Distribution of the product and resampling methods.Multivariate Behavioral Research, 39, 99–128.

McManis, M. H., Bradley, M. M., Berg, W. K., Cuthbert, B. N., & Lang, P. J. (2001).Emotional reactions in children: Verbal, physiological and behavioral responsesto affective pictures. Psychophysiology, 38, 222–231.

Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sex differences in antisocialbehaviour: Conduct disorder, delinquency and violence in the Dunedin longitudinalstudy. Cambridge: Cambridge University Press.

Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirecteffects in simple mediation models. Behavioral Research Methods, Instruments,and Computers, 36, 717–731.

Taylor, S. E., Klein, L. C., Lewis, B. P., Gruenewald, T. L., Gurung, R. A. R., & Updegraff,J. A. (2000). Biobehavioral responses to stress in females: Tend-and-befriend,not fight-or-flight. Psychological Review, 107, 411–429.

Timmers, M., Fischer, A. H., & Manstead, A. S. R. (1998). Gender differences in motivesfor regulating emotions. Personality and Social Psychology Bulletin, 24, 974–986.

Whiteside, S. P., & Lynam, D. R. (2001). The Five Factor model and impulsivity: Usinga structural model of personality to understand impulsivity. Personality andIndividual Differences, 30, 669–689.

Zuckerman, M. (1994). Behavioral expressions and biosocial bases of sensation seeking.New York: Cambridge University Press.

Zuckerman, M., & Kuhlman, D. M. (n.d.). Norms for the Zuckerman–Kuhlmanpersonality questionnaire (ZKPQ). Newark, DE: University of Delaware.

Zuckerman, M., & Kuhlman, D. M. (2000). Personality and risk-taking: Commonbiosocial factors. Journal of Personality, 68, 999–1029.

Zuckerman, M., Kuhlman, D., Joireman, J., Teta, P., & Kraft, M. (1993). A comparisonof three structural models for personality: The big three, the big five andthe alternative big five. Journal of Personality and Social Psychology, 65, 757–768.