You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303886869

Disentangling functions of online aggression: The Cyber-Aggression Typology


Questionnaire (CATQ)

Article in Aggressive Behavior · June 2016


DOI: 10.1002/ab.21663

CITATIONS READS

60 6,704

3 authors, including:

Kevin Runions
Telethon Kids Institute
80 PUBLICATIONS 3,102 CITATIONS

SEE PROFILE

All content following this page was uploaded by Kevin Runions on 03 April 2018.

The user has requested enhancement of the downloaded file.


AGGRESSIVE BEHAVIOR
Volume 9999, pages 1–11 (2016)

Disentangling Functions of Online Aggression: The


Cyber-Aggression Typology Questionnaire (CATQ)
 re
Kevin C. Runions1*, Michal Bak2, and The se Shaw1
1
Telethon Kids Institute, Faculty of Medicine, Telethon Kids Institute, Dentistry and Health Sciences, University of Western
Australia, Perth, Western Australia, Australia
2
University of Western Ontario, London, Ontario, Canada
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
Aggression in online contexts has received much attention over the last decade, yet there is a need for measures identifying the
proximal psychological drivers of cyber-aggressive behavior. The purpose of this study was to present data on the newly developed
Cyber-Aggression Typology Questionnaire (CATQ) designed to distinguish between four distinct types of cyber-aggression on
dimensions of motivational valence and self-control. A sample 314 undergraduate students participated in the study. The results
confirmed the predicted four-factor structure providing evidence for distinct and independent impulsive-aversive, controlled-
aversive, impulsive-appetitive, and controlled-appetitive cyber-aggression types. Further analyses with the Berlin Cyberbullying
Questionnaire, Reactive Proactive Aggression Questionnaire, and the Behavior Inhibition and Activation Systems Scale provide
support for convergent and divergent validity. Understanding the motivations facilitating cyber-aggressive behavior could aid
researchers in the development of new prevention and intervention strategies that focus on individual differences in maladaptive
proximal drivers of aggression. Aggr. Behav. 9999:1–11, 2016. © 2016 Wiley Periodicals, Inc.
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..

Keywords: cyber-aggression; reactive aggression; proactive aggression; motivation; scale development

INTRODUCTION been applied to cyber-aggression. To address this gap,


our study tested the validity and reliability of a newly
As a new field of study, cyberbullying research
developed measure derived from the quadripartite
provides an important foundation for intervention and
typology proposed for general violent aggression by
prevention development. Heterogeneity in why online
Howard (2011). This measure aims to assess heteroge-
aggression (commonly referred to as cyberbullying)
neity in the function of cyber-aggression, and thereby
arises has the potential to obstruct intervention efforts.
provide more targeted research to guide intervention
For example, prevention of deliberate, goal-driven
development.
cyberbullying may fail to deter those who are aggressing
in response to provocation; prevention of revenge-
WHAT DRIVES CYBER-AGGRESSION?
driven cyber-aggression may not reach someone
spontaneously “trolling,” just for fun. But little research Relatively little research has examined motivations
attention has been paid to diversity in the proximal, for cyberbullying or cyber-aggression more broadly.
functional, drivers of cyber-aggression. As with aggres- Raskauskas and Stolz (2007) examined open-ended
sion more broadly, cyber-aggression may arise in responses about why the adolescents’ surveyed thought
different ways, reflecting different functions. However,
instrumentation developed to measure cyberbullying 
Correspondence to: Kevin Runions, Faculty of Medicine, Telethon Kids
and cyber-aggression has failed to capture heterogeneity Institute, Dentistry and Health Sciences, University of Western
in why cyber-aggression arises, instead focusing on the Australia, Perth, P.O. Box 855, Western Australia 6872, Australia.
range of modes or manifestations (e.g., flaming, cyber- E-mail: kevin.runions@telethonkids.org.au
stalking, masquerading; Willard, 2007). In aggression Received 25 September 2015; Revised 28 April 2016; Accepted 9 May
research, this would be considered a focus on the form at 2016
the expense of the function. Indeed, even basic DOI: 10.1002/ab.21663
distinctions in human aggression functions—between Published online XX Month Year in Wiley Online Library
proactive and reactive aggression—have only recently (wileyonlinelibrary.com).

© 2016 Wiley Periodicals, Inc.


2 Runions et al.

young people engaged in cyberbullying. The most More sensitive measures directly examining the
common responses were for recreational purposes function of cyber-aggressive behavior, and not simply
and for revenge. Recently, Compton, Campbell, and offline reactive or proactive aggression, indicate a more
Mergler (2014) interviewed students, parents, and complex picture. In Canada, Law and her colleagues
teachers about their perceptions of motivations for (Law, Shapka, Domene, & Gagne, 2012) modified the
cyberbullying, after being provided with a definition of RPQ to reflect online aggression, but found no factor
bullying. Amongst the key themes for cyberbullying analytic support for a proactive-reactive distinction. A
was the pursuit of power and status, fun-seeking or relief single-factor solution was found, leading Law et al.
of boredom, and use of cyber-aggression as a safe means (2012) to conclude that “participants did not differenti-
of retaliating. A qualitative study of college students ate between proactive and reactive reasons for being
focused on the use of online aggression to achieve social aggressive online” (p. 667). It is possible, however, that
control, to obtain valued social resources, and for this reflects Type 2 error, and the modified instrument
entertainment, as exemplified by trolling (Rafferty & failed to capture diversity in reasons for cyber-
Vander Ven, 2014). Consistent to these few studies, and aggression. In a later study using the same modified
much less commonly noted in research in offline online RPQ measure (Shapka & Law, 2013), two-factors
aggression, are the notions of fun-seeking, revenge were found, and reactive motives were stronger
and retaliation. predictors of general cyber-aggression than were
Revenge and fun-seeking motivations for aggression proactive motivations, contrasting with the finding of
are notable in their poor fit with the reactive-proactive Calvete et al. (2010).
aggression paradigm that dominates developmental Studying the motives behind online aggression can be
psychological research on the function of aggression valuable for intervening and preventing acts of aggres-
(Dodge & Coie, 1987; Dodge, 1991). Reactive aggres- sion and bullying. The pioneering research on reactive
sion is conceived as an impulsive aggressive response in and proactive functions of aggression, however, has left
the context of retaliation for some (perceived or real) an ambiguous picture that does not lend itself to clear
insult or attack. Unlike revenge, reactive aggression is application in prevention or intervention efforts. We
commonly conceived as non-deliberated. Proactive suggest that a re-conceptualization of the motives or
aggression is generally conceived as unprovoked acts functions of aggression in general is required to make
of aggression enacted with deliberation. In distinction further progress in this area. Specifically, we argue that
from the fun-seeking discussed in research on cyber- the reactive-proactive distinction bears a core concep-
aggression motives, proactive aggression is deliberated, tual confound that hinders its further utility, including its
not impulsive. utility in understanding cyber-aggression.
Following Dooley, Pyzalski, and Cross (2009), who
first explicitly discussed cyberbullying in terms of
A QUADRIPARTITE MODEL OF THE FUNCTION
reactive and proactive aggression, a small body of
OF CYBER-AGGRESSION
pioneering research has examined cyber-aggression
through the reactive-proactive lens. Calvete, Orue, Howard (2011) has argued that the reactive-proactive
Estevez, Villardon, and Padilla (2010) published the distinction is limited by its confounding of regulatory
first empirical investigation of proactive and reactive processes and motivational goals. Certainly many—
functions in cyber-aggression. Using a measure of perhaps even most—aggressive retaliatory responses are
cyberbullying and the Reactive-Proactive Aggression enacted impulsively, and many unprovoked aggressive
Questionnaire (RPQ; Raine et al., 2006), they found acts may be committed after deliberation. But deliber-
that only self-reported proactive aggression predicted ated aggressive acts may also be taken in retaliation to
cyber-aggression when controlling for reactive provocation—that is to say, revenge as commonly
aggression in a sample of 12–17-year-old Spanish conceived. Similarly, impulsive acts of aggression that
adolescents. In a similar study with American and are motivated by appetitive motivational goals, such as
Singaporean youth, proactive aggression was uniquely the immediate elicitation of positive affect (or “fun”), are
associated with cyberbullying, after partialling out the poorly captured in the traditional reactive-proactive
influence of reactive aggression (Ang, Huan, & Florell, distinction. Traditional measures of reactive and proac-
2014). These studies replicate findings for face-to-face tive aggression conflate these functional drivers by
bullying (Fontaine, 2007; Salmivalli, Lagerspetz, assuming all retaliatory aggression is impulsively
Bj€ €
orkqvist, Osterman, & Kaukiainen, 1996), which conducted (and thus subsumed under the reactive
could lead to the conclusion that cyberbullying is an aggression banner), and that all appetitive aggression
expression of proactive aggression, regardless of how is deliberately conducted (and thus subsumed under the
it is measured. proactive aggression category).

Aggr. Behav.
Cyber-Aggression Typology Questionnaire 3

As a response to this conceptual and methodological recruitment of capacities for self-control. This quadrant
limitation, Howard (2011) has proposed a quadripartite maps most clearly to traditional conceptualization and
model of aggression based on two orthogonal dimen- measurement of reactive aggression. The controlled-
sions—motivational goals (appetitive vs. aversive) and aversive quadrant, by contrast, also arises as a means to
regulatory control (impulsive vs. deliberated). As exorcise uncomfortable emotions (e.g., shame, embar-
Howard (2009, 2011) has reviewed, impulsive acts rassment) arising from perceived provocation. But here,
arise in the context of minimal or subconscious cognitive capacities for self-control are exercised, and the
appraisal of an environmental cue prompting a strong production of aggression is more likely to take the
affective response and a strong action tendency or urge form of calculated payback, or revenge. This distin-
to act (Baumeister, Vohs, DeWall, & Zhang, 2007). In guishes it from controlled-appetitive aggression, where
the production of impulsive behavior, that action self-control is also exerted, but in the interest of
impulse is acted upon without heed of consequences, generating positive affect via deliberated efforts. This
especially in the long-term. Rather, the behavior is quadrant most closely maps onto traditional conceptu-
engaged in for the immediate emotional pay-off. In alizations of “proactive” aggression, with its focus on
contrast, deliberated acts of aggression are conducted via effortful control in the pursuit of some environmental
the recruitment of self-control capacities in service of reward, such as attaining social status through bullying
some goal. (e.g., Vaillancourt, Hymel, & McDougall, 2003). The
The second dimension in the quadripartite model impulsive-appetitive quadrant reflects spontaneously
reflects the nature of this emotional pay-off. Appetitive produced behavior, conducted without heed of long-
motives are aimed at obtaining a positive emotional state term consequences, arising from the immediate positive
for the aggressor. Aversive motives are aimed at the affect generated by the aggressive act. These impulsive-
removal of negative emotional states, including embar- appetitive motives are commonly noted in the self-
rassment, shame or guilt. We believe this includes the generated accounts of violent offenders (Howard, 2011).
painful experiences of having been victimized or Bjørnebekk and Howard (2012) developed a measure
harassed by others (or perceiving victimization). In (the Angry Aggression Scales, AAS) assessing the four
this sense, aversive motives may often be seen as types of aggression suggested by this typology,
retaliatory, as they may often arise in the context of specifically for use with youth who have a diagnosed
provocation. conduct disorder. Using a sample of 101 young people
In postulating two orthogonal dimensions (appetitive with serious conduct problems and 101 non-aggressive
vs. aversive; impulsive vs. controlled), this motive and control youth, their study found evidence supporting the
self-control model produces four distinct quadrants (see four-factor solution mapping on to the four types of
Fig. 1). The impulsive-aversive quadrant reflects acts, aggression.
often in reaction to perceived provocation, shame To assess how this quadripartite typology mapped onto
elicitation, and other aversive emotional states, that the traditional reactive-proactive distinction, Bjørnebekk
are spontaneously produced or arise with little or no and Howard (2012) correlated factor scores for each
quadrant and a teacher-rating of reactive and proactive
aggression amongst the deviant youth group (Roland &
Idsøe, 2001). As expected, the impulsive-aversive scale
was significantly correlated with reactive aggression
scores. But the controlled-aversive scale was also
significantly associated with reactive aggression. The
appetitive subscales were not significantly correlated with
reactive aggression, providing evidence of convergent
and divergent validity of scores on the AAS. Teacher-
rated proactive aggression was correlated with both
impulsive and controlled appetitive aggression, as might
be expected, but also with impulsive-aversive aggression.
In the same study, Bjørnebekk and Howard (2012)
also examined correlations with self-rated behavioral
inhibition and behavioral activation systems (BIS/BAS).
This distinction between behavioral inhibition and
Fig. 1. Quadripartite typology of aggression, from Howard’s (2010)
activation systems was proposed by Gray (1987) to
Quadripartite violence typology, distinguishing motivational valence account for hypothesized neural systems involved in
of aggression and recruitment of self-control capacities. regulating attention to novel environmental cues

Aggr. Behav.
4 Runions et al.

potentially signaling danger and/or punishment (BIS), As evidence of convergent validity, we first predicted
and neural systems regulating approach behavior to that impulsive-aversive cyber-aggression would be
rewarding behavior (BAS). Carver and White’s (1994) associated with established measures of reactive aggres-
operationalization of these systems involves three sub- sion as reactive aggression is traditionally operationalized
scales for BAS, namely Fun Seeking, Drive, and Reward as aggression in response to aversive stimuli and
responsiveness. BAS Fun Seeking includes items conducted in an uncontrolled manner. We also expected
emphasizing the pursuit of potentially rewarding it to be associated a relative lack of behavioral inhibition.
conditions in a spontaneous manner. The BAS Drive Second, as proactive aggression is traditionally
subscale measures a willingness to pursue desired goals operationalized as a premeditated form of aggression
in a persistent manner. The BAS Reward Responsive- without provocation, we predicted that controlled-
ness subscale focuses on positive affective responses to appetitive cyber-aggression would be most strongly
realized or anticipated reward. Bjørnebekk and Howard correlated with traditional measures of proactive
(2012) found that the impulsive-appetitive subscale was aggression generally. Due to its calculated, deliberate,
positively correlated only with BAS Fun Seeking. reward-seeking nature, we also predicted it would be
Contrary to expectations, impulsive-aversive aggression associated with the BAS Drive and BAS Reward-
also correlated with Fun Seeking, but also showed a Responsiveness scales.
significant positive correlation with BIS or behavioral Our third hypothesis posited that as provocation can be
inhibition. The two controlled types of aggression were responded to via planned retaliatory actions, controlled-
both positively correlated with BAS Drive. Controlled- aversive cyber-aggression was likely to be associated
aversive aggression was also correlated with behavioral with both proactive and reactive aggression as tradition-
inhibition, whereas controlled-appetitive aggression was ally operationalized. We also expected it to be associated
correlated with BAS Reward Responsiveness. This with victimization experiences and as controlled actions
network of associations provides evidence of convergent with BAS Drive scores.
and divergent validity for the quadripartite model. Finally, as aggression may take place without advance
Following Howard, Runions (2013) examined the planning and in the service of immediate positive
relevance for a quadripartite model of cyber-aggression affective consequences for the aggressor, our fourth
and cyberbullying and presented novel hypotheses about hypothesis predicted that impulsive-appetitive cyber-
the co-action of personal characteristics and aspects of aggression would be associated with both proactive (due
the online context as drivers of distinct types of cyber- to its reward-driven nature) and reactive (due to its
aggression. Those hypotheses, however, presume that impulsive nature) aggression. We also predicted that it
the quadripartite model is relevant in an online context, would be associated with fun seeking behavior, and not
but no measures have yet been developed to examine related to victimization experiences.
these distinct types of cyber-aggression.
METHODS
THE PRESENT STUDY Participants
Currently, research on cyber-aggression is without University students at a small Canadian university
tools for measuring diverse functions of cyber-aggres- were invited to participate in the study in an e-mail,
sion and cyberbullying. As such, our understanding of which also provided them with a link to access the web
why cyber-aggression arises is constrained, placing survey. A wide range of the university’s departments
limits on intervention planning and evaluation. The aim were involved in the distribution of study invitation e-
of the present study was to present data on a new Cyber- mails. These e-mail invitations were sent to an
Aggression Typology Questionnaire (CATQ). This undisclosed total number of students enrolled in first
conceptually derived scale was designed to capture year courses through the university departments and
Aversive versus Appetitive motives and Impulsive faculties. As a result, we are unable to present response
versus Controlled modes of expressing those motives. rates for this convenience sample. In total, 314 students
Our goal was to present evidence for the predicted four- (63.4% women) participated. The final sample com-
factor structure, and test for convergent and divergent prised students from a range of disciplines, including arts
validity via the Reactive-Proactive Aggression Ques- (11%), science and mathematics (12.2%), engineering
tionnaire (RPQ; Raine et al., 2006), a measure of and computer sciences (23.2%), business and commerce
cybervictimization, and the BIS/BAS Scales (Carver & (13.1%), psychology (17.1%), other social sciences
White, 1994). We were also interested in examining how (20.4%), with a small set still undecided (3.0%). Almost
the four types of cyber-aggression were related to half were 18 years of age (46.5%, n ¼ 146), with 29.9%
reported experiences of cyber-victimization. (n ¼ 94) 19 year olds, 7.0% (n ¼ 22) 17 year olds, and the

Aggr. Behav.
Cyber-Aggression Typology Questionnaire 5

remaining 17.5% (n ¼ 55) were 20 years or over. We did Procedure


not collect data on ethnicity or socio-economic status. As
Items for the Cyber-Aggression Typology Question-
to their experience with online communication, 97.3%
naire (CATQ) were generated by adapting items from
indicated that they were users of social networks, 82.4%
the Angry Aggression Scale (Bjørnebekk & Howard,
reported they have been victims of some online harm at
2012) and applying them to the online context, and by
some point in their past, and 73.8% indicated that they
generation of items inspired by the conceptual frame-
had harmed someone else online, even if accidental.
work proposed by Runions (2013). As noted above,
Measures recruitment was conducted via email distributed through
Cyber-agression and cyber-victimization. university departments, with a link to the secure online
The Berlin Cyberbullying-Cybervictimisation Question- survey service FluidSurvey. The notices indicated that
naire (BCyQ; Schultze-Krumbholz & Scheithauer, the study was testing a new instrument about internet and
2009) is a 17-item index assessing distinct categories social network use. There were no restrictions or
developed in light of Willard’s (2007) proposed seven exclusion criteria deployed. Where possible, research
categories of cyber-aggressive behavior. Cronbach’s assistants addressed students in key courses prior to
alpha for the cyber-aggression subscale in this study lectures to encourage participation. A script was
sample was .87, and for the cyber-victimization scale prepared for these addresses. Participants were entered
was .89. into a draw for one of three $100 iTunes vouchers. All
Reactive and proactive aggression. The 23 aspects of the study were approved by the university’s
item Reactive-Proactive Questionnaire (Raine et al., institutional human ethics review board.
2006) consists of 11 items assessing reactive aggression A formative version of the scale was developed and
(e.g., “reacted angrily when provoked by others”) and 12 piloted with a sample of 37 university undergraduate
items assessing proactive aggression (e.g., “used force to students, aged 18–19. In this version, impulsive-
obtain money or things from others”). Its psychometric appetitive, controlled-appetitive, impulsive-aversive,
properties have been well established in the original and and controlled-aversive forms of cyber-aggression
translated versions (e.g., Cima, Raine, Meesters, & were assessed with 21, 13, 17, and 13 items,
Popma, 2013). In the current study, both reactive and respectively. With this preliminary scale data, we
proactive scales showed acceptable internal consisten- conducted two exploratory principal components anal-
cies, a ¼ .77 and .76, respectively. yses with varimax and equimax rotations, on the items
Behavioral inhibition and behavioral activa- that were predicted to form distinct subscales. In light of
tion. The 20-item BIS/BAS scales (Carver & White, the very small sample size for PCA analyses, we adopted
1994) were used to provide an operationalization of a very conservative cutoff of .70 for factor loadings, and
Gray’s (1987) BIS/BAS theory. BIS items reflect retained only those items that loaded above that
contexts that contain signals of potential punishment threshold on a factor and below .40 on all other factors
(8 items; e.g., “if I think something unpleasant is going to identified when subjected to both varimax and equimax
happen I usually get pretty “worked up”). Behavioral rotation. Only those items that met this criteria were used
Activation is assessed via three subscales: BAS Reward- in a second phase, described below.
Responsiveness (five items; e.g., “When I get something The retained items were then presented via online
I want, I feel excited and energized), BAS Drive (four survey to a new sample of participants, along with the
items; e.g., “When I want something, I usually go all-out other measures and items assessing basic demographics.
to get it”), and BAS Fun-Seeking (four items; e.g., “I will (Following collection of demographic data, we counter-
often do things for no other reason but that they might be balanced the order of the CATQ scale, the BIS/BAS
fun”). All items are rated on a 4-point Likert scale scales, and the Reactive-Proactive Questionnaire.) These
(1 ¼ not at all true of me; 4 ¼ very true of me). Validity items were then examined via Confirmatory Factor
and reliability have been well-established for these Analysis in MPlus 6 to test construct validity. We
scales (Carver & White, 1994). For our sample, internal compared our four-factor model against two competing
consistency was good, with a Cronbach alpha coefficient two-factor models reflecting controlled versus impulsive
of .81, .72, .77 for BIS, Reward-Responsiveness, Drive, cyber-aggression and appetitive versus aversive cyber-
respectively. Alpha for Fun-Seeking was suboptimal aggression. Scores on the four subscales were calculated
(a ¼ .65). This did not appear to be resulting from any as mean values of the final items. Gender invariance was
single item, and all items were retained for analyses. assessed for these categorical items through simultaneous
Given the importance of this construct for our hypothe- testing of the invariance of the factor loadings and
ses, however, we have used it in our study, and advise thresholds across gender groups as recommended
caution in interpreting relationships with this construct. (Muthen & Muthen, 1998–2009). The Satorra–Bentler

Aggr. Behav.
6 Runions et al.

scaled chi-square difference test was utilized to compare retaliatory items and appetitive items. The second
the fit of the models assuming and not assuming comparator model combined all impulsive and con-
invariance (Muthen & Muthen, 1998–2009, 2012). These trolled items. Neither two-factor models fit the data
residualized mean subscale scores were examined in adequately, and the four-factor model showed best fit
relation to the other study variables to assess convergent across both the chi-square and the approximate fit
and divergent validity, as described in the hypothesis. indices (see Table III) and significantly improved fit.
Difference testing via MPlus indicated that the four-
RESULTS factor model showed significantly better fit than both the
appetitive versus aversive two-factor model (x2 ¼ 91.7,
Testing of CATQ Construct Validity
df ¼ 5, P < .001), and the impulsive versus controlled
The development of the CATQ aimed to produce a two-factor model, (x2 ¼ 119.7, df ¼ 5, P < .001). This
measure of cyber-aggression that was sensitive to not added evidence for a four-factor structure to the data.
only a reactive-proactive distinction in function of The factor structure was invariant across gender groups,
aggression, but a conceptually clearer distinction Satorra–Bentler scaled chi-square difference test TRd
between cyber-aggression driven by appetitive versus ¼ 43.7, df ¼ 50, P ¼ .722.
aversive motives, wherein self-control is or is not Table II reports mean values and standard deviations
recruited. As such, construct validity was assessed via a for the CATQ variables. Notable positive skew was
four-factor confirmatory factor analysis. observed for all subscales (range 1.29–2.22), and inverse
CFA was conducted based on the four motivationally transformations (Tabachnick & Fidell, 2013) were
distinct categories of cyber-aggression, which included conducted for each, which reduced skew values (range
12 impulsive-aversive items, six controlled-aversive 0.44–1.03). To control for the shared variation between
items, six controlled-appetitive items, and five impul- the subscale scores, standardized residual scores were
sive-appetitive items. As with most measures assessing calculated for each subscale by regressing the target
aggression in a non-clinical, non-antisocial population, subscale (inverse-transformed) scores onto the other
the distribution of the ordinal cyber-aggression items three subscales (also inverse transformed).
showed substantial positive skew. Given this, we used
Convergent and Divergent Validity
the weighted least squares mean variance estimator
(WLSMV), recommended for ordinal skewed data To examine convergent and divergent validity we
(Muthen & Muthen, 1998–2009). Fit statistics values examined correlations of the standardized residual
for adequate model fit were RMSEA <.06, and TLI and CATQ scores with the BCyQ (perpetration and
CFI both >.95 (Hu & Bentler, 2009), as well as the victimization), the RPQ, and the BIS/BAS scales.
weighted root mean residual (WRMR) <.90 (Schreiber, Examination of the correlations between the CATQ
Nora, Stage, Barlow, & King, 2006). We also examined subscales and our covariates provide support for
the standardized model results for each item to ensure convergent and divergent validity. To further investigate
factor loadings loaded highly on the scale. All items these forms of validity, we examined the residualized
showed loadings well above our criterion of .4 or above score associations with cyber-victimization as measured
(see Table I). We examined the modification indices, and by the BCyQ, although we present raw score associa-
as none were substantively larger than the rest (all <.20), tions as well (see Table IV). Although we did not
there was little evidence to drop items or modify the hypothesize the magnitude and directions of associa-
model. All fit indices indicated good fit for the four- tions between the CATQ subscales and cyberbullying
factor model, and the subscale factor loading values (as measured by the BCyQ), we were interested to
were in line with the predicted model. As would be investigate these further. Raw correlations between
expected, the four factors were inter-correlated CATQ subscales and cyberbullying as indexed by the
(Table II.). The two types of aversive cyber-aggression BCyQ ranged from .40–.60 (see Table IV).
were correlated at .91, and the two types of appetitive Examination of these residualized correlations indi-
cyber-aggression at .82. Both impulsive forms of cyber- cate that Impulsive-aversive cyber-aggression was
aggression were correlated at .66, and controlled forms significantly correlated with reactive aggression
of cyber-aggression at .89. Finally, the cross-diagonals (r ¼ .20, P < .001) and with cyber-victimization
show that impulsive-aversive was correlated with (r ¼ .16, P < .05), as predicted. However, the correlation
controlled-appetitive at .83, and impulsive-appetitive with the BIS or behavioral inhibition scores was not
was correlated with controlled-aversive at .66. significant. Additional non-predicted, small but signifi-
For a more stringent test of construct validity, we also cant associations with BAS Drive (r ¼ .12, P < .05) and
compared the four-factor model with two competing BAS Reward-Responsiveness (r ¼ .12, P < .05) were
two-factor model interpretations. The first combined all noted.

Aggr. Behav.
Cyber-Aggression Typology Questionnaire 7

TABLE I. Items, Item Descriptives and Item Loadings for CATQ Factor 1 (Impulsive-Aversive), Factor 2 (Controlled-Aversive),
Factor 3 (Controlled-Appetitive), and Factor 4 (Impulsive-Appetitive)
Item F1 F2 F3 F4

If someone tries to hurt me, I will use an ICT device to immediately get back at them 0.86
If I get teased or threatened, I get angry easily and strike back online right away 0.86
I use ICT to get back at someone as soon as they post a hurtful message about me 0.85
If someone makes me angry online I quickly post mean texts and messages online 0.82
If someone makes fun of me on the internet, I get frustrated and respond angrily online right away 0.81
I overreact before I have a chance to think about the consequences when someone says something mean online 0.81
If I see a message online that gets me angry, I react too quickly and then regret the way I responde 0.79
If someone tries to cyberbully me, I quickly lash back with something online 0.79
If someone says something online to hurt me, I post something back right away to get back at them 0.75
If somebody criticizes me online or in a text, I often react aggressively without thinking of the consequences 0.71
I hastily respond to something written online and regret it later 0.67
I respond very quickly to a message or post that is disrespectful to me 0.57
If someone does something to hurt me, I would get back at them in my own time by using my ICT device(s) 0.90
If someone tries to hurt me, I will use my ICT device(s) to get back at them in my own time 0.89
I get back at people who make fun of me on the internet because their posts hurt more the more I think about them 0.84
I like using my ICT device(s) to plan my revenge when I feel angry at someone 0.76
If I need to get revenge on someone, I would rather strike back using my ICT device(s) where I can plan out how to do it 0.72
If I see a mean message about me on my ICT, it bothers me more and more when I think about it, and I try to get even 0.62
If I don’t like someone, I use the internet to turn others against them 0.92
Sometimes I’ll team up with my friends to bring someone down online 0.84
Sometimes I can be mean to people online to get what I want 0.82
When I don’t like a person, I use the internet to make them feel like they do not belong in my group 0.79
I pretend to be someone else online to ruin somebody else’s friendships 0.76
I have at times used the internet to make someone look like bad 0.75
I get carried away having fun online and others think I’m being a cyberbully or a troll 0.90
I make fun of people I don’t know on the internet without thinking about whether they will see it or not 0.83
If I’m having fun and joking online, I don’t care if someone’s feelings get hurt 0.81
I repeatedly annoy people online because I think it’s funny 0.77
Joking online is so much fun that I don’t worry about whether someone might be bothered by what I say 0.69

Controlled-appetitive cyber-aggression correlated sig- correlated with either proactive or reactive aggression
nificantly with proactive aggression (r ¼ .18, P < .01) or any of the measures.
and BAS Drive (r ¼ .14, P < .05), as predicted. A The impulsive-appetitive subscale behaved largely as
significant negative correlation with BAS Reward- predicted. It was correlated with proactive aggression
Responsiveness is counter to the expected positive (r ¼ .17, P < .01), BAS Fun-Seeking (r ¼ .22, P < .001),
association. This form of cyber-aggression was moder- and not with cybervictimization (r ¼ .04, ns). However,
ately correlated with both cyberbullying (r ¼ .28, the anticipated association with reactive aggression was
P < .001) and cybervictimization (r ¼ .24, P < .001). not found (r ¼ .06, ns). Also noted were a small negative
Unexpectedly, the residualized scores for controlled- correlation with the BIS and a positive correlation with
aversive cyber-aggression were not significantly cyberbullying (r ¼ .14, P < .05).

DISCUSSION
TABLE II. Descriptives and Correlations Between CATQ To date, cyber-aggression and cyberbullying have
Subscales as Factors and as Final Variables (n ¼ 317) predominantly been assessed with checklists and scales
Variables 1 2 3 4 M (SD) that enumerate the different forms or means of engaging
in aggressive behavior via new information and
Impulsive-aversive — 0.91 0.83 0.66 1.35 (0.48) communication technologies. The present study pro-
Controlled-aversive 0.67 — 0.89 0.66 1.34 (0.51)
Controlled-appetitive 0.66 0.63 — 0.82 1.28 (0.45)
vided a new conceptually driven direction for the
Impulsive-appetitive 0.47 0.45 0.62 — 1.46 (0.57) assessment of cyber-aggression. Based on Howard’s
(2011) quadripartite model of aggression, and Runions’
Note. Values above the diagonal reflect correlations between factors in
CFA analyses (MPlus); values below the diagonal reflect correlations
(2013) extension of that model to cyber-aggression, this
between final analysis variable (mean of items with inverse transforma- new measure aimed to disentangle the motive for
tion; SPSS). aggressing online (appetitive vs. aversive) from the

Aggr. Behav.
8 Runions et al.

TABLE III. Goodness of fit statistics for CFA models


x2 df CFI TLI RMSEA WRMR

4 factor 178.58 81 0.95 0.99 0.061 0.905


2 factor (controlled and impulsive) 280.19 68 0.89 0.97 0.097 1.277
2 factor (appetitive and aversive) 231.1 68 0.92 0.97 0.085 1.161
CFI, comparative fit index; TLI, Tucker–Lewis index; RMSEA, root mean squared error of approximation; WRMR, weighted root mean square residual.

degree of impulsivity or self-control recruited in the proactive, not reactive, aggression. The impulsivity
production of aggressive behavior. The present study assessed by this subscale does not reflect uncontrolled
was the first to provide data indicating a valid factor anger, but rather “getting carried away” and going too
structure and nomological network for the CATQ. far, which may diverge from the type of self-control
Results of this first study involving undergraduate failure assessed by traditional reactive aggression items.
university students support the validity of the CATQ. The controlled-aversive cyber-aggression was not
Confirmatory factor analysis provided support for the associated with either reactive or proactive aggression,
hypothesized factor structure, with distinct and indepen- once the other CATQ subscales were controlled for.
dent scales assessing impulsive-aversive, controlled- Cyber-aggression based on a deliberated response to
aversive, controlled-appetitive, and impulsive-appetitive perceived provocations does not fit neatly with
subscales. Runions (2013) had proposed that cyber- traditional conceptualizations of reactive or proactive
aggression may arise differentially based on motives and aggression, and consequently, the more closely aligned
capacities for self-control, and our data provide support for CATQ subscales may dominate the shared variance
the hypothesis that the four subscales of the CATQ reflect between reactive and proactive measures. This raises a
distinct types of cyber-aggression. We precited that broader question of whether reactive and proactive
aversive modes of cyber-aggression would arise in the scales have been systematically missing out on
context and be associated with cyber-victimization. This potentially important type of aggression motivated by
was the case for impulsive-aversive cyber-aggression, but the alleviation of averse emotions, fueling wrath and
not controlled-aversive. Controlled-appetitive cyber- driving revenge. Frey and colleagues (Frey, Pearson, &
aggression, however, showed the strongest correlation Cohen, 2015) have explored the importance of consid-
with cyber-victimization, suggesting that this mode of ering retaliation motives for bullying. We suggest that
aggression may be part of a vicious cycle of perpetration the CATQ may provide a means to address this research
and (from the target) retaliation. Finally, as predicted, priority for cyber-aggression more broadly. Indeed,
impulsive-appetitive cyber-aggression was not related to given that the residualized controlled-aversive subscale
cyber-victimization. scores did not significantly correlate with any of the
The present results provide support, building on that of other study variables, this mode of cyber-aggression is
Bjornnebekk and Howard (2012), for the hypothesis that especially in need of further study.
the reactive-proactive aggression distinction is limited The analyses with the BIS/BAS scales provided
by its conflation of motive and self-control. The limited support for the CATQ and its conceptual
quadripartite model clarifies that not all retaliatory (or grounding. The impulsive appetitive mode of aggression
“reactive”) aggression is impulsive and not all appetitive was positively related with BAS Fun-Seeking. This
(or “proactive”) aggression is controlled. Aversive relationship was based on the idea that for some young
aggression may be enacted impulsively or in a more people, cyber-aggression may arise out of a desire for
deliberated manner. When enacted impulsively, it is novel, potentially (emotionally) rewarding experiences
conceptually most akin to reactive aggression. Appeti- without thinking about the consequences.
tive-aggression may be enacted in a controlled, The BAS Drive subscale correlated with controlled-
deliberated manner, or impulsively, in a spontaneous appetitive cyber-aggression, as expected in reflecting the
pursuit of immediate thrill and fun. As predicted, “persistent pursuit of goals” (Carver & White, 1994,
impulsive-aversive cyber-aggression was uniquely p.322) of those who score highly on the BAS scale.
related to reactive aggression and controlled-appetitive However, the lack of correlation with the other form of
cyber-aggression was uniquely related to proactive controlled aggression, namely controlled-aversive, may
aggression. In disentangling the conflation, we were indicate that the pursuit of goals does not extend to goals
particularly interested in relationships of the other two aimed at eliminating averse emotional states or
subscales to reactive and proactive aggression. Impul- achieving revenge. Perhaps this is medium-specific,
sive-appetitive aggression, was associated only with and people who choose to use ICT devices.

Aggr. Behav.
Cyber-Aggression Typology Questionnaire 9

Residual

.26

.22
.17
Impulsive-aversive cyber-aggression was related to

.14
.04

.06

.04
.02
Impulsive-appetitive
the BAS dimensions of Drive and Reward-responsive-
ness, indicating a potential role of approach motivation
dysregulation in the production of retaliatory responses
to perceived provocation that are enacted without the
.49 recruitment of self-control capacities. Impulsive-appeti-
.36
.45
.30
.28

.27
.15
Raw

tive cyber-aggression was the only subscale to uniquely

-.03
TABLE IV. Raw and Residualized Correlations Between CATQ Factors, Cyberbullying, Reactive/Proactive Aggression, and BIS-BAS (n ¼ 317)

relate to low BIS activity. This suggests that this mode of


cyber-aggression is particularly a function of low levels
of responsivity to punishment, and may more generally
reflect low levels of fear and/or anxiety. Although
Residual

further research is required to better test this notion, it


.28
.24
.18

-.14
.12

BCyQ, Berlin cyberbullying questionnaire; RPQ, reactive proactive aggression questionnaire; BIS/BAS, behavioral inhibition and activation systems scale.
.06
.03

-.04
Controlled-appetitive

may be that some degree of fear/anxiety—reflected here


in BIS scores—may deter young people from engaging
in impulsive-appetitive cyber-aggression.
Limitations
.60
.51
.49
.34

.21

Our research has several notable limitations. First, the


Raw

.14
.12

.13

sample size for our preliminary analyses was very small


for exploratory factor analysis or principal components
analysis. Some statisticians have argued that smaller
sample sizes can be acceptable if communalities are
high, model error low, and only a small number of
Residual

.01
.02
.02
.04
.05
.05
.11
.12

factors are expected for the final solution (Preacher &


Controlled-aversive

MacCallum, 2002). Our efforts to impose conservative


criteria for factor loadings was ad hoc, and may have
resulted in an invalid solution. We believe this limitation
is mitigated by the fact that we were adapting a scale that
.40
.35
.36
.27

had good established validity and by the results of the


Raw

.04
.08
.08
.03

second phase, that providing evidence via CFA that the


quadripartite factor structure did indeed fit the data well.
Our sample size provided a good estimate of CFA, but
was somewhat overpowered for our correlational analyses.
Consequently some of our significant correlations are
Residual

.28

potentially trivial in magnitude, with Pearson r values as low


.13

.13
.12
.12
.02

.06

.10
Impulsive-aversive

as 0.12 statistically significant, but arguably of very limited


importance. For current purposes, we were primarily
concerned that our four factors operated as they would be
expected based on our predicted nomological network.
In recruiting a convenience sample of university
.46
.48
.41
.42

.20
Raw

.16

students, our study may have introduced bias that we


.01
.02

cannot control. For example, we cannot establish a


response rate, as we do not know exactly how many
students received the invitation e-mails. Further research
should aim for a representative, or at least well-
BIS/BAS reward-responsiveness

characterized sample. Another limitation of this sampling


BIS/BAS behavioral inhibition

strategy is that the severity of cyber-aggression is likely to


BCyQ cyber-victimization
RPQ proactive aggression
RPQ reactive aggression

be low overall: future research should aim to examine


BIS/BAS fun-seeking
BCyQ cyber-bullying

cyber-aggressive motives in samples with greater overall


levels of cyber-aggression, possibly younger adolescents.
BIS/BAS drive

P < .001.

Future Directions
P < .01.
P < .05.

The present study provides promising support for a




questionnaire that assesses a new typology of cyber-





Aggr. Behav.
10 Runions et al.

aggression that is developed from a theoretically different types of cyber-aggression. Such investigations
driven framework and thereby holds potential to would provide important evidence for researchers
further our understanding of the diversity of types of testing cyberbullying intervention and prevention strat-
online aggression, any or all of which may be egies. Such program evaluation would benefit from a
mislabeled as cyberbullying. Additional research is tool that recognizes that “cyberbullying” is not a single
required with larger sample sizes that is representative monolithic phenomenon and that the motives and
of the population in question, drawn from a wider age capacities for self-control behind cyber-aggression
range of participants that is better reflective of the matter.
population of young people engaging online to test
whether the factor structure confirmed here is consis- REFERENCES
tent across samples. Given that capacities for behav-
Ang, R. P., Huan, V. S., & Florell, D. (2014). Understanding the
ioral and emotional self-regulation continue to develop relationship between proactive and reactive aggression, and cyber-
through adolescence into adulthood (Steinberg et al., bullying across United States and Singapore adolescent samples.
2008), it may be that this factor structure is not Journal of Interpersonal Violence, 29, 237–254. doi: 10.1177/
consistent with young adolescents, for example. 0886260513505149
Baumeister, R. F., Vohs, K. D., DeWall, C. N., & Zhang, L. (2007). How
Alternately, the structure may hold, but mean differ-
emotion shapes behavior: Feedback, anticipation, and reflection,
ences in subscale scores from those of young adults rather than direct causation. Personality and Social Psychology
may be observed. Review, 11, 167–203. doi: 10.1177/1088868307301033
Further testing is also required to better establish the Bjørnebekk, G., & Howard, R. (2012). Validation of a motivation-based
validity and reliability of the CATQ. In particular, test- typology of angry aggression among antisocial youths in Norway.
Behavioral Science and the Law, 30, 167–180. doi: 10.1002/bsl.2007
retest reliability needs to be established to determine the
Calvete, E., Orue, I., Estevez, A., Villard on, L., & Padilla, P. (2010).
extent to which these items assess consistent behavioral Cyberbullying in adolescents: Modalities and aggressors’ profile.
tendencies. Further tests of convergent and divergent Computers in Human Behavior, 26, 1128–1135. doi: 10.1016/j.
validity are also possible. For example, a direct measure chb.2010.03.017
of impulsivity or self-regulation capacity may help Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral
activation, and affective responses to impending reward and
elaborate the nomological network of these subscales,
punishment: The BIS/BAS scales. Journal of Personality and Social
and in theory would predict the two impulsive modes. Psychology, 67, 319–333. doi: 10.1037/0022-3514.67.2.319
Controlled-aversive cyber-aggression may be predicted Cima, M., Raine, A., Meesters, C., & Popma, A. (2013). Validation of the
by anger rumination tendencies and/or self-efficacy for Dutch Reactive Proactive Questionnaire (RPQ): Differential corre-
written communication over face-to-face interaction lates of reactive and proactive aggression from childhood to
adulthood. Aggressive Behavior, 39, 99–113. doi: 10.1002/ab.21458
(Runions, 2013). Impulsive-aversive cyber-aggression
Compton, L., Campbell, M., & Mergler, A. (2014). Teacher, parent and
may be predicted by hostile attribution tendencies, student perceptions of the motives of cyberbullies. Social Psychology
particularly those assessed via real-time methodologies of Education, 17, 383–400. doi: 10.1007/s11218-014-9254-x
(Yaros, Lochman, Rosenbaum, & Jimenez-Camargo, Dodge, K. A. (1991). The structure and function of reactive and proactive
2014), and/or trait measures of anger. Other hypotheses aggression. In D. J. Pepler, & K. H. Rubin (Eds.), The development and
treatment of childhood aggression. (pp. 201–218). Hillside, NJ:
to test may focus on the online context, and test the
Erlbaum.
differential sensitivity of impulsive-appetitive cyber- Dodge, K. A., & Coie, J. D. (1987). Social-information-process factors in
aggression to online social network contexts, in which a reactive and proactive aggression in children’s peer groups. Journal of
peer audience is implicit, versus private, text-message Personality and Social Psychology, 53, 1146–1158. doi: 10.1037/
delivery of cyber-aggression (Runions, 2013). Con- 0022-3514.53.6.1146
Dooley, J. J., Py_zalski, J., & Cross, D. (2009). Cyberbullying versus face-
trolled-appetitive cyber-aggression may be sensitive to
to-face bullying: A theoretical and conceptual review. Zeitschrift f€ ur
online contexts in which the lack of nonverbal cues may Psychologie, 217, 182–188. doi: 10.1027/0044-3409.217.4.182
result in a failure to activate affective empathy, and place Fontaine, R. G. (2007). Toward a conceptual framework of instrumental
excessive burdens on less efficient cognitive empathic antisocial decision making and behavior in youth. Clinical Psychology
processes, which may be less mature in young Review, 27, 655–675. doi: 10.1016/j.cpr.2007.01.012
Frey, K. S., Pearson, C. R., & Cohen, D. (2015). Revenge is seductive, if
adolescents.
not sweet: Why friends matter for prevention efforts. Journal of
The establishment of a reliable and valid scale that Applied Developmental Psychology, 37, 25–35. doi: 10.1016/j.
captures the diversity of function of cyber-aggression appdev.2014.08.002
provides the means of gaining a deeper understanding of Gray, J. A. (1987). The neuropsychology of anxiety. Oxford: Oxford
the distinct individual differences and situational cues University Press.
Howard, R. C. (2009). The neurobiology of affective dyscontrol:
that facilitate cyber-aggression. The CATQ may provide
Implications for understanding &lsquo dangerous M. severe person-
important insight into cyberbullying intervention ality disorder. In M. McMurran, & R. Howard (Eds.), Personality,
research, as different health promotion messages and personality disorder and violence. (pp. 157–174). Chichester: Wiley-
intervention strategies may differentially influence Blackwell.

Aggr. Behav.
Cyber-Aggression Typology Questionnaire 11

Howard, R. C. (2011). The quest for excitement: A missing link between Salmivalli, C., Lagerspetz, K., Bj€ €
orkqvist, K., Osterman, K., &
personality disorder and violence? Journal of Forensic Psychiatry and Kaukiainen, A. (1996). Bullying as a group process: Participant roles
Psychology, 22, 692–705. doi: 10.1080/14789949.2011.617540 and their relations to social status within the group. Aggressive
Hu, L-T., & Bentler, P. M. (2009). Cutoff criteria for fit indexes in Behavior, 22, 1–15. doi: 10.1002/(sici)1098-2337(1996)22:1<1::aid-
covariance structure analysis: Conventional criteria versus new ab1>3.0.co;2-t
alternatives. Structural Equation Modeling: A Multidisciplinary Schreiber, J. B., Nora, A., Stage, F. K., Barlow, E. A., & King, J. (2006).
Journal, 6, 1–55. doi: 10.1080/10705519909540118 Reporting structural equation modeling and confirmatory factor
Law, D. M., Shapka, J. D., Domene, J. F., & Gagne, M. H. (2012). Are analysis results: A review. Journal of Educational Research, 99,
cyberbullies really bullies? An investigation of reactive and proactive 323–338. doi: 10.3200/JOER.99.6.323-338
online aggression. Computers in Human Behavior, 28, 664–672. doi: Schultze-Krumbholz, A., & Scheithauer, H. (2009). Social-behavioral
10.1016/j.chb.2011.11.013 correlates of cyberbullying in an German student sample. Zeitschrift
Muthen, L. K., & Muthen, B. O. (1998–2009). MPlus user’s guide f€
ur Psychologie, 217, 224–226. doi: 10.1027/0044-3409
(5th ed.). Los Angeles, CA: Muthen & Muthen. Shapka, J. D., & Law, D. M. (2013). Does one size fit all? Ethnic
Muthen, B. O., & Muthen, L. K. (2012). Chi-square difference testing differences in parenting behavior and motivations for adolescent
using the Satorra-Bentler scaled chi-square. Retrieved January 2012, engagement in cyberbullying. Journal of Youth and Adolescence, 42,
from www.statmodel.com/chidiff.shtml 723–738. doi: 10.1007/s10964-013-9928-2
Preacher, K. J., & MacCallum, R. C. (2002). Exploratory factor analysis in Steinberg, L., Albert, D., Cauffman, E., Banich, E., Graham, S., &
behavior genetics research: Factor recovery with small sample sizes. Woolard, J. (2008). Age differences in sensation seeking and
Behavior Genetics, 32, 153–161. impulsivity as indexed by behavior and self-report: Evidence for a
Rafferty, R., & Vander Ven, T. (2014). “I hate everything about you”: A dual systems model. Developmental Psychology, 44, 1764–1778. doi:
qualitative examination of cyberbullying and on-line aggression in a 10.1037/a0012955
college sample. Deviant Behavior, 35, 364–377. doi: 10.1080/ Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics
01639625.2013.849171 (6th ed.). Boston: Pearson.
Raine, A., Dodge, K. A., Loeber, R., Gatzke-Kopp, L., Lynam, D., Vaillancourt, T., Hymel, S., & McDougall, P. (2003). Bullying is
Reynolds, C., & Liu, J. (2006). The reactive-proactive aggression power: Implications for school-based intervention strategies.
questionnaire: Differential correlates of reactive and proactive Journal of Applied School Psychology, 19, 157–176. doi: 10.1300/
aggression in adolescent boys. Aggressive Behavior, 32, 159–171. J008v19n02_10
doi: 10.1002/ab.20115 Willard, N. E. (2007). Cyberbullying and cyberthreats: Responding to the
Roland, E., & Idsøe, T. (2001). Aggression and bullying. Aggressive challenge of online social aggression, threats, and distress.
Behavior, 27, 446–462. doi: 10.1002/ab.1029 Champaign, IL: Research Press.
Runions, K. C. (2013). Toward a conceptual model of motive and self- Yaros, A., Lochman, J. E., Rosenbaum, J., & Jimenez-Camargo, L. A.
control in cyber-aggression: Rage, revenge, reward and recreation. (2014). Real-time hostile attribution measurement and aggression in
Journal of Adolescence, 42, 751–771. doi: 10.1007/s10964-013-9936-2 children. Aggressive Behavior, 40, 409–420. doi: 10.1002/ab.21532

Aggr. Behav.

View publication stats

You might also like