Introduction

Pathological gambling (PG) is an impulse control disorder (American Psychiatric Association, 1994) and is comorbid with bipolar spectrum disorders, attention-deficit/hyperactivity disorder, obsessive-compulsive disorder and substance abuse (Dell’Osso, Allen, & Hollander, 2005; Petry, Stinson, & Grant, 2005). The point and life-time prevalence of PG has been reported to be as high as 1.4 and 5.1%, respectively. The prevalence seems to be increasing owing to the spread of legal gambling (Petry & Armentano, 1999)

Various factors have been implicated in the aetiology, development and maintenance of PG. Beside familial and sociological components, attention has been drawn to individual factors such as biology/biochemistry, cognition and psychological states (Raylu & Oei, 2002)

Within the individual factors, personality disorders and personality traits have been the focus of interest in recent years. About 90% of pathological gamblers have at least one personality disorder (Black & Moyer, 1998; Blaszczynski & Steel, 1998) most of them belonging to Cluster B in DSM-IV (American Psychiatric Association, 1994) (i.e. borderline, histrionic and narcissistic personality disorders) (Blaszczynski & Steel, 1998).

As far as personality traits are concerned, there is a consensus that sensation seeking and impulsivity may act as risk factors for gambling and developing PG. Other traits include psychoticism and neuroticism (Raylu & Oei, 2002).

Sensation-seeking is a disposition to seek new and complex feelings and to take risks to achieve this goal (Raylu & Oei, 2002). High-sensation seekers are particularly prone to participate in risky activities such as gambling (Coventry & Brown, 1993; Zuckerman, 1999). It is, however, difficult to determine whether sensation seeking induces gambling behaviour or whether gambling increases a tendency to be a sensation seeker (for a review, see Raylu & Oei, 2002).

Impulsivity, i.e. acting without thought or self-control (Raylu & Oei, 2002), is another major characteristic of PG (Carlton & Manowitz, 1994; Moran, 1970; Steel & Blaszczynski, 1998). Compared to sensation seeking, impulsivity has been considered to be more fundamental and to have a specific role in the gambler’s chasing behaviour. It has been suggested that PG and substance abuse problems develop at the same time during adolescence and that both are related to a deficit in impulse control (Vitaro, Ferland, Jacques, & Ladouceur, 1998). Impulsivity has been considered to favour the drive to continue gambling despite continuous losses while sensation seeking sooner serves as a motivating factor (Zuckerman, 1999).

Overall, the data on personality characteristics in PG are inconclusive. A plausible reason is that different subgroups have been recruited in the studies (Dickerson & O’Connor, 2006).

In a previous study, using the Tridimensional Personality Questionnaire (TPQ) (Kim & Grant, 2001), high levels of novelty seeking (NS), impulsiveness and extravagance were found in pathological gamblers compared to controls. Thus, there is recent evidence for a role of temperament factors but less for character factors in PG.

Clarification of the balance between temperament (automatic emotional responses to experience that are stable throughout life) and character (self-concepts and individual differences in goals and values, which influence the meaning of what is experienced in life) (Cloninger, Przybeck, Svrakic, & Wetzel, 1994) can lead to a better understanding of the role of personality factors in PG. Therefore, we have investigated pathological gamblers using the Temperament and Character Inventory (TCI) (Cloninger et al., 1994).

Material and Methods

In a catchment area comprising about 460,000 inhabitants, active pathological gamblers were recruited for the study through advertisements in local newspapers. Forty individuals (33 males and 7 females) meeting the DSM-IV criteria for PG (American Psychiatric Association, 1994) were included after obtaining their informed consent. The mean age was 34.9 ± 10.5 years. Also, the mean age of onset was 26.2 ± 9.6 years and the mean period of gambling was 9.4 ± 8.1 years. Reported types of gambling were lotteries (3; 7.9%), pools (9; 23.7%) card games (4; 10.5%) roulette (3; 7.9%), fruit machines one-armed bandites (24; 63.2%), wagering (2; 5.3%), bingo (8; 21.1%) and horse racing (13; 34.2%).

All gamblers were subjected to a semistructured interview using the Structured Clinical Interview of DSM (SCID-I) (First, Gibbon, Spitzer, & Williams, 1997a). The interview was directed towards affective disorders, anxiety disorders and substance use disorders. The gamblers also filled in the SCID-II questionnaire for personality disorders (First, Gibbon, Spitzer, Williams, & Benjamin, 1997b) and the complete TCI (Cloninger et al., 1994). TCI measures four dimensions of temperament and three dimensions of character. The temperament dimensions are NS, harm avoidance (HA), reward dependence (RD) and persistence (PS). The character dimensions are self-directedness (SD), cooperativeness (CO), and self-transcendence (ST). Detailed descriptions of each of the seven dimensions are available elsewhere (Cloninger et al. 1994; Cloninger, 1987; Cloninger, Svrakic, & Przybeck, 1993).

Reported ongoing abuse of or dependence on illegal drugs disqualified from participation. In the case of pathological gamblers, occasional intake of bensodiazepines (but not use of neuroleptics or antidepressants) was accepted. Smoking was allowed. None of the pathological gamblers were undergoing psychotherapy.

Complete TCI data were obtained from 31 pathological male and seven pathological female gamblers (mean age 35.4 ± 10.4 years). Seventy-six controls matched for age and gender were chosen from a sample consisting of randomly recruited volunteers from the normal population (Richter, Brändström, & Przybeck, 2001).

The StatView 5 (SAS Institute Inc., Cary, NC, USA) program was used. All variables were considered to be normally distributed. In repeated ANOVAs, gamblers versus controls were used as the between-group factor. The TCI scores for higher and lower- order dimensions were used in turn as the dependent variable. The Bonferroni correction was not used so as not to increase the risk of false negative results. Effect size (difference between the means/root mean square of the two standard deviations) was used to compare differences (Cohen, 1988; Rosnow & Rosenthal, 1996). Cohen (1988) defined effect sizes as small (d ≥ 0.2), medium (d ≥ 0.5) and large (d ≥ 0.8).

Approval for the study was given by the Ethics Committee of the Linköping University Hospital. All subjects gave their informed consent after having been fully informed about the study.

Results

Among pathological gamblers, 53% had an affective disorder, 40% an anxiety disorder and 53% a form of substance use disorder. Six of them (16%) had attempted suicide.

Data on higher order dimensions of temperament and character are shown in Table 1. In temperament, pathological gamblers scored higher on NS and HA than controls. The differences were moderate (effect sizes 0.75 and 0.79). In character, pathological gamblers scored lower than controls on SD and CO. The differences were large and medium, respectively (effect sizes −1.46 and −0.67). Furthermore, pathological gamblers scored higher than controls on ST, and the difference was medium (effect size 0.51).

Table 1. Higher-order dimensions of temperament and character (TCI) for gamblers and controls
Full size table

Comparisons were also made concerning the individual TCI subscales of temperament (Table 2). Pathological gamblers scored lower than controls on NS1 (more rigid), and the difference was small (effect size −0.49). Furthermore, they were also more impulsive (NS2), more extravagant (NS3) and more disorganised (NS4). Those differences were medium, large and large (effect sizes 0.57, 1.28 and 0.86). Pathological gamblers scored higher than controls on HA1 (more anticipatory worry), HA3 (more shy) and HA4 (had more fatigability and asthenia). Those differences were medium (effect sizes 0.71, 0.61 and 0.64). They were also lower on RD4 (more independent) than controls but the difference was small (effect size −0.44).

Table 2. Lower-order dimensions (TCI) of temperament for gamblers and controls
Full size table

Temperament and Character Inventory means of lower-order dimensions of character are shown in Table 3. Pathological gamblers scored lower than controls on SD1 (less responsible), SD2 (less purposeful), SD3 (less resourceful), SD4 (less self-acceptance) and SD5 (less impulse control). These differences were large, medium, large, medium and large (effect sizes −1.2, −0.70, −0.92, −0.69 and −1.42). They also scored lower than controls on CO1 (less tolerant), CO2 (less empathic), CO3 (less helpful) and CO5 (more dishonest behaviour). These differences were small, medium, medium and medium (effect sizes −0.47, −0.57, −0.50 and −0.71). Furthermore, the pathological gamblers scored higher on ST1 (self-conscious) compared to controls, and the difference was large (effect size 0.85).

Table 3. Lower-order dimensions of character (TCI) for gamblers and controls
Full size table

The sum of SD and CO in character was lower than -1SD in 13 pathological gamblers (34.2%), and lower than -2SD in 11 of them (29%). Altogether, 24 of 38 (63%) had an immature personality, i.e. character. Of those below -1SD, 12 (32%) had an unspecified profile and one had a borderline profile (3%). In the other group below -2SD, seven (18%) had a schizotypal personality disorder and four (11%) had a non-specified personality disorder.

Discussion

As expected, pathological gamblers scored high on the temperament NS, with high impulsiveness, rigidity, extravagance and disorganised behaviour. These results are in agreement with previous studies (Allcock & Grace, 1988; Carlton & Manowitz, 1994; Kim & Grant, 2001; Moran, 1970; Steel & Blaszczynski, 1998).

Pathological gamblers also scored higher than controls on the temperament factor HA and its subscales. HA was increased (>1SD) in 15 of 38 (40%) pathological gamblers, with increased anticipatory worry (HA1, 36.8%), shyness (HA3, 45%) and especially increased fatigability and asthenia (HA4, 55%). These results are opposite to those of a previous study of pathological gamblers investigated with the TPQ (Kim & Grant, 2001).

PG is often associated with substance abuse and depression (Petry et al., 2005). This is also in agreement with our finding that affective disorders, anxiety disorders and substance use disorders were common among patients with PG (40–53%). High HA has also been shown to be associated with affective disorders, anxiety disorders and substance abuse (for a review see Cloninger et al., 1994). It is also a clinically well-known fact that high HA is common among those patient groups. Consequently, our finding of high HA is not unexpected. HA and its subscales are known to be state-dependent to some degree but the differences compared to controls are too large to be fully explained by this. Therefore, high HA might, along with NS, be a trait-like characteristic in PG.

The most pronounced difference was found in the character factor SD, with SD lower than -1SD in 24 of 38 (63%) patients. Pathological gamblers were generally immature on all subscales of SD, i.e. they were less responsible (SD1, 68%), less purposeful (SD2, 42%), less resourceful (SD3, 58%), and less self-acceptant (SD4, 37%) and had less impulse control (SD5, 58%). They also differed from controls in CO and ST, but these differences were moderate. The most important factors in CO and ST were more dishonest (CO5) and more self-conscious behaviour (ST1).

Fifteen of 24 pathological gamblers who had low SD also had high HA and all with high HA showed low SD. As expected, there was a relationship between high HA and low SD. High HA seems to have a negative influence on the development of character. Thirty-two of 38 patients (84%) were either low on SD, high on impulsivity (NS2) or displayed more dishonest behaviour (CO5). Therefore, personality (both temperament and character) seems to be important for PG behaviour.

We found that character was lower than -1SD in 13 patients (34%) and lower than -2SD in 11 patients (29%). According to Cloninger’s theory, individuals below -2SD have a personality disorder and approximately half of those between -1SD and -2SD also have a personality disorder. Thus, at least 29 or ∼46% of the patients had a personality disorder. In our study, schizotypal and a non-specified personality disorder were most common. These results are in good agreement with previous studies (Blaszczynski & Steel, 1998; Fernandez-Montalvo & Echeburua, 2004) with the exception that schizotypal personality disorder was most common in our study.

Pathological gamblers generally scored high on NS, high on HA, low on SD, low on CO and high on ST. This profile is similar to that of those with mixed drug abuse (P.-O. Nylander et al., unpublished observation). We also found that substance use disorders were common (53%) in PG.

Pathological gamblers seem to be a heterogeneous group. Approximately two-thirds of them displayed immature character with or without high HA in temperament. The other third displayed normal character, the most common personality traits being extravagant behaviour (86%), high impulsivity (36%) and less responsibility (50%).

Our findings underline the importance of measuring both temperament and character when evaluating personality in PG in future studies.