Reactions of Russian adolescents to reward and punishment: a cross-cultural study of the Gray–Wilson Personality Questionnaire

Reactions of Russian adolescents to reward and punishment: a cross-cultural study of the Gray–Wilson Personality Questionnaire

Personality and Individual Differences 30 (2001) 1211±1224 www.elsevier.com/locate/paid Reactions of Russian adolescents to reward and punishment: a...

118KB Sizes 0 Downloads 8 Views

Personality and Individual Differences 30 (2001) 1211±1224

www.elsevier.com/locate/paid

Reactions of Russian adolescents to reward and punishment: a cross-cultural study of the Gray±Wilson Personality Questionnaire Helena R. Slobodskaya a, Margarita V. Safronova a, Gennadij G. Knyazev a, Glenn D. Wilson b,* a

State Research Institute of Physiology, Siberian Branch of the Russian Academy of Medical Sciences, Timakova Str., 4, Novosibirsk, 630117, Russia b Institute of Psychiatry, University of London, De Crespigny Park, London SE5 8AF, UK Received 8 October 1999; received in revised form 1 April 2000

Abstract A translated, shortened version of the Gray±Wilson Personality Questionnaire was administered, along with an abbreviated adult Eysenck Personality Questionnaire, to 251 Russian schoolchildren aged 14±16 years. Girls were signi®cantly higher than boys on passive avoidance, ¯ight, and N, suggesting higher levels of anxiety/emotionality. In accord with British and Japanese results, principal component analysis of individual items largely failed to identify the six theoretical animal learning paradigms on which the GWPQ was constructed. Nor did intercorrelations among scale scores support the a priori three-dimensional structure of the test; instead there was evidence for two major systems underlying reactions to reward and punishment: Inhibition, with strong loadings on passive avoidance, extinction and ¯ight, and Activation, with positive loadings on ®ght and approach and a negative loading on active avoidance. Correlations with Rutter Teacher scores, the Child Behaviour Checklist and Youth Self-Report are also presented and discussed. As predicted, anxiety and emotional problems were associated with Inhibition, while conduct disorder and attention de®cits were associated with Activation. Multiple regression analysis showed that GWPQ scores were superior to scores on this version of the EPQ as predictors of teacher ratings and added signi®cant predictive power to parent and self-report measures of adolescent adjustment. # 2001 Elsevier Science Ltd. All rights reserved.

* Corresponding author. Tel.: +44-20-7848 0254; fax: +44-20-7083497. E-mail address: [email protected] (G.D. Wilson). 0191-8869/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII: S0191-8869(00)00104-5

1212

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1. Introduction The Gray±Wilson Personality Questionnaire (GWPQ; Wilson, Barrett & Gray, 1989) was designed to measure human equivalents of six animal learning paradigms basic to Gray's theory of personality. The six traits measured are approach and active avoidance (together said to comprise an Activation system), passive avoidance and extinction (comprising an Inhibition system) and ®ght and ¯ight (two components of a general emergency Fight/Flight system). Activation was conceived as relating to conditioned signals of impending reward (i.e., incentives) Inhibition to conditioned signals of threatened punishment (or frustrative non-reward) while ®ght-¯ight has recently been added by Gray as a third major system concerned with unconditioned aversive events. These three behavioural systems were conceived as fairly independent, having separate neurological underpinnings that vary in reactivity from one individual to another, thus producing observable personality di€erences. Given the diculty in devising scales that were both balanced for direction of scoring and that did not impinge conceptually on more than one learning paradigm, it is perhaps not surprising that factor analyses both in Britain (Wilson, Gray & Barrett, 1990) and Japan (Wilson, Barrett & Iwawaki, 1995) only partly matched the six trait a priori structure of the GWPQ. Fight, ¯ight, approach and extinction were readily identi®able with the British sample, but only ®ght, ¯ight and active avoidance in Japan. More importantly, the hypothesised pairings of the six traits into three major systems was not supported with either sample. Approach was associated with ®ght rather than active avoidance (indeed approach was negatively related to active avoidance). Passive avoidance was associated with extinction but equally so with ¯ight. Hence results were more supportive of Gray's earlier (1973) theory of personality which posited only two primary independent systems of impulsiveness (`go') and anxiety (`stop'). These might be called activation and inhibition, respectively, but ®ght seems to go with the former and ¯ight with the latter, thus redistributing the third dimension between the original two. Although it might seem that the original structure of the GWPQ should be replaced by one dictated by factor analysis, there are reasons to resist this move. Factor analysis is a useful procedure for checking hypotheses about the structure of a domain but it sometimes selects factors in which psychological content is contaminated with response biases such as direction of scoring e€ects. This can lead to the erroneous conclusion that conceptually opposite traits (like masculinity vs femininity, or neuroticism vs emotional stability) comprise independent factors rather than bipolar opposites. Factor analysis should be a servant, not a master. Furthermore, only by retaining the conceptual categories related to animal learning paradigms can hypotheses derived from theories such as Gray's be tested, e.g. the idea that approach and active avoidance derive from the same (activation) system and are therefore positively related. However, if the factorial structure of the GWPQ is unclear, then other evidence will be necessary to validate the scales, such as experimental data from human analogues of animal learning paradigms or psychophysiological correlates of the scales. The present study was designed to examine the factor structure of the GWPQ among Russian adolescents and to see to what extent the sex di€erences and EPQ correlates observed in Britain and Japan were replicated. In addition, it was thought useful to see if the personality traits relating to characteristic modes of response to experiences of reward and punishment measured by the GWPQ would be predictive of adolescent emotional and behavioural problems assessed by

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1213

self-report, parent and teacher ratings. The expectation was that an over-active punishment system would be associated with anxiety and emotional problems while an over-active reward system, implicated in impulsiveness, would predict conduct and attention disorders. Another suggestion is that attention-de®cit/hyperactivity (ADHD) is characterised by low anxiety, and hence an underactive punishment system (Quay, 1988). 2. Method The GWPQ was translated into Russian after excluding items that were considered inappropriate for school-age children (e.g. those that related to driving cars, changing jobs or feelings on revisiting one's old school). The 96 items remaining (out of the original 120) were administered to 251 Russian schoolchildren, aged 14±16 years (121 boys and 130 girls). They also completed a Russian short form of the Eysenck Personality Questionnaire containing 56 items from the adult EPQ-R (Hanin, Eysenck, Eysenck & Barrett, 1991). Adolescent adjustment was measured by a Russian translation of the Rutter Teacher Questionnaire (RTQ) Scale B2 (Rutter, 1967; Rutter, Tizard & Whitmore, 1970) and Russian short forms of the Child Behaviour Checklist (CBCL) and Youth Self-Report (YSR) developed from Achenbach (1991a,b). The RTQ, CBCL and YSR are widely used, standardised instruments for the study of children's competencies and problems (Ekblad, 1990; Verhulst & Achenbach, 1995) and the adequacy of the Achenbach forms has been demonstrated with Russian children (Carter, Grigorenko & Pauls, 1995; Slobodskaya, Savina & Starkova, 1998; Slobodskaya, 1999). Exact sample sizes vary because some subjects were either missing or added for each questionnaire administration (the EPQ sample was actually larger, consisting of 269 Ss). 3. Results Means, standard deviations and alpha coecients for the six GWPQ scales are shown in Table 1. Means, standard deviations and alpha coecients for the four EPQ scales are shown in Table 2. If these means seem low by comparison with British adult norms it is because they have Table 1 Means, standard deviations, and alphas for GWPQ scalesa Scale

AP AA PA EX FI FL a

Total sample (N=251)

Boys (N=121)

Girls (N=130)

Alpha

M

SD

Alpha

M

SD

Alpha

M

SD

T

p<

0.50 0.34 0.48 0.52 0.59 0.59

16.65 23.62 17.01 15.82 19.68 15.91

5.93 5.02 5.87 5.82 6.28 6.22

0.46 0.38 0.47 0.54 0.54 0.54

15.94 23.44 15.77 15.08 20.18 13.87

5.66 5.21 5.78 5.93 5.96 5.69

0.53 0.31 0.46 0.50 0.63 0.57

17.31 23.78 18.15 16.50 19.21 17.80

6.12 4.86 5.74 5.64 6.56 6.12

1.84 0.53 3.27 1.94 1.22 5.26

ns ns 0.001 0.054 ns 0.001

The Russian GWPQ Scale scores were transformed (*20/16) so that the expected mean would be equivalent to the original version (with 20 items per scale). AP=Approach, AA=Active Avoidance, PA=Passive Avoidance, EX=Extinction, FI=Fight, FL=Flight.

1214

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

Table 2 Means, standard deviations, and alphas for EPQ scales)a Scale

E N P L

Total sample (N=269)

Boys (N=131)

Girls (N=138)

Alpha

M

SD

Alpha

M

SD

Alpha

M

SD

T

p<

0.70 0.74 0.49 0.70

8.51 7.28 0.96 5.19

2.46 3.27 1.19 2.89

0.69 0.73 0.46 0.64

8.17 6.40 0.83 5.04

2.51 3.13 1.09 2.72

0.70 0.72 0.51 0.73

8.84 8.12 1.08 5.34

2.37 3.19 1.27 3.05

2.26 4.48 1.71 0.86

0.05 0.001 ns ns

a

The means are lower than British norms and those of Hanin et al. because they have not been corrected for shortened scale length.

not been corrected for the smaller number of items after scale shortening. Transforming these means according to numbers of items yields scale scores that are fairly equivalent to those reported for Russian adults by Hanin et al. (1991). Because most of the adolescents tested here were under 16 it might seem that the Children's version of the EPQ should have been used. The adult version was chosen because the wording seemed more attuned to the verbal level of the subjects and because a Russian translation was already available. As for the GWPQ, items thought to be inappropriate were omitted. In order to check the equivalence of this short form of the EPQ, and its validity with school-age adolescents, a factor analysis was conducted with a sample of 373 Russian adolescents (including those studied in this paper). This was used to re®ne the scoring system for E, N, P, and L, after which factor congruences of 0.97, 0.95, 0.95 and 0.95, respectively, were obtained with the loadings of Hanin et al. using a Russian adult sample. This supports the belief that the shortened version of the adult EPQ is valid with this adolescent sample, even though no norms are available for use with it. GWPQ alphas are low compared with the British sample but similar to those found with a Japanese sample. This is not surprising given the original requirement that each scale be balanced for direction of scoring (each item being matched with a near-equivalent reversal). In practice, it proved dicult to reverse an `approach' item without an element of `avoidance' creeping in (and vice versa). Combined with the vagaries of translation, it is not surprising that `clean', unidimensional factors corresponding to the six initial learning paradigms failed to emerge. Alphas for EPQ scales (Table 2) are predictably higher, especially the N scale, for which all items are scored in the same direction (i.e., endorsing an item always increases the N score). This perhaps illustrates the danger of seeking high alphas in questionnaire construction at the expense of all other criteria. Alphas are easily in¯ated by repetitive items with marginally di€erent wording, producing what Cattell called `bloated speci®cs' (Boyle, 1991). On the other hand, the GWPQ alphas given above are dangerously low, suggesting that some items probably bear little relationship to the scale of which they form a part and could therefore be dropped as the instrument is further developed. Tables 1 and 2 also show sex di€erences for the GWPQ and EPQ scales. Girls were signi®cantly higher on passive avoidance and ¯ight than boys, while the tendency for girls to be higher on extinction just failed to reach signi®cance. Girls were more emotional (higher on N) and slightly more extraverted (higher on E) than boys. There was no sex di€erence with respect to the P dimension.

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1215

Table 3 shows comparisons of means across the three cultures. Given the large number of comparisons, the statistically signi®cant level of di€erences was adjusted using a Bonferroni correction for 36 separate comparisons (6 scales/two sexes/three cultures, =0.05/36=0.001). There were four signi®cant di€erences among 12 comparisons between British and Russian Ss, three signi®cant di€erences between British and Japanese Ss, and eight signi®cant di€erences between Russian and Japanese Ss. Eight of 18 di€erences were signi®cant for males, and seven for females. The pattern of cross-cultural di€erences on extinction and ¯ight was similar for males and females. Table 4 shows signi®cant correlations among the GWPQ and EPQ scales for boys and girls separately. It is apparent that there is substantial replication of these correlations across the two sexes. Passive avoidance, extinction, and ¯ight show signi®cant correlations one with another Table 3 Mean di€erences in GWPQ scores of British, Russian and Japanese Ssa Males B-R AP AA PA EX FI FL

1.52 0.73 1.63 3.73* 1.99 3.07*

Females R-J 4.72* 2.21* 3.59* 3.78* 1.94 4.46*

B-J

B-R

3.20* 1.48 1.96 0.05 0.05 1.39

0.27 1.91* 0.15 3.81* 0.94 2.05

R-J 0.91 0.54 1.78 3.25* 4.41* 2.60*

B-J 0.64 2.45* 1.63 0.56 3.47* 0.55

a

B Ð British study (Wilson et al., 1989), N=84 males aged 26.718.16 years and 159 females aged 25.468.78 years; J Ð Japanese study (Wilson et al., 1995), N=289 males aged 19.721.32 years and 310 females aged 19.331.03 years; R Ð this study, N=121 males aged 14.500.68 years and 130 females aged 14.450.67 years. Positive value indicates that British Ss scored higher than Russian and Japanese, and Russian Ss scored higher than Japanese. *t-test, p<0.0001. Table 4 Noteworthy intercorrelations among GWPQ and EPQ personality scales (Ns vary between 111 and 138; all correlations above 0.18 are signi®cant beyond the 0.05 level, two-tailed) Boys Passavoidextinction Passavoid¯ight Extinction¯ight PassavoidN FlightN Fightapproach Flightapproach FightP Fight¯ight ApproachP Approachactavoid Flightactavoid

0.39 0.43 0.29 0.35 0.15 0.32 0.27 0.32 0.04 0.31 0.28 0.31

Girls 0.44 0.41 0.32 0.47 0.21 0.29 0.32 0.10 0.19 0.32 0.27 0.03

1216

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

Table 5 Loadings on six Oblimin rotated factors from the GWPQ (Top eight items provided higher 0.30). N=251 Russian adolescents Loading I. STOICISM (6.45% variance) 0.59 0.52 0.50 0.46 0.43 0.42 0.41 0.41 II. HOSTILITY (5.45% variance) 0.54 0.53 0.52 0.49 0.48 0.47 0.40 0.39 III. CAREFULNESS (3.21%variance) 0.52 0.52 0.45 0.45 0.42 0.42 0.38 0.37 IV. OPTIMISM (2.66% variance) 0.52 0.47 0.39 0.38 0.34 0.33 0.30 V. TOLERANCE (2.61% variance) 0.50 0.49 0.43 0.41 0.35 0.35 0.31

Summary of item content

A priori scale

Would panic in an earthquake Remain placid and calm if events go against one Calm if faced with wild animal Stalwart and courageous under pain Lost for words in public speech Could attend severe wounds Flinch with injection Easily turn down desired food when dieting

Flight+ Fight Flight Flight Passavoid+ Passavoid Flight+ Approach-

Paci®cist by nature Involved in ®ghts with other children Prefer to stay out of ®ghts Curse audibly if things go wrong Murder never contemplated Have felt like killing someone Sort it out with someone angry Would hit someone back

Fight Fight+ Fight Fight+ Fight Fight+ Passavoid Fight+

Careful in buying clothes Carry coat/umbrella in case of rain Inclined to buy things impulsively Clean teeth regularly Redouble e€orts if teacher not pleased If unsuccessful at ®rst, try, try again Have felt like killing someone Think carefully before buying

Approach Actavoid+ Approach+ Actavoid+ Actavoid+ Extinction Fight+ Approach

Carry on playing though losing Quickly give up when gambling unsuccessfully Make approach to attractive person Get excited in anticipation when opening presents Avoid talking with a sore throat Pursue attractive person despite rebu€s Would run from ferocious dog

Extinction Extinction+ Approach+ Approach+ Passavoid+ Extinction Flight+

Keep trying to make contact by telephone Hand over valuables to mugger Tolerant of people's pets Optimistic with lotteries and ra‚es Would go through with concert, despite bad feedback Remain calm if nearly hit by car At celebration, agitated waiting for start

Extinction Flight Fight Extinction Extinction Flight Approach+

(continued on next page)

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1217

Table 5 (continued) Loading

Summary of item content

A priori scale

VI. SENSITIVITY (2.45% variance) 0.52 0.46 0.42 0.41 0.40 0.36 0.35 0.35

Avoid company of people who don't like one Fall to pieces if criticised Give up if failed test Easily embarrassed Immune to criticism Stay out of way of people who are angry Yelp with pain if hurt Avoiding talking if throat infected

Extinction+ Passavoid+ Extinction+ Passavoid+ Passavoid Passavoid+ Flight+ Fight

(and tend to go with N), while ®ght, approach and P form another cluster. Approach is negatively related to active avoidance but positively related to ¯ight. Active avoidance correlates with ¯ight only for boys. The 96 items of the shortened GWPQ were subjected to principal components analysis and six factors were extracted, accounting for 22.83% of variance. These were rotated using Oblimin with Kaiser normalisation. The six-factor correlation matrix revealed no correlations between the rotated factors that were above 0.12. The decision to examine only the ®rst six factors was arbitrary in that the eigenvalues were declining progressively with no obvious cut-o€ point. Six factors were chosen to facilitate comparison with the British and Japanese data and because the a priori structure of the GWPQ was based on six components. The ®rst 10 eigenvalues were as follows: 6.19, 5.23, 3.08, 2.55, 2.51, 2.35, 2.21, 2.18, 2.11, 2.02, 1.99, 1.84. The highest loadings on each of the six rotated factors (the top eight provided no lower than 0.30) are shown in Table 5. Factor I, labelled `Stoicism', is similar to the ®rst factor found with the British sample and describes a capacity to remain calm under duress and su€er pain bravely. It straddles various a priori scales but with an emphasis on ¯ight. Factor II labelled `Hostility', could also be called ®ght because virtually all the high loading items come from this a priori scale. This is similar to the second factor found with the British sample, although scored in the opposite direction. Factor III, labelled `Carefulness', has three high loading items from active avoidance and three from approach, again emphasising the diculty in keeping these concepts separate with balanced scales. This factor is also similar to the third factor emerging with the British sample. In traditional personality terms it might be called `Impulsiveness'. Factor IV has been labelled `Optimism', and draws on both the approach and extinction scales. It is similar to that called `Resilience' in the British sample. Factor V has been labelled `Tolerance' but contains such a mixture of di€erent item contents that it is dicult to interpret. Factor VI has been labelled `Sensitivity' and illustrates the diculty in separating passive avoidance from extinction at the level of human personality traits. A principal components analysis of the six scale scores with varimax rotation yielded two main factors accounting for 34 and 24% of variance, respectively. The ®rst loaded on passavoid (0.86), extinction (0.70) and ¯ight (0.68), all rather cowardly, negative strategies for minimising punishment (inhibition?). The second loaded on ®ght (0.78), approach (0.72) and actavoid ( 0.60), which pits bold, positive ways of seeking rewards against active punishment avoidance (activation?).

1218

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

In case the choice of an orthogonal rotation should have made any di€erence to the above results the analysis of scales was repeated using direct oblimin. The factor loadings so produced were virtually identical. Table 6 shows correlations between GWPQ scales and Rutter scores for boys and girls separately. Most strikingly, conduct disorder and hyperactivity are predicted by ®ght scores (especially for boys). They also go with Approach and low levels of active avoidance (boys only). Emotional symptoms are linked with passive avoidance, low active avoidance (boys only) and low levels of ¯ight (girls only). The Child Behaviour Checklist and Youth Self-Reports (Tables 7 and 8) provide a generally similar picture. Although the sample sizes obtained with the CBCL are quite small, a signi®cant pattern emerges: anxiety, depression and somatic complaints (emotional problems) go with high active avoidance and low levels of ®ght, approach and ¯ight, whereas aggression and delinquency (behavioural problems) are more aligned with high approach and ®ght, and low levels of passive avoidance. Note, however, that the former association is apparent only with girls and the latter only with boys. Table 6 Correlations between GWPQ scales and Rutter Teacher scores for girls (boys in parentheses)a GWPQ scales

[Rutter scores] Total deviance

AP AA PA EX FI FL a

Conduct disorder

0.20 (0.27*) 0.02 ( 0.32**) 0.08 (0.02) 0.08 (0.04) 0.14 (0.45**) 0.28** (0.10)

Emotional symptoms

0.15 (0.32**) 0.00 ( 0.33**) 0.01 ( 0.07) 0.00 (0.02) 0.29** (0.57**) 0.20 (0.06)

Hyperactivity

0.21 (0.09) 0.01 ( 0.22*) 0.18 (0.25*) 0.08 (0.14) 0.10 (0.03) 0.21* (0.13)

0.18 (0.30**) 0.00 ( 0.33**) 0.08 ( 0.01) 0.03 (0.06) 0.23* (0.55**) 0.17 (0.04)

Girls, n=85; boys, n=86. *p<0.05, two-tailed. **p<0.01, two-tailed.

Table 7 Correlations between GWPQ and Russian short form of the CBCL scalesa CBCL scores

[GWPQ scales] AP

Somatic complaints Anxious/depressed Attention problems Delinquent behaviour Aggressive behaviour Emotional problems Behavioural problems Total problems a

0.25 (0.11) 0.34* (0.26) 0.09 (0.41*) 0.06 (0.39*) 0.07 (0.32) 0.34* (0.20) 0.07 (0.39*) 0.21 (0.38*)

AA 0.45** (0.10) 0.39* (0.02) 0.31 ( 0.19) 0.22 ( 0.14) 0.10 ( 0.23) 0.46** (0.07) 0.17 ( 0.21) 0.44** ( 0.09)

PA 0.01 0.07 0.02 0.19 0.03 0.01 0.08 0.04

EX (0.18) ( 0.10) ( 0.10) ( 0.39*) ( 0.24) ( 0.01) ( 0.35*) ( 0.17)

0.01 0.08 0.25 0.18 0.22 0.02 0.21 0.14

FI (0.11) (0.29) (0.18) ( 0.08) ( 0.06) (0.25) ( 0.08) (0.16)

0.43** ( 0.12) 0.37* ( 0.05) 0.15 (0.20) 0.09 (0.43**) 0.02 (0.42**) 0.42** ( 0.11) 0.04 (0.47**) 0.31 (0.16)

FL 0.41** (0.21) 0.41** (0.13) 0.16 (0.15) 0.27 ( 0.23) 0.23 ( 0.17) 0.38* (0.19) 0.27 ( 0.22) 0.40** (0.07)

Girls, upper line, n=40; boys (in parentheses), lower line, n=38. *p<0.05, two-tailed; **p<0.01, two-tailed.

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1219

Table 8 Correlations between GWPQ and Russian short form of the YSR scalesa YSR scores

[GWPQ scales] AP

Somatic Complaints Anxious Depressed Attention problems Delinquent behaviour Aggressive behaviour Emotional problems Behavioural problems Total problems a

0.00 (0.01) 0.05 ( 0.06) 0.05 (0.06) 0.15 (0.19*) 0.21* (0.23*) 0.27** (0.23*) 0.00 (0.00) 0.27** (0.26*) 0.15 (0.17)

AA 0.13 ( 0.09) 0.07 ( 0.01) 0.08 ( 0.06) 0.32** ( 0.40**) 0.14 ( 0.15) 0.24** ( 0.26**) 0.07 ( 0.07) 0.22* ( 0.25**) 0.20* ( 0.26**)

PA

EX

0.16 (0.18) 0.07 0.26** (0.19*)0.00 0.08 (0.36*) 0.10 0.10 (0.16) 0.14 0.09 ( 0.14) 0.12 0.04 ( 0.10) 0.04 0.21* (0.31*) 0.08 0.02 ( 0.13) 0.08 0.14 (0.18) 0.11

FI (0.02) (0.15) (0.27**) (0.22*) ( 0.07) (0.02) (0.18) ( 0.02) (0.16)

0.03 ( 0.11) 0.07 ( 0.27**) 0.02 ( 0.11) 0.22* (0.02) 0.37** (0.33**) 0.37** (0.25**) 0.05 ( 0.22*) 0.42** (0.32**) 0.19* (0.01)

FL 0.09 (0.18) 0.24** (0.22*) 0.03 (0.24*) 0.09 (0.24*) 0.15 (0.00) 0.10 (0.03) 0.06 (0.27*) 0.14 (0.02) 0.12 (0.24*)

Girls, upper line, n=119; boys (in parentheses), lower line, n=109. *p<0.05, two-tailed; **p<0.01, two-tailed.

Correlations with the YSR are lower but sample sizes are much larger, resulting in higher levels of signi®cance. This time anxiety and total emotional problems are associated with passive avoidance (boys and girls). This pattern is more consistent with results for Rutter Teacher scores. Anxiety is also associated with ¯ight (boys and girls) and low levels of ®ght (boys only), while depression is associated with passive avoidance, extinction and ¯ight (boys only). Aggression, delinquency and attention problems are associated with ®ght and approach, as well as low levels of active avoidance. Because of the signi®cant correlations between the GWPQ and EPQ scales, partial correlations between the GWPQ scales and adjustment measures controlling for the EPQ scales were examined. For the RTQ, 12 of 15 signi®cant bivariate correlations stayed and six appeared, resulting in 18 partial correlations signi®cant at p<0.05. For girls, four signi®cant correlations emerged with approach and two with ¯ight, while for boys, two bivariate correlations with approach lost their signi®cance. For the short CBCL, 15 of 22 signi®cant bivariate correlations stayed, and four appeared, resulting in 19 signi®cant partial correlations. Of those that lost signi®cance, three were with approach and four with ®ght. In addition, three signi®cant correlations emerged with ¯ight. For the short YSR, only 13 partial correlations were signi®cant. All seven correlations with approach, ®ve of eight with active avoidance, and ®ve of 10 with ®ght lost their signi®cance, while all correlations with ¯ight stayed. Although the values of the correlation coecients and their signi®cance level after controlling for the EPQ scales changed, the patterns of bivariate and partial correlation matrices were very similar. Finally, Table 9 employs step-wise multiple regression analyses to explore the contribution of sex, age, EPQ and GWPQ scores to the prediction of RTQ, CBCL and YSR scores. For teachers' ratings, GWPQ scales predicted adolescents' adjustment better than EPQ scales. Fight accounted for 7.8% of variance in total deviance and 17.8% of variance in conduct disorder scores, while passive avoidance and ¯ight were the only signi®cant predictors of emotional symptoms, accounting for 6.5% of variance. For parent's ratings, the contribution of GWPQ scales was comparable to EPQ scales. Passive avoidance was the best predictor of delinquent behaviour,

1220

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

Table 9 Multiple regression of adjustment measures on GWPQ and EPQ scales Dependent variable

Step

Predictor variables

RTQ scores Total deviance Conduct disorder

1

FI

1 2 3 1 2 1 2 3 4 5

Emotional symptoms Hyperactivity

CBCL scores Anxious/depressed Attention problems Delinquent behaviour Aggressive behaviour Emotional problems Behavioural problems YSR scores Somatic complaints Anxious Depressed Attention problems Delinquent behaviour Aggressive behaviour

R2, %



F

P<

7.75

+0.278

14.19

0.0002

FI Sex E PA FL FI Sex PA E FL

17.81 19.89 21.74 4.31 6.53 14.21 18.59 20.83 23.32 25.80

+0.422 0.144 +0.144 +0.208 0.164 +0.377 0.209 +0.160 +0.170 0.184

36.62 20.85 15.46 7.61 5.87 28.00 19.18 14.65 12.62 11.48

0.0001 0.0001 0.0001 0.005 0.005 0.0001 0.0001 0.0001 0.0001 0.0001

1 1 1 2 1 1 2 1 2

N P PA E L N AA E FL

6.07 6.88 7.21 13.75 5.35 7.86 14.00 5.78 12.09

+0.246 +0.262 0.268 +0.256 0.231 +0.280 +0.248 +0.240 0.253

4.72 5.40 5.67 5.74 4.13 6.23 5.86 4.48 4.95

0.05 0.05 0.02 0.005 0.05 0.02 0.005 0.05 0.01

1 2 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 4 5

N L N FL FI N Sex EX AA N Sex FI L E L Sex FI AA E

10.30 11.88 16.46 19.60 21.05 17.20 19.82 21.68 12.64 18.10 19.65 11.49 15.92 17.69 12.52 16.89 19.78 21.85 24.00

+0.321 0.126 +0.407 +0.183 0.123 +0.415 +0.168 +0.140 0.355 +0.234 +0.129 +0.339 0.229 +0.136 0.354 +0.209 +0.185 0.149 +0.153

25.83 15.10 44.33 27.29 19.82 46.74 27.69 20.58 32.54 24.76 18.18 29.20 21.21 15.98 32.21 22.76 18.33 15.52 13.96

0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

(continued on next page)

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1221

Table 9 (continued) Dependent variable

Step

Predictor variables

R2, %



F

P<

Emotional problems

1 2 3 1 2 3 4 5 1 2 3 4

N FL Age L FI Sex E AA N L Sex AA

23.93 25.67 26.98 15.14 19.75 23.58 25.46 27.22 14.55 20.91 23.48 25.81

+0.489 +0.136 0.115 0.389 +0.234 +0.196 +0.142 0.139 +0.381 0.252 +0.166 0.157

70.79 38.67 27.46 40.13 27.57 22.93 18.95 16.53 38.30 29.61 22.80 19.31

0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

Behavioural problems

Total problems

accounting for 7.2% of variance, while active avoidance was the second predictor of emotional problems after N, accounting for 7.2% of variance. E and ¯ight predicted behavioural problems, accounting for 6% of variance each. For self-ratings, N was the best predictor of somatic complaints, anxiety, depression and emotional problems, while ®ght, ¯ight and extinction contributed an additional 2±3%. Fight was the best predictor of delinquent behaviour, accounting for 11.5% of variance and the second predictor of behavioural problems after L, accounting for 4.6% of variance. Sex, age and all personality scales with the exception of approach made some contribution to the prediction of adjustment scores. 4. Discussion The present results con®rm some of the ®ndings using the GWPQ in British and Japanese society. Firstly, approach and active avoidance seem unlikely to re¯ect the operation of the same brain systems. As measured by self-report in human subjects they are negatively, rather than positively, related. Fight and ¯ight were positively correlated, but only slightly in girls and not at all in boys, so again it could be said that they cohere so poorly at the trait level that it is not useful to think of them as constituting a uni®ed personality dimension. Passive avoidance and extinction are positively related (as Gray's theory would predict) but ¯ight also goes along with this cluster. The factor analysis of the scale scores con®rms that Gray's earlier two-factor (anxiety and impulsiveness) conceptualisation ®ts the data more comfortably than the three-factor extension of the theory. Sensitivity to punishment (whether conditioned signals or unconditioned aversion) appears as one major dimension, while sensitivity to reward (incorporating both appetitive impulses and hostile forms of assertiveness) emerges as the other. These two clusters do seem to have some independence that can be sustained even using scales that are balanced for direction of scoring. As noted above, it would be all too easy to develop independent scales of approach and avoidance by `blind' use of factor analysis in which endorsement bias is confounded with item content. (For a fuller discussion of this methodological issue see Wilson, 1975).

1222

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

The correlations between EPQ and GWPQ traits are generally in accord with expectations (Table 4). N goes with ¯ight and passive avoidance (anxiety or inhibition traits) and P goes with ®ght and approach (impulsive or activation traits). One might have expected E to go with activation also, as Gray (1973) would have supposed, but over the years it has become clear that Impulsiveness as a primary trait is more aligned with P than E. After the introduction of a third factor (P) to Eysenck's system of personality description, E became increasingly focused around sociability. Therefore it is not surprising that P has more relevance to Gray's reinforcement sensitivity theory than E (Diaz & Pickering, 1993). The British and Japanese data showed women to be higher on the sensitivity to punishment scales than men (particularly ¯ight and active avoidance). This was true also for the Russian data except that passive avoidance achieved statistical signi®cance rather than active avoidance. This re¯ects the well-established ®nding that females score higher on anxiety or N scales than men (Eysenck, Barrett, Wilson & Jackson, 1992). One might equally expect that, with their reputation for aggressiveness and higher P scores, men would produce higher scores on the ®ght and approach (activation) scales. This was true for the Japanese sample, but not for either British men or the Russian adolescents tested here. It is not immediately clear why this should be so, unless one supposes that the feminist message that women should assert themselves and express their anger openly has taken hold more in Europe than Japan (which retains more traditions in society). In fact, decades of a Soviet ideology that promoted equality and full-time work for women may have reduced gender di€erences by comparison even with Western society; Slobodskaya (1994) describes research with Cattell's 16PF which failed to show some `traditional' sex di€erences observed in the US. Such an interpretation is supported by the cross-cultural comparison of means (Table 3) where it is clear that Japanese females stand out as low on ®ght by comparison with British and Russian females, whereas Japanese males are higher on approach. Contrary to the ®ndings with adults (Hanin et al., 1991) Russian girls were more extraverted than boys. Evidence for gender di€erences in E in the adolescent years is equivocal; in some studies girls score higher, in others boys, and some studies report no gender di€erences (Francis, 1996). These contradictory results do not seem to be related to culture but may be accounted for in terms of di€erent time courses of the social development of boys and girls (Anthony, 1973; Eysenck, 1996). The same is probably true with respect to the failure to ®nd any sex di€erences in P within this adolescent sample (whereas males were higher than females in the Hanin et al. study of Russian adults). Since P may have links with testosterone (Zuckerman, 1991) it is possible that sex di€erentiation only emerges beyond puberty into early adulthood. Another ®nding of interest in this study concerns the relationships between GWPQ scores and adjustment diculties. It was hypothesised that sensitivity to punishment (inhibition) would be related to emotional disturbances, such as anxiety and depression, while sensitivity to reward (activation) would predict behavioural problems such as delinquency and hyperactivity/attention de®cit. This is precisely what was observed, both as measured by self-report and by teacher and parent ratings, con®rming the likelihood that adolescent adjustment problems are partly rooted in brain systems basic to innate individual di€erences in reactions to reward and punishment. Since it is already known that EPQ scores predict emotional and conduct disorders in children (e.g., Tranah, Hartnett & Yule, 1998) the question could reasonably be asked as to whether the GWPQ correlates of adolescent adjustment could be accounted for by EPQ scores. This is

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

1223

important because, given its factorial impurity, the importance of the GWPQ stands or falls on its predictive validity. If it is unable to add anything by way of prediction to the EPQ then it would surely have to be abandoned as an independent personality measure. Results of the partial correlation analysis and multiple regression (Table 9) show that when EPQ scores are statistically controlled the GWPQ does indeed retain predictive power. All of the Rutter teacher ratings are better predicted by GWPQ scores than EPQ scores, as well as delinquency on the CBCL and YSR and attention problems on the YSR. On many other adjustment measures GWPQ scores add signi®cantly to the prediction of EPQ scales. N is generally the best predictor of self-rated emotional problems but this is not surprising given the similar wording of many EPQ and YSR items (worry a lot, feel guilty, nervous, lonely, etc.). Hence this study supports the predictive validity of the GWPQ and encourages further work using this instrument in domains where individual di€erences in reactions to reward and punishment might be relevant. Acknowledgements We are grateful to Dr Paul Barrett for conducting the analysis of factor congruence between the salient factor loadings in our study and those of Hanin et al. (1991). References Achenbach, T. M. (1991a). Manual for the Youth Self-Report and 1991 pro®le. Burlington: University of Vermont, Department of Psychiatry. Achenbach, T. M. (1991b). Manual for the Child Behaviour Checklist/4-18 and 1991 pro®le. Burlington: University of Vermont, Department of Psychiatry. Anthony, W. (1973). The development of extraversion, of ability, and of the relation between them. British Journal of Educational Psychology, 43, 223±227. Boyle, G. J. (1991). Does item homogeneity indicate internal consistency or item redundancy in psychometric scales. Personality and Individual Di€erences, 12, 291±294. Carter, A. S., Grigorenko, E. L., & Pauls, D. L. (1995). A Russian adaptation of the Child Behaviour Checklist: Psychometric properties and associations with child and maternal a€ective symptomatology and family functioning. Journal of Abnormal Child Psychology, 23, 661±684. Diaz, A., & Pickering, A. D. (1993). The relationship between Gray's and Eysenck's personality spaces. Personality and Individual Di€erences, 15, 297±305. Ekblad, S. (1990). The Children's Behaviour Questionnaire for completion by parents and teachers in a Chinese sample. Journal of Child Psychology and Psychiatry, 31, 775±791. Eysenck, H. J. (1996). Personality and the experimental study of education. European Journal of Personality, 10, 427±439. Eysenck, H. J., Barrett, P. T., Wilson, G. D., & Jackson, C. (1992). Primary trait measurement of the 21 components of the P-E-N system. European Journal of Psychological Assessment, 8, 109±117. Francis, L. J. (1996). The development of an abbreviated form of the Revised Junior Eysenck Personality Questionnaire (JEPQR-A) among 13±15 year olds. Personality and Individual Di€erences, 21, 835±844. Gray, J. A. (1973). Causal theories of personality and how to test them. In J. R. Royce, Multivariate analysis and psychological theory (pp. ±). New York: Academic Press. Hanin, Y., Eysenck, S. B. G., Eysenck, H. J., & Barrett, P. T. (1991). A cross-cultural study of personality: Russia and England. Personality and Individual Di€erences, 12, 265±271. Quay, H. C. (1988). The behavioural reward and inhibition system in childhood behaviour disorder. In L. M. Bloomingdale, Attention-De®cit Disorder: New research in attention, treatment and psychopharmocology. Oxford: Pergamon Press.

1224

H.R. Slobodskaya et al. / Personality and Individual Di€erences 30 (2001) 1211±1224

Rutter, M. (1967). A children's behaviour questionnaire for completion by teachers: preliminary ®ndings. Journal of Child Psychology and Psychiatry, 8, 1±12. Rutter, M., Tizard, J., & Whitmore, K. (1970). Education, health and behaviour. London: Longman. Slobodskaya, H. R. (1994). Child development: The modern scienti®c views and the Russian reality. In H. R. Slobodskaya, Ideal and real childhood: collections of articles of contemporary Western scientists (pp. 10±32). Novosibirsk: Sibirsky Khronograf. Slobodskaya, H. R. (1999). Competence, emotional and behavioural problems in Russian adolescents. European Child and Adolescent Psychiatry, 8, 173±180. Slobodskaya, H. R., Savina, N. N., & Starkova, N. V. (1998). Russian short forms of the Teacher Temperament Questionnaire and Child Behaviour Checklist. Paper presented at Ninth European Conference on Personality, University of Surrey, Guildford, June. Tranah, T., Hartnett, P., & Yule, W. (1998). Conduct disorder and personality. Personality and Individual Di€erences, 24, 741±745. Verhulst, F. C., & Achenbach, T. M. (1995). Empirically based assessment and taxonomy of psychopathology: crosscultural applications. A Review. European Child and Adolescent Psychiatry, 4, 61±76. Wilson, G. D. (1975). The ``catchphrase'' approach to attitude measurement. Personality and Individual Di€erences, 6, 31±37. Wilson, G. D., Barrett, P. T., & Gray, J. A. (1989). Human reactions to reward and punishment: A questionnaire examination of Gray's personality theory. British Journal of Psychology, 80, 509±515. Wilson, G. D., Barrett, P. T., & Iwawaki, S. (1995). Japanese reactions to reward and punishment: A cross-cultural personality study. Personality and Individual Di€erences, 19, 109±112. Wilson, G. D., Gray, J. A., & Barrett, P. T. (1990). A factor analysis of the Gray±Wilson Personality Questionnaire. Personality and Individual Di€erences, 10, 1037±1045. Zuckerman, M. (1991). Psychobiology of Personality. New York: Cambridge University Press.