parental attachment and its relationship to contextualised trait expression and mean-corrected...

6
Parental attachment and its relationship to contextualised trait expression and mean-corrected cross-context trait variability Oliver Robinson a,, Gordon R.T. Wright b,1 , Elaine Kendall a,2 a Dept. of Psychology and Counselling, University of Greenwich, Avery Hill Road, London, SE9 2UG, UK b School of Psychology, Birkbeck College, University of London, Malet Street, London, WC1E 7HX, UK article info Article history: Received 12 July 2010 Received in revised form 4 November 2010 Accepted 13 November 2010 Available online 30 December 2010 Keywords: Personality Social context Parental attachment Intra-individual variability Big Five abstract The study explored relationships between attachment to parents, personality traits and cross-context trait variability in adulthood. The 209 adult participants completed the Parental Attachment Questionnaire (PAQ), and rated themselves on a Big Five trait inventory to describe their personality as it manifests when with parents, friends and work colleagues. Indices of intra-individual variability were corrected to negate correlations between trait variability and trait means. It was found that weaker parental attachment was related to higher levels of cross-context variability in Agreeableness, Openness to Experience and Emotional Stability (Neuroticism). Correlations between an overall index of cross- context personality variability and trait means were moderated by context. PAQ scores were linked to all five trait means in the parent context, but not to trait means in other contexts. The findings support a link between the attachment bond with parents and particular personality traits in respect to both cross-context variability and mean level. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Personality is to a degree socially malleable; a variety of studies have found that adults, when asked to rate their personality as it manifests in different social contexts, report significant mean-level variability between these contexts, both retrospectively (Block, 1961; Donahue & Harary, 1998; Robinson, 2009) and when they rate behaviour while in the contexts in question (Heller, Watson, Komar, Min, & Perunovic, 2007). This ‘cross-context variability’ is paradoxically systematic; people generally rate themselves as gen- erally more extraverted, agreeable and open when with friends, more conscientious at work and more neurotic around parents (Donahue & Harary, 1998; Roberts & Donahue, 1994; Robinson, 2009). The extent to which people vary their personality traits across social situations can be captured in a numerical index, and various studies have demonstrated that such indices are related to other dispositional variables and mental health indicators. Sheldon, Ryan, Rawsthorne, and Ilardi (1997) found that higher cross- context variability was related to lower authenticity and lower self-esteem. Other studies have found a link between high cross- context personality variability and anxiety (Diehl, Hastings, & Stanton, 2001), depression (Lutz & Ross, 2003); lower life- satisfaction (Suh, 2002), lower role-satisfaction (Donahue, Robins, Roberts, & John, 1993) and poorer physical health (Cross, Gore, & Morris, 2003). The picture is one of personality variability being an undesirable characteristic. However, recently this correlative relation between cross- context variability and measures of wellbeing was called into question, when Baird, Le, and Lucas (2006) highlighted sources of error within cross-context personality variability indices. They firstly criticised the most common index calculation, which involves running a Primary Components factor analysis on each person’s data. To calculate this index, the variance that cannot be accounted for by the first factor (and therefore is not shared across the contexts) is used as an index of variability for each participant. This calculation was shown by Baird et al. to be flawed, as the resulting index reflects not only cross-context variability, but also cross-trait variability: if trait means are more variable then the index is inflated. Thus the index cannot be said to be a reflection of just the cross-context trait malleability that it is intended to capture. Baird et al. (2006) suggested that a preferable index of cross- context trait variability is derived from cross-context standard deviations for each participant. The standard deviation of a trait’s scores across the different contexts indicates how much variability a person shows across those contexts. However this index, while 0191-8869/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.paid.2010.11.024 Corresponding author. Tel.: +44 (0) 208 331 9630. E-mail addresses: [email protected] (O. Robinson), [email protected] (G.R.T. Wright), [email protected] (E. Kendall). 1 Tel.: +44 (0) 207 631 6000. 2 Tel.: +44 (0) 208 331 9630. Personality and Individual Differences 50 (2011) 547–552 Contents lists available at ScienceDirect Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid

Upload: oliver-robinson

Post on 11-Sep-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Personality and Individual Differences 50 (2011) 547–552

Contents lists available at ScienceDirect

Personality and Individual Differences

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

Parental attachment and its relationship to contextualised trait expressionand mean-corrected cross-context trait variability

Oliver Robinson a,⇑, Gordon R.T. Wright b,1, Elaine Kendall a,2

a Dept. of Psychology and Counselling, University of Greenwich, Avery Hill Road, London, SE9 2UG, UKb School of Psychology, Birkbeck College, University of London, Malet Street, London, WC1E 7HX, UK

a r t i c l e i n f o

Article history:Received 12 July 2010Received in revised form 4 November 2010Accepted 13 November 2010Available online 30 December 2010

Keywords:PersonalitySocial contextParental attachmentIntra-individual variabilityBig Five

0191-8869/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.paid.2010.11.024

⇑ Corresponding author. Tel.: +44 (0) 208 331 9630E-mail addresses: [email protected] (O. Rob

(G.R.T. Wright), [email protected] (E. Kendall).1 Tel.: +44 (0) 207 631 6000.2 Tel.: +44 (0) 208 331 9630.

a b s t r a c t

The study explored relationships between attachment to parents, personality traits and cross-contexttrait variability in adulthood. The 209 adult participants completed the Parental AttachmentQuestionnaire (PAQ), and rated themselves on a Big Five trait inventory to describe their personality asit manifests when with parents, friends and work colleagues. Indices of intra-individual variability werecorrected to negate correlations between trait variability and trait means. It was found that weakerparental attachment was related to higher levels of cross-context variability in Agreeableness, Opennessto Experience and Emotional Stability (Neuroticism). Correlations between an overall index of cross-context personality variability and trait means were moderated by context. PAQ scores were linked toall five trait means in the parent context, but not to trait means in other contexts. The findings supporta link between the attachment bond with parents and particular personality traits in respect to bothcross-context variability and mean level.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Personality is to a degree socially malleable; a variety of studieshave found that adults, when asked to rate their personality as itmanifests in different social contexts, report significant mean-levelvariability between these contexts, both retrospectively (Block,1961; Donahue & Harary, 1998; Robinson, 2009) and when theyrate behaviour while in the contexts in question (Heller, Watson,Komar, Min, & Perunovic, 2007). This ‘cross-context variability’ isparadoxically systematic; people generally rate themselves as gen-erally more extraverted, agreeable and open when with friends,more conscientious at work and more neurotic around parents(Donahue & Harary, 1998; Roberts & Donahue, 1994; Robinson,2009).

The extent to which people vary their personality traits acrosssocial situations can be captured in a numerical index, and variousstudies have demonstrated that such indices are related to otherdispositional variables and mental health indicators. Sheldon,Ryan, Rawsthorne, and Ilardi (1997) found that higher cross-context variability was related to lower authenticity and lower

ll rights reserved.

.inson), [email protected]

self-esteem. Other studies have found a link between high cross-context personality variability and anxiety (Diehl, Hastings, &Stanton, 2001), depression (Lutz & Ross, 2003); lower life-satisfaction (Suh, 2002), lower role-satisfaction (Donahue, Robins,Roberts, & John, 1993) and poorer physical health (Cross, Gore, &Morris, 2003). The picture is one of personality variability beingan undesirable characteristic.

However, recently this correlative relation between cross-context variability and measures of wellbeing was called intoquestion, when Baird, Le, and Lucas (2006) highlighted sources oferror within cross-context personality variability indices. Theyfirstly criticised the most common index calculation, whichinvolves running a Primary Components factor analysis on eachperson’s data. To calculate this index, the variance that cannot beaccounted for by the first factor (and therefore is not shared acrossthe contexts) is used as an index of variability for each participant.This calculation was shown by Baird et al. to be flawed, as theresulting index reflects not only cross-context variability, but alsocross-trait variability: if trait means are more variable then theindex is inflated. Thus the index cannot be said to be a reflectionof just the cross-context trait malleability that it is intendedto capture.

Baird et al. (2006) suggested that a preferable index of cross-context trait variability is derived from cross-context standarddeviations for each participant. The standard deviation of a trait’sscores across the different contexts indicates how much variabilitya person shows across those contexts. However this index, while

548 O. Robinson et al. / Personality and Individual Differences 50 (2011) 547–552

preferable to the factor analytic approach, can still be affected bytrait means: means and standard deviations on a scale areinter-dependent – high and low means permit smaller standarddeviations, while larger standard deviations are possible withmeans towards the middle of a scale. If a distribution is skewed,then a standard deviation measure can become indirectly a mea-sure of trait means. For example, if a distribution is skewed to-wards the low and middle levels of a trait, then across thesample lower means will show lower standard deviations, andhigher means will show higher standard deviations.

Baird et al. (2006) suggested a simple process for correctingcross-context standard deviations so that they are not correlatedwith means. For each personality item; they suggested running aregression using the standard deviation of each item as the DV,and the mean and square of the mean as IVs. Residuals from theregression quantify the amount of variance in the item standarddeviation that is not accounted for by the mean. The residuals ofall items are then averaged to produce the mean-corrected vari-ability index. Baird et al. (2006) found that this mean-correctedmeasure of personality consistency was found not to correlate withwellbeing, suggesting that the previously found association be-tween personality variability and wellbeing may be an artefact ofthe link between wellbeing and trait means. However it has notyet been established whether or not a mean-corrected variabilityindex may link to other variables beyond wellbeing measures suchas attachment.

The current study investigated the link between intra-individual personality trait variability, contextualised trait expres-sion, trait means and parental attachment. The rationale forstudying the link between parental relationships and cross-context personality variability is found in the construct of thefalse self or persona. Querstret and Robinson (2010) found thatindividuals who show high levels of cross-context trait variabilitydescribe the motivation for adapting their personality to be theuse of social personae or ‘false selves’ – public fronts that morphpersonality into a socially acceptable or impressive form for a par-ticular context, and they frequently linked this to problematicfamily relationships. Theorists have suggested that false selves,and the socially variable personality patterns that result fromtheir use, may result from particular attachments with parents.For example, Winnicott (1960) suggested that individuals whohave a problematic or insecure attachment to their parents aschildren may have an ongoing fear of rejection in adult life andso habitually adapt their personality in order to be accepted byothers and to minimise the possibility of not being sociallyaccepted. This suggestion has empirical support from a varietyof areas; Block (1961) found that those who were more variablein the social presentation of personality described higher levelsof familial tension; a study by Schoenrock (1991) evidenced arelationship between higher familial conflict and higher self-monitoring (the tendency to monitor social cues and changebehaviour to fit the context); and Robinson (2008) found thatcross-context personality variability was related to high levels ofneuroticism when with parents, but not to neuroticism in theother social contexts measured. A more recent study by Gillath,Sesko, Shaver, and Chun (2010) found that inauthenticity in youngadults (the tendency to conceal thoughts, feelings and prefer-ences) was related to insecure attachment to partners. Howeverthe link between parental attachment and personality variabilityremains an unchartered empirical territory.

1.1. Aims and hypotheses

The aim of the current study was to explore the predictedlink between cross-context trait variability and parental attach-ment, while also looking at trait means and employing a modi-

fied version of the mean-correction process suggested by Bairdet al. (2006) to avoid conflation of cross-context variance andmean trait levels. The study investigated cross-context varia-bility at the level of personality as a whole and at the level ofeach trait.

Three hypotheses were made based on existing research(Robinson, 2008, 2009). It was predicted that: (1) Personality traitswould show significant mean variability across social contexts. (2)Secondly, it was predicted that personality variability and parentalattachment would relate to personality traits in a context-specificmanner, and show the strongest link with the ‘with parents’context. (3) Thirdly, it was predicted that those with a strongattachment to parents would vary their personality traits less thanthose with a weaker parental attachment. (4) Finally, it waspredicted that both trait means and trait variability indices wouldaccount for incremental attachment variance.

2. Method

2.1. Participants

209 adults (81 males and 128 females) from London partici-pated in the study. Participants ranged in age from 18 to 59(mean = 29.20, SD = 10.72). These individuals were recruited froman adult education college (N = 77), an undergraduate cohort at aLondon University (N = 80) and several workplaces in South EastLondon (N = 52).

2.2. Measures

2.2.1. The Ten-Item Personality Inventory – three contexts (TIPI-3C)The TIPI-3C is a self-report measure that assesses how the Big

Five traits manifest in three life contexts; with parents, withfriends and with work colleagues (Robinson, 2009). The measurewas adapted from the Ten-Item Personality Inventory, or TIPI(Gosling, Rentfrow, & Swann, 2003), a very brief measure of the fivefactor model of personality that has been shown to have predictivevalidity in a variety of settings (e.g., Robinson, Demetre, & Corney,2010), test–retest reliability of r = 0.72 and convergent correlationswith a number of longer Big Five inventories (Gosling et al., 2003).The 10 items consist of pairs of similar trait adjectives that arerated on a 7-point Likert scale ranging from 1 (disagree strongly)to 7 (agree strongly). Each item is responded to three times, witheach time corresponding to how personality is with either parents,friends or work colleagues. Cronbach Alphas in the TIPI-3C showedgood internal consistency (Extraversion, a = 0.65, Openness,a = 0.77, Conscientiousness, a = 0.80, Emotional Stability, a = 0.70and Agreeableness, a = 0.66). The TIPI refers to EmotionalStability (ES) rather than Neuroticism, and the current paper doesthe same.

2.2.2. The Parental Attachment Questionnaire (PAQ)The Parental Attachment Questionnaire assesses parental

attachment in adolescents and adults (Kenny, 1987). It is com-prised of 55 items rated on a five-point Likert rating scale (1 notat all – 5 very much). Kenny (1987) reports overall Cronbach alphasof a = 0.93 and a = 0.95 for male and female samples respectively.The version used in the current study gives a combined attachmentrating for both parents. Items on the questionnaire are divided intothree subscales: (1) Affective Quality of Relationship with Parents, (2)Parents as Facilitators of Independence and (3) Parents as a Source ofSupport. The present study found Cronbach alphas of a = 0.95,a = 0.87 and a = 0.86 for the above subscales respectively.Responses to items for each subscale were averaged to obtain thesubscale score. Subscale scores were summed for a total attach-ment score.

O. Robinson et al. / Personality and Individual Differences 50 (2011) 547–552 549

2.3. Procedure

Questionnaires were distributed to participants at participatingcolleges and workplaces. Prospective participants were informedthat in order to participate they need to be in regular contact withat least one parent, and be in part-time or full-time work. Partici-pants then signed a consent form and gave a personal code foridentifying their data at a later date if required. They gave theirage and gender before completing the TIPI-3C and PAQ. Participa-tion in the study was voluntary and there was no credit gained forstudents by participating.

2.3.1. Calculating mean-corrected cross-context variability index (MC-PVI)

For each of the ten items in the TIPI-3C (e.g., item 1 ‘‘extra-verted, enthusiastic’’), participants gave three ratings to describethemselves with parents, work colleagues and friends. The stan-dard deviation of scores in these three settings was calculated foreach item. Then the standard deviations for the two TIPI items thatcontributed to each trait were averaged, to give a trait standarddeviation. This was followed by a modified version of the Bairdet al. (2006) means-correction process; in their procedure the stan-dard deviation for each trait was regressed on the trait mean andsquare of the mean – the saved residual for each participant be-came their mean-corrected variability index for that trait. Theresiduals for all five traits are averaged to give an overall mean-correct personality variability index (MC-PVI). However, this pro-cedure uses two highly correlated predictor variables (trait meanand square of the trait mean), which leads to possible multicollin-earity in the regression. We compared indices with and withoutthe square of trait means as a predictor, and they correlated witheach other at r = 0.98 while having correlations with parentalattachment within r = 0.01 of each other. Therefore it was decidedto use the mean-correction process without including the square ofthe mean, thus avoiding multicollinearity concerns.

3. Results

Due to the number of statistical tests conducted, all inferentialanalyses were required to meet the p < 0.01 level to be consideredsignificant.

3.1. Cross-context variability and correlative consistency in the BigFive traits

Figure 1 shows the sample means for the five traits across thethree contexts. Extraversion, Openness and Agreeableness arehighest with friends, while Emotional Stability and Consciousnessare highest in the work context. A five (traits) x three (contexts) re-peated measures ANOVA was performed revealing significant maineffects for both trait (F4,832 = 9.260, p < 0.001, partial g2 = 0.043)and context (F2,416 = 36.192, p < 0.001, partial g2 = 0.148) as wellas a significant interaction of trait and context (F8,1664 = 18.715,p < 0.001, partial g2 = 0.116). This shows systematic variability,with trait-specific patterns, across all three contexts for each ofthe trait scores.

3.2. Personality variability and parental attachment: relationshipswith contextualised Big Five trait scores

It was predicted that personality variability would have a con-text-specific relationship with the Big Five traits as reported inthe three social settings. The top half of Table 1 presents thecorrelations between overall mean-corrected personality variabil-ity and contextualised traits. The figures suggest that personality

variability is related to being less extraverted at work, moreextraverted with friends, being more open with friends, and withbeing more prone to negative emotions and less agreeable whenwith parents. This pattern of small to medium effect sizes iscomprehensible within the theoretical framework that informsthis paper, and will be interpreted in the discussion.

Correlations were calculated between total parental attachmentand each of the contextualised Big Five trait scores (see Table 1,bottom half). Significant positive correlations were found for alltraits within the ‘‘with parents’’ context with small to moderate ef-fect sizes, but no significant results were found in either the ‘‘withwork colleagues’’ or ‘‘with friends’’ scenarios, showing a clear linkbetween the strength of parental attachment and the expression ofpersonality traits when with parents.

3.3. Relationships between trait variability and parental attachment

It was predicted that mean-corrected variability for all traitswould correlate negatively with parental attachment, i.e., thatthose with weaker parental attachments would vary their person-ality more. However, cross-context variability in just three traitsrelated to parental attachment; Agreeableness (moderate effectsize), Openness to Experience (small effect size) and EmotionalStability (moderate effect size). Mean-corrected variability inExtraversion and Conscientiousness was not linked to attachment.These findings (shown in Table 2) suggest that those with astronger attachment to parents vary their personality less acrosscontexts, as predicted, but that this effect is trait-specific.

The Parental Attachment Questionnaire is comprised of threesubscales: Affective Quality of Relationship (Mean = 3.86, SD =0.78), Parents as Facilitators of Independence (Mean = 3.66, SD =0.76), and Parents as Source of Support (Mean = 3.34, SD = 0.78). Inorder to explore the respective predictive contribution of thesesubscales to MC-PVI, a regression was conducted with the subscaletotals as IVs (Enter method), and MC-PVI as DV. It was found thatthe Affective subscale was the only significant predictor (b =�0.31) in the resulting model (R2 = 0.40, F(3205) = 13.22), whileIndependence (b = �0.14) and Support (b = 0.01) subscales werenot significant predictors.

3.4. Trait means and trait variability as predictors of parentalattachment

Although the current study is based on a theoretical suppositionthat parental attachment may lead to personality variability, pre-diction from the latter to the former is also possible and potentiallyuseful. It was hypothesised that both cross-context trait means andvariability indices would account for variance in parental attach-ment. In order to test this hypothesis, a regression was conductedin which total parental attachment was regressed onto all fivecross-context trait means and all five means-corrected trait stan-dard deviations. The resulting model is shown in Table 3. AdjustedR2 was 0.33, suggesting that if generalized to the population, 33% ofthe variability in parental attachment can be accounted for by theBig Five trait means and variability indices. Those predictors thatexceeded the 0.01 significance level are variability in EmotionalStability, variability in Agreeableness, variability in Openness,and mean level of Agreeableness.

4. Discussion

The current study hypothesised that personality traits wouldvary significantly across work, parent and friend contexts. This pre-diction was supported, and the rank-order of the means for the BigFive traits across the contexts of with parents, friends and work

Fig. 1. Sample means for the Big Five traits, rated for ‘with parents’, ‘with work colleagues’, and ‘with friends’ contexts (N = 209). Non-overlapping confidence intervals ofmeans (95% level) were shown between at least two contexts for all traits. Overlapping CIs were between ‘with parents’ and ‘with work colleagues’ for Openness andAgreeableness, between ‘with parents’ and ‘with friends’ for Conscientiousness and between the ‘with work colleagues’ and ‘with friends’ contexts for Emotional Stability.

Table 1Significant correlations between mean-corrected personality variability index (MC-PVI) and Big Five traits in contexts; and correlations between total parental attachment andBig Five traits in contexts.

Extraversion Openness Conscientiousness Emotional Stability Agreeableness

MC-PVI Parent �0.13 �0.07 �0.12 �0.22** �0.23**

Work �0.24** �0.09 0.10 0.13 0.01Friends 0.22** 0.30*** 0.03 0.05 0.15

Attachment Parent 0.38*** 0.24*** 0.21*** 0.45*** 0.42***

Work 0.08 0.10 0.14 0.03 0.15Friends 0.03 �0.03 0.16 0.11 0.10

** Significant correlations at p < 0.01.*** Significant correlations at p < 0.001.

Table 2Correlations between mean-corrected trait cross-context variability indices; overall mean-corrected personality variability (MC-PVI) and total parental attachment.

E variability A variability C variability ES variability O variability MC-PVI

PAQ: total attachment �0.08 �0.33*** �0.15 �0.38*** �0.29*** �0.39***

*** Significant correlations at p < 0.001.

550 O. Robinson et al. / Personality and Individual Differences 50 (2011) 547–552

colleagues is the same as in Robinson (2009), showing the robustnature of cross-context variation patterns for all five traits at thelevel of total sample.

The second hypothesis was that mean-corrected personalityvariability and parental attachment would relate to the personalitytraits in a context-specific manner, and show the strongest linkwith the parent context. Personality variability was found to relateto being less emotionally stable and agreeable with parents, sup-porting the findings of Robinson (2008). This can be interpretedthrough the framework of social personae and false selves. A falseself may originate from problematic home environments, in which

case it is linked to ongoing emotional and antagonistic problems athome. Future research that explores the links between attachment,personality variability and experiences of authenticity (e.g., Wood,Linley, Maltby, Baliousis, & Joseph, 2008; Lopez & Rice, 2006) couldhelp to view whether experiences of falsity/authenticity show thesame pattern across contexts as the trait variability patterns foundin this study.

Also part of the second hypothesis was the prediction thatattachment would relate to personality traits in the parent contextbut not in the work and friends contexts. This was supported – thedata suggest that attachment links to all traits in the parent con-

Table 3Individual predictors within a regression using parental attachment as DV; and traitmeans and corrected standard deviations as IVs (Enter method).

B Standard error B b

ES – variability �0.95 0.20 �0.30***

O – variability �0.75 0.23 �0.21**

A – mean 0.46 0.15 0.20**

A – variability �0.69 0.26 �0.17**

ES – mean 0.25 0.12 0.13*

E – mean 0.24 0.13 0.12C – mean 0.21 0.11 0.11E – variability 0.22 0.20 0.07O – mean 0.12 0.13 0.06C – variability 0.19 0.25 0.05

Note: E = Extraversion; A = Agreeableness; O = Openness to Experience; ES = Emo-tional Stability.R2 = 0.37; F(10,198) = 11.45; p < 0.001; adjusted R2 = 0.33; p < 0.001.

* p < 0.05.** p < 0.01.

*** p < 0.001.

O. Robinson et al. / Personality and Individual Differences 50 (2011) 547–552 551

text and none in the friend or work contexts. Participants withstronger parental attachment were in general more extraverted,more agreeable, more emotionally stable, more conscientious andmore open to experience with parents. These findings suggest thatthe link between the Big Five traits and attachment that has beenreported by previous research (e.g., Noftle & Shaver, 2006) may infact be strongly moderated by social context and this could be ta-ken into account in future research exploring the link betweenattachment and personality.

Previous papers on personality variability other than the Bairdet al. (2006) study have opted for a single global index of cross-context variability, and avoided a consideration of trait-specificvariability, yet this study supports the importance of consideringtrait-specific cross-context variability. The third prediction wasthat those who showed more social variability in specific traitswould show weaker parental attachment. In the analysis, it wasfound that mean-corrected variability in three of the five traits(ES, A and O) was linked to parental attachment. In the regressionmodel, variability in these three traits also emerged as significantpredictors. Mean-corrected ES variability and O variability werestronger predictors of parental attachment than any trait means,suggesting that personality trait variability may be as importantto consider in the interface between attachment and personalityas any raw or averaged trait levels.

The prediction that those with stronger parental attachmentwould show less cross-context personality variability was sup-ported, and regression analysis showed that the affective dimen-sion of parental attachment is the key component of therelationship to MC-PVI. The hypothesis was derived from the the-oretical assumption that a variable personality is indicative of theemployment of false selves and their origin in problematic care-giver relationships within the family. While the study cannot con-clusively confirm if this is indeed the mechanism of personalityvariability, the fact that a hypothesis based on this theory receivedsupport provides indirect evidence for its validity, and suggeststhat continued exploration of this area is a worthwhile endeavour.Another possible interpretation of the negative relationship be-tween personality variability and parental relationship is that itis an artefact of age; the proposition would be that as individualsgrow older they develop weaker attachment to their parents andtheir personalities become more consistent, thus the negative cor-relation. However the data in this study do not bear this interpre-tation out – neither parental attachment nor personality variabilityis correlated with age.

Interpreting personality variability as indicative of the use ofpersonae or false selves raises the question of why mean-correctedPVI does not seem to link directly to emotional problems (e.g.,

negative affect or life satisfaction – Baird et al., 2006). We speculatethat this may be because the moderate use of personae, shown by amiddle-range PVI score, is in fact not a pathological or maladaptivestate, but is an adaptive sensitivity to social context (e.g., Gergen,1972). However, research conducted on those who are one SD ormore above PVI mean (Querstret & Robinson, 2010) found that ahigh level of PVI does link to emotional and identity problems, suchas social anxiety, fear of rejection and held-in negative affect. Insummary, it may be that while parental attachment relates tothe full spectrum of PVI, pathology only relates to very high PVI,hence a correlation with the former but not the latter.

4.1. Limitations and future research

This exploratory study links the two previously unconnected re-search domains of attachment and intra-individual personalityvariability. Cross-context variability indices from past researchhave been shown to be mean-dependent and thus reflective of traitmeans as well as trait variability. The mean-correction process thatBaird et al. (2006) devised has not yet been widely applied to newresearch. While they found that it removed any significant link be-tween personality variability and wellbeing (a finding later sup-ported by LaGuardia and Ryan (2007)), it was found in thecurrent study that employing this process (with or without thesquare of the mean in the calculation) does not remove a link withparental attachment, it indeed clarifies it by showing which traitsare linked to parental attachment, independent of their mean lev-els. Baird et al. also found that personality variability is a generaltendency common to all Big Five traits, as did this study (seeFig. 1). However, the link between cross-context variability andattachment is not generic to all traits – Extraversion and Conscien-tiousness are not implicated in this relationship.

A way of developing a more in-depth analysis of personalityvariability would be to focus on Big Five trait facets and their mod-ulation across contexts. The TIPI items employed in the currentstudy give an assessment at the trait level. Gosling et al. (2003)found strong to modest correlation effect sizes between TIPI dataand all respective NEO-PI-R trait facets, with certain facets show-ing stronger loadings to the TIPI, but a facet-level critique is notpossible within the scope of the paper. Future research that em-ploys the NEO-PI-R to explore contextual change in personalitycould explore the facet-level of analysis for a more fine-grainedconsideration of cross-context personality variability and its rela-tionship to other variables.

A limitation of the research was that the PAQ measured con-temporary parental attachment, yet the theory underlying thehypothesis explicitly suggests that it is parental attachment duringchildhood that links to adult personality integration and consis-tency. The PAQ has the advantage of being less reliant on memory,however the authors are already conducting follow-up researchusing retrospective childhood parental-care measure, to assesswhether that also relates to adult personality variability. This willmore closely align to false-self theory, but conversely will be proneto the effects of long-term recall. If an effect between retrospectiveparental care and adult personality variability is found, it wouldsuggest that a longitudinal study linking childhood environmentfactors with adult personality coherence would be a worthwhileendeavour. Further research may also look to explore the age-PVIrelation, as although the current study used a sample spanning18–59 years of age, the age distribution was skewed towards lowerages, with a mean of 29. A more normally distributed age rangemay still demonstrate a link between PVI and age that was notapparent in this study.

The limitations of self-report data in studying personality vari-ability have been outlined by Heller et al. (2007) – for examplethey point out that the process of filling out a personality inventory

552 O. Robinson et al. / Personality and Individual Differences 50 (2011) 547–552

multiple times may imply an expectation of variability. They con-clude that despite such limitations, self-report data converge withbehavioural data on the issue of cross-context inconsistencypatterns to a substantial degree and are therefore valid. The self-report findings from the current exploratory study suggest thatthe link between attachment and social personality variability isan important and fruitful domain for ongoing enquiry, and in linewith Heller et al.’s (2007) vision for future research into contextu-alized personality, it is hoped that studies using observational andexperience-sampling measures will follow.

References

Baird, B. M., Le, K., & Lucas, R. E. (2006). On the nature of intraindividual personalityvariability: Reliability, validity and associations with well-being. Journal ofPersonality and Social Psychology, 3, 512–527.

Block, J. (1961). Ego identity, role variability and adjustment. Journal of ConsultingPsychology, 25, 392–397.

Cross, S. E., Gore, J. S., & Morris, M. L. (2003). The relational-interdependent self-construal, self-concept consistency and well-being. Journal of Personality andSocial Psychology, 85, 933–944.

Diehl, M., Hastings, C. T., & Stanton, J. M. (2001). Self-concept differentiation acrossthe adult life span. Psychology and Aging, 16, 643–654.

Donahue, E. M., & Harary, K. (1998). The patterned inconsistency of traits: Mappingthe differential effects of social roles on self-perceptions of the Big Five.Personality and Social Psychology Bulletin, 24, 610–619.

Donahue, E. M., Robins, R. W., Roberts, B. W., & John, O. P. (1993). The divided self:Concurrent and longitudinal effects of psychological adjustment and socialroles on self-concept differentiation. Journal of Personality and Social Psychology,64, 834–846.

Gergen, K. (1972). Multiple identity: The healthy, happy human being wears manymasks. Psychology Today, 31–65.

Gillath, O., Sesko, A. K., Shaver, P. R., & Chun, D. S. (2010). Attachment, authenticityand honesty: Dispositional and experimentally induced security can reduceself- and other-deception. Journal of Personality and Social Psychology, 98,841–855.

Gosling, S. D., Rentfrow, P. J., & Swann, W. B. Jr., (2003). A very brief measure of theBig-Five personality domains. Journal of Research in Personality, 37, 504–528.

Heller, D., Watson, D., Komar, J. A., Min, J., & Perunovic, W. Q. E. (2007).Contextualized personality: Traditional and new assessment procedures.Journal of Personality, 75, 1229–1253.

Kenny, M. E. (1987). The extent and function of parental attachment among first-year college students. Journal of Youth and Adolescence, 16, 17–27.

LaGuardia, J. G., & Ryan, R. M. (2007). Why identities fluctuate: Variability in traitsas a function of situational variations in autonomy support. Journal ofPersonality, 75(6), 1205–1228.

Lopez, F. G., & Rice, K. G. (2006). Preliminary development and validation of ameasure of relationship authenticity. Journal of Counselling Psychology, 53,362–371.

Lutz, C. R., & Ross, S. R. (2003). Elaboration versus fragmentation: Distinguishingbetween self-complexity and self-concept differentiation. Journal of Social andClinical Psychology, 22, 537–560.

Noftle, E. E., & Shaver, P. R. (2006). Attachment dimensions and the Big Fivepersonality traits: Associations and comparative ability to predict relationshipquality. Journal of Research in Personality, 40, 179–208.

Querstret, D., & Robinson, O. (2010). Putting on an act? Exploring motivations andexperiences of high cross-context personality variability in young adults. BPSAnnual Conference, Stratford-upon-Avon.

Roberts, B. W., & Donahue, E. M. (1994). One personality, multiple selves:Integrating personality and social roles. Journal of Personality, 62, 201–218.

Robinson, O. C. (2008). Personality inconsistency and problems with parents:Uncovering the roots of the false self. BPS Clinical Psychology Conference, London.

Robinson, O. C. (2009). On the social malleability of traits: Variability andconsistency in Big 5 trait expression across three interpersonal contexts.Journal of Individual Differences, 30(4), 201–208.

Robinson, O., Demetre, J. D., & Corney, R. H. (2010). Personality and retirement:Exploring the links between the Big Five personality traits, reasons forretirement, and the experience of being retired. Personality and IndividualDifferences, 48, 792–797.

Schoenrock, C. J. (1991). Family functioning and self-monitoring in late adolescence.Doctoral dissertation, Texas Tech University. Available from: http://etd.lib.ttu.edu/theses/available/etd-02262009-31295006972029/unrestricted/31295006972029.pdf Retrieved 20.10.10.

Sheldon, K. M., Ryan, R. M., Rawsthorne, L. J., & Ilardi, B. (1997). Trait self and trueself: Cross-role variation in the Big-Five personality traits and its relations withpsychological authenticity and subjective well-being. Journal of Personality andSocial Psychology, 73, 1380–1393.

Suh, E. M. (2002). Culture, identity consistency, and subjective well-being. Journal ofPersonality and Social Psychology, 83, 1378–1391.

Winnicott, D. W. (1960). Ego distortion in terms of true and false self. In Thematurational processes and the facilitating environment (pp. 140–152).

Wood, A. M., Linley, P. A., Maltby, J., Baliousis, M., & Joseph, S. (2008). The authenticpersonality: A theoretical and empirical conceptualization and thedevelopment of the authenticity scale. Journal of Counseling Psychology, 55,385–399.