1The growing instability of unions is leading to increased diversity in kin relationships and in ties between family members. Intergenerational relationships are affected by these transformations which may both strengthen or weaken the links between children, parents and grandparents. When parents separate, is one side of the family preferred over the other, or do the parents seek to maintain a balance in their children’s relationships with the two family lineages ? Alexandre Pillonel, Cornelia Hummel and Ivan deCarlo examine this question using data from a statistical survey of adolescents in Switzerland. They show that grandchild-grandparent relationships can take a wide variety of forms. The authors observe that the parents’ separation does not necessarily disrupt the grandchildren’s relationship with one or other side of the family, and that the matrilateral bias observed when parents are separated also exists when the family is intact.

2Intergenerational relationships within a family, especially those between grandparents and grandchildren have often been studied in the context of contemporary family transformations (Attias-Donfut et al., 2002 ; Kellerhals and Widmer, 2007 ; Bonvalet and Lelièvre, 2012). Studying the impact of divorce on these relationships provides a means to explore the ties linking two generations in a context of sociological debate on the relational family where relations of affinity take precedence over statutory relations (Hummel and Perrenoud, 2009).

3Numerous studies have observed the phenomenon – now classic in sociology of the family – of the matrilaterality of kin relations, i.e. an imbalance between family lines in favour of the maternal side (Matthews and Sprey, 1985 ; Pitrou, 1992 ; Coenen-Huther et al., 1994 ; Déchaux, 1994 and 2007 ; Uhlenberg and Hammill, 1998 ; Chan and Elder, 2000 ; Hammer et al., 2001). The consequences of parental divorce on the relations between children and their ascendants has been documented by several studies in Europe and the United States. They observe a weakening of ties between children and their father, and with the paternal kin more generally (Johnson, 1983 ; Matthews and Sprey, 1984 ; Cherlin and Furstenberg, 1986 and 1992 ; Kivett, 1991 ; Creasey, 1993 ; Martin, 1997 ; Villeneuve-Gokalp, 2000 ; Cadolle, 2000 ; Kellerhals et al., 2001 ; Mueller and Elder, 2003), and a (re)centring on maternal kin following union dissolution.

4Other studies show that the impact of divorce on relations between grandparents and grandchildren may be less clearcut, however, for example when the children’s age is taken into account (Matthews and Sprey, 1985 ; Cogswell and Henry, 1995 ; Cooney and Smith, 1996). As grandchildren grow older, their relationship with grandparents becomes more independent because less influenced by the intermediate generation.

5This independence may, to a certain extent, protect grandparent relationships from the conflicts between the intermediate and elder generations following divorce or separation. A study by Cogswell and Henry (1995) on 327 students aged 18-22 concerning their perception of their grandparents’ role and of their relationship with them shows that after controlling for the effect of geographical distance, perceptions are identical for young people from divorced and intact families, with no lineage effect. The authors highlight the importance of grandparents for young people, independently of the union status of the intermediate generation.

6Lussier, Deater-Deckard, Dunn and Davies (2002), for their part, stress the importance of the grandparent figure and its role for maintaining a sense of closeness at times of family change due to divorce or remarriage. The impact of parental separation on relations with grandparents is also examined by Douglas and Ferguson (2003) who highlight the stability of the configuration. According to these authors, the nature and style of this relationship is established before the separation of the intermediate generation, and is not fundamentally altered by this event. As the maternal grandparents are often closer during the years when the parental union is still intact, the weaker ties with the paternal grandparents after divorce is a consequence of this earlier situation.

7Chan and Elder (2000) reach similar conclusions based on a longitudinal study of data from the Iowa Youth and Families Project. They argue that each parent gives preference to his or her own lineage, but that bias is always stronger for women. Matrilateral bias observed after separation is thus the result of a long-term process initiated when the family is still intact. In this sense, the stronger ties between the mother and her parents also give rise to a closer relationship between the young generation and the maternal grandparents, who are a resource that can be mobilized at a time of family crisis such as a divorce. In this case, post-divorce matrilaterality reflects the combination of a lineage stability effect and custodial parent effect (this parent most often being the mother).

8The research cited above reveals two viewpoints on the economics of family relationships that result in two diverging explanations of post-divorce matrilaterality. A first set of analyses is based on the affirmation that family relationships within marriage are managed under a principle of equity. As this equity is disrupted by separation, the balance shifts in favour of the family of one or other of the former partners – generally that of the maternal side because the mother is most often the custodial parent. A second set considers that matrilateral bias is often already at play when the family is still intact. The effect of parental separation is moderate, therefore, and the fact that the mother has custody of the children simply reinforces a pre-existing matrilateral bias.

9As our data are cross-sectional, we cannot study the transformation of inter-generational ties before and after union dissolution. We can, however, explore the grandparent-grandchild relationship by comparing situations where the parents live together, and those where they live apart. Our central research questions are the following :

• What differences, in terms of laterality, are observed in grandparent-grandchild relationships according to the parents’ union status ?
• When the family is intact, are relations with grandparents characterized by equity between the paternal and maternal lines, or is there a matrilateral bias ?
• Is matrilateral bias stronger in separated than in intact families ?

I – Data and approach

10The question of grandparent-grandchild relationships was studied by comparing intact and separated families through secondary analysis of data from the survey entitled “Enfants, adolescents et leurs grands-parents dans une société en mutation” [1] (Children, adolescents and their grandparents in a changing society) conducted in Switzerland in 2004. The study covered 685 girls and boys aged 12-16, [2] attending a state school [3] and living in Geneva, in Zurich or in four towns of the canton of Valais. The respondents were asked about their relationship with all their living grandparents (one questionnaire per living grandparent). A total of 1,353 questionnaires were then used for the analysis. The sample thus comprises 1,353 dyads [4] for which the relationship between the adolescent respondent and one of his or her grandparents can be described.

11Note that the data were self-reported accounts of the young people’s relationships with their grandparents, and not the reverse. Analyses from this viewpoint provide an opportunity to develop a unique and original approach to the question of inter-generational relations in Switzerland. Some of the responses given were invalid for certain indicators, [5] resulting in slight variations in sample size in the analyses presented below.

12In our sample, 79.1% of the adolescents lived in an intact family, 20.5% in a separated family. [6] These proportions are slightly different from the Swiss average : in 2000, 13.2% of children aged 13 born to a married couple had divorced parents (Wanner, 2006). The over-representation in our sample of children with separated parents is probably due to the fact that respondents were all from urban areas, where divorce rates are higher in Switzerland. [7] The sample also had a small majority of girls (52.1%).

13To assess any laterality in the grandparent-grandchild relationships, we estimated the strength of the relationship by frequency of contact and by its emotional intensity as measured by the adolescent using five criteria : general importance of the relationship, financial support, help with school work, emotional support, and grandparents’ availability.

14The frequency of face-to-face or telephone contact (factual dimension), determined by the question “How much contact do you have with your grandmother/grandfather ?” was measured using four response categories : once a week or more ; once a month ; two or three times a year ; rarely or never. The adolescent was also asked about who initiated the face-to-face contact : contact initiated by parents, grandparents or the adolescent, very frequently, frequently, rarely or never.

15The emotional intensity (expressive dimension) was determined by the question “Is your relationship with your grandfather/grandmother important to you ?” and was expressed via four response categories : very important, quite important, not very important, not at all important. Next, four aspects of this subjective importance were explored (“In your view, in what areas is the role of grandmother/grandfather important ?”) : financial support, emotional support when things are going badly, help with schoolwork and availability if needed, again with four response categories.

16We decided to keep these original response categories for the bivariate analyses ; they were recoded as a dichotomous variable in the successive multivariate analyses for two reasons : first, these measures are very unequally distributed over the response categories for frequent contact and strong importance of the relationship (cf. section II) ; second, we believe that frequent contact and considering the relationship to be very important are pertinent indicators of the strength of the relationship. For frequency of contact, the “once a week or more” category is compared against the three other categories and, for the importance of the relationship, the “very important” category against the three others.

17As the data are cross-sectional, we are comparing two different situations at a given moment in time, i.e. the relationship between adolescents and grandparents when the adolescent’s parents are together and when they are separated. This strategy is applied in the bivariate analysis presented in the next section. The limits of this strategy, inherent to the cross-sectional nature of the data, are partly counterbalanced by the greater detail obtained in the analyses of the dyadic relationships between adolescents and each of their grandparents presented later in this article. The dyad-by-dyad analysis also ties in well with the research protocol whereby each respondent filled in a questionnaire for each grandparent ; this limits the comparison between different grandparents so that attention can focus on the importance of types of support received from just one grandparent at a time. With multivariate models (logistic regressions) it is also possible to control for a whole series of factors such as respondent’s sex, presence or absence of other grandparents, distance between respondents’ and grandparents’ homes, grandparents’ health, age and place of residence (urban/rural). Last, for the effects on face-to-face contact, the person who initiated the encounter is introduced as control variable. In the full sample, we observe that it is the adolescent’s parents who initiate the encounter in the majority of cases.

II – Results

18A descriptive analysis of the indicators and a series of bivariate analyses on the complete sample are given here to reveal variations in laterality effects on the various indicators for adolescents whose parents are together and those whose parents are separated. We then present multivariate analyses on four sub-samples corresponding to each of the four possible dyads by grandparents’ sex and lineage. The structure of this presentation has a dual objective.

19By associating bivariate and multivariate analyses, which tend to contradict each other in certain respects, the aim is to show that each dyadic relationship between the adolescents and their various grandparents must be observed separately. This approach enables us to distinguish much more clearly the relative impact of parental separation on each of these relationships. If the relationship with one of the grandparents is weakened, strengthened or unaffected by parental separation, this does not imply that the same is true for another grandparent, even if he/she belongs to the same lineage. Moreover, when the results are interpreted, a finer distinction can be made between gender and lineage effects.

20The second aim is to demonstrate the existence of matrilateral bias, both when the parents are together and when they are separated. The variations in the intensity of this bias by parents’ union status suggest that separation does not systematically penalize the paternal lineage.

1 – Bivariate analyses : the couple, matrix of matrilaterality

21Before comparing the relationships between grandparents and adolescents by the parents’ union status, it is worth highlighting the importance of these relationships for the young generation. In general terms, they are considered to be very important or quite important for a majority of adolescents (87%). The leading factor is the relatively vague notion of grandparent availability : 76% of adolescents consider that it is important for their grandmother/grandfather to be “on hand” for them, and to be available when needed. [8] This is followed by help with schoolwork (57%) and emotional support when things are going badly (55%). A minority of grandchildren expect their grandparents to provide financial support. In the majority of cases, it is the parents who initiate face-to-face encounters : 51% of meetings are always or very often initiated by parents, versus 27% by grandparents and 24% by the adolescents themselves. [9]

22Parental separation affects the factual and expressive dimensions of the grandparent-grandchild relationship in different ways. The frequency of telephone contact, the general importance of the relationship, but also help with schoolwork are given lower scores by adolescents whose parents are separated (Table 1). Among these respondents, 25.2% report talking to their grandparents on the phone once a week or more ; among respondents whose parents are still together, the proportion is 33.6%. Similar variations are observed for the “very important” response category when assessing the importance of the relationship (41.3% versus 54.9% when the parents’ union is intact) and the importance of help with schoolwork (22.6% versus 28.2%). By contrast, no statistically significant effect of union status is observed for the frequency of face-to-face contact, for the importance of emotional and financial support, or the importance of grandparents’ availability. Last, in none of the observed cases does parental separation increase either the frequency of the various forms of contact or the importance of the expressive dimensions of the relationship as reported by the adolescent. When the separated family situation differs from that of the intact family in several respects, the effects all seem to follow the direction of a weakening of the grandparent-grandchild relationship. A more detailed analysis is needed to determine whether this partial weakening affects one lineage more than the other – in this case the paternal lineage, as reported in a part of the literature.

Table 1

Distributions of indicators of contact and of importance of the grandparent-grandchild relationship by parents’ union status, total sample (%)

Contact Importance Face-to-face (n = 1,267) By phone (n = 1,241) In general (n = 1,269) Financial support (n = 1,270) Emotional support (n = 1,272) Help with schoolwork (n = 1,271) Availability (n = 1,266) Parents separated Parents separated Parents separated Parents separated Parents separated Parents separated Parents separated Yes No Total Yes No Total Yes No Total Yes No Total Yes No Total Yes No Total Yes No Total Rarely/never 9.7 13.4 12.6 25.2 20.8 21.7 Not at all important 4.5 1.9 2.4 29.5 29.9 29.8 18.2 18.7 18.6 19.5 17.2 17.6 11.5 10.2 10.5 2-3 times per year 25.5 29.1 28.3 15.5 14.0 14.3 Not very important 10.6 7.0 7.7 42.2 41.9 42.0 29.5 22.5 23.9 28.8 21.2 22.7 13.1 10.6 11.1 Once a month 28.6 24.6 25.4 34.1 31.6 32.2 Quite important 43.6 36.2 37.7 19.4 17.2 17.6 28.7 31.8 31.1 29.2 33.4 32.6 31.2 35.0 34.2 Once a week or more 36.3 32.9 33.6 25.2 33.6 31.8 Very important 41.3 54.9 52.1 8.9 11.0 10.6 23.6 27.0 26.3 22.6 28.2 27.1 44.2 44.1 44.2 Total 100 100 100 100 100 100 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Cramer’s V 0.06 0.08* Cramer’s V 0.13*** 0.03 0.07 0.09** 0.04

Distributions of indicators of contact and of importance of the grandparent-grandchild relationship by parents’ union status, total sample (%)

Statistical significance : * p < 0.1 ; ** p < 0.05 ; *** p < 0.01.

23Table 2 shows the distributions of the various indicators of the strength of the relationship by lineage, for the sub-sample of respondents whose parents were together, followed by that of respondents whose parents were separated. When the parents’ union is intact, matrilateral bias is observed, not only for the two measures of frequency of contact, but also for the general importance of the relationship and for emotional support. Among the adolescent respondents, 28.6% reported face-to-face contact with their paternal grandparents once a week or more, versus 36.8% for maternal grandparents. The same pattern is observed for frequency of phone contact, which is higher for the maternal than for paternal grandparents (37.4% versus 29.3%). While 57.7% of respondents considered their relationship with their maternal grandparents to be very important, the proportion was 50.6% for the relationship with paternal grandparents. For emotional support, the proportions were 29.7% and 24.0%, respectively, in favour of the maternal line. Note that the strength of these relationships is practically identical to that observed for the entire sample, without distinguishing by parents’ union status. At this stage of the analysis, we can thus confirm that laterality exists within the couple, always in favour of the maternal line. Matrilaterality is not, therefore, a consequence of separation.

Table 2

Distributions of indicators by parents’ union status and grandparents’ lineage (%)

Sub-sample with parents in an intact union Contact Importance Face-to-face (n = 1,008) By phone (n = 983) In general (n = 1,005) Financial support (n = 1,012) Emotional support (n = 1,014) Help with schoolwork (n = 1,014) Availability (n = 1,006) Lineage Lineage Lineage Lineage Lineage Lineage Lineage Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Rarely/never 15.2 11.8 13.4 23.6 18.2 20.8 Not at all important 3.3 1.5 2.3 28.4 31.3 29.9 20.6 17.1 18.7 16.6 17.7 17.2 11.4 9.2 10.2 2-3 times per year 31.8 26.6 29.1 15.2 13.0 14.0 Not very important 7.6 6.2 6.8 41.9 41.9 41.9 21.7 23.2 22.5 22.5 20.1 21.2 11.2 10.1 10.6 Once a month 24.4 24.8 24.6 31.9 31.4 31.6 Quite important 38.5 34.7 36.5 18.6 15.9 17.2 33.7 30.1 31.8 33.2 33.6 33.4 35.8 34.3 35.0 Once a week or more 28.6 36.8 32.9 29.3 37.4 33.6 Very important 50.6 57.7 54.4 11.0 10.9 11.0 24.0 29.7 27.0 27.7 28.6 28.2 41.5 46.4 44.1 Total 100 100 100 100 100 100 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Cramer’s V 0.10** 0.10** Cramer’s V 0.08** 0.04 0.08* 0.03 0.06 Sub-sample with separated parents Contact Importance Face-to-face (n = 259) By phone (n = 258) In general (n = 264) Financial support (n = 258) Emotional support (n = 258) Help with schoolwork (n = 257) Availability (n = 260) Lineage Lineage Lineage Lineage Lineage Lineage Lineage Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Pat. Mat. Total Rarely/never 11.1 8.6 9.7 24.8 25.5 25.2 Not at all important 4.5 4.6 4.5 30.6 28.7 29.5 18.3 18.1 18.2 15.6 22.3 19.5 13.8 9.9 11.5 2-3 times per year 26.9 24.5 25.5 23.9 9.4 15.5 Not very important 13.5 8.5 10.6 41.7 42.7 42.2 33.0 26.8 29.5 28.4 29.1 28.8 12.8 13.2 13.1 Once a month 30.6 27.2 28.6 29.4 37.6 34.1 Quite important 46.8 41.2 43.6 17.6 20.7 19.4 22.9 32.9 28.7 30.3 28.4 29.2 33.0 29.8 31.2 Once a week or more 31.5 39.7 36.3 22.0 27.5 25.2 Very important 35.1 45.8 41.3 10.2 8.0 8.9 25.7 22.1 23.6 25.7 20.3 22.6 40.4 47.0 44.2 Total 100 100 100 100 100 100 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Cramer’s V 0.09 0.20** Cramer’s V 0.12 0.05 0.11 0.10 0.08

Distributions of indicators by parents’ union status and grandparents’ lineage (%)

Statistical significance : * p < 0.1 ; ** p < 0.05 ; *** p < 0.01.

24Looking at adolescents whose parents are separated (second part of Table 2), we again observe matrilateral bias on the indicator measuring the frequency of phone contact, with 22.0% of respondents reporting phone contact with paternal grandparents once a week or more, versus 27.5% with maternal grandparents. But this bias disappears for indicators measuring the closeness of the relationship. The matrilaterality observed when the parents’ union is intact disappears when only adolescents with separated parents are considered. They do not distinguish between maternal and paternal sides when judging the general importance of the relationship or the importance of emotional support. The comparison of these two sub-samples sheds new light on matrilateral bias when parents are separated : this bias is not stronger when the parents are separated, but is in fact weaker when it no longer influences certain dimensions measuring the strength of the relationship between the adolescent and his/her grandparents. Rather than being a fatal disadvantage for the paternal side, conjugal separation may thus provide an opportunity to renegotiate relationships between the lineages.

2 – Multivariate analyses : matrilaterality as revealed by dyad analysis

25The observed differences between these two situations – intact union and separation – suggest variations in matrilateral bias in family relationships when the parents are separated. But what form do these variations take ? Is a decrease in matrilateral bias the only change observed, or is there a tendency for matrilateral bias to be “counterbalanced” by a strengthening of ties with the paternal line ?

Dyadic relationships

26To better understand the complexity of inter-generational relationships, each grandchild-grandparent relationship should be studied separately (Whitbeck et al., 1993). We therefore opted for a comparative analysis between dyads to explore more fully the relationships between young people and their grandparents by lineage and union status. Each dyad is evaluated in a logistic regression model, and four sub-samples representing each of the grandchild-grandparent dyads are analysed in a series of multivariate models. [10] Table 3 shows the models for paternal grandparents and Table 4 the models for maternal grandparents.

27Moreover, as the literature offers numerous examples of other effects liable to influence relationships between adolescents and grandparents, we included a series of control variables in our logistic regression models to reinforce the analyses. Geographical distance is often highlighted as a strong explanatory factor of imbalance in contacts between the paternal and maternal sides : the greater the distance, the more difficult it is to organize frequent contacts (Cherlin and Furstenberg, 1986). Other potential factors are the health and age of the elder generation. Adolescents may be less inclined to maintain frequent contact with a grandparent in very poor health. [11]

28It is also important to take account of the grandparents’ and grandchildren’s gender. Applying the notion of gynocentrism (Déchaux, 2007) specific to the cognatic model of western kinship, [12] we assume that young girls are closer to their maternal grandmother. The effect of the number of available grandparents (i.e. living grandparents in either lineage) is also included in our analysis models to control for the possibility that relations with a grandparent are close simply because he or she is the only one still alive. In such cases, the affection which in other circumstances would be divided among several grandparents, is focused on just one person. The same question arises in relation to lineage : if one lineage is absent, relationships with the grandparents of the other lineage may be especially close. We will also examine a “couple effect” suggested by the qualitative part of the study (Hummel and Perrenoud, 2009) : regular contact with the maternal grandmother may also be of indirect advantage to the maternal grandfather.

29The survey used for this analysis took place in relatively different urban contexts. Geneva and Zurich are two of Switzerland’s largest urban centres, while the towns in the Alpine canton of Valais are smaller and located in a rural region. It is important to check whether the place of residence affects relations between adolescents and their grandparents, since we know that the divorce rate is below average in rural regions, so behaviours may be affected by different normative representations of the family.

30Finally, for the models assessing the effects of face-to-face contact, we also introduce the control variable for the person who initiates the encounter. [13]

The different effects of lateralization

31After introducing the various control variables in the logistic regression models we can identify the three possible patterns of matrilateral bias when the specific effect of parents’ union status on each dyad is compared : stable matrilateral bias, reduction or even disappearance of matrilateral bias, and emergence of patrilateral bias for certain dimensions of the relationship (this latter being an unlikely and purely theoretical possibility).

32The hypothesis of stable matrilateral bias is observed for one indicator of the factual dimension of the relationship, namely the frequency of phone contact (Tables 3 and 4). For this indicator, none of the dyadic relationships are affected by the union status of the adolescent’s parents. Hence, as suggested by the bivariate analyses on this indicator, matrilateral bias observed in intact unions is also observed after separation. In general terms, it should be pointed out that separation does not strengthen matrilateral bias for any of our indicators (Table 4). At best, we can suggest that the maternal line appears to conserve its advantages whatever the parents’ union status.

Table 3

Logistic regression models on indicators for paternal grandfather and paternal grandmother

Contact Importance Face-to-face By phone In general Financial support Emotional support Help with schoolwork Availability Sub-sample of relationship between grandchild and paternal grandfather n 232 228 232 230 232 233 231 Parents separated (ref. together) 2.27 0.61 0.43** 0.98 0.35** 0.61 0.73 Maternal grandmother alive 4 (ref. paternal grandfather alive) 1.00 0.67 0.82 0.91 0.88 0.47* 0.67 Maternal grandfather alive 1.37 1.13 1.27 1.10 0.85 0.86 0.63 Paternal grandmother alive 1.77 1.53 0.95 1.07 1.13 1.36 1.26 Distance between homes (ref. same district/village) same canton 0.60 1.63 0.40** 1.22 0.68 1.05 1.38 elsewhere 0.03*** 1.43 0.69 1.85* 1.27 1.14 1.70 Grandparent’s health (ref. good) poor 0.09*** 0.20** 0.22*** 0.27* 0.36* 0.49 0.26** fair 0.47* 0.80 0.38*** 0.99 0.68 0.79 0.30*** Grandparent’s age (continuous variable) 1.03 0.97* 1.00 0.98 0.97 1.00 0.98 Grandchild’s sex (ref. male) female 0.61 0.72 0.95 1.34 1.34 1.13 2.01* Urban/rural context (ref. Geneva/Zurich) Valais 0.47* 0.98 0.58* 1.15 1.54 2.08** 1.04 Frequency of initiating encounter (ref. rarely) * by parents 1.06 by grandparent 1.45* by adolescent 0.92 Constant 1.35 16.40 5.47 1.48 8.94 2.05 17.12 R² 0.56 0.12 0.17 0.06 0.12 0.09 0.14 Sub-sample of relationship between grandchild and paternal grandmother n 351 348 351 350 352 352 531 Parents separated (ref. together) 2.14* 0.70 0.62* 0.89 0.96 0.82 0.71 Maternal grandmother alive 4 (ref. maternal grandmother alive) 0.85 0.89 1.33 1.07 0.92 0.81 1.00 Maternal grandfather alive 1.31 0.89 0.92 1.09 1.02 1.04 0.70 Paternal grandmother alive 1.36 0.94 0.99 1.14 0.87 0.73 0.71 Distance between homes (ref. same district/village) same canton 0.50 0.86 0.65 0.95 0.74 0.67 0.66 elsewhere 0.02*** 0.85 1.03 1.13 1.51 1.03 0.70 Grandparent’s health (ref. good) poor 1.59 0.53 0.74 0.44** 0.79 0.91 0.49 fair 0.61 0.60* 0.47** 0.57** 0.61* 0.34*** Grandparent’s age (continuous variable) 0.98 0.97* 0.95** 1.00 0.96** 0.98 0.98 Grandchild’s sex (ref. male) female 0.90 1.01 0.82 0.96 1.60** 1.24 1.97** Urban/rural context (ref. Geneva/Zurich) Valais 0.45** 1.20 0.58** 0.81 1.28 2.25*** 1.02 Frequency of initiating encounter (ref. rarely) * by parents 1.01 by grandparent 1.55** by adolescent 1.19 Constant 36.55* 28.11** 49.09** 0.43 17.55* 5.57 34.42* R² 0.57 0.06 0.12 0.04 0.08 0.08 0.11

Logistic regression models on indicators for paternal grandfather and paternal grandmother

Note : The coefficients are odds ratios. For the importance of financial support, the “fair” and “poor” categories of the health variable were grouped to increase the sample size.
Statistical significance : * p < 0.1 ; ** p < 0.05 ; *** p < 0.01.

Table 4

Logistic regression models on indicators for maternal grandfather and maternal grandmother

Contact Importance Face-to face By phone In general Financial support Emotional support Help with schoolwork Availability Sub-sample of relationship between grandchild and maternal grandfather n 283 272 283 285 284 287 282 Parents separated (ref. together) 0.61 0.96 0.46** 1.10 0.81* 0.53** 0.62 Maternal grandmother alive 4 (ref. paternal grandfather alive) 3.66** 2.29** 1.84 1.21 1.30* 1.20 1.20 Paternal grandfather alive 1.48 1.95** 0.88 1.40 1.37 1.01 0.89 Paternal grandmother alive 0.68 0.98 1.19 0.93 1.18 0.95 1.17 Distance between homes (ref. same district/village) same canton 0.72 0.91 1.44 1.90* 1.66 1.30 1.40 elsewhere 0.02*** 0.71 0.97 1.25 1.08 0.97 0.63 Grandparent’s health (ref. good) poor 1.20 0.33** 0.32** 0.80 0.79 0.74 0.39* fair 1.10 0.59 0.53* 0.82 0.49** 0.71 0.51* Grandparent’s age (continuous variable) 1.01 1.04** 1.01 0.99 1.00 1.00 0.97 Grandchild’s sex (ref. male) female 1.08 1.25 1.14 1.82** 1.16 1.36 1.77* Urban/rural context (ref. Geneva/Zurich) Valais 0.54 1.12 0.71 0.92 1.24 2.49*** 1.08 Frequency of initiating encounter (ref. rarely) * by parents 1.02 by grandparent 1.29 by adolescent 1.11 Constant 3.06 0.04* 0.45 0.23 0.42 0.69 31.04* R² 0.59 0.11 0.11 0.05 0.06 0.09 0.09 Sub-sample of relationship between grandchild and maternal grandmother n 401 393 403 405 404 399 402 Parents separated (ref. together) 1.32 0.81 0.72 1.13 0.83 0.50** 0.74 Maternal grandfather alive (ref. maternal grandmother alive) 0.99 1.24 0.97 0.88 1.09 1.14 1.31 Paternal grandfather alive 1.26 1.57* 1.03 1.37 1.09 0.85 0.92 Paternal grandmother alive 1.12 0.75 1.10 0.85 1.19 1.18 1.16 Distance between homes (ref. same district/village) same canton 0.54 1.40 0.73 1.08 0.87 0.70 0.60 elsewhere 0.02*** 0.94 0.67 1.02 1.02 1.27 0.54* Grandparent’s health (ref. good) poor 0.49 0.33** 0.80 0.47 0.62 0.26*** 0.26*** fair 0.66 0.68 0.58** 0.91 0.67 0.62* 0.74 Grandparent’s age (continuous variable) 1.02 1.02 0.98 0.99 0.97 0.99 0.99 Grandchild’s sex (ref. male) female 1.44 1.11 1.39 0.92 2.19*** 1.04 2.41*** Urban/rural context (ref. Geneva/Zurich) Valais 0.81 1.10 0.61** 0.83 1.22 2.64*** 1.05 Frequency of initiating encounter (ref. rarely) * by parents 1.16 by grandparent 1.36 by adolescent 1.12 Constant 5.32 0.49 9.93* 0.99 6.55 1.95 11.34 R² 0.60 0.05 0.06 0.02 0.08 0.12 0.11

Logistic regression models on indicators for maternal grandfather and maternal grandmother

Note : The coefficients are odds ratios.
Statistical significance : * p < 0.1 ; ** p < 0.05 ; *** p < 0.01.

33The hypothesis of a reduction in matrilateral bias is confirmed when we look at the results of the logistic regression for the importance of help with schoolwork. We see that having separated parents affects the importance of this type of support, but only with respect to the maternal grandparents : the odds ratio is 0.53 for the maternal grandfather and 0.50 for the maternal grandmother (Table 4). These results thus show that both the maternal grandmother and the maternal grandfather are half as likely to be considered by the adolescents as “very important” for support with schoolwork when their parents are separated. Yet this difference is not counterbalanced by greater closeness with their paternal equivalents (Table 3). So there is no change in the lateralization of relationships and we may posit that an advantage acquired by the maternal grandparents when the adolescent’s parents are together is lost when they are separated.

34We have already pointed out that parental separation in itself is never synonymous with a strengthening of bonds. While this remains true when we observe the two dyads of the maternal line (Table 4), we note an exception for the adolescent’s paternal grandparents (Table 3) : adolescents with separated parents have around twice as much face-to-face contact with the paternal grandmother as those whose parents are together (odds ratio 2.14). There is no reduction in face-to-face contact on the maternal side, however (Table 4).

35These results are clearly consistent with Chan and Elder’s hypothesis (2000). Matrilaterality takes root within the couple and there appears to be no principle of equity in the management of family relationships. In addition, these analyses reveal both the stability of matrilaterality whatever the union status and, in certain cases, a weakening or “neutralization” of matrilateral bias when the parents are separated. This neutralization occurs when the relationship becomes stronger on the paternal side, without a corresponding weakening on the maternal side. It is a situation of this type that is described above for contacts with the paternal grandmother.

Primacy of the maternal grandmother, and adolescents’ gendered expectations

36Dyads can be compared to distinguish gender effects from lineage effects. The results obtained for the four dyads on the general importance of the relationship show that the only grandparent not affected by the parents’ union status is the maternal grandmother (Table 4). A statistically significant odds ratio of less than 1 is obtained for the three other dyads (0.43 for the paternal grandfather, 0.46 for the maternal grandfather, and 0.62 for the paternal grandmother – Tables 3 and 4), but for the maternal grandmother the odds ratio is no longer significant. Given that the paternal grandmother is also less concerned by this effect than the grandfathers of both lineages, it is clear that we are measuring as much a gender effect as a lineage effect. This gender effect is even more visible for the indicator that measures the importance of emotional support. In this case, parental separation lowers the perceived importance of this factor only when the adolescent refers to his or her grandfathers independently of their lineage. For the maternal grandfather, the odds ratio is 0.81, and for the paternal grandfather it is 0.35 (Tables 3 and 4). So we see that this gender effect either weakens or strengthens matrilateral bias. In the case of emotional support, the fact that support from both maternal and paternal grandmothers is considered equally important by adolescents tends to cloud or even reduce such bias.

37Overall, the maternal grandmother appears to gain substantially from the conjunction of gender and lineage effects. For all the indicators indicating the strength of the relationship, the bond between adolescents and their maternal grandmother is only marginally influenced by parental separation. This finding is confirmed by other indices : for the indicator of the importance of emotional support, we see that the control variables such as health, age or geographical distance are factors that weaken the relationship in at least one of the three other dyads, but have no effect in relation to the maternal grandmother. If, for example, we compare a fair or poor state of health with a good one, the odds ratio is 0.36 for the paternal grandfather, 0.57 for the paternal grandmother, and 0.49 for the maternal grandfather, but the effect disappears for the maternal grandmother (Tables 3 and 4).

38The effects introduced by the number of available (i.e. living) grandparents confirm this privileged position of the maternal grandmother. Observing the logistic regression on the importance of support with schoolwork, we can show that when the maternal grandmother is alive, the likelihood that help with schoolwork provided by the paternal grandfather will be considered important is halved, with an odds ratio of 0.47 (Table 3). The maternal grandmother’s privileged position is also expressed by the knock-on benefits for her partner. This is particularly visible for the two indicators of frequency of contact (Table 4). For the maternal grandfather, the likelihood of seeing his grandchildren quadruples for face-to-face contact (3.66) and doubles for phone contact (2.29) when the maternal grandmother is alive.

39The paternal grandfather offers a mirror image of the privileged maternal grandmother. It is for him that the various control variables have the strongest effect : health, age, geographical distance, place of residence or the fact that the other grandparents are alive.

40It is important to point out that the gender effects are observed in terms of not only the grandparents’ sex, but also that of the adolescent. In our study, the adolescent girls produce a gender effect in the relationships with their grandparents. The importance of emotional support expected of the elder generation is significantly linked to the sex of the young generation : for the maternal grandmother, adolescent girls are twice as likely as boys to consider this type of support important, with an odds ratio of 2.19, versus 1.60 for the paternal grandmother. The sex of the young generation is not important in relation to expectations of emotional support from grandfathers. It does play a role, however in relation to financial support, but solely for maternal grandfathers. Adolescent girls are almost twice as likely to consider this type of support as very important (1.81). So young girls seem to turn to their grandmothers for emotional support and to their grandfathers for questions of money. Boys do not seem to make a distinction in terms of gender roles. Might this reflect the transmission of a gendered habitus whereby the duties of managing the relational dimension of family life are assigned to women ?

Indirect matrilaterality bias

41Our logistic regression models reveal the inertia of matrilateral bias. It would also be interesting to discover the less obvious ways in which this bias is expressed. For example, the distribution of the effects of the control variables across the four dyads suggests the existence of “indirect” matrilateral bias, since these variables are more detrimental to relationships with the paternal lineage. Comparing the results obtained across the four dyads for the general importance of the relationship shows that geographical distance, health, age and place of residence penalize the paternal line. These effects tend to decrease or disappear for the maternal line. Take the example of geographical distance. The simple fact of living not in the same municipality but in the same canton more than halves (0.40) the likelihood that adolescents will consider their relationship with the paternal grandparent to be very important (Table 3). Likewise, when the grandparents are advanced in age, frequency of phone contact with the adolescents decreases slightly, but only for the two paternal grandparents.

42When the person who initiated the encounter is taken into account, frequency of face-to-face contact tends to increase if the grandparents are the initiators, but for the paternal line only. The odds ratio when the paternal grandfather is the initiator is 1.45 (Table 3), indicating a higher frequency of contact with his grandchild, but the other dyadic relationships are not affected when the grandparents are the initiators. Grandparental investment in their relationship with grandchildren appears to be all the more important in the paternal side of the family, as if the relative disadvantage of this side compared to the maternal side had to be counteracted by more active efforts to stay in touch.

43It would appear, therefore, that matrilateral bias affecting relationships between adolescents and their grandparents does not really reflect a certain form of orientation in relationship preferences, but rather the malleability or the lineage weighting of effects linked to age, geographical distance, health, the adolescent’s gender or the existence of other grandparents. It would thus appear that the laterality exhibited by the adolescent is not purely elective, i.e. produced by a choice, but also the outcome of a series of constraining factors that weigh even more heavily on the paternal line.

III – Discussion

44What answers can this study provide to the initial questions :

• How are adolescents’ relationships with their grandparents affected by the union status of the intermediate generation ?
• To what extent is conjugal separation a factor of lateralization of inter-generational relationships ?

45While these analyses suggest that parental separation does indeed affect the fabric of family relationships, the effects are distributed very unequally between grandmothers and grandfathers. So while there is no overall effect of separation on family relationships as a whole, there are indeed specific effects which vary not only according to the grandparent figure considered, but also according to the indicator observed. The factual and expressive dimensions of the relationship are not affected in the same way when the parents are separated, so no general rule about lateralization can be established with regard to post-separation relationships. This suggests that relationships are reoriented in different ways after a separation, and several interpretations are possible.

46We have successfully shown that matrilateral bias is not concomitant with parental separation and that it begins when the parents’ union is intact. Our analyses have shown two effects of parental separation, depending on the chosen indicator : first, the inertia of matrilateral bias and second, to a lesser extent, a counterbalancing of this bias on the paternal side. These conclusions thus strengthen the hypothesis of a matrilateral advantage when the parents’ union is intact and suggest that adolescents have little influence on the pattern of relationships already formed by their parents. However, the neutralization effects observed on certain variables when the parents are separated suggests a potential margin of flexibility among adolescents : after their parents have separated, they may feel less obliged by family norms, so their relationships with certain grandparents may be more based on personal affinity. The fact that adolescents more frequently take the initiative to meet their respective grandparents when their parents are separated, while their parents do so less frequently, suggests that separation may give the adolescent more freedom in the management of family relationships. In a wider context, Attias-Donfut et al. (2002), referring to forms of intrafamily solidarity, note the transition from a normative rationale to one of greater personal affinity, and this finding appears to apply here too.

47In parallel, we have also highlighted a form of indirect matrilaterality expressed via an imbalanced distribution across the paternal and maternal lines of the effects specific to the control variables, as well as gender effects relative to the sex of both the grandparents and the adolescents, especially for young girls influenced by gendered representations and practices. Combining the results for separation and those attributable to the control variables has highlighted two contrasting figures : that of the maternal grandmother, and that of the paternal grandfather. The maternal grandmother’s advantage doubtless stems from the combination of gender and lineage effects which provide a certain protection from family crises such as separation. Last, when the parents are separated, the effects in terms of laterality are observed mainly in the expressive dimension of the relationship. These effects appear to stem more from the representations specific to the adolescent than from more frequent contact with one or other grandparent.

48It is important to mention the elements which lend pertinence to this analysis model. The diversity of lineage effects across dyads when the parents are separated calls for differentiated treatment of intergenerational relationships. By identifying one-directional effects in terms of laterality, the literature dealing with this question would do well to avoid using the metaphor of balance, whereby if one lineage is favoured, then the other one is necessarily penalized. If relationships are weak on one side of the family, this does not necessarily mean that contacts are frequent on the other side. Likewise, a decrease in matrilateral bias, with regard to frequency of contact for example, does not necessarily imply an increase in contacts on the paternal side. For this reason, our dyad approach is especially useful for shedding light on the effects on both lineages.

49There are limits to our approach, however. The most obvious one is its cross-sectional nature, which does not allow us to explore the dynamics of these relationships and which necessarily narrows the scope of comparison between two distinct situations (parents who are together or separated). Dyad analysis reduces this bias, however, since dyads are compared with each other, independently of the parents’ union status. While the analyses are more robust when the factual dimension of the grandchild-grandparent relationship is considered, more precise measures of the expressive dimension of the relationship would provide more insight into the extent and the nature of the phenomenon observed. Moreover, as the analysis is based on secondary data, we cannot clearly identify relationship patterns which depend, for example, on the adolescent’s age, or on the grandparents’ symbolic or identity resources. We believe that these findings are significant, however, mainly because they shed a unique and original light on the bonds linking two generations in Switzerland.

50Can these findings be interpreted as a sign, within our societies, of changing kin relationships in situations of conjugal separation ? The work of Catherine Villeneuve-Gokalp suggests that this is the case. She shows that between 1986 and 1994 in France, the ties between fathers and their children became stronger after parental separation. This may stem from “a change in recent years in the way parental roles are perceived : the idea that the parental couple should survive the conjugal couple has gradually gained ground in France” (Villeneuve-Gokalp, 2000). In Switzerland, the norm of the “good divorce” (Théry, 1993), with amicable sharing of parenting roles, may also be gaining ground, notably thanks to legislative change [14] and to the actions of professional psychologists and mediators, not forgetting the influence of the media and the numerous self-help books available to parents (Lambert, 2009). In this respect, continuity of relationships with the father might also benefit the paternal grandparents.

51Cogswell and Henry (1995) rightly point out that in studies of grandparenthood, variables such as time since parental separation and grandchildren’s age at the time of separation are often lacking. As our study focuses on adolescents, the question of long-term patterns in grandparent relationships is of particular interest to us. [15] Studying support received by young adults from divorced and reconstituted families, Sylvie Cadolle (2004, 2005) shows that contacts between paternal grandparents and grandchildren are liable to vary over time. Of course, the maternal grandparents are a strong presence during the grandchildren’s childhood – most children being raised by their mother – but the paternal grandparents (grandmother especially) are able to “weather” the effects of separation (sometimes despite loss of contact between the children and their father) and maintain ties that can be revived when the grandchildren grow older. If the parents separate when the grandchildren are still young, we can hypothesize that a certain number of grandparents will “come back” into their lives when they grow up. If this is indeed the case, the paternal grandparents may not systematically lose out after their son’s separation, but may enter, or re-enter, their grandchildren’s lives as time goes by. To examine this hypothesis in more depth, we need to explore relationship trajectories over long periods, taking account of time since separation, grandchildren’s age and their degree of autonomy in family relationships.