Copyright Sociological Research Online, 2002


Charlotte Aull Davies and Nickie Charles (2002) 'The Piano in the Parlour: Methodological Issues in the Conduct of a Restudy'
Sociological Research Online, vol. 7, no. 2, <>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 9/5/2002      Accepted: 31/7/2002      Published: 31/8/2002


Although re-studies are relatively rare in sociological research, they can be very valuable resources for understanding social change. However, they also raise methodological questions about the validity of comparison given the inevitable changes in both social and analytical contexts during the period between the original and the re- study. This paper considers such methodological issues with reference to a major restudy in the area of family and household research. By looking at some of the details of research design in this re-study, we argue that reflexive consideration of relevant changes makes possible an examined and modified research design that retains much of the original and alters the remainder in ways that still allow for meaningful comparison.

Comparison; Family; Household; Methodology; Re-study; Social Change


Longitudinal studies in which data are collected from a specified panel or cohort at pre-determined intervals usually covering years if not decades are well known in sociological research (e.g., The British Household Panel Study; The National Child Development Study). Furthermore regularly repeated surveys of the same population base but not specifically the same cohort of individuals, of which official government censuses are prime examples, are well-known and generally accepted forms of data gathering which are subsequently widely used in sociological analyses. Re-studies, on the other hand, are considerably less common than these and while they share many methodological issues with longitudinal and other types of repeated research, they also raise some questions which are unique to themselves. By a re-study we mean a deliberate intent to repeat insofar as possible a previous research study using the same research design and methods to investigate similar theoretical concerns usually with the goal of better understanding social change. Re-studies differ from longitudinal studies in that they are determined upon post hoc, rather than being part of the original research design. Thus whereas longitudinal studies have some built-in continuity, which frequently includes continuity of personnel, re-studies often have to be redesigned with access only to the published findings of the original study rather than to raw data or to more intimate knowledge of how it was actually planned and executed (e.g. Phillipson et al. 1998). In order to be a candidate for a re-study, the original research obviously must have been significant and influential in a theoretical area that remains of interest after considerable time has passed. It also will be an area in which there is evidence of significant social change during the intervening period so that a re-study may be expected to provide new findings and insight into the nature and effects of such change.

Re-studies are relatively rare in sociological research, with the two Banbury community studies being a notable exception (Stacey 1960; Stacey et al. 1975; also cf. Crow and Allan 1994: 194-5 on community re-studies). They have received considerably more attention in anthropology, where a researcher may return to a particular location or group of people decades after they were first studied and the original ethnography produced. The most notable of such re-studies have seriously criticised or even claimed to undermine completely the original study (e.g., Freeman 1983 and Mead 1943 [1928]; Lewis 1953 and Redfield 1930). However, in other cases the re-study has expanded and modified the original findings, often by accessing previously unresearched aspects of the original location (e.g., Malinowski 1922 and Weiner 1988; also cf. Garbett 1967), while still recognising the validity of the original study. Re-studies in this vein are more informative about social change since they tend to reveal processes of change rather than asserting a major analytical discontinuity between two studies. Our re-study is of this latter type in that we intend to build on the findings of a now classic study undertaken forty years ago - The Family and Social Change: A Study of Family and Kinship in a South Wales Town (Rosser & Harris 1965)1. We will use this study as a basis for comparison to evaluate the effects on families and households of major social changes in the 40 years that have elapsed since it was undertaken. The original study included both a large-scale survey and ethnographic research. We will replicate both of these design elements in the re- study. This will provide greater robustness to the findings of the re-study and strengthen the validity of comparison with the original, since the survey can help overcome some of the difficulties of reliability introduced into purely ethnographic re- studies (cf. Davies 1999: 87-90) by the considerably greater effect the researcher has on data collected by ethnographic as opposed to survey methods. However, we do not propose an uncritical adoption of the methods of the original study, nor an overly facile acceptance of the validity of any comparison with its findings. We fully recognise the difficulties inherent in such an undertaking. One of the authors of the Banbury re-study argues that before undertaking a re-study, researchers should consider both the social and sociological changes in the intervening period and reflect on the relationship between the object of their re-study and the account of it produced by the original study (Bell 1977). This paper develops such a reflexive approach to what we refer to as the Rosser and Harris re-study by examining the various methodological issues it raises and how they can be addressed.

Re-studies in Family and Household Research

The area of family and household research in the post-World War II period produced a series of what have come to be regarded as classic sociological studies (Bott 1957; Rosser & Harris 1965; Sheldon 1948; Townsend 1957; Wilmott & Young 1960;Young & Wilmott 1957). These studies were concerned with the role of families and kin in the support of individuals, particularly as they age, under conditions of social change, in particular the growth of the welfare state, which appeared to be undermining the importance of families. Most of the studies found that extended families continued to have an important role to play in social life most notably in providing support for their members at times of crisis and major transitions and into old age. One of the last of these studies was undertaken in the early 1960s in Swansea in south Wales by Rosser and Harris (1965) specifically as a comparison with the Young and Wilmott study in Bethnal Green (1957). The major findings of the Rosser and Harris study confirmed earlier conclusions regarding the continuing importance of extended families but also pointed to indications of change in three specific areas which if continued would be likely to affect the role of family and kin in social life. Specifically they found changes underway in the following directions: They reported a lessening of the 'degree of domesticity' (Rosser and Harris 1965: 208) of women that was likely to reduce frequency of family contacts, which were disproportionately maintained by women. Second, they noted increasing mobility, primarily geographic but also occupational mobility, that seemed likely to reduce the frequency of contacts within the family as well as the number of kin relationships that could be maintained (1965: 209-218). And finally, they used their generational data to observe that the size of individuals' kin networks was decreasing through the twentieth century (ibid.: 172-5). Thus the Rosser and Harris study clearly recognised a number of changes apparently underway that suggests the value of a re-investigation to determine whether such changes have had the expected effect on the social role of families and kinship networks. Underlying these changes are profound changes in women's economic and social position, particularly their increasing participation in paid employment. In 1951, 27 per cent of the female population in Swansea were in employment, an increase of only three percentage points from 1911 and 1921 (ibid.: 77). But by 1991 women's economic activity rate had risen to 62 per cent in Swansea, with 53 per cent of employed women being in full-time employment (Welsh Office 1995: 109, 13). It is women's full-time employment rather than their employment per se which appears to be most significant in affecting the maintenance of family contacts (McGlone, Park and Roberts 1999). Another development that was not considered in the original study is the proliferation in family and household types since the 1960s. A recent (1994) ESRC programme that looking at population and household change found much greater diversity of family and household types including, in addition to the married couple with dependent children: single person households; cohabiting couples with and without children; various forms of reconstituted families as a consequence of divorce and remarriage, as well as death of a partner; lone parents - divorced, widowed and never married; gay and lesbian couples; and pensioners (McRae 1999). Clearly given this variety any re-study that attempted to use the same defining criteria for family and household as in the 1960s would simply miss much of the significant data about families in south Wales at the start of the twenty-first century. On the other hand, the relationship between these new family types and the trends identified by Rosser and Harris as likely to affect the social role of extended families is of considerable theoretical interest. It is important to ask both whether these new family and household forms are to some degree a result of these trends and/or are innovative ways of responding to their expected consequences in terms of undermining the support individuals throughout their lifetimes offer to and receive from their families. This presents a major methodological challenge for the re-study in that we want to retain the design and methods of the original but need to be able to incorporate data about these various family and household forms which the original research design clearly did not do.

The Original Research Design and Contextual Change

What was the original research design? Although the Rosser and Harris project was not a re-study, it was intended to evaluate the prior Bethnal Green findings regarding the continued social vitality of extended families but to do so in an urban setting that incorporated much greater social variation than the entirely urban working-class locality of Bethnal Green. This would enable it not only to test the findings but also to extend and refine them in terms of class and ethnic differences. The study consisted of two principal components: a major survey of the adult population of Swansea; and ethnographic research in two localities. The survey included the key Bethnal Green questions about frequency of contact with and distance of residence from various categories of the respondent's kin. It also collected data that enabled subsequent analysis to break new ground in its development of a social class classification that took account of two generations as well as self-assessment. A somewhat similar procedure was used in determining ethnicity. Maintaining the comparability of the ethnographic research between the two studies proved more difficult, due to the much greater heterogeneity of the survey area. 'Swansea was, in socio-geographic terms a loose federation of topographically distinct and dispersed "urban villages" rather than, as was Bethnal Green, a single densely populated area without internal boundaries' (C. Harris, personal communication). This implied that participant observation 'in the community' was simply not feasible, and 'in the event Colin [Rosser] went off and did some ethnographic work in Morriston and I in Sketty' (ibid.). This ethnographic work consisted of 'semi-structured interviewing with target groups and participant observation in specific localities' (ibid.).

The ESRC funded project is specifically to carry out a re-study of the Rosser and Harris study four decades on as a means of investigating the effects of social change on families and their role in supporting individuals at various stages in their lives. This paper discusses the various methodological challenges that such a re-study faces and how they are being addressed. We fully accept that such challenges are substantial. Indeed another recent re-study of three other classic studies of the same genre (Sheldon 1948; Townsend 1957; Young & Willmott 1957), while concluding that replication was impossible in any case due to the considerable differences between the original studies, further suggested that it was not even desirable both because 'social science methods have moved on and different approaches are now available' and also because 'repeating particular questions at different time periods raises issues of comparability' (Phillipson et al. 1998: 264). However, a major strength of our project is its intent to analyse social changes and their effects on families not just on the basis of contemporary data but through comparison with a highly influential, substantial and significant study undertaken in the same geographical location, with a similar population sample, and with similar theoretical concerns 40 years previously. Thus it is very important for us to retain the validity of this comparison from our overall research design down to the details of specific methods adopted.

However, we maintain that a re- study does not mean exact replication. Clearly this would be impossible. Contexts have altered so that the same question may have significantly altered meanings for respondents as well as being interpreted very differently by analysts. Nevertheless available knowledge about the nature of various contextual changes can be used to make necessary adaptations while retaining as much similarity as is feasible and still meaningful. This approach is not without precedent. Even regularly repeated surveys are routinely altered, with questions being dropped as they become irrelevant, new questions added, others altered - all without seriously undermining the comparability between surveys. Similar changes are normally built into longitudinal studies. We have identified five main areas in which contextual change provides methodological challenges for this re-study and this paper discusses these challenges in turn and some of the specific measures the Rosser and Harris re-study project is adopting to meet them. The five areas, moving from the comparatively trivial to the most serious, are:

  1. Changes in methods and methodological expectations;
  2. Technological changes in communication and transportation that affect the nature of social contacts;
  3. Changes in analytical contexts, in particular the treatment of social class and gender;
  4. Changes in the social context, such as the distribution and nature of cultural and ethnic groups; and
  5. A combination of social changes that have led to increased variety of household types and associated changes in theoretical treatment of families.

Changes in Methods and Methodology

The original research design, as already noted, was for a large-scale survey of approximately 2000 individuals (a random sample of 1 in 50, drawn from the electoral register) and ethnographic research in two of Swansea's 'urban villages'. The re-study will replicate the survey insofar as possible and undertake ethnographic research in four locations. The sampling for the survey will be carried out in the same way, with a random sample from the electoral register of 1 in 57 producing a sample of approximately the same size as in the original study. There have been some boundary changes in the intervening 40 years but these are accommodated fairly easily. The most problematic aspect of the survey is in the content of the questionnaire and issues raised there will be addressed in subsequent sections. In terms of more general considerations about how methods per se have changed since the early 1960s, the most important change for research based on large-scale social surveys is the widespread availability of computer software specifically designed for statistical analysis of the data they produce. Thus we will be able to produce the data in a variety of composite forms much more rapidly than was possible with the laborious although very ingenious techniques (involving punched cards and knitting needles) developed in the original study. This means we will also be able to have a much higher proportion of the collected data immediately accessible and to apply more sophisticated statistical tools when appropriate. However, although we hope we will be able to push the data analysis further than was possible in the 1960s, such capabilities do not undermine the potential or value of comparisons with the original study.

Ethnographic research on the other hand has undergone a much more thorough critique since the 1960s with resulting expectations about fieldwork practice and reporting conventions being quite considerably altered. We are somewhat hampered in any case by having less direct feedback on how the original ethnographic research was carried out. Furthermore ethnographic research, and particularly participant observation, is more profoundly affected by the ethnographer's own involvement in the research process. The original study conducted ethnographic research in two localities: 'we selected Morriston as an old industrial community representative of traditional Welsh chapel culture, . . . [and] Sketty as a middle-class residential area' (1965: 39), although the initial intention had been to sample three additional localities. The re-study is carrying out ethnographic research in four localities: Morriston, which retains much of its 1960s character; Mumbles, whose middle-class character is probably closer to that of 1960s Sketty than is the Sketty of today; a large post-war housing estate with severe economic deprivation and associated serious social problems; and an area near the town centre with a comparatively large non-white ethnic population. This research will be undertaken by three ethnographers who will endeavour to retain as much similarity as feasible in their approaches. This will be achieved in terms of an agreed schedule for ethnographic interviews as well as frequent meetings to compare findings. Although close replication of method is not possible for the ethnographic part of the project, the original study provides a wealth of ethnographically derived material (e.g., 1965: 6- 12, 82-5, 152-4, 231-2) which will be used to enable considerable similarity in the type of data collected.

Technological Changes

Since a major focus of the original study was on contacts maintained with members of a person's extended family, the main technological changes that will be of concern in the study are changes in means of communication and transportation that are likely to affect these contacts, either positively or negatively. Thus improved transportation and increases in car ownership not only make it possible to travel greater distances to see family members but also facilitate accepting employment farther away from one's residence as well as one's family of origin. In the original questionnaire, data were collected about when and where individual family members, representing different categories of relations, were last seen. The options from which respondents selected for the time ranged from the previous 24 hours to over a year and for the location from the respondent's home to the individual relative's home to the home of other relatives and an unspecified location ('elsewhere'). Our questionnaire has retained these questions in their original form. But we have added two additional questions asking when the individual relative was last contacted and how they were contacted. This final question provides responses that allow us to collect data about the use of new technologies for retaining family contacts, with options of 'telephone', 'e-mail' and 'other'. We also collect data across different households about the availability of new technologies, in particular internet access and mobile phones since clearly this will be a determinant factor in their use and may vary systematically by household type as well as by social class.

On the whole these technological changes are fairly easily accommodated into the research design without seriously compromising comparability with the original study. The degree to which the support of extended family is provided via new communication technologies rather than through direct contact, as well as information about whether this kind of support is associated with particular family types or social categories, will be a significant and valid area of comparison. Our ethnographic data will allow us to fill out our understanding of how such technologies are employed in this regard.

Changes in Analytical Contexts

An essential element in a study of families is determination of social class, an important independent variable. Our survey is based on a sample of individuals, but collects data about the households in which they reside, their extended families and those of their current partner or spouse, if there is one, who resides in the same household. The original study developed a sophisticated measure of social class that broke new ground in incorporating subjective assessments of social class into the measure, as well as attempting to make use of cultural indicators of class. Basically Rosser and Harris developed a measure based on occupations of fathers and sons over two generations and self-assessment (Rosser and Harris 1965: 99-108). The resulting class position was then checked for consistency with various economic and cultural indicators of class, including education, income, house ownership or tenancy and house type, and, somewhat more problematically, prestige possessions (cars, televisions, telephones and pianos), holidays, attitudes to football and rugby, membership of clubs and attitudes about Sunday opening of pubs. These indicators were found to be distributed as expected across the different social classes generated by the measure that had been developed for the study (1965: 106-14). The main difficulty with this procedure from a contemporary perspective is the lack of attention to gender in the determination of social class. Thus only the father's occupation was even collected in the original survey and the head of household was assumed to be male with married women given the class position of their husbands. Furthermore, several of the cultural indicators of class were likely to be gender-specific and were predominantly associated with men, as was implicitly acknowledged in the wording of one such question: 'Do you (or does your husband) follow Rugby or Soccer?'

In the subsequent four decades social theorists have engaged in a major critique of social class determinations, particularly debating the way in which gender should be incorporated into these determinations (e.g., Acker 1973; Charles 1990; Delphy 1981; Goldthorpe 1983, 1984; Stanworth 1984), which must be taken into account in our analysis. Women's greatly increased participation in paid employment, noted above, clearly has implications for the way social class is defined and measured. The necessary changes in the questionnaire itself, to enable us to collect data relevant for these issues, are comparatively minor. Essentially we ensure that occupational data are collected for women as well as for men, unlike the original study, and we will not assign cohabiting or married women to class on the basis of their partner's occupation. In addition we are careful not to assume that there is a head of household, either male or female, whereas the original questionnaire presumed that there was a male head of household. We do, however, ask respondents whether they would say that there is a main breadwinner in the household and, if so, who it is (see Table 1 in Appendix 1). We do this in order to find out how the provider role is shared and whether or not it is gendered.

At the same time, direct comparability of findings about the effects of social class on extended family relationships would appear to be compromised if different methods (male-centred in the original and gender-sensitive in the re-study) for determination of class are employed. However, given the comparative rapidity with which results can be generated using computer-based analysis, our intention is to apply both methods of social class determination to the data collected in the survey. This will allow us to compare the results of the two surveys based on the same method of social class determination, but in addition we will be able to consider the differences in the explanatory power of social class when it has been assigned by gender-sensitive methods. Thus we will have two levels of comparison: the first, between the original study and the re-study based on the same method for determining social class; and the second, between the two methods in terms of the kinds of explanations each generates when applied to the re-study data. This latter comparison will also allow some informed speculation as to whether and how the findings of the original study may have been affected by its use of a male-centred determination of social class.

The cultural indicators are more problematic in general, given the normally high variability of the content of class (and ethnic) cultures. For instance, does the assumed association of rugby with middle- class, and soccer with working-class culture still hold? Clearly television is no longer a 'prestige possession', and it is unlikely that the piano in the parlour has the same meaning as 40 years ago, when it was very much an expected feature of respectable Welsh Nonconformist households. Nevertheless these cultural indicators were used in the original study only as an additional confirmation of the social class assignments rather than as one of the determinants. Thus they are not central to the analysis, but a comparison between the strength of their association with social class in the two studies will provide interesting information about cultural change. Furthermore, if they are unconfirmed when a gender-sensitive measure of social class is constructed, this is a useful finding whose implications can be explored, yet which does not undermine the alternative analysis. The use of cultural indicators of class raises similar analytical questions to those associated with the use of cultural indicators for ethnic groups which are discussed further in the next section.

Changes in the Social Context

In addition to the substantial changes in household types, which are discussed below, there are a number of other significant changes in the social context which will affect findings and have had to be reflected in the redesigned questionnaire. Of these the most important have to do with levels of unemployment and questions of ethnicity. When the original study was carried out in the early 1960s there was virtually full (male) employment in Swansea, as in the rest of Britain. Thus questions about employment and occupation simply asked whether respondents were employed or not and what their occupation was. Data on occupation were also gathered for those who were retired but not for those who were out of the labour market for other reasons, such as unemployment or family responsibilities. We have taken account of increasing levels of unemployment and women's increased participation in the workforce by asking for current employment status and for current and past (if out of the labour market) occupational status. These changes are important in assessing women's occupational category, since they may be a housewife at the time of interview, but have a clear occupational status based on previous employment and training. They will also ensure that we obtain information about the occupations of those who are unemployed. There is a high incidence of long-term unemployment in this area of declining traditional heavy industries, and we wish to explore its effects on families and whether it tends to pass from one generation to the next. Thus the re-designed questionnaire collects data about this and provides a basis for assessing the degree to which long-term parental unemployment may be related to a similar outcome for the next generation. This type of question is new to the re-study simply because it was not seen as an issue in the Swansea of the early 1960s.

There are two aspects to the changes that affect determination of ethnicity. In the first place, the minority ethnic population in Swansea has grown fairly substantially since the early 1960s. While still a small percentage (1.6% in 1991), it marks an important change since the original study, and since it is concentrated in a couple of Swansea's 'urban villages', it has had a significant impact on these specific areas. Given our sampling procedure working from the electoral register, we expect to obtain a subset of respondents from these minority ethnic groups that is approximately the same percentage of the sample as these groups are in the total population. Unfortunately even if this is the case the numbers will be quite low (estimate of 32) making statistical procedures invalid in many instances. We think it is important that this population is adequately represented in the findings and intend to ensure this by organising additional questionnaires to make up the numbers and by carrying out one of the ethnographic studies in an area with a high minority ethnic population. Another consideration is to ensure that the content of the questionnaire is not ethnocentric in ways that would make it difficult or impossible for individuals from these groups to respond sensibly. The main questions where this is problematic are those about religious affiliation or attendance, and these have been altered, removing implicit assumptions that equated religion with Christianity (see Table 2 in Appendix 1).

After some discussion, we decided to retain questions about where respondents celebrate Christmas since the secular dimensions of this particular holiday are so pervasive. We have also made the collection and coding of information about location of the residence of various categories of relatives more open to those from elsewhere by expanding available responses (see Table 3 in Appendix 1).

The only ethnic distinction that the original study dealt with was that between 'Welsh' and 'non-Welsh', also referred to as between 'Welsh' and 'English' (Rosser and Harris 1965: 247, 255). The method developed to assign ethnicity to a respondent paralleled that developed for social class in that it looked at a particular indicator over two generations. However, unlike the procedure developed for social class, it did not incorporate any element of self-assessment of ethnic identity. Furthermore, the task was perceived as one of identifying those who were culturally Welsh, and the single indicator that was selected was Welsh-speaking ability, taking both respondents and their parents into account (1965: 117-126). This meant first that Rosser and Harris explicitly rejected place of birth as their criterion for Welshness, asserting that 'whatever its significance in determining "nationality" . . . [it] is of little use in determining an individual's cultural allegiance' (1965: 118). It also meant that, in contrast to their determination of social class, which used positive attributes for each of the two possible categories, Welshness was determined by the presence or absence of the single indicator, the Welsh-speaking ability of respondents or their parents. The 'non- Welsh' category, which made up over half the sample, thus was a residual category characterised as 'a very mixed bag of English, Irish, and Scots immigrants (or their descendants) together with a substantial . . . proportion of anglicized Welsh' (1965: 121). Certainly questions can be raised about the face validity of this approach even for Swansea in the 1960s, given the large percentage of the population of Wales, mainly in the old industrial areas of south Wales, whose self- identity is strongly Welsh but whose first language is English (cf. Balsom 1985 for a discussion of this population based on survey data from 1979).

Nevertheless it must be acknowledged that the original survey did not aspire to locate those who self- identified as Welsh but rather to define a collectivity which the researchers themselves judged to be culturally Welsh. Hence, as with their determination of class, they examined how their 'Welsh' and 'non-Welsh' collectivities stood up when checked against various other cultural criteria, for example, attitudes about compulsory teaching of Welsh in schools and Sunday opening of pubs, and religious affiliation and attendance. Thus the 'Welsh' were found to favour Welsh being taught in schools (82 per cent compared to 64 per cent of the 'non-Welsh') and to disapprove of Sunday opening (58 per cent to 41 per cent). They were also more likely to belong to a Welsh Nonconfomist denomination and to attend chapel or church. Our questionnaire will retain most of the questions that allow us to compare the distribution of these cultural indicators with the original study, although we deemed the question about Sunday opening to be so irrelevant to contemporary concerns as not to warrant inclusion and, as already noted, the questions about religion are inclusive of non-Christian religions. We expect that analysis will provide interesting findings about the malleability of cultural criteria as well as the direction of cultural change.

Even had the original study incorporated some form of self-assessment of Welsh identity, it is very unlikely that a re-study would be able to make valid direct comparisons due to far-reaching social and political changes in the intervening period that have inevitably affected ideas about the nature of Welsh identity and the significance of the Welsh language for this identity. With respect to the principal indicator adopted, namely Welsh-speaking ability, the social and economic pressures that have undermined the transmission of Welsh between generations since at least the late nineteenth century have continued although in modified form. But they have been to some degree countered by the results of political actions that have increasingly provided the language official status and a greater public presence, as well as by the success of locally based movements for Welsh-medium education. In addition, the establishment in 1999 of the National Assembly for Wales with its advocacy of a concept of Welsh citizenship based on residence further complicates the question of Welsh identity, tending to weaken its association with cultural indicators.

The re-study will collect the same data regarding Welsh-speaking ability of respondents and their parents, which will allow us to generate the 'Welsh' and 'non-Welsh' categories on the same basis as in the original study. It should be noted that the original survey made a significant improvement on the Welsh-language question in the official Census of the time by allowing respondents to select whether they were fluent or partially fluent Welsh speakers or non-Welsh speaking, a distinction our questionnaire will retain. We will thus be able to make a direct comparison with the results of the original survey. Based on the Welsh-language statistics from the official censuses since 1961, we expect to find a dramatic reduction in the relative size of the 'Welsh' category compared to that of the original study. For comparison with the original and to provide an alternative independent variable based on ethnicity, we will also introduce a number of other language and ethnicity questions to establish a fuller picture of the cultural and linguistic characteristics of our sample. In the first place, we recognise the greater linguistic diversity due to the presence of other ethnic groups by problematising not only Welsh-speaking ability but also English-speaking. Second, we collect data not only about language-speaking ability but also about language use within the household. Third, we collect data about attendance at Welsh-medium schools by respondents, their children and their parents. Finally, we incorporate a question about ethnic self-identity. This proved a difficult question to formulate so as to fit our particular requirements. There are a number of models for questions of this type, mainly developed for surveys of political opinions and voting intentions (cf. Wyn Jones & Trystan 1999), but these concentrate on either Welsh/English or Welsh/British distinctions which we felt were both too directive, and too exclusive for our purposes. The ideal solution would be to provide an open response to a question analogous to the one used for self-assessment of class, namely, 'What social class would you say you are?' Thus we considered asking 'How would you define your ethnicity?' However, the general understanding of the meaning of 'social class' is considerably more consistent with sociological interpretations than is the general understanding of 'ethnicity', which often is identified with 'race'. On the other hand, asking about 'cultural identity' seemed imprecise and even more ambiguous than ethnicity. Considering our main interest is to identify a Welsh collectivity, based in part on self-definition, and also to locate other self-defined ethnic collectivities, we opted for the following format:

Do you consider yourself to be Welsh? [Yes/No]
Do you consider you have any other cultural, ethnic or national identities?
IF YES, What are they?

These questions will allow us to locate respondents who self-identify as Welsh and compare this to the collectivity that is identified by the language data, as employed in the original survey. It is also open to multiple ethnic and cultural identities, encouraging respondents to consider them and allowing self-determination as to how they are labeled. We feel that this will enable us to construct a sophisticated picture of the ethnic variation within our sample. Ethnographic research will then allow us to explore further the meanings of this ethnic and cultural variation, and how it may be related to household types and extended family ties.

Changes in Family and Household Types

As mentioned earlier, the area of social change with probably the most far-reaching implications for the conduct of this re-study is the proliferation in family and household types since the 1960s. The re- study will look at how these various household types are related to independent variables such as class, ethnicity and generation. And it will also consider whether and how the frequency and nature of extended family ties are affected by these new family and household types. There are two main concerns in accommodating this area of social change. The first is to ensure that data collection is adequate to capture this variety, without inadvertently making assumptions that force respondents into inappropriate categories. The second has to do with questions that arise as the result of the increased variability within the dataset and the consequent smaller size of some analytical subcategories.

In order to ensure the openness of the questionnaire to this variety, we have had to alter the wording of many of the questions so as to eliminate the underlying assumption, in the original, of the standard household as based on a married couple with dependent children. Thus throughout we refer to 'partner or spouse' and to 'cohabitation or marriage'. When we collect data on long-term relationships, we preface this by referring to 'marriage - or any relationships like a marriage - ' and are careful to avoid language that assumes a heterosexual basis for such a relationship. Our definition of a household, which is consistent with the original study, is taken from the official census: one person living alone; or a group of people (not necessarily related) living at the same address with common housekeeping - sharing either a living room or a sitting room or at least one meal a day. This is clearly open to all sorts of household compositions, and the first table in the questionnaire, as in the original, collects data about all the household residents.

The orginal questionnaire also contained tables for data about various categories of relatives: parents, children and siblings of respondents and their spouses; maternal and paternal aunts and uncles; and an open category for relatives (cousins, for example) who were important to the respondent but not specified elsewhere. We have retained these tables but supplemented them so as to include varieties of reconstituted families. Thus we collect data on step-parents and step-children, half- and step-siblings; and in the open category we prompt the respondent with the examples of ex-partners and ex-in-laws as well as grandchildren and cousins. Taken together these additions should ensure that our data will reflect the variability in households and families that has become increasingly common since the 1960s without sacrificing comparability.

The ethnographic studies are particularly important in enabling us to flesh out our data on family and household types. The core of the four ethnographic studies is semi-structured interviewing based on an open but agreed interview schedule which allows us to ensure an acceptable level of consistency and comparability between the data collected by different ethnographers. In these interviews we ask our informants to explore more fully the meaning of family for themselves. We ask, for example, who they consider to be family and how their families contribute to their own sense of self. We also adopt a procedure (cf. Phillipson et al. 2000: 29-31) in which informants are shown a diagram consisting of three concentric circles and asked to place in the inner circle those who are so important to them that they could not imagine life without them, in the next circle those who are important but somewhat less central to their lives and so forth. This procedure is designed both to give us more information about the kinds of emotional support people give and receive and the subjective meaning of that support as well as about the prevalence and nature of close 'family-like' ties with non-relatives and how these are interpreted. The ethnographic interviews are supplemented by participant observation to enable us to verify and contextualise the data collected by other methods.

The second concern raised by the inclusion in the re-study of this greatly increased variability in household types is a primarily technical consideration. Analysis of the data in the original study was accomplished almost entirely by means of multivariate analysis based on contingency tables. The principal independent variables were social class and cultural group, but it was often necessary to control for other variables, such as religion and locality. Furthermore, in examining different aspects of family behaviour, it was also important to take into account how this was affected by their position in a cycle from formation to dissolution. Nearly all the respondents in the original study (1725 of the 1962 completed interviews) could be placed in one of four family phases based on a married couple model - from 'home-making' through 'procreation' (with the birth of the first child) and 'dispersion' (as grown children marry and leave home) to the 'final' phase, which ends the cycle with the death of the original partners (1965: 165). When these phases were also linked to household composition, the number of respondents included in the cross-tabulation remained high at 1525 (1965: 167). Clearly the data in the re-study will yield a much greater variety of household types, and the family phases model will be inapplicable for larger numbers of them. Furthermore, household type itself becomes a significant additional variable, which greatly increases the number of cells in contingency tables that are based on it and reduces the number of available cases for tables that attempt to control for it. Thus it is likely that for some parts of the analysis the number of cases in individual cells will be too low to allow analysis to proceed. We will address this difficulty in part by completing extra questionnaires for particular subcategories, as has already been noted regarding ethnic groups. These additional respondents will be identified in the course of our ethnographic research. In addition, we will supplement analyses by means of contingency tables with other multivariate designs, such as factorial and covariate designs, which will be less affected by the small size of the subcategories being examined. While such changes in the forms of analysis from the original should allow us to incorporate the much greater variability in household types which contemporary studies of family and kin must accommodate, it does not appear to compromise comparability with the original study.


Re-studies provide a potentially rich resource for sociological analysis, particularly for enhancing our understanding of social change, a resource which has not been adequately developed. Certainly there are often pragmatic reasons, such as lack of information about the details of the original research design, that mean close replication cannot be achieved (e.g., Phillipson et al. 1998; 2000). The Rosser and Harris re- study is fortunate in having the active co-operation of one of the original research team, Professor C. C. Harris, as well as access to the original survey instrument.

Society changes, as does the way it is conceptualised, and both kinds of change affect the conduct of re-studies. Since the original Rosser and Harris study was carried out in 1961, the range and distribution of family forms have changed dramatically. Equally significant for research on families and households are changes in patterns of employment (especially women's greatly increased participation in paid employment), the increasing insecurity of employment and the growth of a more ethnically varied population. Cultural content is also notoriously variable as are the meanings of cultural markers - for example, the association of certain consumer goods (the piano in the parlour) with particular social groups (middle-class, Welsh Nonconformist), so self-evident in the 1960s, is puzzling or amusing to researchers in 2001. And political changes, such as the establishment of the Welsh Assembly, also affect the way cultural and ethnic identities are understood both by respondents and researchers. In addition to these historical and cultural changes, sociological analysis has moved on, so that families, gender relations, class and ethnicity are all conceptualised differently than they were in the 1960s. Clearly such conceptual changes are much more important than other technical developments in data analysis, although these may facilitate some aspects of the re-study. All of these factors must be considered early on in the conduct of a re-study and the methodological issues they raise addressed. However, if a re-study is approached in such a reflexive manner, so that the research design is replicated where meaningful and altered after careful consideration of social and analytical changes where necessary, it can be a very worthwhile and valuable contribution to social analysis - one that is likely to increase our understanding of both the original object of study and the sociological account of it, as well as the contemporary object of the re-study and the intervening processes by which one was transformed into the other.

Appendix 1

Table 1 Determination of Head of Household

Table 2 Determination of Religious Affiliation

Table 3 Determination of Place of Residence of Relatives


This re-study is a 3-year research project funded by the ESRC (Social Change, Family Formation and Kin Relationships R000238454). It is being carried out in the School of Social Sciences and International Development and the National Centre for Public Policy at the University of Wales Swansea by Professor Nickie Charles, Professor C. C. Harris, Dr Charlotte Aull Davies, Dr Bettina Becker and Professor Michael Sullivan.


ACKER, J. (1973) 'Women and Social Stratification: A Case of Intellectual Sexism', American Journal of Sociology, Vol. 78, No. 4, pp. 936 - 945.

BALSOM, D. (1985) 'The Three-Wales Model' in J. Osmond (editor) The National Question Again. Llandysul: Gomer.

BELL, C. (1977) 'Reflections on the Banbury Restudy' in C. Bell and H. Newby (editors) Doing Sociological Research. London: George Allen & Unwin.

BOTT, E. (1957) Family and Social Network. London: Tavistock.

CHARLES, N. (1990) 'Women and Class - A Problematic Relationship?', The Sociological Review, Vol. 38, No. 1, pp. 43 - 89.

CROW, G. and ALLAN, G. (1994) Community Life: An Introduction to Local Social Relations. New York: Harvester Wheatsheaf.

DAVIES, C. A. (1999) Reflexive Ethnography. London: Routledge.

DELPHY, C. (1981) 'Women in Stratification Studies' in H. Roberts (editor) Doing Feminist Research. London: Routledge.

FREEMAN, D. (1983). Margaret Mead and Samoa. Cambridge, Mass: Harvard UP.

GARBETT, G.K. (1967) 'The Restudy as a Technique for the Examination of Social Change' in D.G. Jongmans and P.C.W. Gutkind (editors) Anthropologists in the Field. Assen: Van Gorcum.

GOLDTHORPE, J. (1983) 'Women and Class Analysis: In Defence of the Conventional View', Sociology Vol. 17, No. 4, pp. 465 - 488.

GOLDTHORPE, J. (1984) 'Women and Class Analysis: A Reply to the Replies', Sociology Vol. 18, No. 4, pp. 491 - 499.

LEWIS, O. (1953) 'Tepoztlan Restudied', Rural Sociology Vol. 18, No. 2, pp. 121 - 136.

MALINOWSKI, B. (1922) Argonauts of the Western Pacific. London: Routledge & Kegan Paul.

MCGLONE, F., PARK, A. and ROBERTS, C. (1999) 'Kinship and Friendship: Attitudes and Behaviour in Britain, 1986-1995' in S. McRae (editor) Changing Britain: Families and Households in the 1990s. Oxford: Oxford UP.

MCRAE, S. (1999) 'Introduction: Family and Household Change in Britain' in S. McRae (editor) Changing Britain: Families and Households in the 1990s. Oxford: Oxford UP.

MEAD, M. (1943 [1928]) Coming of Age in Samoa. Harmondsworth: Penguin.

PHILLIPSON, C., BERNARD, M., PHILLIPS, J. and OGG, J. (1998) 'The Family and Community Life of Older People: Household Composition and Social Networks in Three Urban Areas', Ageing and Society Vol. 18, No. 3, pp. 259 - 289.

PHILLIPSON, C., BERNARD, M., PHILLIPS, J. and OGG, J. (2000) The Family and Community Life of Older People: Social Networks and Social Support in Three Urban Areas. London: Routledge.

REDFIELD, R. (1930) Tepoztlan - A Mexican Village. Chicago: University of Chicago Press.

ROSSER, C. and HARRIS, C. (1965) The Family and Social Change. London: Routledge & Kegan Paul.

SHELDON, S. (1948) The Social Medicine of Old Age. Oxford: Oxford UP.

STACEY, M. (1960) Tradition and Change: A Study of Banbury. London: OUP.

STACEY, M., BATSTONE, E., BELL, C. and MURCOTT, A. (1975) Power, Persistence and Change: A Second Study of Banbury. London: Routledge & Kegan Paul.

STANWORTH, M. (1984) 'Women and class analysis: a reply to John Goldthorpe' Sociology 18 (2): 159-70.

TOWNSEND, P. (1957) The Family Life of Old People. London: Routledge & Kegan Paul.

WEINER, A. B. (1988) The Trobrianders of Papua New Guinea. New York: Holt, Rinehart & Winston.

WELSH OFFICE (1995) Welsh Economic Trends, No. 16. Cardiff: The Welsh Office.

WILMOTT, P. and YOUNG, M. (1960) Family and Class in a London Suburb. London: Routledge & Kegan Paul.

WYN JONES, R. and TRYSTAN, D. (1999) 'The 1997 Welsh Referendum Vote' in B. Taylor & K. Thomson (editors) Scotland and Wales: Nations Again?. Cardiff: University of Wales Press.

YOUNG, M. and WILMOTT, P. (1957) Family and Kinship in East London. London: Routledge & Kegan Paul.

Copyright Sociological Research Online, 2002