Food studies and the integration of multiple methods

The study of food and eating draws evidence from many different disciplines using many different methods. This paper argues that this should be viewed positively; the idea that there is one best method, or suite of methods, to which every social scientist should be committed is at odds with both processes of knowledge formation and the complexity of alimentary life. Complex research questions necessitate multiple sources and methods, the greatest challenge being to fashion and to justify an integrated interpretation of different types of data. The paper examines these issues with reference to debates about the use of mixed methods, which have mostly focused on the reconciliation within a single study of quantitative and qualitative data. A broader remit is advocated, paying attention to protocols for integrating multiple methods. The argument is illustrated with reference to the design and analysis of a study which collected evidence about the activity of eating out. The paper speculates about whether deliberate and purposeful use of multiple methods may be a key to generating explanations, and ultimately theories, which transcend disciplinary boundaries.


Introduction: food studies, theories and methods
This paper reflects on the status of methods in the social sciences with specific reference to the field of food studies.Developing new methods and refining old ones is obviously worthwhile, but I contend that social scientists when discussing methods pay too much attention to data collection and protocols for data manipulation and too little to procedures for interpretation and analysis.The consequence is a tendency to over-estimate the value of methodological purity.An alternative is purposefully to combine different types of method.This is the apparent objective of recent widespread advocacy of mixed methods.However, I express strong reservations regarding the codification of mixed methods as a new methodological paradigm, preferring greater flexibility in the pragmatic application of complementary methods.This can be especially valuable for food studies because of its multidisciplinary constitution.Also the social science of food and eating remains highly fragmented and consequently is theoretically weak.The field spans agronomy, chemistry, medicine, the physiology of taste, nutrition, marketing, logistics, culture, theology, hospitality, history and the social sciences.Probably no one has adequate command of all these areas, and thankfully few scientific questions would require the contribution of all, although perhaps an explanation of the spread of obesity might.However, many narrower questions would also benefit from insights from a range of perspectives and disciplines.
Explanations of patterns, changes and problems associated with food consumption require different types of expertise which are typically concentrated within the boundaries of different disciplines.But while the boundaries of disciplines serve usefully to protect and enhance scientific theory, they are often not helpful in explaining the sources of practical problems like obesity, taste or food choice.Such explanations will usually depend on the application of more than one method.My basic contention is that food studies can, and should, make deliberate and purposeful use of multiple methods and data sources.However, a major problem arises of how to devise suitable protocols for achieving a combined interpretation of data generated through different procedures theoretical and disciplinary traditions.At the extreme, this would apply to the union of experimental evidence in a psychologist's laboratory and the ethnographic fieldwork of the anthropologist, or perhaps the econometrician's equations and the interview transcripts of the sociologist.For the present, much less ambitiously, I explore merely types of research design which permit and encourage the coherent use of findings derived from more than one type of data..However, that does not entail endorsement of mixed methods as currently discussed in the literature, even though the aspiration towards synthesis is most welcome.
The main illustration is an example of the use of multiple, complementary methods in an investigation of the practice of eating out which involved a research design employing quantitative and qualitative evidence.The example presents some previously unreported empirical findings relevant to a substantive interpretation of the phenomenon of eating out as well as illustrating some principles of method.One Warde Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 53 implication is that greater benefit would be obtained from refining protocols for combining methods than from honing any single method or technical instrument.
Another implication is that interdisciplinary collaboration might best be enhanced through the alignment of different methods.Whether the existence of disciplines is a boon or an encumbrance is much disputed.Inter-disciplinarity is increasingly recommended.This is particularly the case when agencies commissioning research find themselves wishing to intervene to resolve practical problems -a not uncommon occurrence with respect to the sponsoring of food research.Policymakers generally do not find it helpful, when for example they perceive a crisis of obesity, to be offered intelligence and advice in terms of mutually exclusive disciplinary perspectives.Yet they often are because disciplines have their own particular scientific agendas and foci of attention.Disciplines tend to be committed to incommensurable theories, which have been formed over time in relation to particular substantive interests.Their key concepts bracket out those forces, processes and facts which are of no theoretical interest.Disciplinary institutionalisation and rivalry militates against theoretical synthesis and results in high degree of concentration on explaining specific categories of phenomena.However, there appear to be fewer barriers to the sharing of social scientific methods.Social research methods texts and articles are rarely dedicated to the practices of specific disciplines.If methods travel better than theories they may provide grounds for convergence in explanation.

Disciplines, methodologies and methods
Disciplines typically favour some methodological techniques over others.Methods have typically been nurtured, invented and developed in relation to explanatory problems of interest to specific disciplines.Consequently a strong connection between method and theory is often apparent.Connectedness has sometimes been enunciated as an epistemological and methodological principle.In the last quarter of the 20 th century it was maintained widely in the field of Social Research 1 that qualitative and quantitative methods were incommensurable because they did not share the same abstract ontological and epistemological assumptions about social action and social processes. 2Thus were founded versions of methodological purism which rendered qualitative and quantitative approaches antagonistic and mutually exclusive.Disagreement was intense. 3 Morgan (2007) sketches the background as one where advocates of the qualitative, or 'metaphysical', 4 approach challenged the dominant positivistic and quantitative orthodoxy in social science methodology.Insistence on there being a radical disjuncture led to the formulation and promotion of a new 'paradigm', founded in the interpretivist tradition of social science and committed exclusively to qualitative methodology.Within the domains of research influenced by Sociology, researchers became attached to one or the other, almost entirely irrespective of the range of relevant research questions envisaged.
While methodological purism -the contention that there is a best method -remains surprisingly rife, 5 the climate of opinion has softened and there is probably a good deal of pragmatic acceptance, or at least tolerance, of mixing methods.Alan Bryman (2006b), a constant supporter of the judicious deployment of complementary methods, welcomed the cessation of hostilities in the 'paradigm wars' as the basis of a healthier climate for investigation.Claims to methodological exclusivity on grounds of ontology and epistemology are now rare.Indeed, methods seem very amenable to transfer across the boundaries of disciplines and theoretical schools.One reason is that methods have path dependent trajectories which are relatively autonomous of disciplines.This is itself probably a result of the specialisation of scholarship about methods; describing, developing and justifying procedures for data analysis has become a full-time occupational role.Focus on methodical procedures irrespective of substantive area or theoretical proclivity is demonstrated, for example, when methods textbooks draw examples from many different domains or disciplines.Whether researchers should allow methodologists to prescribe how to use methods is itself a debatable matter, but techniques of investigation have proved portable.So while imminent spontaneous convergence is to be expected neither over the questions that sociologists, economists and psychologists ask, nor over their core theoretical commitments, methods are inherently more susceptible to transfer.Perhaps, therefore, the relatively loose connection between theory and methods could be turned to advantage and in the case of food studies impetus _____________ 3 The basis of the dispute was the view that studies of numerical distributions, often labelled positivism, and studies of expressions of meaning, interpretivism.There are both historical and philosophical grounds for asserting that version of positivism and interpretivism each adopt mutually incompatible accounts of structure and action at the level of abstract social theory.Whether they must be severed in the same way at the level of substantive explanation is much less certain.In relation to analysis in search of practical or policy interventions sociologists of different types and scholars from different disciplines are less argumentative. 4The label probably reveals aspects of Morgan's preferred understanding of the debate. 5The idea of a best method probably makes sense only on the basis of a prior theoretical commitment.Disciplines with greater theoretical consensus are likely therefore be most adamant about the superiority of their (currently) preferred methods.Some of the associated issues have been explicitly addressed in discussions of 'mixed methods'.While it is now easy to find explicit advocacy of the use of mixed methods in research design only a quarter of a century ago their recommendation was considered highly controversial.Julia Brannen's (1992) Mixing Methods, for instance, was introduced against a backdrop of division and hostility between proponents of either qualitative or quantitative research.Mutual incompatibility was identified and defended at the levels of epistemology, middle-range theory, and methods and techniques (Brannen, 1992:3).The effects were momentous.Abbott  (2001: xvi), illustrating fractal distinction and the principle of self-similarity in sociology joked: if we take any group of sociologists and lock them in a room they will argue and at once differentiate themselves into positivists and interpretivists.But if we separate those two groups and lock them in separate rooms, those two groups will each in turn divide over exactly the same issue'.
In doing so, while pointing to the widely advertised dispute between advocates of quantitative and qualitative approaches, he also suggested compatibilities.Currently, majority opinion rejects arguments for incommensurability.Symonds and Gorard  (2012: 5-9), for example, argue plausibly that none of the elements of the research process -qualities of data, collection tools, sampling, data types, criteria of validity and bias, method of analysis -are exclusive to quantitative or qualitative paradigms.Examples abound of numerical data being analysed in an interpretative manner, as with Multiple Correspondence Analysis, and textual material with statistical algorithms, as with content analysis.
As confidence in the incommensurability thesis has declined, two positions have transpired. 6One is the promotion of a new, transcendent, third way, a selfconsciously constructed and promoted alternative research paradigm called variously 'mixed' or 'mixed methods' research (Johnson et al, 2007; Greene, 2008).Emerging in the 1990s, mixed methods has a variety of formulations but central to each is the transcendence of the division between qualitative and quantitative approaches.With a long history, one well known thread being reflection on the methodological virtues of triangulation, many contributions have attempted to formalise 'mixed' research (eg Cresswell and Plano Clark, 2007; Tashakkori and Teddlie,  2003).These have had impact to the extent that some commentators see mixed methods as not only highly fashionable but increasingly a new orthodoxy; Green  and Preston (2005: 168)   7 The other position implicitly favours reliance on improvisation and invention of new research tools developed in the light of the desirability of incorporating multiple methods into research designs.However, the notion of a paradigm is rejected; the very idea of there being entities that could sensibly be deemed qualitative or quantitative 'paradigms' is deemed fallacious (Gorard, 2004; Morgan, 2007).Indeed, pretensions to paradigmatic status may be greeted with scepticism, as might suggestions that all research designs should deploy mixed methods; I would not recommend devoting much time to the gushing and babbling stream of methodological reflection which now surrounds the notion of mixed methods.However, the underlying imperative to consider carefully how to integrate findings generated by many different methods deserves very serious consideration.Sound reasons for mixing methods have emanated from the discussion.
Fielding (2012) considers that there are three principal reasons for mixing methods: 'illustration, convergent validation, and analytic density'.Illustration is a matter of embellishing accounts with data from different sources, which may be informative or decorative without being methodologically profound.Of the second he is very properly wary; he echoes criticisms of expositions of triangulation wherein a dubious capacity for mutual verification is attributed to the bringing of different methods together.The third is the most important and the basis for integration or synthesis: The really compelling argument for mixed methods designs remains their benefits for sophisticated analytical conceptualization.As Maxwell (2010) argues, the real quantitative/qualitative distinction is not between number and text but between understanding the world by a theory of variance featuring variables and correlations and understanding the world by a theory of process in terms of events and interactions.Put that way, it is clear that both are essential.Rather than mixing because there is something intrinsic or distinctive about quantitative data or qualitative data we mix so as to integrate the two fundamental ways of thinking about social phenomena.(Fielding 2012; 125): He goes on to note, however, that demonstrations of how to integrate data are sparse.
_____________ 7 Since 2007 it has a journal devoted exclusively to it -Journal of Mixed Methods Research, whose inaugural editorial was entitled 'The new era of mixed methods' (Tashakkori  and Cresswell, 2007), it is the subject of other special editions (eg Green and Preston, 2005), and there are handbooks of mixed method research which attend to procedure, design and analysis (eg Cresswell and Plano Clark, 2007; Tashakkori, A. and E. Teddlie, 2003).Fielding (2012: 124) quoting Ivankova and Kawamura (2010) note that articles featuring mixed methods, as reported in various bibliographic databases, increased from 10 in 2000 to 243 in 2008.2014, 51, Núm. 1 51-72   57 Tacitly, most explanations have always depended on multiple methods. 8Multiple methods are implicitly endorsed when we accept, make appeal to or quote findings which derive from techniques and methods which we ourselves do not use -and in many instances would be incapable of using.Theories, concepts and established knowledge have been derived and crafted from research using multifarious methods with the circumstances and specific methods of their discovery being quite forgotten.For it is almost impossible to give a sociological analysis or interpretation of any phenomenon without referring to mundane schemes of classification, statistical distributions, social and institutional settings, practical understandings and common utterances.Nevertheless, only recently has the appropriate procedure for dealing with this attracted attention, and it has come in the wake of the promotion of research designs based on mixed-method strategy.Bryman (2006a) conducted a content analysis of 232 social science articles, published between 1994 and 2003, which claimed to deploy mixed methods.Mixed methods clearly were becoming fashionable.This almost always meant the combination of quantitative and qualitative methods, which has been how methodology specialists have constructed the problem, although it seems to me that there are other possible combinations. 9Combining surveys with interviews was the most common strategy.Bryman's review very usefully isolated a substantial number of different justifications for the use of mixed methods (pp.105-107), although he concluded that the studies reported very often did not proceed on the basis of the rationale announced.Justifications for adoption of mixed methods included triangulation, completeness, process, discovery, better explanation and credibility.Bryman proposed that mixed methods research strategies should be better clarified in advance of data collection, while also noting opportunistic benefits arising when 'researchers discover uses of the ensuing findings that they had not anticipated' (p.110).Overall he argues persuasively that while use of mixed methods has become fairly common, established criteria for good design or evaluation of quality are few.

Política y Sociedad
Bryman found few convincing examples of applications of mixed methods. 10In most instances the relationship between the findings were simply additive; viewing a multi-faceted reality from two or more angles, each method offers its particular _____________ 8 This seems to be accepted by several of the more sanguine commentaries on mixed methods (Fielding, 2012; Howe, 2012; Morgan, 2007).The debate may well be seen in retrospect as much ado about very little. 9Other potential combinations include an author and the text that she has composed, the Key Informant and the minutes of the management meetings of his organization, a television programme and the audience at its viewing, food diaries and a medical examination, a comprehensive corpus of recipes and the particular selection in significant cookery books. 10It might be said that good examples might be uncommon in journal articles because lengthy reporting is required to give an adequate account of the results of mixed method investigations.
Warde Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 58 mode of understanding which is juxtaposed against the others, rather than being subjected to any form of technical integration.I will refer to this tentative distinction as one between 'additive' and 'integrated' multiple method designs.If few research projects have explicitly used integrated mixed method designs, they seem especially rare within food studies.In the absence of ready to hand examples of inquiries which draw evenly upon several disciplinary traditions and which might demonstrate the potentiality for integration of both theory and method, I will present an example from a study drawing upon recent developments in the sociology of culture, where exercises in the combination of qualitative and quantitative techniques have proved successful (Bennett et al, 2009; Elliott et al, 2010; Miles and  Sullivan, 2013; Cveticanin, 2012; Warde et al, 2008).The example derives only from within a single discipline, Sociology, although it is a reasonable testing ground not only because it is currently comparatively undogmatic as regards theory or doctrine but also because many of its controversies find echoes between and within other social sciences.In Bryman's (2006) sample of articles in social science journals Sociology was the discipline most prevalent in claiming use of mixed methods (the next most common being Social Psychology and Management Studies) .I describe the experience of a modest use of multiple methods in a recent study which obtained data on eating out and hospitality in Britain.This introduces, as an example of integrated multiple method, some previously unreported data of substantive relevance to the understanding of eating out.

An example: explaining eating out
The increased incidence of eating out is one of the most significant trends in contemporary European foodways.Any answer to the question 'Why are people eating away from home more often than before?' will necessarily take account of many contributory factors.Urban concentration of populations makes them easier to provision.Greater affluence generates additional personal financial resources and increases the potential market for commercial hospitality.There is increasing range of provision in maturing service economies which employ sophisticated advertising techniques.More married women are in employment than earlier, which both increases income and reduces time available for domestic work.Some relationship exists between the distribution of cooking skills and propensity to purchase readymade meals.A strong relationship exists also between levels of personal mobilitycommuting to work, leisure travel, visiting distant kin -and the injunction to eat at regular intervals.Aestheticisation of daily life impacts upon the food sector and broadens the range and desire for new culinary experience.To enjoy eating, particularly in the company of others, without personal responsibility for maintaining a flow of food to the table, is both privilege and treat.All of these factors have some relevance.Some are matters of social structural change, others aspects of a personal calculus of individual well-being.The social scientific interest lies partly in weighing the relative importance of each, and partly in determining how they are config-ured together.It seems incontrovertible that the necessary evidence can only be obtained from data collected and analysed in accordance with the procedures of several different methods.Yet there are few sources of authoritative instruction about how to draw these different types of findings together or about how to demonstrate their mutual compatibility. 11 want to report on a study which addressed the problem of combining the findings of different methods explicitly at the point of research design, where several methods were specified in the proposal and their mutual inter-linkage assured in a substantive manner.A study conducted with several colleagues, called 'Cultural Capital and Social Exclusion' (CCSE), 12 a partial replication of Pierre Bourdieu's  Distinction (1984), went beyond multiple methods by obtaining data about the same subjects using two (occasionally three) different techniques (Bennett et al, 2009;  also Silva 2005 and Thomson, 2004 for technical details).After having completed a random sample survey about cultural consumption and taste, some of the respondents were interviewed in much greater depth.By using the technique of Multiple Correspondence Analysis (MCA) to document the survey results, identifying symbolically significant cultural items along four principal axes of differentiation, a geometric representation of the probable coincidence of practices and tastes is obtained.Thereafter, using SPAD software, an identical space can also be drawn to position every respondent on each axis; technically referred to as the cloud of individuals, this located every individual in relation to all others by their possession of cultural capital.Individuals could thus be identified both within the population and within a structure of national taste using two methods, one typically considered quantitative, the other qualitative.Interviewing people who had already answered the survey gave the option of aligning the findings of two different methods using a single individual as the bridge between characteristic behaviours captured relative to a national population and the detail of personal tastes and practices.
The study paid relatively little attention to food habits, but the survey included questions about what restaurants people liked and disliked.In face-to-face semistructured interviews everyone was asked to elaborate on their responses to those questions and how they entertained guests at home.
The MCA graphs showed that tastes in eating out contributed in moderate degree to variation across the sample.On each of the first three axes one or more item was significant (Bennett et al, 2009: 45-52 gives full details).The first axis (see Figure 1) 13 showed that those with greater volumes of capital ate out frequently, _____________ 11 There is on the other hand a strong pragmatic streak in the social science community which accomplishes this through its writing, where rhetorical strategies overcome, often by overlooking, the potential hazards.
12 ESRC project Cultural Capital and Social Exclusion: A Critical Investigation. 13To read the figure, (black) diamonds refer to participation, (red) squares to expressions of taste.For participation suffix 0 means never or rarely, 2 means a lot, 1 sometimes.For Warde Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 60 while those with little capital went to restaurants rarely.As regards taste, the former group were likely to name fish and chip restaurants as the least liked, while the latter were the most likely to nominate these as favourite.However the numbers offering these expressions of taste were small; MCA often isolates as significant items which are not very common.On the second axis, high significance attached to eating in French restaurants; younger people tended to like them least of all, and it was older people who were most likely to nominate them as favourite, although there was also a class dimension.The third axis, not shown here, 14 revealed that men were particularly attracted to Indian restaurants.)Occupational class positions were laid out monotonically along the first axis, and while restaurant preferences contributed only moderately to that axis, when crosstabulated significant class differences were apparent (see Table 1).Thus for instance 14 percent of the professional-executive class nominated French restaurants as their favourite, compared with 3 percent of the working class.The equivalent figures for liking French restaurants least were 3 per cent and 7 per cent respectively.LeRoux et al (2008).Class is allocated on the basis of respondent's current or most recent occupation.
Interviews reveal some reasons for these preferences and give additional confidence that French cuisine plays a symbolic role related to the possession of cultural capital and social distinction.Thus Jennie, 15 a woman in her 40s with a high volume of capital, talked about reasons for her liking of French restaurants.'Well we lived in France, I like the rich sauces … I like my meat rare.She continued when asked 'was it living in France that … generated the interest in French cuisine or did it come before?'She continued: Well I grew up with quite a limited cuisine, and I think probably Italian would be the first influence and I don't mean pasta and pizza I mean again the good sauces, and the rich sauces and then I liked French as well and I like the fresh vegetables and so on.What came first?I think probably we watched some TV programmes, maybe found some, things like steak Diane, I mean _____________ Warde Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 62 these things were, that was supposedly French, whether it actually was or not I don't know.Coq au vin; I remember experimenting with coq au vin with one of my first cookery books so I think that interest came before we went to live in France.... I think their approach to vegetables and fresh cooking, I like that.And wine as well, red wine.
Such opportunities for learning are rarely available to the working class, and in the absence of any personal experience, and in the light of the reputation of French restaurants, they may well nominate them as least favourite.One interviewee admitted as much.Having nominated French restaurants as his least favourite, in interview explained that he had never been to one: 'I'm not qualified really to say that to be honest, but I had to give an answer so I figured it's one, let me put one that I haven't been to'.Elaborating he said, 'there's no real reason, that's no reflection on the French or anything, purely because I'm not, again it's something that I've not been to enough to really make a judgement'.One source of the alignment between social position and taste is thus rendered meaningful.
In addition, interviews contributed to greater certainty about the role of cultural omnivorousness associated with patterns of consumption in the British middle class.Perhaps the most striking feature of Table 1 is the strength of preferences for 'foreign' cuisine among professionals, two thirds of whom selected one of the four non-British cuisines listed.Earlier research in the mid-1990s (Warde and Martens, 2000;  Warde, Olsen and Martens, 1999) had shown a propensity for those in higher social class positions to like a wider array of foreign restaurants, a finding not inconsistent with Table 1, but not proven by it.Examining transcripts of interviews (and also of focus groups) confirmed that the quantitative evidence is neither random nor arbitrary.Not only did interviewees with high 'institutional cultural capital' ( i.e. those with post-secondary school qualifications) very often select foreign favourites, and disproportionately French and Italian, but they also introduced other cuisines into discussion.Thus Maria explained, 'I love Spanish food.I adore Spanish food because I used to like French food but I've got food allergies and I can't eat cream and things like that and butter.So French food's out of the window unless I cook it myself, but Spanish food to me, they just take very natural, very wholesome ingredients and just turn it into a taste explosion.' Evidence was also collected on dislikes.Table 2 shows evidence from the survey indicating a class distribution to dislikes, but here most opprobrium is attached to fast food and vegetarian restaurants.Those with high cultural capital amassed predictable widely-circulating arguments against fast food.The reasons for avoiding vegetarian restaurants were probably more interesting.Vegetarian food incites reference to a highly symbolic issue, the value of meat.Britain has a significant proportion of people who consider themselves vegetarian, surveys have been recording around 7 per cent during the last twenty years, although the definition of being vegetarian will often include eating fish and sometimes also everything except red meat.Eight interviewees pronounced themselves least inclined to go to a Warde Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 63 vegetarian restaurant. 16Vegetarian is an apparently acceptable thing to dislike irrespective of experience.Thus Hilda, a young working class mother, who registered vegetarian as the least attractive option, said: Agh I've never ate like quorn-mince or had tofu or soya or whatever.And I know its not, probably I wouldn't mind the taste of it in my mouth, but I don't know if it would be very nice There is considerable nuance among the discursive responses and intimations that reasons of sociability might come before those of taste.So Terry (mid-30s, white, male, with two children, and having few educational qualifications) asked about his least favourite place said, 'It would have to be a vegetarian restaurant because I'm not a vegetarian.I love my meat.I love my meat', and as his preferred place to eat he nominated 'McDonalds, especially for the boys'.Nevertheless he added, 'if it's like everyone else was there and everyone else … to keep the peace then obviously you just follow along'.Edie, 29 years old and a clerk and recently a mother, said something similar.After saying that she and husband Joe didn't much like Indian restaurants (because they provided 'mushy food, with 'strong spices' and you couldn't 'see what's in it'), she identified the vegetarian restaurant as the least appealing, explaining 'We both like meat, I think every meal we have meat in some _____________ 16 Interviews were held with partners as well as survey respondents, amounting to 43 interviews in total.The proportion of interviewees liking vegetarian restaurants least was thus roughly similar to that for the nationally random sample surveyed.

Warde
Food studies and the integration of multiple methods Política y Sociedad 2014, 51, Núm. 1 51-72 64 form.'However, she continued, 'if some friends were going, we'd go along, but we wouldn't choose to go.
This preparedness by working class interviewees to go when with friends to a place otherwise unappealing and unattractive turned out to be an expression of distinctive patterns of sociability.A further demonstration related to entertaining at home.Generosity and conviviality were often in evidence among the working class respondents.They are least likely to stand on ceremony, perhaps helped by their propensity mostly to entertain other family members rather than friends or colleagues.Most, although not all, of the formal events reported were described by persons of high cultural capital.Those with low cultural capital were more prone to informal arrangements.Households of higher social status often took an aesthetic interest in food and its presentation, especially with reference to restaurants but also to some extent their domestic entertaining.Interestingly, there was no obvious compulsion for the middle class to entertain, formally or otherwise.Caroline, aged about 30, only invited people round for drinks and Seren, in her 50s and living alone, said 'I don't do that much cooking now, you know?', and describing the last time she had visitors she prepared steak, potatoes and vegetables but noted that she 'can't be arsed peeling vegetables these days'.By comparison, some working class respondents were much more generous, sociable, concerned to entertain, and even explicitly to impress, their friends.Hilda was not exceptional among working class women in describing elaborate but informal provision for family and friends, and Margaret (a working class mother of three, aged 30) said 'I like to impress' (Figure 2).
Although it is impossible to demonstrate systematically with the data to hand, it is reassuring when analysing the survey data to be able to visit different worlds of meanings and dispositions (apparent through detailed descriptions and justifications of conduct) among people located in different social positions in the space of lifestyles (see Silva et al, 2009).Conversely, social location and context adds to the richness of biographical accounts of individuals about their tastes.Eating preferences can be located in a wider cultural context and the inter-relationship of personal and social attributes can be determined.While homologies between tastes in eating and tastes in other cultural domains were not strong, the association in the space of lifestyles of French restaurants with a taste for opera, Impressionist painting and orchestral concerts (as shown in the lower right quadrant of Figure 1) does reveal a shared propensity for established culture among the richest and most highly educated sections of the population.At the same time it is worth remarking that the capacity to make sense of both individual testimonies and survey results in relation to the practice of eating out is in significant part the result of having worked through many other sources and many other interpretations which put this particular project, this modular investigation so to speak, in context of a more comprehensive programme of research.The degree of synthesis which has been accomplished depends on the fruits of many other modules of inquiry.

Discussion
Most methodological literature focuses on the proper implementation of individual methods, with techniques for data collection and data manipulation attracting far more attention than interpretation and explanation of findings.Methods, no matter how rigorous, inventive or systematic, only take social science to a point prior to the formulation of descriptions and explanations.Social researchers typically expect too much of their methods.They also become unduly attached to their favourites.There are career, disciplinary and practical reasons for becoming highly attached to the invention, adoption, promotion and defence of familiar methodological instruments for dealing with particular types of data.The institutional practices of the scientific field and the imperatives of the scientific career are partly responsible.
Many scientists make their reputations on the basis of studies using one particular method which they continue to use more or less exclusively throughout their active careers.Attachment to methods may also be a weapon in the rivalrous promotion of disciplinary interests, with all that that means for the flourishing of professional associations and the allocation of research funding.Such aspects of the social organization of science militate in favour of the fetishisation of method and sanctification of methodological purism.Not, of course, that I am suggesting other than that appropriate, reliable, replicable and robust methodical procedures should be the platform upon which data analysis should be performed.Far from it.However, the problem seems not to be less a need for more or better methods than for better final analysis.I find the methods at my disposal mostly adequate to the types of question that I can imagine posing.What is lacking, rather, as a result of the impulse to specialise in single methods, is focused attention on their role in the explanation and interpretation of social phenomena.The real challenge for the sociology of eating is to clarify the processes appropriate to combining findings retrieved from different methods.
This should not be too daunting a prospect since for most of the time interpretation and explanation requires that we cite and integrate the results of several different methods.Methodological purism derives its plausibility from a specific model of scientific process, usually involving discrete experiments or focused tests of single hypotheses directed towards cumulative generalizable knowledge about discrete relationships between measured items or variables.Therein, the ceteris paribus clause plays a crucial role; that everything else be held equal is essential and is achieved by isolation of the minimum possible sources of variation, so as to be able to estimate the probability of a particular relationship holding between variables.Such a set of procedures is appropriate for relatively few explanatory puzzles.Sociology, or any other science of institutions, will rarely be able to do this, nor would it want to.Not many of the issues central to food studies are likely to be resolved in such a manner.
In one sense there was nothing very special about the use of a mix of methods in the CCSE study, for most studies use several methods simultaneously.Historical accounts, for instance, typically build up a composite descriptive picture from multiple sources of sequences of events and forces which condition and steer the actions to be accounted for.Many details, gleaned from multiple sources, are marshalled into as coherent and persuasive a story as possible.Whether they deserve the label 'mixed method' is debatable. 17However, the decision to interview people who had already responded to a survey did give a valuable and rare option to align the findings of two different methods by way of a single individual.The bridge between methods in the study was the person whose characteristic behaviours were _____________ captured relative to a national population and who also reported in detail on their own tastes and practices.Starting interviews by asking what some of their survey answers meant to the respondent proved a neat point of entry to explore sensemaking, reasons and justifications (as well as to evaluate the meaningfulness of answers recorded in the survey (Silva et al, 2008)).The capacity of the software to move back and forth between the space of lifestyles and the cloud of individuals permitted mutual illumination of patterned activity and personal reasoning.As Fielding (2012) describes, some new and valuable tools for data collection and processing have already built-in capacities for the combined analysis of information obtained from different methods.In the CCSE case, the degree to which an individual is typical of a social group or category (ie inhabitants of similar locations in social space) can be ascertained, as can their key differences from members of other groups.Sociologically it is possible to bring together persons, social positions and situations.
A primary feature of this inquiry was that a unit of analysis, identical in its characteristics, was 'measured' in more ways than one.Two different methods, tapping different properties, were applied to the same person.This is preferable to the more widely adopted procedure -often adopted for reasons of practicability or limitation of resources -of using separate instruments on separate samples. 18In the latter instance the unit of analysis is similar, but not identical.The benefits of inquiring about an identical unit include: first, a more secure link between different levels of social order; second, a more thorough and rounded analysis of both the individual and the group; and third, a more thorough understanding of the practice involved because both distribution and process are examined.An account with 'analytic density' involves knowing about the social position and the situations in which performances occur.
My illustration employs individuals as the bridge between methods.It demonstrates the productive use of different types of data about the same person.This might be obtained either at different points in time or with respect to different facets of experience.However, not all schools of Sociology are prepared to take the individual as either the starting point or the end point of analysis.The logic would apply equally well to supra-individual entities, like organizations or even nationstates, agencies which establish their identities in juxtaposition against others of the same type but which nevertheless have specific characteristics not reducible to that relative position.
A yet more radical alternative is suggested by developments in the study of consumption where a theoretical commitment to employing practices as the principal unit of social analysis has begun to influence empirical research designs (Halkier et   _____________  al, 2011; McMeekin and Southerton, 2012; Shove et al, 2012).In many ways such an approach, based on depicting differentiated and institutionalised practice, calls even more loudly for the integration of multiple methods.A thorough analysis of the practice of eating out (considered as a specific mode of the more general practice of eating) might be best served by explicit investment in a more systematic linking of sources of evidence about shared understandings, common procedures and standards of adequacy (see Warde, 2004).In such a venture the performance would serve as the bridge, the link between aggregate formations of custom and conduct and positioned instances of behaviour.

Conclusion
Complex research questions usually necessitate multiple sources and methods.The ensuing challenge is to fashion and justify an integrated interpretation of different types of data.As a matter of fact, most social scientific accounts employ several sorts of methods to process varied data.However, justifications of procedure and validity of explanations tend to be couched in terms of the correct implementation of a single method.The in-depth interview study brackets off as background the knowledge derived from surveys and descriptions of institutional settings and pays no respect to the methodological techniques upon which the latter rely.The predominant template of scientific procedure remains the isolated experiment that features in laboratory science. 19Outside of the laboratory a highly restricted and limited focus is almost impossible to achieve, and explanation of observed events and processes usually depends upon bringing together relevant knowledge generated using different types of method.In this respect, descriptions of method often avoid discussing the most common and most problematic aspect of social scientific analysis.
The gist of this paper has been advocacy of research designs which explicitly seek to exploit the potential of multiple methods.Arguably there is no option.Multiple methods are always tacitly or implicitly in use -they are very normal.A broad and deep understanding of eating out would require much wider range of evidence than could be gleaned from the combination of Multiple Correspondence Analysis and some interviews.A thorough understanding would draw upon many sources of evidence -textual, numerical, testimonial and observational.My central example has been one where the detail of particular instances is located in the social aggregate, such that their mutual influence can be examined.This is helpful in _____________ 19 The meaningfulness and efficacy of such experiments, where the sole focus of attention is precisely the minute details of whatever interaction is under the microscope, depend upon there having been a long series of similar investigations and a tradition of theory and interpretation, usually implicit, which gives meaning to the findings.sociological explanation because it allows for mutual interaction between the individual and the aggregate level, seeing the aggregate level as an arena of social interdependence and inter-subjectivity which transpires only in part as emergent properties of individual action.This makes it possible to use the evidence of individual experience to give a meaningful foundation to prevalent, collectively orchestrated and collectively experienced, concurrent and sequential situations.Just as panel data is ideally to be preferred for explanatory purposes to repeated crosssectional surveys, interviewing people who have been surveyed is more instructive than asking questions of subjects sampled independently.In both designs, then, a single person becomes the bridge that connects one data source to another.The particularity of the understanding and rationale for behaviour of an individual can be contextualised by the cultural characteristics of the social groups to which they belong;20 and the shared experience that group membership and joint participation generate can be identified from the framing of a personal account.Designs which deal with the same individual, or other identical unit of analysis, across different time points and different types of experience are especially powerful.
Multiple methods are not always, and perhaps even not usually, necessary.My selective review of the burgeoning discussion of mixed methods would suggest that there is no advantage to according it paradigmatic status in the methodological pantheon.The virtues of mixing methods lie in their capacity to address explanatory puzzles more thoroughly, with greater breath and depth, to permit what Fielding (2012) called 'analytic density'.Indeed, progress in research technique appears to come more from adaptation and improvisation by researchers in response to problems of explanation than from grandiose formalizations of methodology-centred paradigms.Often this might be achieved without actually integrating methods -the perspectives afforded by findings from different methods may often be sufficient.This is largely because, in the generation of explanations, interpretation of evidence is more important than methodical procedures.What is mostly lacking in the literature on research process is extensive and helpful guidance about the transformation of reliable evidence into credible explanation.
It is no simple matter to extrapolate from the use of integrated mixed methods in Sociology to cross-disciplinary collaboration.All problems are magnified if viewed in an inter-disciplinary perspective.However, the points made might potentially be scaled up to address the bigger question of complementarity at inter-disciplinary level.The barriers between disciplines are higher, because of theoretical commitments and the specific intrinsic properties of the phenomena which constitute the primary substantive focus of each.Almost by definition inter-disciplinary projects do not share theories.To work from within the parameters of a theory may sometimes preclude some methods, but usually many alternative methods can be consist-_____________ ently and productively applied to a given research question.Nonetheless, while some methods are more prevalent and more highly valued in particular disciplines (ethnography in Anthropology, or randomised control trials in Medicine), methods flow easily across disciplinary boundaries.The importation into a discipline of a new method often proves to be a primary source of innovation -social network analysis, sequence analysis and discourse analysis have circulated promiscuously in the recent past.Methods, it would seem, are not a principal barrier to interdisciplinary understanding.The fact that disciplines are wedded less firmly to particular techniques than to their theories offers a promising point of leverage for progress in food studies.Greater tolerance for multiple methods might be a precursor or herald of a thawing of the more icy boundaries between disciplines.In principle, the logic of mixed methods should be transferable, resulting in better integrated explanations.In conclusion, I contend that the social sciences are generally very well equipped with methods for investigation of social phenomena, but are often deficient when it comes to delivering insightful explanation.Currently, innovation in technique and procedure is required less than explicit protocols for interpretation which come into effect after the empirical investigation of distributions, associations, discursive themes, clusters or structures has been completed.

Figure1:
Figure1: Multiple Correspondence Analysis, cultural tastes and participation of random sample of UK population, 2003 (N=1564): Axis 1 and 2, with variables contributing to axis 1 indicated.

WardeFood
Figure 2: MCA: Social position of survey respondents in cultural space and identification of qualitative interviewees

Table 1 :
Type of restaurant liked most, by social class (percentages)

Table 2 :
Type of restaurant least liked, by social class (percentages, by column)