Data, methods and interpretation in analyses of cultural consumption: A reply to Peterson and Wuggenig

Poetics 35 (2007) 317 329 www.elsevier.com/locate/poetic Discussion Data, methods and interpretation in analyses of cultural consumption: A reply to Peterson and Wuggenig Tak Wing Chan a, *, John H. Goldthorpe b a Department of Sociology, University of Oxford, United Kingdom b Nuffield College, University of Oxford, United Kingdom Available online 6 August 2007 Peterson s and Wuggenig s critical comments refer specifically to our paper Social stratification and cultural consumption: the Visual Arts in England (Chan and Goldthorpe, 2007c). However, their criticisms could, at least in some of their more general aspects, be taken to apply to all of the contributions to the special issue of Poetics in which our paper appeared, and of which we were Guest Editors. We would therefore wish this reply to be in turn understood as having a wider reference than to our paper alone, and we shall in fact have occasion to refer, if only briefly, to the work of several of our colleagues. We consider Peterson s and Wuggenig s commentaries together under the following three headings: data, analytical methods and interpretation, although as regards methods we are chiefly concerned with Wuggenig s position. 1. Data Both Peterson and Wuggenig seek to show that our paper is flawed through its reliance on the secondary analysis of data of a kind that is inadequate to our substantive concerns. As a preliminary point here, we may note that Peterson (p. 301) begins by contrasting ourapproachwiththatofbourdieu (1984) which, in his view, represents a classic case of a purpose-built questionnaire. The contrast is unfortunate, for while Bourdieu s questionnaire is indeed purpose-built, his survey is technically flawed both as regards sampling 1 and as we * Corresponding author. E-mail address: tw.chan@sociology.ox.ac.uk (T.W. Chan). 1 Bourdieu s sample appears to consist of 692 men and women, interviewed in 1963 in Paris, Lille and a small provincial town, supplemented by a further 595 interviewed (it is not clear where) in 1967 1968. But this is not in fact a sample of any specified population; nor do we know how its members were chosen. What we are told (Bourdieu, 1984,p. 505) is that it contained approximately equal numbers of Parisians and provincials or, in other words, in relation to the French population as a whole, Parisians were heavily over-represented. In addition, the upper and middle classes were also over-represented so as to give an adequate sample of each of their fractions. Correspondingly, then, the working class and its fractions are under-represented, while peasants and farm workers were entirely excluded from the analysis. It is not clear to us that Bourdieu did any weighting in his analyses. 0304-422X/$ see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.poetic.2007.06.004

318 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 discuss further below as regards the kinds of data that were collected. It may also be relevant to add that the data of the survey were never placed in the public domain and, so far as we can ascertain, have now disappeared, so that, unlike the data-sets that we and our colleagues use, never were, or will be, available for secondary analysis and the critical scrutiny that this entails. In respect to Peterson and Wuggenig s more specific claims of weaknesses in our data, these can for the most part be shown to be (i) unduly strained, if not simply mistaken, or (ii) to refer to weaknesses that we did in fact recognize and allow for in developing our substantive arguments, or (iii) to be irrelevant to our concerns. For example, Peterson makes a series of criticisms of the wording of the survey questions on cultural consumption on which our analyses are based. Most seriously, he maintains that three of the five items concerned confound visual arts with other cultural forms. Thus, in the case of the question on visiting a craft exhibition, he suggests that an exhibition so described could as well be devoted to music, dancing or theatricals as to items of visual interest. It may be that American English here differs from English English, but we have difficulty in taking this criticism seriously. In everyday English usage, craft refers to the skilled production of some kind of physical object that is open to visual inspection a piece of furniture, clothing, pottery, etc.; and we see no reason to doubt that the word was understood in this sense both by those formulating the question in the Arts in England survey that we use and by the vast majority of those responding to it. Peterson objects on similar grounds to the questions on visiting a cultural festival or a museum or art gallery. However, we explicitly acknowledge the problems that here arise (pp. 173 174). Furthermore, in the former case, we point out (n. 6) that excluding this item from our analyses does not have any substantial effect on our results, while, in the latter case, we take care not to advance any interpretation of our findings that could be seriously compromised by the linking of museums and galleries (again, a matter to which we return below). Peterson also raises doubts about the further question on attendance at any event including video or electronic art, on the grounds that while apparently designed to tap an elite activity, it turns out that this is the most frequent form of consumption of those whom we label as inactives perhaps because of confusion with simply watching videos at home. But, if any such confusion did occur, it could only have been to a very limited extent. It has to be noted that, under our latent class model, the probability of inactives responding positively to this item is still only 0.092 (see our Table 3); and the relevant contrasts then to be made are with the corresponding probabilities of 0.252 and 0.632, respectively for those consumers whom we label as paucivores and omnivores. 2 Finally, in this connection, Peterson suggests that we should have tested the validity of the five survey items on which we rely by considering the correlation of responses to them, since in this way their credibility might be greatly increased (p. 302). But what Peterson here fails toseeisthatwedoinfactfollowsucha multiple indicator approach and on more appropriate lines than those he would propose through our latent class modelling (cf. also the use of this technique by our colleagues, Alderson et al. (2007)). And the results of this 2 Peterson (p. 302, n. 2) instead compares the probability of inactives having attended an event including video or electronic art with the probabilities of their having been active in other ways, and claims that the former is high relative to all the latter. But in fact the difference with their probability of having visited a museum or gallery (0.071) is of doubtful significance and, in any event, the only sensible conclusion to be drawn from these probabilities overall is that essentially the inactives are... inactive.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 319 modelling (our Table 2) indicate that all of the association between the responses to the five questions we use can be adequately captured by just three latent classes those that then give rise to our typology of consumers. In other words, and as we remark, this is a result that points to a very strong patterning in the responses to our indicator items, and one that should then by his own criterion lead Peterson to significantly change his views as to their validity. Apart from his concerns about question wording, Peterson is also critical of our focus, via the Arts in England Survey, on data on individuals cultural consumption, to the neglect of data on their cultural tastes. And in this regard he is apparently, though somewhat confusedly, joined by Wuggenig, who, under a heading of Measures of Art Consumption, proceeds to reprove us for not having more measures of taste in the manner of Bourdieu. In response, the first point we would make is the obvious one that cultural taste and cultural consumption are not the same thing. And, we would add, it is in turn a dangerous procedure to seek to infer one from the other and, especially, consumption from expressions of taste. 3 As we observe in our paper (n. 3), Bourdieu s survey does in general contain far more items concerning tastes than actual consumption; and indeed in the case of the visual arts, it includes only a single question as compared to our five that relates to the latter: i.e. question 24 on possible visits to four named Paris museums or galleries or to any museum (sic) in the respondent s own city or region (Bourdieu, 1984, pp. 516 517; or see the list of questions summarized by Wuggenig, p. 311). In our view, it is a major flaw in the design of Bourdieu s survey that such an imbalance between items concerning taste and consumption should arise. And, far more serious, is then Bourdieu s tendency to elide what should be a crucial distinction, so that in his discussion of his empirical findings there are recurrent though largely covert inferences from taste to consumption, for which little warrant can in fact exist. Second, we would repeat what was made abundantly clear in our paper: that the focus of our own work is on cultural consumption as a form of social action, and not on cultural tastes or on what Peterson refers to as the cultural self-constructions of respondents. 4 Thus, for us, it cannot be an objection, as Peterson would seem to suppose (p. 304), that measures of attendance and participation [as distinct from those of taste] are filtered through the constraints of time, money, availability, opportunity and the like. If one is concerned with the way in which cultural consumption is socially stratified, then the effects of such constraints are of obvious interest as also, on the other hand, are aspects of cultural consumption that may be engaged in less as an expression of individual taste than out of status motivations (bored executives sitting in hospitality boxes at the opera). Third, while we would agree with Peterson that respondents reports of their cultural consumption could be prone to some bias in the direction of what they take to be socially desirable answers, we fail to see why exactly the same problem does not arise in the case of questions of taste. But there is then the difference that while little check can be made on the validity of expressions of 3 Oddly, Peterson cites the very finding from Silva (2006) that we ourselves note (p. 171) in order to underline this point, but appears then not to grasp its significance: i.e. that persons expressing preferences for landscapes, still lifes or portraits are far less likely to go to museums and galleries to see the kind of art that they say they like than are those expressing preferences for Renaissance art or Impressionism. 4 As we recognise in our paper, a quite legitimate interest may indeed exist in studying cultural tastes per se; but then the implications for actual consumption must be seen as a separate question. Or, again, tastes may be reasonably inferred from consumption following a revealed preferences approach but with due allowance then being made for the possibility of status-rather than taste-motivated consumption that we go on to note in the text (cf., the contributions to the special issue of Katz-Gerro et al. (2007) and Coulangeon and Lemel (2007), respectively).

320 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 taste, reports obtained from members of a representative national sample on their cultural participation can often be set against attendance figures for different kinds of cultural venue. And, as we remark (n. 3), data on cultural consumption derived from the Arts in England Survey would appear to be essentially consistent with such figures of this kind as are available. We would then see it as quite unnecessarily extreme to treat the data we use as being informative only about respondents cultural self-constructions and not about the facts of their cultural consumption. 5 2. Analytical methods In this section, as previously indicated, we deal primarily with comments made by Wuggenig. In our paper (n. 1) we did in fact anticipate a Bourdeusien response which would have the usual aim of seeking to protect the prophet against any impious empirical critique. Wuggenig fully lives up to our expectations. The main objection that he raises against our work (p. 307) is that Bourdieu s homology assumptions cannot be tested using mainstream or Anglo-Saxon social science methods in particular, regression techniques which, we are told, Bourdieu had already refuted in the 1970s. Rather, the only appropriate quantitative technique to use is some form of correspondence analysis (CA) or of multiple correspondence analysis (MCA). In addition, we have of course not reconstructed Bourdieu s position correctly. To begin with, we would point to the very dubious nature of any aprioriclaim that a particular theoretical position can only be empirically evaluated through the use of one particular technique of data analysis. It must surely be open to researchers to use any technique that can produce results that bear on testable hypotheses deriving from a theory. If the results appear negative, defenders of the theory can of course always seek to show that the results themselves are questionable, that their implications for the theory have been misunderstood, and so on. But all such arguments have then to be conducted at the level of specific instances rather than at the level of vague generalities of the kind in which Wuggenig for the most part engages. In the case that here concerns us, we would note, to begin with, that while Bourdieu may well have rejected mainstream approaches of Anglo-Saxon quantitative sociology, this is scarcely to say that he has refuted them. We do not know where Wuggenig believes this refutation is to be found but, from our own reading of Bourdieu, his remarks on this topic are, to say the least, far from convincing. Furthermore, we would question the attempts of Wuggenig and others to represent CA and MCA as techniques that are in some way set apart from the Anglo-Saxon mainstream of multivariate data analysis in the social sciences, and that in turn have some special affinity with a distinctive relational or field theoretic or neo-structuralist sociology, as pioneered and exemplified by Bourdieu. All such attempts seem to us little more than exercises in mystification. As Wuggenig at one point acknowledges (p. 307), CA and MCA are anchored in the work of Anglo-Saxon statisticians of the 1930s and 1940s; and, although later notably developed and widely applied in France, it needs to be stressed that they remain a part valuable but in no sense 5 Wuggenig (p. 306) seeks to question our findings on cultural consumption in the visual arts as being inconsistent with England s recent upward mobility in this field. However, his remarks are to a large extent either irrelevant to issues of consumption among the population at large (e.g. those concerning the support structure of theoreticians, critics etc. or London s position in the international art market) or are ill-informed. For example, although, as Wuggenig says, Tate Modern is a success story so far as attendances go, this is far less clearly the case with various other museums and galleries, including several opened in the golden years of millenium or national lottery capital funding.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 321 exceptional of what would by now best be regarded as an entirely international, or global, corpus of techniques of data analysis. 6 Moreover, like all such techniques, they have their advantages and disadvantages that need to be assessed in relation to the purposes of their application in any particular instance. Thus, as regards the social stratification of cultural consumption, the most obvious use of CA or MCA is for purposes of general description. For example, our colleagues Coulangeon and Lemel (2007) apply multiple correspondence analysis, quite explicitly, in this way in their study of musical consumption in contemporary France. It enables them to display graphically, on the one hand, how the consumption of different musical genres are associated (or disassociated) and, on the other, how different complexes of consumption are associated with a range of stratification variables. And on this basis they can then draw interesting conclusions about the degree of omnivorisation of musical consumption in France and about the degree and pattern of its stratification. However, there is no reason to see their analyses and Coulangeon and Lemel make no claims to this effect as being in any way set apart from those of the variable sociology that Wuggenig is so keen to disparage; nor yet to suppose that the relations among variables that they are able to reveal through MCA are any more relational than those that would be shown up by, say, latent class analysis or regression. 7 Moreover, Lemel and Coulangeon clearly recognise the limitations of their analyses. In particular, they note (2007:106) that while MCA can give a good picture of the gross effects of stratification variables on musical consumption, it is not adequate to analysing their net effects i.e. the effects of these variables considered independently of each other. Thus, while MCA can show that a particular pattern of musical consumption is associated with, say, an advantaged class position, high social status, high education and high income, it can give little indication of whether these associations are all of comparable importance or whether, perhaps, one is dominant, another spurious, and so on. To illustrate this point, we have carried out some multiple correspondence analysis on our own visual arts data. Fig. 1 displays the results graphically: my, vy, ey, cry and cuy refer to positive responses to the five visual arts items of visiting (1) a museum or art gallery, (2) an event involving video or electronic art, (3) an exhibition or collection of art, photography or sculpture, (4) a craft exhibition and (5) a cultural festival, respectively. Similarly, mn, vn, en, crn and cun refer to negative responses to the same five items. It can be seen that the Yes replies are clustered together and, to an even greater extent, so are the five No replies. Furthermore, the Yes cluster is located to the right of the No cluster. At the same time, more advantaged social groups are also found to the right of less advantaged groups. Thus, for 6 There is a slight difference among statisticians on the relative weights they give to description vis-à-vis inference, or between exploratory and confirmatory analysis. But such difference should not be overdrawn. See the discussion in van der Heijden et al. (1989) for ways in which CA and log-linear analysis complement each other in the analysis of contingency tables. Similarly, Gower and Hand (1996, chap. 4) point out that MCA can be regarded as the categorical counterpart of principal component analysis. Goodman (1996) brings out the affinities among a whole range of methods for treating the non-independence of cross-classified data, including various forms of correspondence analysis, by showing how they can in fact all be seen as special cases of a single, more general method of analysis. He also uses his approach to reveal some errors quite frequently made in the interpretation of the graphical displays produced by CA. 7 It is true that regression analysis involves a distinction between a dependent variable (e.g. cultural consumption) and independent variables (e.g. our social stratification variables), while all variables are treated, at least formally, on the same footing in CA and MCA. However, this does not imply any deep philosophical difference between the methods, because, pace Bourdieu, all that CA, MCA or, for that matter, regression analysis does is to reveal any associations that might exist between variables.

322 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 Fig. 1. Multiple correspondence analysis of visual arts consumption data with education, class, income and status. Note: Labels for the categories of the four stratification variables of education, class, income and education are explained in footnotes 8, 9, 10 and 11, respectively. example, we distinguish six levels of educational attainment (in descending order): degree, sub-degree tertiary qualifications, A-levels, O-levels, CSE and none. 8 In Fig. 1 degree is found to the right of sub-degree tertiary qualifications, which in turn is found to the right of A-levels, and so on. The same is true for social class, 9 income, 10 and social status. 11 Thus, it would seem reasonable to infer from Fig. 1 that there is a social gradient in cultural consumption. Respondents in more advantaged positions, defined in terms of either education, income, class or status, are more likely to participate in the five types of visual arts activities than those in less advantaged position. 12 This view is consistent with the bivariate associations that we report in our paper (Chan and Goldthorpe, 2007c, Table 4, Fig. 1). 13 However, consistently with the 8 In Fig. 1, e1 is degree, e2 sub-degree tertiary qualifications, e3 A-levels, e4 O-levels, e5 CSE and e6 is none. 9 We use the seven-fold NS SEC classification: c1 is higher managers and professionals, c2 lower managers and professionals, c3 intermediate employees, c4 small employers and own-account workers, c5 lower supervisors and technicians, c6 semi-routine workers, and c7 routine workers. 10 We distinguish five income quintiles: i1 refers to the top quintile and i5 to the bottom quintile. 11 In Chan and Goldthorpe (2004), we note that our status scale can be categorised into four broad status bands: s1 refers to the top status band and s4 to the bottom status band. 12 While the interpretation suggested for the horizontal dimension of Fig. 1 seems plausible if rather imprecise how the vertical dimension should be interpreted is less clear. However, this dimension is certainly structured more by the stratification variables than by the visual arts items, as the responses to the latter, both positive and negative, are found in the centre of the dimension. In other words, it is the contrast between certain categories of the stratification variables that chiefly defines the vertical dimension. We have repeated our MCA with additional sociodemographic variables such as gender, age and region of residence. This does not change the results reported above. But we do not include the graphical output of that exercise here because it is even more cluttered and less readable than Fig. 1. Also, we have not included the row variables of our data in Fig. 1, because to do so would mean representing each respondent as a point in the plot. Since there are over three thousand respondents in the sample, the resulting plot would be literally unreadable. 13 We report bivariate associations between latent class membership on the one hand, and class and status on the other. Similar bivariate associations with education and income can be produced.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 323 observation of Coulangeon and Lemel, what is hard to determine from Fig. 1 is the net associations between the various stratification variables, on the one hand, and cultural consumption, on the other. This turns out to be a very germane point. From our latent class and regression analyses, we are able to show that it is in fact education and status that primarily predict visual arts consumption. 14 Controlling for education, status and other socio-demographic variables, class effects disappear, 15 and income is only significant for the contrast between paucivores and inactives (Chan and Goldthorpe, 2007c, p. 182, Table 7 esp.). Furthermore, based on our regression analyses, we are able to estimate the substantive magnitude of the effects of education and status, and thus gain a sense of their relative importance in shaping visual arts consumption (Chan and Goldthorpe, 2007c, p. 184 and Fig. 2). Is it possible to make assessments of this kind with CA or MCA? We very much doubt it. There is a further limitation of CA and MCAwhich bears directly on the substantive issues that concerns us. As noted above, an inspection of Fig. 1 would suggest that there is a social gradient of cultural consumption. This might give the impression that Bourdieu s homology argument is supported by our data. However, as our latent class analysis shows (Chan and Goldthorpe, 2007c, Table 3), as compared with those individuals who do not consume what might be construed as high culture in this domain (i.e. visit a museum or art gallery, an exhibition or collection of art, photography or sculpture or an event involving video or electronic art ), those who do consume in these ways are in fact more likely to consume more popular forms as well (i.e. attend a cultural festival or craft exhibition ). In other words, there is no evidence of an elite who consume high culture only, while shunning popular culture. Further, having identified the three latent classes of omnivores, paucivores and inactives, we are able to show that at least large minorities of the most advantaged social classes or the highest status groups are visual arts inactives (Chan and Goldthorpe, 2007c, Tables 4 5, Fig. 1). So, yes, there is a social gradient in visual arts consumption. But this is not to say that people in advantaged social positions are all omnivorous visual arts consumers. These observations are inconsistent with the homology argument, as we argue further in Section 3 below. At this point, we simply note that it is not clear to us how one could read off such understandings from the graphical output of MCA as shown in Fig. 1. To repeat, CA and MCA have their place in the international corpus of techniques of data analysis, a place that is not bounded by supposed national traditions. We do not wish to underestimate the analytical value of CA and MCA. But neither do we overestimate it. What is key is whether, in any particular instance, they yield more insights than other techniques. In the 14 When, as is typically the case, quite specific effects of education on cultural consumption are revealed, it is by no means strange, as Peterson implies (p. 304), to pose the question of how far, in this context, education should then itself be considered as a stratification variable. Nor is it correct for Peterson to say that in our paper this issue is raised without elaboration or explanation. See pp. 182 and n. 15, and also the fuller discussion in our previously published work (Chan and Goldthorpe, 2007a, 2007b). 15 In the comparative research project from which all the papers in the special issue of Poetics derive, a special focus of interest as we remark in our Introduction does in fact lie in distinguishing the effects of class and status on cultural consumption, in contrast with the use of synthetic notions of socio-economic status, as in much American literature, and with approaches inspired by Bourdieu s attempt to overcome the Weberian opposition of class and status. In the outcome, the preponderant effect of individual s status over class is rather consistently demonstrated; and, where appropriate data are available, the further importance emerges of parental status (Bukodi, 2007; Katz-Gerro et al., 2007) and of status of spouse (Kraaykamp et al., 2007). A valuable by-product of the project has been the creation of social status scales for each of the seven national societies involved, following the approach pioneered by Laumann (1966); Laumann (1973).

324 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 present case, we would argue that MCA is clearly less appropriate than the latent class analysis and multinomial logistic regression that we use in our paper. We would also emphasize in this connection that our approach does not imply, as Wuggenig would suppose (p. 309), treating reified variables like agents, or accepting a Newtonian logic of regression, or believing that causal explanations can be directly cranked out of statistical analyses. The idea is, rather, to obtain as refined an understanding as possible of the statistically demonstrable relationships that arise in the patterning of cultural consumption and in its social stratification, so that, in the next stage of the research process, explicit (and testable) accounts can be advanced of the mechanisms that give rise to these relationships in terms, ultimately, of the action and interaction of the individuals involved. Wuggenig seems sadly out of touch so far as developments in mechanism-based theory and causal explanation in sociology are concerned (see e.g. Hedström and Swedberg, 1998, vol. I, chs. 6 and 9 esp.; Goldthorpe, 2007, vol. I, chs. 6 and 9 esp.). 16 The foregoing considerations then bear directly on the objectives that we pursue in our paper. Insofar as we are concerned with Bourdieu, we take him seriously as not simply making what Wuggenig calls homology assumptions as regards cultural stratification but as in fact advancing a homology argument that involves substantive propositions. And we takehim also to be suggesting a mechanism through which the homology is created and sustained. That is, the process that Wuggenig himself (p. 310) represents as crucial to Bourdieu s theoretical position whereby the habitus mediates between the distinctive class conditions in which it is formed and the distinctive lifestyles cultural taste and consumption included through which it is expressed. In other words, far from incorrectly reconstructing Bourdieu, as Wuggenig maintains, we focus precisely on what we, along with Wuggenig, would see as his central claims regarding the form and the dynamics of the relationship between social and cultural hierarchies. And we then ask how well these claims fare in the light of the results that our empirical analyses produce. We begin with a latent class analysis that shows how our five indicators of cultural consumption in thevisual arts are interrelated. To repeat, it turns out that what they indicate can in fact be captured by just three, well-differentiated patterns of consumption. Using the method of modal assignment, we then move from the three latent classes to three types of consumer, and finally we use multinomial logistic regression to examine how the probabilities of individuals conforming to one or other of these types depend on their class, status, income and education, independently considered, as well as on other socio-demographic variables. In the outcome, the results thus obtained do not appear supportive of Bourdieu s position nor in fact of what we label as rival individualisation arguments, nor, in all respects, of the omnivore univore argument as pioneered by Peterson. In conclusion, we return to these interpretations of our results and show that they are robust to the objections that Wuggenig and Peterson raise against them. 3. Interpretations As a preliminary here, we would yet again stress that our prime and quite explicit concern is with the testing of the arguments referred to above and not, as Wuggenig, especially, seems to 16 Wuggenig makes reference to the paper by Martin (2003) on field theory. While we do not share Wuggenig s enthusiasm for this paper, we would at all events agree with Martin (2003, p. 12) that field theories may be seen as provisional theories that we are happy to replace when adequate knowledge of mechanisms is gained.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 325 assume (p. 306), with providing some general descriptive account or (to use his word) portrait of the culture of the visual arts in contemporary England. 17 It is therefore beside the point for our critics to maintain that our data are insufficient for us to give such an account. 18 What they need to show is that our indicators of consumption in the visual arts are deficient in such specific ways as to undermine the conclusions that we do seek to draw from our empirical analyses; and this, we believe, they fail to do. Three sets of conclusions are chiefly in question, which we consider in turn: i.e. those relating to whether or not, in the case of the visual arts, there exists (i) a culturally exclusive group of consumers who can be taken as representative of a dominant class, (ii) a group of omnivorous consumers, and (iii) a substantial proportion of the population who are inactives or essentially non-consumers. Our analyses, we argue, provide little evidence of a homology between cultural and social stratification: i.e. there is little indication that consumers in more advantaged class, or high status, positions typically consume high art forms while, moreover, displaying aesthetic distance from other forms. Wuggenig seeks to challenge this conclusion on two grounds (pp. 314 315): first, our data, in both their range and quality, are inadequate to sustain it; and second, it is not in any event necessary to the neo-structuralist homology argument that fine art consumption should be typical of a dominant class. Such consumption, it seems, could be quite minoritarian provided only that it is more frequent in this class than in others. On the first point, Wuggenig fails to see that the relatively crude nature of our data here tends in fact to strengthen our argument. Take, for example, our main finding that among groups that would clearly form part of Bourdieu s dominant class such as higher-level professionals and managers substantial minorities are, in our typology, inactives so far as the visual arts are concerned while most of the remainder are only paucivores. This finding could only have emerged more clearly if we had been able, say, to separate art gallery and art museum visits from visits to other kinds of museum and had then discounted the latter; or again if we had data on the frequency of visits to galleries, exhibitions and collections etc. and then discounted infrequent visitors. 19 On Wuggenig s second point, we would note, to begin with, that he provides no textual reference to support this, quite surprising, reinterpretation of Bourdieu. Furthermore, it is in any event one clearly inconsistent with the crucial role that is to be given to the habitus in mediating between class conditions and lifestyles. Since members of the dominant class presumably share 17 Wuggenig further suggests that it would have been better if we had combined our results for all three of the cultural domains that we study in a single analysis. This we have in fact done in a paper, available from the authors on request, that is to be published elsewhere. But its findings, we suspect, will be no more congenial to Wuggenig than those of the paper presently under discussion which concentrates on the visual arts so as to complement two earlier papers (Chan and Goldthorpe, 2005, 2007b) that concentrate on theatre, dance and cinema and on music, respectively. 18 Other contributors to the special issue are also concerned to make their position clear in a similar way. That is, they recognise that the data available to them are of clearly limited value for descriptive purposes, and especially at the individual level; but they still seek and are in fact able to use these data to establish, at an aggregate level, well-defined patterns of cultural consumption and their social bases. For example, in the particular case of book readership, see Bukodi (2007), Kraaykamp et al. (2007) and Torche (2007). 19 To be sure, we might perhaps find that those individuals who were more frequent visitors were also more likely to concentrate on galleries and art museums; but these individuals, even if taken as forming a new type of consumer, would of necessity still be less numerous than those we count as museum and gallery attenders on the basis of the data that we have.

326 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 the same class conditions that are the matrix of their habitus, why should it be that only a small minority then have that system of dispositions that results in fine arts consumption playing a significant part in their lifestyles? 20 Finally, in the present connection, we may note Peterson s claim that we do in fact have evidence though we fail to recognise it of an exclusive elite of highbrow snobs (or highbrow univores ) in the case of those individuals we label as paucivores. This is so, Peterson argues (p. 305) because the activity in which they are most likely to participate is visiting an exhibition or collection of art, photography or sculpture, while they are very unlikely to attend more middlebrow craft exhibitions or cultural festivals. However, it seems that Peterson has simply misread the relevant table (Table 3). The paucivores most likely activity is in fact visiting a museum or art gallery, with a probability, under our latent class model of 0.809, as against a probability of only 0.416 of visiting exhibitions or collections. And while paucivores are indeed infrequent attenders at craft exhibitions or cultural festivals, they are also far less likely than our omnivores to visit exhibitions or collections or again events involving more avant garde video or electronic art. In other words, paucivore consumption in the visual arts, far from being elitist, is best understood as following a modest and, as we put it, seemingly rather unadventurous pattern. Turning next to the question of omnivores, we conclude that a type of omnivorous consumer can indeed be identified in the visual arts, although those conforming to this type are not very numerous. Moreover, while omnivores might perhaps be regarded as in some sense a cultural elite, if not of course an aesthetically exclusive one, they can in no way be seen as a social elite contrary to homology arguments. While omnivores are more likely to be graduates than are paucivores, they tend at the same time to be less well placed economically, being more likely than paucivores to be found in the lower than in the higher levels of the salariat. Peterson, it seems, has no objections in this regard since our omnivores do have high probabilities of engaging both in activities that he accepts as clearly highbrow (attending exhibitions and collections) and in those he sees as middlebrow (attending craft fairs and cultural festivals). But Wuggenig contends (p. 313) that, because of the nature of our data, we are unable to provide a specific description of the omnivorous lifestyle with respect to art. Here again, however, Wuggenig seems off the point. It was not our concern to provide a description of the omnivorous or of any other lifestyle, but rather to test, in the domain of the visual arts, at least one version of the omnivore univore argument, and in particular as against homology arguments. Our data may well not be adequate to give a full characterisation of a cultural lifestyle but they still cover a fairly wide range of visual arts consumption (and, to repeat, clearly wider than that covered by Bourdieu), and one that Wuggenig can in fact only impugn (p. 313) by seeking to impose his own, eminently disputable, aesthetic judgements for example, that photography is still not a fully legitimate art form or that electronic art has failed as such. 20 And again Wuggenig s attempts to buttress his arguments with references to the current British visual arts scene are not impressive. For example, he claims that recent soaring prices for fine art indicate the existence of a cultural elite endowed with the economic capital to make its purchases possible. However, this neglects (i) that some buyers are institutions, not individuals; and (ii) that while the individuals involved may form some kind of rich and superrich elite, they are still far too small in number to be equated with Bourdieu s dominant class (whichever of the various representations of this one chooses to take) and, further, that rather than forming an elite within the British population, as Wuggenig supposes, they are in fact quite international.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 327 What then emerges is evidence clearly in favor of the omnivore univore argument, at least in Peterson s original formulation of it. 21 Omnivorousness is indicated rather than any tendency towards aesthetic distancing in that it is those respondents with the highest propensity to consume high art forms, that is, to visit art exhibitions and collections and to attend more avantgarde art events who also show the highest propensity to engage in other forms of cultural consumption in this domain and including visiting craft fairs which Wuggenig himself (p. 313) deems to be an art form that definitely does not belong to high art. Finally, then, we come to the matter of inactives or virtual non-consumers in the domain of the visual arts. While our findings lend support to the omnivore univore argument in revealing a group of omnivores, they also call this argument into question in failing to show up a body of univores analogous, say, to the pop-and-rock univores that we (Chan and Goldthorpe, 2007b), and others, have been able to contrast with omnivores in the domain of music. In the visual arts omnivores seem, rather, to be complemented by a sizable number of paucivores and a yet larger number of inactives. What, we suggest, underlies this finding is that in the visual arts there are not popular genres that have the same extensive institutional basis or wide dissemination through the media as does popular music. Both of our critics seek to challenge the position we here take up. One objection raised by Wuggenig is that we would have achieved a different result if we had included cinema in the visual arts. Indeed, we would. But we have elsewhere (Chan and Goldthorpe, 2005) treated cinema-going as a form of univorous consumption within a different context: that is, as one engaged in by a substantial number of people who have only a very low probability of attending a range of live theatre events unlike omnivores in this domain who are both cinema- and theatregoers. This is, in our view, a far more appropriate way in which to view cinema. The move that Wuggenig proposes seems to be aimed more at avoiding the problem of popular genres in the visual arts, by means of a rather strained redefinition of the field, than at promoting an understanding of it. A more serious point, made by both Peterson and Wuggenig, is that our data refer only to consumption in the visual arts within institutional settings and that we have no information about such consumption as may occur either in the home or in the street or other public places. Thus, they suggest, many of those whom we label as inactives may in fact, as Peterson (p. 305) puts it, have rich visual aesthetic lives that they pursue outside of museums and galleries. However, our critics here fail to acknowledge that this is an issue that we quite explicitly discuss. Indeed, we state (p. 187) that our conclusions regarding inactives might need to be modified if analyses of cultural consumption in the visual arts of the kind we have here presented could be complemented by studies made in domestic and public settings. At the same time, though, we give grounds for believing that an interest in art in the home or in public places is likely to be correlated with, ratherthandisconnectedfrom,visitingmuseums and galleries. So while, then, we do recognize the possibility of our supposed inactives leading rich, visual aesthetic lives, we have strong, and reasoned, doubts about whether this is in fact the case. Peterson and Wuggenig choose, as of course they may, to take a different view. But, rather than implying that we simply neglect the issue and that our questioning of the omnivore univore argument in the domain of the visual arts is thus undermined, they need 21 As we note in our paper (p. 171), we do not find helpful Peterson s more recent 2005 attempt to elaborate his position by in effect crossing the omnivore univore and highbrow lowbrow distinctions. This threatens to lose the original insight that omnivorousness implies an openness to appreciating everything in contrast to cultural consumption which is based on rigid rules of exclusion (Peterson and Kern, 1996, p. 904).

328 T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 to accept that, in the present state of research, their own position lacks any supporting evidence. In sum, we believe that the conclusions that we draw on the basis of our data analyses hold good, with due regard being taken of the qualifications and reservations that we ourselves express. And in turn, we see no reason to modify the empirically grounded criticisms that we then develop, specifically in relation to the visual arts, of more general arguments on the social stratification of cultural consumption, and, in particular, of the homology and omnivore univore arguments. Peterson and Wuggenig understandably seek to defend ideas to which they have evident commitments. But despite their efforts and, we think, they do often appear to be trying rather too hard the criticisms stand. Acknowledgements While we are grateful to Art Alderson, Yannick Lemel, Florencia Torche and Wout Ultee for the helpful advice and comments that they have given us in preparing this reply, we should make it clear that we alone carry responsibility for its content. The multiple correspondence analysis reported in this paper was carried out with the free statistical software R (R Development Core Team, 2005). We are grateful to the many contributors of R for making this wonderful program available to us, and others, for free. Our research was supported by a ESRIAHRC research grant under their cultures of consumption research programme, award number RES-154-25-0006. References Alderson, A.S., Junisbai, A., Heacock, I., 2007. Social status and cultural consumption in the United States. Poetics. Bourdieu, P., 1984. Distinction: A Social Critique of the Judgement of Taste. Routledge & Kegan Paul, London. Bukodi, E., 2007. Social stratification and cultural consumption in Hungary: book readership. Poetics. Chan, T.W., Goldthorpe, J.H., 2004. Is there a status order in contemporary British society? Evidence from the occupational structure of friendship. European Sociological Review 20 (5), 383 401. Chan, T.W., Goldthorpe, J.H., 2005. The social stratification of theatre, dance and cinema attendance. Cultural Trends 14 (3), 193 212. Chan, T.W., Goldthorpe, J.H., 2007a. Social status and newspaper readership. American Journal of Sociology 112 (4), 1095 1134. Chan, T.W., Goldthorpe, J.H., 2007b. Social stratification and cultural consumption: music in England. European Sociological Review 23 (1), 1 19. Chan, T.W., Goldthorpe, J.H., 2007c. Social stratification and cultural consumption: the visual arts in England. Poetics 35, 168 190. Coulangeon, P., Lemel, Y., 2007. Is Distinction really outdated? Questioning the meaning of the omnivorization of musical taste in contemporary France. Poetics. Goldthorpe, J.H., 2007. On Sociology, second ed., vol. 1. Stanford University Press, Stanford. Goodman, L.A., 1996. A single general method for the analysis of cross-classified data: reconciliation and synthesis of some methods of Pearson, Yule, and Fisher, and also some methods of correspondence analysis and association analysis. Journal of the American Statistical Association 91 (433), 408 428. Gower, J.C., Hand, D.J., 1996. Biplots. Chapman & Hall, London. Hedström, P., Swedberg, R. (Eds.), 1998. Social Mechanisms: An Analytical Approach to Social Theory. Cambridge University Press, Cambridge. Katz-Gerro, T., Raz, S., Yaish, M., 2007. Class, status, and the intergenerational transmission of musical tastes in Israel. Poetics. Kraaykamp, G., van Eijck, K., Ultee, W., van Rees, K., 2007. Status and media use in the Netherlands: Do partners affect media tastes? Poetics. Laumann, E.O., 1966. Prestige and Association in an Urban Community. Bobbs-Merrill, Indianapolis. Laumann, E.O., 1973. Bonds of Pluralism. Wiley, New York.

T.W. Chan, J.H. Goldthorpe / Poetics 35 (2007) 317 329 329 Martin, J.L., 2003. What is field theory? American Journal of Sociology 109 (1), 1 49. Peterson, R.A., 2005. Problems in comparative research: the example of omnivorousness. Poetics 33, 257 282. Peterson, R.A., Kern, R.M., 1996. Changing highbrow taste: from snob to omnivore. American Sociological Review 61 (5), 900 907. R Development Core Team, 2005. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, ISBN 3-900051-07-0. Silva, E.B., 2006. Distinction through visual art. Cultural Trends 15 (2 3), 141 158. Torche, F., 2007. Social status and cultural consumption: the case of reading in Chile. Poetics. van der Heijden, P.G.M., de Falguerolles, A., de Leeuw, J., 1989. A combined approach to contingency table analysis using correspondence analysis and log-linear analysis. Journal of the Royal Statistical Society, Series C, Applied statistics 38 (2), 249 292. Tak Wing Chan is a university lecturer in Sociology at the University of Oxford, where he is also a fellow and tutor of New College. His research interests are social stratification and mobility, and the life course, especially family formation and dissolution. John H Goldthorpe is an Emeritus Fellow of Nuffield College, Oxford and a Fellow of the British Academy. His main area of research is social stratification. His two current research projects are concerned with problems of education-base meritocracy and with social status and cultural consumption. He is also interested in the closer integration of sociological research and theory. The second, two-volume, edition of his book, On Sociology, is published by Stanford University Press.