Describing Description (and Keeping Causality): The Case of Academic Articles on Food and Eating

Introduction

Recently, Savage and Burrows (2007) argued that there is an impending empirical crisis in the social sciences, which is in part driven by the increasing processes of digitization and the practices involved in ‘knowing capitalism’ (Thrift, 2005). Their article triggered a number of direct responses (e.g. Crompton, 2008; Webber, 2009) and a renewed interest in the importance of considering sociological methodological practices as part of the emergent systems of knowledge production. It also renewed an interest in description (e.g. Savage, 2009) and it is this particular aspect of their argument that this article taps into. By providing a description of the changing focus of discussions about food and eating in peer-reviewed academic articles, this article provides an illustrative example of the kind of sociological description that Savage and Burrows (2007: 896) advocate, which is to ‘link narrative, numbers, and images’.

This article, therefore, starts where Savage and Burrows ends. More precisely, whilst the ethos of the argument here strongly supports Savage and Burrows’ ultimate message that sociologists must engage with the ubiquitous processes of digitization and that this entails rethinking social research, it departs from their proposal in its conclusion. Whereas they conclude that sociologists should ‘abandon a sole focus on causality (which we are very bad at) and analysis and embrace instead an interest in description and classification’ (2007: 896), here the position is that causality needs to remain firmly on the sociologist’s table of activities. There are three main reasons for this. First, it is shown that in order to make sense of any description, it is necessary to draw upon an existing causal framework. Second, from a more abstract level, drawing on the work of Dorothy Smith and Ian Hacking, it is argued that description depends on the possibilities of description, which in turn reflect the possibilities of action. Third, there are very good ethical reasons why description and causality need to be considered together from a sociological perspective. In conclusion, the question is not whether to reject causality in favour of description, but rather what kinds of descriptions help to explore causality and how description and causality can work together to enhance sociology itself.

The illustrative example used to explicate the argument is that of the changing academic focus in peer-reviewed articles about food and eating between the late 1950s and early 2010. After all, over the past 20 years, there has been an explosion of work relating to the sociology of food and eating practices (Caplan, 1997; Cheng et al., 2007; Germov and Williams, 1999; Mennell et al., 1992; Mintz and Du Bois, 2002; Murcott, 1998; Warde and Martens, 2000). However, so far, little attention has been paid to the changes within this emerging field of study. In order to explore the ways in which academic discussions on food and eating themselves have changed over time, and as a way of taking full advantage of the digitization of academic knowledge, the electronic bibliographic database, CSA Illumina Social Sciences, was used to extract and examine the contents of what had been written and when. This experimental way of exploring the enormous food literature began as part of a wider ESRC project, ¹ but ended in this article about description in sociology in general for reasons which will become clearer throughout the discussion.

Sociological Abstracts is one of the databases that many sociologists will be familiar with; this is one of the 20 databases hosted within the CSA Illumina Social Science collection. Most university libraries subscribe to some of these and this study had access to the following: ASSIA: Social Sciences Index and Abstracts; Criminal Justice Abstracts; ERIC; CSA Linguistics and Language Behaviour Abstracts; PAIS International; PILOTS Database; Social Services Abstracts; and Sociological Abstracts. (For the purposes of this article, when referring to ‘all databases’, this is the list of databases that is being referred to.) Collectively, then, at the time of the study, this allowed access to over 4,003,000 records of journal articles, book reviews, conference proceedings and magazines articles. No matter how big the collection of records is, though, it never includes everything, not just because of accidental omissions, but also because of the commercial and political issues involved in which publishers can and do participate within CSA Illumina. The exact details of which journals and publishers feed into this database are unfortunately not publically available. ² However, CSA Illumina is owned by ProQuest, which is owned by the Cambridge Information Group (CIG), the large American information services corporation. There is a sense, therefore, that the records are slightly biased towards American publishers. However, this assumption is necessarily tentative, since so many publishers are themselves owned by an increasingly global set of parent companies. Whilst not knowing the exact story of the production of the database is arguably a nuisance, presumably many an academic article has also been written without authors digging much deeper either. This being the case, in exploiting the database’s search facility, the following words were employed to narrow down the sample of articles: ‘food*’ and ‘eat*’, where the asterisk represents a ‘wild card’ to include any words beginning with ‘eat’ or ‘food’ (e.g. ‘eaters’, ‘foods’, etc.). The search was also restricted to ‘journal articles only’, which were ‘peer reviewed articles’ and ‘English only’; all duplicates were removed to avoid double counting. This generated a list of records, of which the title, abstract and key words were extracted for further exploration. In other words, the final sample of text consisted only of article titles, abstracts and key words.

The free online tool Wordle was used (see Feinberg, 2010) to explore the extracted text. The website (www.wordle.net) states: ‘Wordle is a toy for generating “word clouds” from text that you provide. The clouds give greater prominence to words that appear more frequently in the source text. You can tweak your clouds with different fonts, layouts, and color schemes.’ This web facility was considered best for three main reasons. First, its simplicity and speed outshines many other commercial or open source text-mining software tools. ³ Many packages require data to be entered in a particular format, whether that be in columns, Excel or rich text format, etc., whereas Wordle doesn’t, so it was merely a matter of ‘copying’ and ‘pasting’ the text into the web page. Second, whereas most text-mining software packages require the user to know something about the text beforehand, Wordle requires no prior assumptions. For example, typically with text-mining software, key words can be searched, coded, grouped, mapped, but they rely on the user to identify terms or phrases, or at the very least provide a tabular framework that helps to group the text into, say, types of answer, in the case of a questionnaire. Finally, the simple ‘tweaks’ are powerful and ultimately allow a relatively sophisticated analysis to be conducted. For example, by adding ‘~’ between words, it is possible to include phrases and terms (e.g. ‘social~policy’); it is possible to amend the number of words that appear in the final visualization, with the default 150 words being the maximum number possible; and actual counts can also be seen.

Wordle was used here in two main ways. ⁴ First, to account for the problems of (not) knowing the exact details of the sample, different databases were compared (using the same search terms) (see Table 1). This produced an image of four Wordles showing the ‘top 50’ words from the sample of text extracted from ‘all CSA databases’ and ‘Sociological Abstracts only’ using ‘eat*’ and ‘food*’ as key search terms (see Figure 1). As can be seen, the different databases do generate slightly different images, suggesting as one might expect, that authors write different kinds of articles for different kinds of journals. However, what is interesting is that, although the databases do pull in different work, the major words remain, and, in the case of the ‘eat*’ sample of text, they remain somewhat stable, providing greater confidence in the validity of those particular coincidences of words. That is, Wordles created from text based on articles with ‘eat*’ in the title still generate ‘disordered’ and ‘disorder’ and ‘women’ as key words irrespective of whether Sociological Abstracts or CSA Illumina databases were used, telling us something meaningful about the generally ‘disordered’ and gendered focus of the discussions within the articles about eating. Second, and more directly relevant to the argument presented here relating to description in general, Wordle was used to explore change over time by providing an image for each decade, ⁵ providing a chronological series of images, much like a story board of the changing coincidence of words in articles about food and eating over time (see Figures 2–5). The final images are based on the top 50 words, mainly to highlight key comparative differences and similarities between the different images.

OPEN IN VIEWER

Table 1.

Total number of articles generated by particular key words across the different databases.

Search term in title	All databases		Sociological Abstracts only
	Total number of records from all journals	Total number of records from peer-reviewed journals	Total number of records from all journals	Total number of records from peer-reviewed journals
‘food*’	6683	3711	1056	903
‘eat*’	3846	2667	445	392

OPEN IN VIEWER

Figure 1.

Top 50 words in peer-reviewed articles with ‘food*’ and ‘eat*’ in the title across different databases, 1950–2009.

OPEN IN VIEWER

Figure 2.

Total number of peer-reviewed articles with ‘food*’ and ‘eat*’ in the title per decade.

OPEN IN VIEWER

Figure 3.

Changing ‘food*’ articles over time.

OPEN IN VIEWER

Figure 4.

Changing ‘eat*’ articles 1950–2009.

OPEN IN VIEWER

Figure 5.

Exploring ‘food*’ versus ‘eat*’ 1950–2009.

In effect, Wordle provides a visual frequency count of the most often recurring words. However, the largest words are those that appear most frequently relative to all other words within which they are situated, so comparing the absolute size of words across Wordle images is problematic. Likewise, the frequency of words does not necessarily tell us anything about the meanings behind these words, although the coincidence of words goes some way towards that (Weber, 1990). Whilst these matters are not unimportant, since the Wordle images provide a description which links ‘narrative, numbers and images’, they nevertheless serve as concrete empirical material with which to think through the implications of Savage and Burrows’ conclusion. Indeed, as is now illustrated, it is precisely because of the challenge involved in making sense of this particular temporal description of the changing coincidence of words that it is argued that their argument is fundamentally problematic for three main reasons, each reason forming a distinct section of the discussion that follows.

Describing Description – and Drawing on a Causal Framework

Although the Wordle images are in themselves interesting, other than comparing them with one another and saying that x image is different or similar to y image, or that in a particular decade a word appeared more or less frequently relative to the others around it compared with another decade, the extent to which we can draw anything meaningful without going beyond them is actually very limited. Indeed, it is suggested that only three things can be confidently said about the images in and of themselves. First, what emerges more than anything from the images is that, even though food and eating are related to one another insofar as food is eaten, these words generate different kinds of academic work. Academic writing about ‘eating’, the images suggest, is primarily associated with eating behaviours and disorders, which are also typically associated with women and a focus on the body. Second, in contrast to the food images, ‘eating’ is gendered throughout, is associated with concerns with being ‘obese’ and ‘overweight’ in the 1970s, and thereafter becomes increasingly ‘disordered’. Third, the very tentative – emphasis on tentative – story of ‘food’ based on the images is one that starts off in the 1950s as being primarily associated with countries and income and shifts from being about agriculture and global food policy in the 1960s and 1970s, with ‘production’ remaining a key feature throughout and in particular in the 1980s, then consumption and security become more prominent in the 1990s, until finally in the first decade of the 2000s, there is another slight shift with the emphasis on children and health and the importance of the social.

Thus, on the face of it, authors write in particular ways for a particular audience and the words that they have used to tell their stories in this example have changed both in frequency and coincidence over time. However, going further than these rather basic observations imputes meaning and narrative to the images. Piecing together a story about why these images are as they are weaves a tale that demands explanatory modes of understanding and immediately requires stepping foot into the realm of causality. Description instigates questions about why it is the way it is. These questions need to remain present rather than be thrown aside and abandoned. Description provides the soil from where causal modes of inquiry can germinate and grow. That description acts in this way is a good thing sociologically, and not, as Savage and Burrows suggest, something we should be delimiting. Rather than abandoning causality, therefore, here it is argued that rich sociological work comes after description, which positively begs for causal modes of understanding.

After all, these Wordle images raise more questions than they answer. Whilst this might have more to do with Wordle or the sample used than with the act of description in general, the point about description raising further questions still stands. This is precisely why Savage and Burrows’ call to abandon causality is problematic; they are ignoring the fact that description, as one anonymous reviewer put it, ‘sets hares running’. Yet contrary to Savage and Burrows’ position, the strength of description arguably lies precisely in interrogating how and why it is the way it is and grappling with the multiple hypothetical explanations that are triggered. Indeed, we might say that description actually serves as a ‘hypotheses-generating procedure’ (Abt, 1987: 78).

For example, in this illustrative example, the differences between the ‘food’ images versus the ‘eating’ ones might lie in the way that ‘food’ is associated with intertwined series of systems of production, consumption, development, health and children, whereas eating is pathologized, disordered and predominantly about women, the body, and disorders. Alternatively, the differences between the ‘food’ and ‘eat’ images may be explained by one being derived from a noun (food) and the other a verb (eat). As Langacker (1987) argues, nouns and verbs are semantically distinct precisely in the way that they each construe contrasting imagery. A verb is temporal, or at least more so than a noun (1987). Paradoxically, therefore, the relative continuity of the ‘eating’ images may point to the atemporality of the act of eating, which always and everywhere involves a relationship between an individual (of any kind) and food as an object (of any kind). This still would not explain why the relationship is supposedly more and more ‘disordered’. It is difficult to discount the possibility that the different images have as much to do with the ontology of the words themselves as the epistemology of the images produced by them. Another possible reason why the Wordle images are as they are might be due to the arbitrary way of ‘cutting up’ time in these descriptions. A different story might have been produced had the time periods been otherwise (Abbott, 2003). Or perhaps the images merely reflect the (middle-)class discourses of food and eating, whereby the authors of the articles might be aggregately assumed to be middle class by virtue of the fact that they are writing within peer-reviewed journals. This echoes Darmon’s (2009: 718; original emphasis) work on ‘class practices, class dispositions, and class habitus not as causes, but, rather, as social conditions of possibility (or social conditions of likelihood) of the anorexic career’. In other words, it is possible to consider the academic author as an intentional actor situated within the social, with the content of the article seen as a direct consequence of the dynamics of being in the social world at the time it was written; we return to this issue shortly.

There are likely to be other ways of reading these Wordle images. Suffice to say, however, without further research it is difficult to know which possible hypothetical explanation is most valid and reliable. As Abbott (2003: 52) summarizes, ‘Every description seems contestable; all require interpretative frameworks’ – herein lies a real problem of what Savage and Burrows imply. The difficulty with any description tends to be in confirming the validity of the competing interpretations and the extent to which validity claims are accepted (Flyvbjerg, 2001: 139). The difference or similarity between one image and another is rendered meaningful only through recourse to theory, which is not actually depicted.

Questioning description by drawing on causality arguably avoids the trouble that radical relativism gets us into whereby every description is valid in and of itself. Deliberately engaging with the hypothetical explanations behind a description potentially permits a way of investigating the validity of the description itself. Whilst it is tempting to see the Wordle images as a reflection of the changing discourses within academic articles on food and eating, as has been suggested, it would be misleading to think of them as a description of the changing sociological significance of food and eating in society more generally. After all, there are at least the possible reasons above to think that there is more going on behind the images themselves. This is not to say they are substantively meaningless or tell us nothing about the history of food and eating literature more generally. Rather it is to say that every description needs to be tentatively approached with regards to understanding what it is and why it is the way it is. Contrary to Savage and Burrows, therefore, the position here is that description and causality need to go hand in hand in order to assess the validity of either.

Describing Description – and the Possibilities for Action

The second main reason why Savage and Burrows’ argument is problematic is that, since every description is conducive to interrogation about how and why it came to be that way, abandoning causality actually misses the real sociological potential intrinsic to any description. Indeed, another way of thinking about any description is to consider it within a wider story of change. Description can tell us something about the emergent ‘traces’ being produced in and by the social world. The term ‘traces’ follows Byrne’s (2002) approach to interpreting quantitative variables – as alternative expressions ‘of the real systems that compose the world’, much like the tracks in the sand left behind for the animal tracker to follow. As Abbott (2003: 51; my translation) puts it, also using the analogy of the animal tracker, ‘description follows action like a tracker follows the path of its prey, but only managing to see its tail disappearing round the corner of the path’. What becomes key to distinguishing descriptions, according to Abbott (2003), is the temporal ordering of the descriptions themselves. Smith (1981: 335) takes this further and argues that ‘The ordering of the descriptive procedure thus becomes the ordering attributed to courses of action and to the social relations of the setting itself.’ That is, description always and necessarily reflects something about the social organization of the setting in which it was constructed and the sequential courses of action that have gone into that particular construction. Vice versa, the sequential order of description may itself feed into how that setting and those actions are construed.

Following Smith (1981), we might say that the focus in the Wordle images is on words and terms that are preserved, but their uses and their meanings change. The words are used within a particular setting – that is, academic peer-reviewed articles within a particular time and place. Conversely, in exploring the ways in which the meaning of the words changes over time, the setting itself is necessarily described as well. We have, then, a description within a description, or rather a description of a description; more accurately still, we have a description of multiple descriptions over time. Furthermore, it is possible to conceive that the authors of the peer-reviewed articles were themselves individually and collectively describing, critiquing, commenting on the social world in which they were situated.

After all, the Wordle images are ultimately descriptions of the most frequent words social science academic authors were using in relation to what they were writing about. Academic articles are just documents; they may be a particular type of document, but they nevertheless remain in the realm of the social and are written by individual actors who are necessarily situated in time and space (see Scott, 1990). What becomes important to interpreting the sample of text used to build the images are the ‘unstated meaning structures’ (Cicourel, 1964) as potential sources of information about past social change. Viewed like that – and this is the key point – description becomes part of a mode of production in which context and meaning are inscribed.

Consequently, the images may be seen as windows through which social (academic) commentators were implicitly capturing what was going on around them. Furthermore, the images can be said to depict the changing intentional actions of the authors’ descriptions in particular social settings that change over time. As Smith (1981: 336) notes: ‘The social organization is always necessarily “present” in the description, and the description depends upon it though it does not explicate it … This organization is already “in” the description.’ In turn, much like the dried-up water stains left on windows, which remind us, even on the sunniest days, that it rained, it is possible to reinterpret both Smith’s and Savage and Burrows’ call for description as a ‘trace’ of some contemporary epistemological, methodological and substantive concerns that sociologists may be experiencing (in the present). (Of course, this article is no exception and also contributes to the ‘traces’ left behind.)

This leaves us in a rather precarious predicament: the possibilities of description are always and necessarily recursively co-constructed not only by the setting(s) in which they emerge, but also by the possibilities for action within that setting. As Hacking (2002: 48) puts it, drawing on the work of Elizabeth Anscombe:

… intentional action is action under description. So there have to be descriptions. If we can show that descriptions change, some dropping in and some dropping out, then there is simply a change in what we can (as a matter of logic) do or not do.

He explains this further:

… what things are doing, and indeed what camels are doing, does not depend on how we describe them. But some of the things that we ourselves do are intimately connected to our descriptions. […] What is curious about human action is that by and large what I am deliberately doing depends on the possibilities of description. To repeat, this is a tautological inference from what is now a philosopher’s commonplace, that all intentional acts are acts under a description. Hence if new modes of description come into being, new possibilities for action come into being in consequence. (2002: 108)

Hence, descriptions are not only necessary forms of practice, they are also intrinsically linked to intentional action, which is itself co-constructed through the possibilities of description. Put simply, were it not for the fact that it was possible to relatively easily extract hundreds of articles with ‘food*’ and ‘eat*’ in the titles and then to provide a Wordle image to depict something about the most frequent words, it would not have been possible to say anything about the ways in which authors have intentionally approached their work on food and eating. Moreover, were it not for authors writing about food and eating, this whole exercise would have been impossible. The fact that authors intentionally locate their food writings in discussions about the general social structures, whereas writing about eating is located within disordered and gendered discourses, is interesting in itself. That food and eating generate such different Wordles over time says as much about the social setting of the publishing industry itself as it is does about where the expectations of funding for particular social concerns lie. Hence, by approaching description at this more abstract level and deliberately going beyond the description in and of itself, we can ponder more meaningful issues about the conditions in which new possibilities for action may exist. This is entirely compatible with a critical sociological enterprise.

Describing Description – and the Ethics of Sociological Research

The third and final reason why Savage and Burrows’ argument is questionable is simple but central. Whilst they urge us to do more description, they fail to spell out what we are to do once it is produced. Their answer is to abandon causality because it is difficult and we are not very good at it, and to continue therefore to focus on description alone. But sociologically, this is a mistake. Are we really going to just describe, say, the inequalities we observe in the world without any recourse to critique them or change them? If so, then Savage and Burrows are absolutely correct in arguing that there is an empirical crisis in sociology, and it is far more grievous than they suggest. On the contrary, there are appropriate ethical reasons for going beyond description and investigating the causality behind its production. We should and must pose causal questions about why something is as it is.

Let us return one last time to these Wordle images before bringing this article to a close. The extreme disparity between the food and eating coincidence of words is alarming. We all need to eat, and food gets eaten, so why might eating food generate such different images? Why are the eating images so ‘disordered’ compared with the food ones? The gendered nature of these images raises ethical concerns about food, eating and gender in society. The images raise the more general question about the extent to which social scientists are merely narrating a particular kind of description emanating from particular phenomena. ⁶ Equally, though, we should query the extent to which it is sufficient to do this given the increasing coincidence of words relating to the ‘disordered’ nature of ‘women’ ‘eating’ that is couched within these images.

If, as has been suggested, descriptions are forms of practice intrinsically linked to intentional action, which is itself dependent on the possibilities of description of the social setting in which that description was produced, then the question is not whether or not to abandon causality in favour of description, but rather to think strategically about the kinds of descriptions that might be produced in order to explicitly learn about the social setting, and indeed the possibilities of intentional action within that setting. In addition, to understand which modes of description have led to new possibilities of action, it is important to know something about how those modes of description have come into being in the first place. This not only requires description, but it is also a causal narrative about how and why a description has come to be that way. Given that this is the case, it is difficult to fathom how description could ever be fully severed from causality or what the benefits of doing so would be to anyone. Abandoning causality for description is problematic for ethical reasons if none other.

Conclusion

What this empirical exercise has sought to illustrate is that interpreting description invites us to keep a foot in a causal frame of reference and, in turn, to explore the social setting and the international course of action that has gone into making that description possible in the first place. Indeed, reprising the three key points suggested here, we might say that keeping descriptive and causal modes of inquiry together arguably places fuzzy parameters as to which descriptions are more or less useful and valid. Moreover, the utility of a description is itself gauged by the extent to which it facilitates knowing the social setting in which that description was produced in the first place, and, importantly, to question the possibilities of intentional action within the setting that are inscribed within it. Furthermore, since there are very good ethical reasons why it is necessary to maintain causal questions about why sociological descriptions are as they are, thinking strategically about the kinds of descriptions sociologists may want to produce at all is essential.

The issue, then, is not whether or not to reject causality in favour of description, but rather which kind(s) of descriptions help to adequately explore causality and, conversely, which kind of causality helps us to make sense of the descriptions that emerge (see Byrne and Uprichard, in press, on this issue). Descriptions that ‘link narrative, numbers and images’ are powerful tools, especially where new technologies facilitate the speedy assembly of such descriptions. Producing communicable findings that are of interest at local, national and global levels is necessary (Flyvbjerg, 2001: 166). This is where these kinds of descriptions may have a niche within the kind of social science that Flyvbjerg (2001: 166) argues is ‘done in public for the public’ and serves ‘as eyes and ears in our ongoing efforts at understanding the present and deliberating about the future’. Nevertheless, the extent to which these kinds of description are useful in and of themselves is questionable.

Perhaps a core issue underpinning the ‘empirical crisis’ that Savage and Burrows raise is a sociological lethargy. This may be overstating things somewhat; no doubt readers will disagree. However, it does seem that some of the debates about the nature of the discipline or the empirical and/or methodological crisis within it are tapping into a more grievous concern about the lack of confidence – or rather lack of hope – that sociology (as a discipline) can actually tackle the problems that it has so successfully helped to describe (e.g. inequalities, segregation, power differences, etc.). Yet if this is the case, then surely the notion of causality is worth holding on to. We need to continue trying to understand why things are as they are – i.e. engage with causality. Of course, descriptions can be sufficiently powerful in themselves precisely because they impact on meanings and values (Engel, 2002; Rorty, 1979), but this is also because they impact on the causal models that serve to hold the existing status quo in the first place. Destabilizing causal mechanisms, therefore, is another way of thinking about the kinds of sociological description that might be sought after (Cartwright, 2007; Smith, 1981).

Savage and Burrows are right in concluding that sociology needs to rethink its methodological focus in an age of ‘knowing capitalism’ (Thrift, 2005). They are also right in arguing that there is a need to engage with the rapid proliferation of digitized data that lends itself to new forms of description. Description is certainly necessary, and Savage and Burrows are spot on in shaking the discipline awake with that particular call. However, they are wrong in suggesting that this means abandoning a focus on causality; the baby must not be thrown out with the bathwater. Description and causality go hand in hand. They are complementary and intrinsically linked and any attempt to sever that interconnection will only result in a poorer social science and a greater ‘empirical crisis’. A sociology without descriptive and causal empirical orientations would be one without recourse to re-describe the world differently and one, therefore, that we should surely seek to avoid.