The validity of a grounded theory approach to research on democratization

Introduction

This paper attempts to demonstrate that a grounded theory textual analysis can yield results that have the same validity as other more conventional methods, such as interview and archival research. The method was applied to analyse the stenographic records of three communist party conferences held in the Soviet Union, Hungary and Yugoslavia in 1988. In each of these states, such all-party conferences had not been convened since 1941, 1957 and 1973, respectively. While provisions in the Party Rules clearly define the powers of a congress (where a new leadership may be elected), the institution of a ‘national’, ‘federal’ or ‘all-party’ conference was described very loosely to provide that the party leadership ‘may convene the conference, between [five-yearly] congresses’. This granted the leadership significant discretion over what could be achieved at a conference. For example, in the case of the Hungarian Socialist Workers’ Party (HSWP) conference, a palace coup took place, which included very significant leadership change (Schöpflin et al., 1988). The Yugoslav (federal) party conference, in contrast, simply did not feature as particularly significant, and while a good deal of press coverage was manufactured, it received little attention. This may be explained by the fact that the federal party, the League of Communists of Yugoslavia (LCY), by this time, had little effective power over the constituent republican parties, and so people generally took little interest in its institutions (Lampe, 1996: 338). The Communist Party of the Soviet Union’s (CPSU) conference in 1988 has been interpreted by some as Gorbachev introducing a democratizing agenda (Brown, 1996: 175–177), while others view the reforms as an attempt to consolidate the party’s authority and increase its control over the state and society (Sakwa, 1990: 129; Hazan, 1990: 185).

There has, however, been no systematic comparative research into these conferences and the speed with which events unfolded and subsequent final party congresses convened across Central and Eastern Europe meant that this particular moment in the liberalization stage in the region was overlooked by existing scholarship. These conferences took place at a time when these leaderships sought to modernize their respective communist parties to try to attract more members and make the one-party state viable. Ostensibly, the purpose of these conferences was to solicit the wider opinion of party members, at an all-party meeting, in order to find solutions to the problems these states faced. Since these regimes had begun liberalizing to a limited extent, it was anticipated that debate at these meetings would display a range of different preferences. Textual analysis of these minutes revealed instead a staged show of support by delegates for the centrist–reformist leaderships that organized the meetings.

The application of textual analysis and the later verification of the results against data from interviews and archives provided an opportunity to evaluate the validity of the method post facto through the application of a triangulation of approaches. This was not the intended aim of the textual analysis, which was instead to glean as much data as possible, in the event that interviews with relevant individuals or access to archives was not possible or revealed little. There were concerns that the pool of relevant interviewees was limited to former high-ranking party members and specific individuals present at these conferences. The interview research was conducted in 2002–2003 ² and since these conferences took place in 1988 it was anticipated that the lapse of time could be an issue, both in tracing relevant interviewees and securing meetings with them to share their memories of the event. The other unknown quantity at the outset was whether or not access would be granted to relevant documentary evidence in the various archives. When it came to the fieldwork stage, however, these two other avenues proved surprisingly fruitful.

Nonetheless, the textual analysis brought findings that better informed the following stages of fieldwork, and also produced some results not replicated in data from archives or interviews. Moreover, it confirmed to me the validity and importance of a grounded theory method (GTM) textual analysis, which would have been informative in its own right had interview and archival sources not been forthcoming.

In some circles, however, textual analysis is considered controversial, in that the results produced are often unverifiable. This tension has been identified in a very different textual analysis research project in terms of subject matter and method, yet what both my project and this one share, broadly, was a use of textual analysis to challenge existing interpretations of political history. Gilbert Shapiro and John Markoff conducted a content analysis of the French revolutionary demands in the Cahiers de doléances of 1789. Anticipating a good deal of opposition to their findings, they devoted half of the book published from their research to thoroughly defending their application of content analysis, acknowledging that for a long time content analysis has been considered a ‘methodological ghetto’ (Shapiro and Markoff 1998: 20–21). They argue that this criticism is unwarranted because the essential procedures of content analysis are very similar to the component processes of nearly all other social science research techniques (Shapiro and Markoff, 1998: 20).

Such scepticism may be best understood by the explanation that certain methods are simply less favoured in particular disciplines, as has been demonstrated by a diverse group of social scientists who investigated the systematic review process. Their research empirically showed that this review process is structured in such a way as to clearly favour quantitative over qualitative publications in the health sciences, which in turn serves to marginalize qualitative research from the ‘mainstream’ (Dixon-Woods et al., 2006). Moreover, in the pursuit of contentious research, the stakes are raised: ‘The lingering hegemony of positivism still makes controversial research suspect… therefore, the data for such studies must be unassailable’ (Charmaz, 2008: 209, citing Fine et al. (2000)). Similarly, while textual analysis and GTM have generally been welcomed in a very broad range of disciplines, as is testified by this journal, among many other publications, there is a tendency to treat its results as more conjectural than those attained through other methods. A number of excellent in-depth and well-informed accounts on how the method has developed already exist, some of which are cited in the following summary of the debate. This serves to frame how the method was applied in the textual analysis of the party conference minutes.

Grounded theory: having one’s cake and eating it?

As is well known to the readership of this journal, the method is commonly attributed to Glaser and Strauss, as launched in their 1967 publication, and has since received a good deal of attention from both critics and supporters.

One of the most scathing critiques of GTM comes from Michael Burawoy, who condemns the method as ‘antitheoretical microempiricism, brilliantly mislabelled as grounded theory’ (Burawoy, 2009: 8). Other criticism questions the explanatory force of theories generated by this method; some doubt whether a theory ‘generated’ from the particular context can be generalized to other cases: ‘This is not what qualitative inquiry is about. To want to do this – to foreground the uniqueness of interpretation, at the same time as developing theory – is to want to have one’s cake and eat it’ (Thomas, 2002: 421).

GTM was originally conceived as a counterweight to the prevalent culture of hypothesis-testing and what Glaser and Strauss saw as the over-emphasis on verification theory, which they believed to be a constraint on the bounds of research. Instead, therefore, they proposed a method that would elicit theory directly from the data under analysis (Howell, 2000: 1). As the method was initially conceived, the main focus was to be on the data, without any presuppositions on the part of the analyst as to what patterns would emerge. The analyst should aim to develop a theory that applies to the example under examination through the comparative empirical analysis of data (Glaser and Strauss, 1967: 3–6). This action of generating a theory from the data ensures that ideas or hypotheses are not only derived from the data, but are worked out in relation to the data as the research progresses. The main underlying idea, as originally articulated, was that the research should be entirely inductive, approached with as few predetermined ideas as possible, and undertaken with all logically deduced a priori hypotheses blocked out (Glaser, 1978: 3).

Since the 1990s, however, GTM underwent a schism, whereby Strauss and his adherents acknowledged the role of deductive processes inherent in the method, in the analyst’s choice of core categories and practice of selective and axial coding. Strauss conceded that these inevitably shaped the emergent theory, which could not, therefore, be considered to have been generated through purely inductive processes. Glaser, on the other hand, maintains his commitment to the method as entirely inductive (Bryant and Charmaz, 2007a : 4–11). To some extent Glaser and Strauss’s original prescription of the method in purely inductive terms has been retrospectively diagnosed as a reaction to the powerful and dogmatic positivist movement that had been prevalent since the 1950s (Clarke, 2005: 3). Indeed, although inductive, the method itself was originally described in positivist terms to put a veneer on the method, which would make it apparently less controversial ‘as a template for doing qualitative research stamped with positivist approval’ (Charmaz, 2008: 205). Some practitioners since then have moved even further away from the model of GTM to reject what they consider to be some of the more positivist procedures prescribed by Glaser and Strauss, as artificial and inappropriate constraints on qualitative inquiry, and to frame a new variant of the method in the tradition of the Chicago school, which they term as ‘constructivist’ GTM (Charmaz, 2008: 204–207). They acknowledge that a researcher’s pre-conceived ideas, knowledge and preferences influence the emergent theory and that in doing so, such practitioners have taken something of a middle ground between induction and deduction, which has been coined as ‘abduction’ (Richardson, 2006). This ‘abductive’ approach was taken in the analysis of the conference publications, and the choice of QSR N6 encouraged a more structured, deductive approach to the analysis, which would appear closer to Anselm Strauss’s position.

Why GTM and QSR N6 were selected for the analysis

The reason I chose GTM to analyse the published minutes of the three party conferences, was because the meetings had not been systematically and comparatively examined before, and secondary literature concerning them was sparse. In selecting the tabula rasa approach of GTM the aim was to generate new theory (rather than challenge existing theory), as others have employed the method (e.g. Kezar, 2005; Prindeville, 2004; Danis and Parkhe, 2002; DeLeon-Granados, 1999).

The coding frame was designed to categorize views expressed in speeches, and record the range of views, as this was expected to act as a barometer of transitional change within each party, defining the extent to which party members were advocating, or sympathetic towards systemic transition or transformation. At the same time, room was allowed for other patterns or salient information to emerge from the analysis. Once a coding frame had been devised, patterns began emerging which revealed similarities between conferences.

QSR N6 was selected for the analysis (900 pages of published text) because it offers the facility to hierarchically structure codes. QSR N6 has been employed by many practitioners of GTM and the method has apparently been shaped by, as well as reflected in, the software. ³ For example, NVIVO, the sister software to Nud*ist, has been named after Strauss’s term in his open coding system ‘in vivo’ – a function in the software which allows the practitioner to generate a code directly from the text. Similarly, the version N*6 has a function ‘in vivo’ which allows the creation of a coding category taken directly from the text. ⁴ While this marriage of theory and software has been well received (Weitzman and Miles, 1995: 18; 238–256), some have expressed concern that it may stifle the necessary creativity required for theory building (Bryant and Charmaz, 2007a: 24, citing Glaser (2003); Walsh, 2002: 529, citing Atkinson, Coffey and Delamont).

This seems to draw on the same argument proposed by Charmaz and others, described earlier, and it is indeed valid criticism that the software itself encourages a positivist approach to structuring the data, which then subsequently shapes the emergent theory. The software facilitates the storage, coding, retrieval and analysis of text and was designed to promote theory-building, by allowing the function of making links between designated coded categories (termed in the software as ‘nodes’) and the option of classifying codes within a structured hierarchy using the terminology associated with a tree. N*6 therefore implicitly encourages axial coding, by imposing a hierarchy on the categories of codes and enticing users to draw links between these to organize them as they consider appropriate. Indeed, in any software where a large number of different coding stripes have been created from the text there is an overwhelming urge to impose order on these in a hierarchical framework, as a way to catch a glimpse at the bigger picture, but also to make sense of the data in a way that is meaningful to the research. As is apparent from the sample of node reports, the imposed hierarchy developed during the three party conference projects is visible. ⁵

The coding frame

The coding system adopted for the textual analysis was Anselm Strauss’s: open, axial and selective coding (Strauss and Corbin, 1990: 31). Although, in the constructivist model of GTM, axial coding has now been discarded, in that it is ‘not a productive research strategy in that it relies far too much on preconceived prescriptions’ (Bryant and Charmaz, 2007a: 9), this was the strategy adopted for this project, in part to deductively test to see whether these conferences could be interpreted as ‘democratizing.’ Since the research project was to be framed within the democratization literature, reference to materials within that body of scholarship was desirable, but as acknowledged, the process by which these core categories ‘emerged’ was not entirely through inductive process. The analysis was therefore of the ‘abductive’ type in that the initial process was inductive, beginning by open-coding, but became more deductive, as I imposed pre-existing categories to capture and structure this data.

Finally, ‘selective coding’ followed, whereby the main core categories were picked out and on this basis a hierarchy was constructed, appending sub-categories to the designated core categories (Strauss, 1987: 24–37). Again, in the case of this project, use was made of the transition and democratization literature. My core category in each case is ‘democratization’ or ‘liberalization’ versus evidence to show that conferences were trying to approve policies that would maintain the status quo. The three sets of conference minutes were coded independently of one another, but useful categories were ‘borrowed’ where evidence of similar patterns occurred. Common core categories emerged, for example, in the evidence of discussion of perceived irregularities surrounding delegate-selection to the conference (who was selected and why), who was selected to speak, policies approved and whether they were legally binding, and references to how the conference was prepared.

The coding process created an exhaustive report of all viewpoints on issues expressed at the conference, revealing trends of opinion and helping identify how the different ‘factions’ were represented at the conference. Moreover, the categorization of opinions and topics discussed at the conference helped assess the extent to which genuine debate or spontaneous discussion could be identified from the conference minutes and the incidence of material which appeared artificially staged.

Some text was coded more than once, being relevant to more than one code, ⁶ while other text coded was of peripheral value to the theory which emerged and was therefore not reported. For example, in the Moscow project, each time another speaker or another person was cited (in whatever capacity – whether to agree or disagree with another speaker at the conference, or to cite figures of historical importance, such as Alexander Pushkin or Vladimir Lenin for the sake of dramatic effect) these persons were listed as a new ‘child node’ under the top level tree node ‘Citation’, resulting in 63 child nodes. Most of these appeared of little consequence, with one or two references to a person made usually in agreement with another speaker to endorse a view they had expressed. References to Gorbachev were numerous and all positive, so again were not reported as they did not appear particularly significant. However, interestingly, in the case of a couple of individuals citation was heavy and negative. For example, in the case of the node ‘Citation’, references to the W Director of the Institute of Economics, USSR Academy of Sciences, Leonid Abalkin, came to be of importance for the textual analysis report. References to Boris Yeltsin were also significant; so much so, in fact, that a new ‘top level node’ (of which there are only 53 in the entire Moscow project, whereas there are over a thousand lower level ‘nodes’) was created for ‘Yeltsin’, as speakers expressed a wider range of views about Yeltsin and devoted more time to this.

Coding was to some extent difficult in the context of the material under examination, because although parties were in the process of liberalizing and individuals in these regimes had developed different preferences by this time, the leadership still expected party unity. Dissenters were therefore under these conditions likely to remain guarded in what they said and how they expressed opinions that deviated from the party line.

Interpretive coding, self-censorship and semantic satiation

This phenomenon of self-restraint is generally referred to as ‘self-censorship’ and denotes an individual’s perception of operating limits, and the practice of using restricted speech to voice opinions for fear of uttering something that could be considered impolitic under conditions where speakers anticipate that such utterances will be penalized (Horváth and Szakolczai, 1992: 217–219; Blyum, 2003: 3). Soviet era citizens dealt with this by using ambiguous language or even non-reference, or in other cases completely re-defined words to convey the desired meaning. By re-defining words and using ambiguous language, people were able to communicate on taboo subjects and promote forbidden ideals without being open to the accusation of breaking the rules (Horváth and Szakolczai, 1992: 217–219). This ambiguity is attested in the conference minutes by delegates at the HSWP conference who demanded that delegates speak plainly. In instances where speakers’ discourse may be constrained, therefore, interpretive coding is helpful as a means of picking up nuance.

At the other end of the spectrum, speech at the conference that could have been interpreted as advocating full-scale regime change had to be treated with some degree of scepticism. Textual analysts have pointed out the methodological flaws of over-interpreting the significance of presence or absence of words. Commonly, words or phrases lose their force and meaning through over-use; this phenomenon has been termed ‘semantic satiation’, which is when a phrase or word is used by a speaker to convey the effect of endorsing a value, but is perhaps received by the speaker’s audience as insincere. Symbols of democracy occur less frequently where democratic procedures are accepted than where they are in dispute (Krippendorf, 1980: 41). ⁷ In the context of the conference minutes, for example, use of the term ‘perestroika’ in a positive context was not necessarily an indication of the speaker’s full affirmation of Gorbachev’s policy of perestroika. János Kádár (the Hungarian First Party Secretary in 1988) freely used the word in his speech without in any sense expressing a desire for change in keeping with Gorbachev’s policy (which promoted personnel change via the removal of the aged party elite, which logically included Kádár’s own resignation). So, the presence of such apparently ‘liberalizing’ terms may not necessarily be evidence of an alignment of policy with Gorbachev. This is borne out by a comparison of the presence of such words in delegates’ speeches. Kádár was one of the speakers least in favour of Gorbachevian alignment at the Hungarian conference, and yet he mentions in his speech the word perestroika three times, democracy four times and glasnost’ four times; he quotes a church leader and Margaret Thatcher in positive terms, although he clearly did not endorse any of these in the policies he supported (HSWP, 1988: 231–239). In contrast to this, some of the most reformist speakers did not use any of these words, perhaps because they had become perceived as meaningless through overuse. This is well illustrated by the speech of the most reformist delegate at the Hungarian conference, a young economics lecturer, who did not express herself using words which, if applying a content analysis dictionary, would indicate a reformist position. However, interpretation can lead us to infer that she was promoting the introduction of a multiparty system (HSWP, 1988: 85–89). Thus, only by examining what is said on the level of sentences and paragraphs can the language be ‘unpacked’ and adequately coded. For example, reference to the need for personnel changes would be an indication of a speaker’s desire for ‘perestroika’ in the sense that Gorbachev meant it.

Similarly, some delegates made very conflicting statements over policy which required interpretation during the coding process. For example, in the case where there was incidence of, say, a vacuous reference in favour of ‘democratization’ in conjunction with a demand that the Communist Party increase its ‘vanguard status in society’, this was interpreted as paying lip-service to the rhetoric of ‘democratization’ though not advocating the introduction of a multiparty system (as instead the strengthening of the CPSU’s role in society was proposed). So, ‘democratization’ in this context may be better interpreted as ‘greater mobilization of the people in favour of the CPSU’. This is how such conflicting statements were interpreted during the coding procedure, for example, from the speech of Sapamurat Niyazov, ⁸ the first secretary of the Turkmen Communist Party. Thus, coding relied to a large extent on interpretation of delegates’ speeches in the context of what they said. Clearly, the need to maintain consistency when undertaking interpretive coding in this way is essential to ensuring the reliability of the coding frame and the ensuing results.

Maintaining reliability through computer-assisted textual analysis

While, as mentioned earlier the use of software risks stifling creativity and downplaying some of the more extraordinary features in the data, the practice of coding using QSR N6 made the entire analysis open to scrutiny. The validity of the codes was maintained by the use of interpretive coding of text and the constant checking of codes and exemplars as the project was developed. The iterative nature of the coding process promoted consistent treatment, while the flexibility of the software aided this process of constant reappraisal as new codes were generated from the text.

In the case of this research project, reliability testing comprised checking my own consistent coding (‘intra’-coder reliability), which was performed through the simple mechanism of checking codes created previously during the next session (and on subsequent occasions) to ensure consistent treatment of the data. Especially, where an instance proved the exception to the rule, such as speakers expressing a divergent view, the text was examined a number of times on different occasions to ensure that the interpretation was valid. As mentioned earlier, the context in which statements were made was taken into account and interpretive coding applied to unpack any apparently contradictory positions adopted by delegates in their speeches over policy. There may have been a tendency, therefore, to err on the side of a rather conservative coding of the text, as my concern was to ensure that my own interpretation of the text shaped the coding frame as little as possible. Nonetheless, the results enabled a subsequently richer discussion with interviewees who were present at these conferences, and patterns that emerged were confirmed at interview, which would not otherwise have been brought to light.

Theoretical sampling

Data collection is prescribed in the GTM formula as an iterative process, whereby data (traditionally collected through interviews) is added to the project until such time that they elicit no new categories or insights to the research. This process of ‘theoretical sampling’ is another way to ensure that the theory is valid in that it is underpinned by a representative (and exhaustive) dataset. While vital, this is also acknowledged to be difficult to effect in practice (Bowen 2008). Similarly, Bryant and Charmaz have expressed concern that there is clearly no certainty that, beyond the dataset collected and the theory based on this, an unexplored exception to the rule does not exist, which, had the researcher happened upon and included, would overturn their theory. At the same time, common sense dictates that this is generally the case in every form of inquiry (Bryant and Charmaz, 2007b: 45). In the context of this project, theoretical sampling was a matter of a priori selection of cases from the possible pool of conferences held in the region in 1988. The cases selected were those that were construed by the existing literature as the most pro-perestroika, and pro-reformist. Hence, it was anticipated that, at these party conferences, there was more likely to be debate about whether or not these states should relinquish the power of the one-party state, and that discussions would extend to the possibility of democratization. These parties each held their respective conferences between May and June 1988. Analogous national party conferences held in Bulgaria (1988) and Romania (1987) were less likely to display any development towards democratization, as these regimes were less liberal.

With respect to data saturation, therefore, while interviews and archival resources might have been included in the textual analysis, it was neither practical nor desirable to include these, as this would have been very time-consuming, and there were concerns, as mentioned earlier, that interviews could not be exhaustive, nor archival access to documents comparable (as each of the archives had quite divergent policies on which documents I could access). This could potentially have weighted some cases with data that could not be replicated in the other cases. Moreover, as it turned out it was good to keep these data separate, as the archival and interview resources told, in part, a very different story, which indicated strong divergence of opinion in the party that was not apparent at these conferences.

The results of the textual analysis

The results showed that delegate preferences expressed at the conferences were predominantly ‘centrist’, and wholeheartedly supported a pro-perestroika approach. The theory generated by this analysis was that these conferences were staged by centrist leaderships specifically to convey this centrist message. This became apparent in the case of the Soviet conference through the appearance of a pattern that the only two speakers who diverged from the common line, Leonid Abalkin and Boris Yeltsin, were systematically castigated by others. Since they were both high profile speakers, their views would have been known in advance, and they appear to have been set up at the event specifically to be knocked down. Abalkin queried whether the leading role of the party could be maintained under conditions of democratization (CPSU: 1988, v.1, 119). Abalkin was denounced by subsequent speakers in each but the last of the sessions that followed (CPSU: 1988, v.1, 122; 136; 156; 203; 222; 261; 334; v.2, 4), and Gorbachev himself reiterated that democratization was indeed feasible within the bounds of the one-party state (CPSU: 1988, v.1, 198), as did others, such as Evgenii Primakov (CPSU: 1988, v.2, 33). As the last speaker in the penultimate session of the conference, Boris Yeltsin pushed the bounds yet further in his own speech, advocating the creation of a second party from the factions that had evolved in the Party. Yeltsin was criticized by five out of eleven of the speakers during the last session of the meeting, and received very little, if any, support from speakers throughout the conference. Given that Yeltsin was very soon after this able to rally sufficient support in the party for his selection as President of the RSFSR, and then soon after his election as President of Russia, suggests that those invited to speak at the conference in 1988, were not a representative cross-section of the Party.

In the case of the Hungarian conference, those with a voice at the conference were also overwhelmingly from the centrist camp led by Károly Grósz, who was subsequently elected at the meeting to replace János Kádár as the party’s First Secretary. Although a small number of more reformist delegates were invited to speak, views expressed rarely diverged from the centrist line. However, as was pointed out at the conference itself, those views expressed were not representative, as delegates had not been selected according to the usual criteria. Although well over half of the delegates advocated some form of ‘democratization’ within the party, from the context it is clear that what they were advocating was far from ‘democratization’ per se. What they proposed varied from maintaining the status quo to major reform of party structures. More than two thirds of the speakers advocated perestroika and just over half demanded more glasnost’. Clearly, what was advocated in delegates’ speeches corresponds closely to CPSU modernization, comprising some liberalization of party structures, a purge of corruption and further economic liberalization. Moreover, at the event itself, widespread personnel changes in the party leadership took place.

At the Yugoslav conference delegates were critical of the leadership and also demanded radical personnel change (LCY, 1988: 28). All speakers representing the respective component federal organizations advocated leadership change (LCY, 1988: 66; 76; 94; 113; 124; 138–139; 141; 145–146; 158; 178; 186–187; 190). LCY conference reports reflected an overwhelming preoccupation with the loss of the federal party’s legitimacy in society, as had speeches at the Soviet and Hungarian conferences. As in the cases of the HSWP and CPSU, those sectors of society identified as most alienated from the LCY were the youth and workers. This is evident in all reports presented at the conference, which demanded measures be taken to combat the loss of faith in the party and re-engage these sectors of society (LCY, 1988: 25; 53; 62; 74; 92; 104; 118; 127–128; 160–161; 187; 189–190). With the exception of the Slovenian and Kosovan reports, all conference reports advocated renewed worker participation in the LC. This confirmed that the LCY’s intention at this time was to affirm the traditional communist party model with the workers as the vanguard of the party. What was most surprising was the apparent agreement among the diverse republican parties, which was certainly not representative of the very divergent views some parties had adopted over issues at this time. At interview, one respondent present at the LCY conference in 1988 stated that only very senior party members were selected as delegates. ⁹ The republican leader who most threatened the unity of the federation, Slobodan Milošević (of the Serbian LC) was conspicuously absent from the conference, and evidently resented taking directives from Boško Krunić, the federal party leader, because he was from Vojvodina. ¹⁰ Milošević strongly disagreed with federal party policy, over maintaining the independent status of Kosovo and Vojvodina, which he and many in his party considered should be subsumed into Serbia, and had already, some months before the conference, taken steps to change the Serbian Constitution accordingly. The Slovenian party, meanwhile, had a much more relaxed attitude than the rest of the republican parties towards the student opposition groups that had evolved and these received tacit support from the Slovenian party leadership (Lampe, 1996: 342–343). Acceptance of such groups was also tacit support for political pluralism, which the LCY conference reiterated that it did not condone, and sought, rather, to strengthen the leading role of the party. If dissent was expressed at the conference over these issues as we might expect, certainly the federal leadership organizing the event tried to minimize this by selecting delegates that they considered would be more conciliatory in their approach towards maintaining the federation, and without the ‘ringleaders’ of dissent, as it were, more junior delegates would be outranked by their colleagues.

Conclusion

This textual analysis was conducted before the interview and archival stages of the research into the conferences, and the results indicated similar efforts by leaderships in these states to broadcast a common message. These events were carefully staged to show all-party approval of a centrist line of reform, which was not necessarily embraced by the majority of members in these parties, and was apparently more the result of careful selection of the delegates invited to speak at these meetings. As one delegate at the Soviet conference quipped,

Comrade communists, delegates of the party conference, I will not keep you for long. I am already the fourth metallurgist and worker that has been allowed to speak today. This fact could serve to show really, either the extent to which the situation in metallurgy has become very problematic, and throws up a lot of issues, or the fact that we are speaking out more heatedly in favour of perestroika. (CPSU, 1988: v.1, 337)

Subsequent data from archives and interviews confirmed the results of the textual analysis, which in turn validates the method applied (Hardman, 2012: 72–268 passim). While archival and interview materials brought to light very different issues concerning these conferences, including evidence that they had been organized in synchronism (Hardman, 2012: 108–147 passim), the staged nature of these meetings and the surprising similarities between the leaderships’ policies, as advocated at these conferences would not have been evident without the textual analysis of the conference publications.

With respect to the question of reproducibility, that is, the validity of the theory generated and whether it is generalizable to other cases (the question of having my cake and eating it), the answer is yes and no. One could argue either way whether or not the research generated a ‘substantive grounded theory’ specific to one particular case, or a ‘formal grounded theory’ that has implications for other data in the same substantive area (Glaser, 2007). Clearly the theory does not apply generally to any conference or meeting. However, the research shows that the theory generated works for the particular institution of the communist party conference during the Gorbachev era. ¹¹ Generically, as an institution the conference (across the fraternal socialist states of Central and Eastern Europe) was described in similar, open-ended terms, which essentially granted the leadership of these parties the right to convene the conference between five-yearly congresses, to discuss pressing issues (Hardman, 2012: 43–44). The lack of formal rules defining the conference leaves open to question how this meeting differed in terms of its organization and representation from the other all-party meeting, the congress, which is defined in more concrete terms and had a decentralized process of delegate selection through regional organizations (Hardman, 2012: 43–49; 55–57; 162–163; 167; 169; 232–233). Textual analysis revealed clear patterns that views expressed appeared un-representative of the party membership’s different preferences at this time. Delegates who spoke at these meetings conveyed a clear message supporting the centrist–reformist leaderships and expressed the need to consolidate the party’s power to maintain the status quo of the party’s leading role in the one-party state. Moreover, Hungarian and Yugoslav conferences indicated a clear alignment with Gorbachev’s policies.

An example of how this model extends to other cases can be illustrated with that of the Polish United Workers’ Party conference, in May 1989, held a year after the three meetings in Hungary, the Soviet Union and Yugoslavia. Since the Polish conference was convened an entire year later, we might expect that this conference would prove an exception to the rule. Archival and interview research into the case of Poland was conducted after the main body of research had been completed. In the case of the Polish fraternal party, delegates’ speeches were not published in their entirety, and records of these were only available in the Polish national archives and comprised more than 500 pages of documentation. Computer-assisted textual analysis performed in the same way as the other cases was therefore not feasible in this instance, and instead required reading the stenographic reports in the archives and taking notes. I was more reliant, therefore, on interview and other archival data. On reading the text there appeared a similar homogeneity with the other conference records, which again suggested that speakers were chosen to convey a specific message. Similarly, the message appeared particularly anachronistic in the context of Poland in May 1989, by which time semi-free elections to the Parliament had already been organized for the next month, and the Party had already begun negotiating with the opposition the terms of power sharing. Although unable to examine the stenographic report with the same rigour as the Hungarian, Soviet and Yugoslav conferences, and evaluate views expressed with the same accuracy, a similar picture emerged. The vast majority of speakers at the conference rallied round Jaruzelski, advocating that the Party retain its leading role. Although Round Table negotiators had already agreed and arranged semi-free elections to the Parliament at which members of Solidarity would stand as candidates in opposition to the Party, speakers at this conference still queried the Party’s recognition of Solidarity in this way, advocating that the Party retain its leading role and consolidate its power (Hardman, 2012: 243–248).

As one interviewee confirmed, delegate selection for the conference was controlled by the leadership, whereas delegate selection to the other all-party meeting, the five yearly congress, was a more decentralized process. The leadership chose, therefore, to convene the party conference, as they had more control over its outcome. ¹² The evidence strongly suggested that the theory generated from the textual analysis of the other three cases would apply equally to the Polish case (Hardman, 2012: 79–92, 197–205).

With respect to evaluating the method of a GTM textual analysis, the validity of its application in this instance and whether the theory it generates has a predictive value, the research confirmed to me that it does have this capacity. This suggests that, in instances where fieldwork or archival research is not feasible, or access is limited, the analysis of relevant published text is a valid and worthwhile exercise, and, moreover, may present patterns that interviewees themselves might not be forthcoming about, or even aware of.