Grounded theory and the PhD – notes for novice researchers

Abstract

Purpose

This paper aims to consider the realities and problematics of applying a grounded theory (GT) approach to research, as a novice, within a mixed methods study during post graduate research. Its intention is to provide the novice user with a framework of considerations and greater awareness of the issues that GT can expose during research activity.

Design/methodology/approach

Using empirical evidence and a comparative approach, the paper compares the efficacy of both the classic Glaserian and Straussian models. It observes the effects of a positivist academic environment upon the choice of approach and its application. This study was specific to design education; however, its reliance upon a social science epistemology results in findings beneficial to research novices across broader disciplines.

Findings

GT presents the novice researcher with several potential pitfalls. Most problematic were the immutable, positivist institutional requirements, researcher a priori knowledge, the reliance upon literature for the research proposal and structure of the proposal itself. These include suspension of the notion that the purist use of either model can be applied in the current academic environment, the need for a close relationship with the data and toleration of a non-linear process with unexpected results.

Originality/value

The practicalities of GT research are often reflected upon by the academy, but use by novice researchers is little considered. The findings from this study provide a novel set of guidelines for use by those embarking on GT research and particularly where the requirements of formal education may cause a conflict.

Citation

Thurlow, L. (2020), "Grounded theory and the PhD – notes for novice researchers", Journal of Humanities and Applied Social Sciences, Vol. 2 No. 4, pp. 257-270. https://doi.org/10.1108/JHASS-05-2020-0079

Introduction

This paper considers the practicalities and problematics of applying grounded theory (GT) as a novice researcher during a mixed methods research project. Presented as a critical review of GT via a case study, it observes postgraduate (PhD) investigation into the nature of sketch inhibition among undergraduates within design higher education. The aim of the study was to build an effective theory of sketch inhibition as along with a set of pedagogic tools for its management in higher education – sketch inhibition is defined as a phenomenon whereby the suffer feels or demonstrates a reluctance or inability to engage with the mark-making aspect of design ideation and development (Author, 2019). This type of problem-solving and conceptual activity is also evident in broader environments including the social sciences, sciences and business. Although often criticised for its lack of formal epistemology (Doherty, 2015; Downs, 2017), the design disciplines have historically borrowed heavily from the social sciences, and an unintended consequence of this study offers methodological insight, not only to design research but also to other disciplines.

The following considers GT as a research approach and methodological framework (deleted phrase) alongside the personality of post graduate research. Its characteristics (both Glaserian and Straussian) are considered by the literature, together with a critical evaluation of their relationship. The benefits and problematics of GT applied within the mainly positivist environment of independent post graduate study are considered – from initial proposal, data gathering and analysis; thesis-writing; and identifying points along the research process where method particular attention or method slurring (Baker et al., 1992) were required. By way of conclusion, the findings from this have been developed into a set of considerations for prospective users of the method, intended as a decision-making tool for novices to GT research.

The personality of grounded theory research – comparison of the schools during project proposal development

Suddaby (2006) and Muratovski (2016) believe GT is the best used to observe a phenomenon where little extant theory is available: it “relies on the absence of an existing theory and its purpose is to set up a new theory,” (Muratovski, 2016, p. 99). Such lack of theory relating to the phenomenon of sketch inhibition drew the research towards an inductive process, and paradigmatically, GT. A constructivist approach was identified as the most appropriate – sketch inhibition being phenomenological and perceived at both the macro-level across discipline and by individual sufferers. GT, (Glaser and Strauss, 1967; Strauss and Corbin, 1990), offering both method and result (Bohm, 2004) in providing understanding (Furniss, 2011) of sketch inhibition was the most appropriate approach for observing the phenomenon.

Based on this initial understanding of GT research, and its apparent suitability for the study, further review of evaluation method was conducted. Being data-driven, GT study demands identification of an area of interest or research question to be investigated but no explicit methodology at the outset – this was perplexing; the antithesis of the requirement for post graduate study. GT and phenomenology paradigms occupying a close relationship within the social sciences, method slurring is often an unavoidable consequence of its use (Baker et al., 1992). The close relationship between research activity and data and the recursive nature of their method were noted by Glaser and Strauss (1967, p. 6): “most hypotheses and concepts not only come from the data, but are systematically worked out in relation to the data during the course of the research.” A notable trait of GT research is the requirement for analysis of data as it is collected, rather than as afterwards – this in contrast to previous activity, the body of data sought in their entirety before analysis begins. Identification and saturation of categories is data-driven (Muratovski, 2016) using an emergent approach to classification: pre-defined categories unnecessary and potentially harmful to the process. Both Glaserian and Straussian approaches to GT use constant comparison – the evaluation of new data against existing categories and development of new categories should these emerge during the process. Locke (1996) observing this recursive process necessary for subsequent growth of the research: “the materializing theory drives ongoing data collection” (Locke, 1996, p. 240).

Both approaches use theoretical sampling – the identification of further sources of data to be evaluated, these reliant upon the development of theory emerging from existing data (Suddaby, 2006). Via the process of constant comparison and purposive sampling, identification of clear categories and the relationships between them emerge from the data:

Categories or codes […] are the basic building blocks of a grounded theory. As they are developed, the same recursive, theory driven, comparative processes are used to surface the links and relationships among the categories to construct a complete theoretical framework (Locke, 1996, p. 241)

This would allow the data to drive the research and obviate the need for a preformulated methodology.

GT’s separation into two individual schools, to include Straussian in 1990 offered the research a choice: that of emergence versus forcing of data. The Glaserian model of emergence relied upon allowing the data to simply appear during analysis characterized by the separateness between researcher and the external world that incorporates their subject matter (Howell, 2013). Locke (1996, p. 241) considered the benefit of this, the Glaserian model favouring a passive, neutral approach, thereby avoiding contamination of pre-conception, providing a “one-way mirror” on the data: “categories emerge upon comparison and properties emerge upon more comparison. And that is all there is to it” (Glaser, 1992, p. 43). Locke (1996, p. 239) suggested the Glaserian model enabled researchers “to use their intellectual imagination and creativity to develop theories related to the areas of enquiry” through the gathering of naturalistic data. Borgatti (2020) suggested theory developed from such activity (deleted phrase) aims to “focus on making implicit belief systems explicit”.

The Straussian model’s (Strauss and Corbin, 1990) requirement of questioning the data to develop theory provides the researcher with a world view through a Constructivist lens, (Guba and Lincoln, 2005). Strauss and Corbin (1990) suggested both induction, whereby data is used to build a picture of a reality, and deduction based upon hypothesis testing, (deleted phrase) were intrinsic to such research and the act of conceptualization by the researcher would by default, involve deduction, suggesting interaction between the two was necessary for theory-building. With a dearth of knowledge around sketch inhibition, the methodological purity of the Glaserian model was attractive, and further considered.

Coding, according to Walker and Myrick (2006, p. 549), “transports researchers and their data from transcript to theory,” observing that both models use the same basic functions: “gather data, code, compare, categorize, theoretically sample, develop a core category, and generate a theory” (p. 550). Glaser’s (1978) approach involves two separate processes – substantive coding (fracturing the gathered data into categories based upon its properties), followed by selective, or theoretical coding; grouping into codes at the conceptual level and allowing the theory to develop as a result (Walker and Myrick, 2006). Strauss and Corbin (1990), coding is more complex, involving open, axial and selective coding, although individual stages are subject to blurring and can be used both sequentially and concurrently (Walker and Myrick, 2006).

Initial, open coding allows reduction of data into concise manageable themes that accurately reflect the phenomenon. Axial coding allows for interpretation of categories to be identified. Muratovski (2016) refers to Leedy and Ormrod’s (2010) questioning (provoking) of data to facilitate this. They ask, “What are the conditions that have given rise to this process? What is the context in which this process is embedded? […] What are the consequences of these strategies?” (Leedy and Ormrod, 2010, p. 143).

The final, selective coding stage involves the bringing together of categories and connections, their development into a storyline to describe the mechanics of the issue. According to Muratovski (2016), this is the point where theory can be developed. This more constructivist approach to coding was criticised by Glaser for being too aggressive, negatively affecting the research outcome. “Strauss’ sampling is controlled by the evolving relevant concepts, and relevance comes from testing out what is looked for, not what is emerging” (Glaser, 1992, p. 103). He believed this caused contamination of the analysis and favoured anticipation over emergence – that it could “force conceptual descriptions” as opposed to enabling the natural emergence of “grand theories,” (Glaser, 1992, p. 8). Borgatti (2020), however, endorsed the Straussian model for consisting of “a set of steps whose careful execution is thought to guarantee a good theory as the outcome”. At this point, the purity of the Galserian model was still considered the most appropriate for the study, enabling a natural emergence of knowledge about sketch inhibition, and an almost effortless development of theory.

The contentious nature of the researcher’s a priori experience illustrated how the Glaserian approach could be problematic to the study. Glaser’s (1992, p. 50) belief that previous knowledge was detrimental to effective theorizing – the researcher should, “just not know as he approaches the data,” was problematic, Suddaby (2006, p. 634) suggesting the negation of researcher experience, agenda and the literature impossible: “the researcher is a blank sheet devoid of experience or knowledge” being unattainable within any research scenario.

The Straussian paradigm allowed for, and even endorsed for their insight, the benefit of the researcher’s a priori experience and exposure to the issues under scrutiny – this including engagement with relevant literature. Potter (2006) believed unintentional researcher influence unavoidable, subsequent knowledge considered to be of a Constructivist epistemology. Baker et al. (1992) believed that fully understanding the realities of social or psycho-social situations within a GT study could only happen through observation, listening, inferring from the literature and reflecting upon one’s own experiences – effectively, everything could be considered data within a GT study – this, via the lens of the researcher. This more structured approach was potentially more manageable than a classic Glaserian style and appeared to provide a robust and justifiable route towards growth of theory (Wacker, 2008). Despite the unease between the two schools, Suddaby (2006, p. 635) maintains that GT offers, “a practical middle ground between a theory-laden view of the world and an unfettered empiricism”.

Using grounded theory to research sketch inhibition

The essentialist concept of measurability being vital to success in education – the award of academic qualifications impossible without it – creates an immutable environment for within which post graduate researchers must function. Such need for measurability requires, by default, a set of criteria to measure against, the research proposal being central to this. With most influence on the study was the requirement for a formal, developed proposal: this was completely at odds with GT and effectively precluded its use purest form – methodological concessions were already being made.

A review of the literature was necessary to frame the scope of the research and with so little literature referring directly to sketch inhibition, a wider search allowed context to be established. This continued for over a year, almost exclusively, to build a research framework robust enough to carry the primary research through to completion of the study. If the Glaserian model were to be observed, the use of the literature review together with empirical data would deem the data already contaminated. The Straussian model, by default, had become the approach to the study.

According to the institutional requirement, the research proposal was submitted. The initial aim was:

An investigation into the reasons for design students and early career designers avoiding manual drawing tools during design development and the proposal of a pedagogical framework to address this.

The objectives were presented, thus, as an investigation into:

• The stages of design development where drawing is used and an investigation into its purpose within creative development.

• Current practice of designers across a range of disciplines regarding their use of drawing techniques during design development.

• Reasons for students choosing not to use drawing as a tool for development and presentation – explicit reasoning.

• An investigation into the use of drawing as a tool for design development within HE – tactic reasoning.

• The position and value of drawing within current frameworks for design education – tacit reasoning (Author, 2015).

This proposal was problematic on several levels. As a statement of intent, it was far too complex. In contradiction to GT it made assumptions about the nature and extent of sketch inhibition and presupposed that it was indeed an issue. In addition to this, and in further contradiction to GT, the proposal required submission of a literature review and proposed methodology for data collection and analysis. The methodology was also submitted for ethical approval: pre-empting methods and samples was required despite there being little data upon which to base their need. Regardless of this, without such approval, the research could not have been conducted.

The formal review

At this point, a full literature review together with the developed methodology for data collection and analysis was required. The standard PhD model demanded the literature review prior to primary research. This again was in conflict with a GT approach. Based upon institutional requirements, the methodology presented for review was as follows:

Semi-structured interviews: Divided into two groups, with those who observed sketch inhibition, i.e. industry and education specialists; and with those who suffered sketch inhibition, i.e. undergraduates of design. The semi-structured approach was considered the most appropriate mode and accordingly, a standard operating procedure had to be developed and a set of questions designed.

Protocol analysis experiment and observation: to identify the symptoms of sketch inhibition among sufferers, a sample of inhibited students would complete an ideation task to be observed and coded. This was based upon similar methodologies of Suwa et al. (1998), Bilda and Gero (2005) and Kim et al. (2010) identified from the literature used to investigate designers’ processes. It was intended that data would be analysed using a coding system based on precedents set by Suwa et al. (1998) and Tang et al. (2011).

NASA TLX questionnaire: to be applied post-protocol analysis experiment to establish participants’ emotional response to the activity to provide data about the soft issues of sufferers.

Questionnaire and Delphi study: once a proposal for sketch inhibition management had been developed, this would be submitted for feedback to interview subjects from Group 1. In addition to this, the Delphi study (Hsu and Sandford, 2007) was intended to produce a normalised set of moderated pedagogic tools for use by educators (Author, 2016).

Getting it wrong

Based upon this proposal, the formal review was passed and progression to a PhD was approved. However, as a piece of GT research, the project was already failing: the methodology up to this point had been driven entirely by institutional requirements and not by the data. The remit of the study was the development of a theory of sketch inhibition and pedagogic framework; however, the proposed tools would not facilitate the constant comparative and purposive sampling essential to achieve this. In fact, the research process had developed into a series of box-ticking exercises to fulfil the requirements of the institution and understanding sketch inhibition had become subordinate to the research proposal. This was completely at odds with the aim and approach of the study and a watershed moment – the GT literature was revisited, the protocol experiment, NASA TLX questionnaire and questionnaire and Delphi study duly scrapped and restructuring of the project undertaken.

Getting it right

Based solely upon the emergence of issues from data, the interview method alone was kept, albeit in a form more reflective of true GT. The semi-structured approach was scrapped, instead, identifying issues to be discussed with subjects based, simply, upon the question, “what do I need to know about sketch inhibition?” This would be applied to the same two groups, i.e.; observers and sufferers of sketch inhibition. From this, further interviews were conducted, data coded immediately after each one, and emergent themes used to inform the next interview, i.e.; adding to the body of issues to be discussed.

The data

The interviews provided both data for the study, and insight into the problematics of conducting GT research. Digression was a common issue, particularly among industry subjects and often difficult to manage: if everything was considered as data within a GT study, to what extent could digression be allowed in case it offered up some new and unexpected insight? This was difficult to resolve – it also resulted in lengthy transcriptions and data extraction that were the most time-consuming part of the study.

Lack of structure during the interviews with students was particularly problematic. It was assumed the unstructured approach favoured in GT studies would elicit breadth and depth of data, but this was not the case: students were simply unaware of what they didn’t know. Lack of maturity and experience may have affected the way subjects responded, and it was evident their understanding of sketching and the design process was somewhat poor. The frustration of trying to tease out responses from some subjects created a tendency to ask leading questions – this had to be carefully monitored to avoid corrupted the data. Data from educators was very high and proved most valuable to the study. Constant comparison and theoretical sampling led to an interview with one subject whose data approved pivotal to the whole study: without using GT, this subject not have been identified.

Always a conundrum for qualitative research, interview sample size was surprisingly simple to establish. Where the literature offered a plethora of notions about this, GT made it simpler: the interviews continued until no further new issues emerged from the data. Instead of an arbitrarily-set sample, constant comparison enabled identification of the point of saturation.

Data management

Depth and breadth of data during GT research is difficult to predict, Fassinger (2005) noting the complexity of data handling as potentially problematic. NVivo software was used throughout the study for storage, management, coding and analysis, thereby mitigating some of the complexity observed by Charmaz (2000). NVivo’s graphic tools enabled visual macro-analysis of the data – this, essential for interrogating the quantity of data generated by the study. Charmaz (2000, p. 520) suggested that such software had a tendency to “unintentionally foster an illusion that interpretive work can be reduced to a set of procedures”. This did not, however, appear problematic: emergent themes rather than software parameters were the driver of data handling.

Data analysis

The coding process, “identifying patterns and discovering theoretical properties in the data,” (Bowen, 2008, p. 144), adhered to the Straussian method, initially developing open coding. Individual nodes were created as they emerged from the data, observing Borgatti's (2020) “nouns and verbs of a conceptual world.” Boyatzis’ (1998, p. 161) definition of a theme was observed as closely as possible; “a pattern in the information that at minimum describes and organises the possible observations and at maximum interprets aspects of the phenomenon” – (process illustrated in Figure 1).

A hierarchy of themes emerged as coding progressed. Meta-themes became structured into parent nodes, for example, “cognitive issues” and “definitions of sketching.” As new interview data were analysed, additional themes emerged and iterative (constant comparison) process of revisiting already coded data to code for new themes was conducted. And so, the number of parent nodes increased, as did child nodes within these. Braun and Clarke’s (2006) method was also observed: to reduce loss of context during coding some of the surrounding data was kept: whole sentences and sometimes paragraphs relevant to the theme were coded to maintain clarity of meaning. Multiple coding also formed part of the constant comparison process – coding data as many times as necessary to ensure it was coded into all nodes it related to. Throughout the coding process, axial coding, using mind mapping techniques, identified further issues within and between themes, according to Walker and Myrick (2006, p. 553), to “understand categories in relationship to other categories and their subcategories” (Figure 2).

Selective coding, “the process of selecting the central or core category, systematically relating it to other categories, validating those relationships and filling in categories that need further refinement and development,” (Strauss and Corbin, 1990, p. 116), began towards the end of the data gathering process. This underpinned the structure of findings and their presentation as a narrative of sketch inhibition (Figure 3).

Theoretical sampling

Where theoretical sampling offered efficiency to the study, the lack of time to research new methodologies was problematic. During coding, the potential benefit of a learning style survey emerged. Responses from the interviews with sufferers of sketch inhibition suggested that there may be a link between inhibition and learning preference or learning difference. As such, a new data gathering methodology was applied. Similarly, the interview data suggested a possible issue among sufferers of inhibition and their employability – the benefit of a longitudinal study emerged.

The findings from the learning preference study were valuable to the study; however, the longitudinal study failed to gather any purposeful data: the GT approach of developing methodology according to emerging need was proving problematic. The fixed timeframe of the study prevented the development of an effective methodology and its application in an effective way. Instead, a rushed study with limited sample, based upon revisiting interview subjects via email was applied, very unsuccessfully.

Thesis structure

The thesis, in traditional PhD study, requires a linear set of content to be presented for examination. A product of the positivist tradition, such structure tends to favour the sciences. This is endorsed by the institution’s Code of Practice for Research Degree Students (De Montfort University, 2018, p. 52), which describes the structure of “a conventional dissertation”. Additionally, the mandatory training modules provided by the doctoral training programme, specifically, Structuring and Completing Your Thesis (De Montfort University, 2020), further validate this, describing the required format for thesis presentation (Figure 4).

Despite pouring through many theses during the course of the study in search of non-traditional formats, these requirements appear to have never been challenged. It was tempting to present the study in a non-linear format truly reflective of Grounded Theory, but too much was at stake and thus a version of the traditional structure was submitted.

Positivist issues for grounded theory research

A typical PhD taking between three and seven years to complete, timeframe is an immutable factor and certainly impinged upon this study. Without the limits of time, a truer reflection of the possibilities of GT research would have been achieved. The study would have continued as long as was necessary, the data and findings growing far beyond those presented in the thesis. Time restrictions were a constant issue – the joy of observing the emergence of a new issue to research, coupled with the lack of available time to investigate a potential methodology was problematic. This was particularly apparent during the learning style questionnaire and longitudinal study. Despite this, it was also an essential mechanism for the study – an ensuing deadline guaranteed to sharpen the mind. Although positivism could be criticised for placing restrictions upon the study it would have looked very different without it – and not necessarily for the better. PhD requirements actually lent a beneficial framework to the research, structure providing helpful boundaries to work within.

Henwood and Pidgeon (2003) believe, “The excitement and challenge of GT is finding a way out of its maze, but there is no one legitimate way out of the maze,” and GT research is certainly nothing if not complex. Fernández and Lehmann (2005) considered creativity an important part of GT research coupled with the need to conceptualise to develop theory from the data. They also believed the researcher should be able to tolerate confusion, and sporadic regression of the research process. These factors were certainly reflective of the study, challenging traditional linear approaches to previous projects.

GT appears to relate closely to Complexity Theory, (Kuhn, 2008; Wang, 2010). Although in its infancy, this could provide a paradigm for the future of design education, and potentially benefit research into creative issues – both approaches able to accommodate complex, creative, non-linear systems and emergence of unexpected data. Despite the methodological and epistemological benefits that could accompany this, such a reality is probably distant, and the shoe-horning of non-positivist endeavours into positivist structure would have to continue.

Conclusion

Novice use of GT can be fraught with complexity and initially perceived as in-compatible with traditional post graduate research. However, a version of such an approach within a finite research structure is possible and very rewarding. The Glaserian model suffers most as a result of institutional requirements. Its purity is compromised from the outset by researcher prior knowledge and the use of literature during the proposal development stage (deleted phrase). The Straussian version is more accepting of the realities of constructivism – and more forgiving of research structure. Despite this, the research proposal and ethical requirements of contemporary research projects have a huge impact upon such a study.

The dual approach of constant comparison and theoretical sampling of both models are invaluable. They enable close observation of the study, both in terms of the data analysis and as a tool for the management of processes. They also support the researcher in dealing with novel and unexpected findings – the greatest joy of conducting research.

Based upon the observations of the case study, a set of considerations is offered for the novice researcher:

Understand what GT is about before you start. Make sure you fully understand its purpose and nature before you embark on research of this type: its remit is theory-building within an area of lack. It may be easier and less stressful to embrace other paradigms with more structured methodologies. The question is how hard do you want to make it for yourself?

It is almost impossible to apply GT in its purest form during post graduate research; identification of an issue to investigate is not enough for many institutions, (or funding bodies). They require certainty, a developed proposal with clear objectives and a methodology early on in the process, this being essential for ethical approval. This will require considerable reference to literature, understanding of GT and possible negotiation over the proposal before starting. At this point, theoretically, your research is no longer GT, but becomes a hybridized, institutionally acceptable form of the approach. Live with this, as there is little you can do.

Know the research will grow – it is not linear. GT research, being data driven, relies on the last piece of data to inform the next activity, (constant comparison and theoretical sampling). This implies a degree of flying by the seat of one’s pants, and allowing yourself to be taken wherever the data dictates. Where other types of research can be planned, GT is different and may lead to heated discussion with supervisors over matters of project management.

Allow the data to drive you at all times. This is almost a mantra when conducting such a project. The urge to lean towards a highly structured proposal is huge, especially during times of isolation and hopelessness that characterize PG research. Keep in mind that theory while covering unchartered territory is never going to be easy.

Being data driven, rather than relying on prescribed samples, data saturation can be easily identified using GT. Theoretical sampling is also efficient for focussing effort where it is required. The close relationship between research data and research activity allows this – and why coding data at the point of collection is so important.

Time will be problematic. Researching entire new methods of data gathering and applying these effectively may be problematic. It is somewhat of a Catch 22 for the researcher: such methods cannot be fully investigated and piloted prior to a GT study, as the data has not guided you there. However, during a GT study, getting to grips with unfamiliar and unexpected methodologies takes time. This can result potentially, in poor application and results of little benefit to the research.

It is not tidy. If you prefer a clear, highly managed approach to research, GT may not be your bag. The snowball effect of research and data growing in different and unexpected directions at the same time can be overwhelming. A pragmatic disposition is required in this situation – the ability to detach necessary to maintain control of the process.

Positivist factors should be embraced – timeframes, deadlines and structure are the antithesis of GT, but without them the novice handler will struggle to maintain focus and momentum.

GT research offers a steep learning curve and the balancing between immersion in the data and maintaining objectivity. If all these factors do not deter the novice researcher, such projects can be creative, exciting and hugely rewarding.

Figures

Figure 1.

Emergence of initial themes from the data according to Strauss’ method of open coding

Figure 2.

Model of axial coding using mind mapping techniques

Figure 3.

Selective coding to build the narrative of sketch inhibition

Figure 4.

Required thesis structure