The grounded theory method provides a logically consistent set of data collection and analysis procedures that can be used in textual analysis. The original aim of the grounded theory method was to develop a research methodology that would systematically derive theories about human behaviour from empirical data. The grounded theory method seeks to discover what is going on. Typically, it is applied to texts obtained by interview, observation or other data collection methods. Explicit in the use of grounded theory is data collection from participants who may have different views of the phenomena being studied, and which must be accommodated in the development of theory. There is an apparent resonance here with Ingarden’s concepts of schematised aspects held in readiness, and transparency of represented objectivities varying between individuals.

Since its original announcement, grounded theory has become an accepted qualitative research technique in information systems (Urquhart, 2001). Its adoption has probably been assisted by its originators’ assertions that it is ‘not bound by either discipline or data collection’ (Glaser, 1992). Grounded theory is concerned with the identification of categories, or properties of categories, as a major issue (Glaser, 1992). While grounded theory was described as having emerged from symbolic interactionism, Annells (1996) classified it as being ontologically based in critical realism. Glaser and Strauss (1967) state ‘our position is not logical; it is phenomenological’. Grounded theory can therefore be considered to be framed from an ontological perspective that is not in conflict with the realism of Ingarden.

At this point it should be noted that, while we are seeking to develop a methodology where all components share a consistent philosophical perspective, the content of the papers being analysed would not be limited to this perspective. The nature of the philosophical perspective used and reported on by researchers in the papers studied would be a value assigned to a category in our analysis.

Reviews and examples of the use of grounded theory in information systems research can be found in Calloway and Ariav (1991), Pidgeon et al (1991), Hughes and Howcroft (2000), Urquhart (2001), Fernandez (2004) and Lings and Lundell (2005). Hughes and Howcroft (2000) review a number of uses and adaptations of the grounded theory method by information systems researchers. After considerable discussion of the implications of the way in which information systems researchers have applied grounded theory, and the views of the originators, Strauss and Corbin, Hughes and Howcroft (2000) argue against the rigid application of the grounded theory method in practice. They also note that ‘[i]f the research community is to mature then it would be of far greater benefit to tell the story as it were, and this should include the researcher’s perspective, actual use of the method, and a reflective evaluation’. Further, they assert that ‘the adoption and diffusion of the method should be welcomed since it represents its usefulness as a pragmatic tool for research’.

Urquhart (2001) provides considerable detail regarding an instance of using grounded theory method to examine client — analyst interaction and behaviour, and also lists a number of IS researchers using the grounded theory method. One instance where grounded theory was used (Orlikowski, 1993) received MISQ’s Best Paper Award in 1993. Fernandez (2004) also provides a background discussion of the grounded theory method, plus a report of the experience of using the method. He agrees with many of the points made by Urquhart (2001). Fernandez (2004) employed the software package ATLAS.ti to assist with his analysis and makes some cautionary points about the limitations and negative aspects of software assisted coding, and in particular that the potential for automatic coding (e.g. coding all occurrences of a word or phrase) can have a negative effect in obscuring discovery by the researcher. This confirms the cautions given by Glaser about hiring coders and the use of automated systems, which remove the analyst from close contact with their data (Glaser 1978).

At this point it should be noted that grounded theory has come to include two divergent approaches. We now distinguish between those two approaches and the degree to which they suit the needs of this research.

In 1992 Barney Glaser published Emergence vs. Forcing: Basics of Grounded Theory Analysis (Glaser, 1992) in which he set out a comprehensive and vigorous attack on the contents of a 1990 book written by his original collaborator, Anselm Strauss, and Strauss’ research partner, Juliet Corbin (Strauss and Corbin, 1990). Glaser considered their book ‘distorts and misconceives grounded theory method, while engaging in a gross neglect of 90% of its important ideas’ (Glaser 1992). In the years since this conflict, researchers have found both approaches useful since their different emphases make them more or less appropriate in specific research settings (Fernandez, 2004). The Glaserian approach is described as abstract conceptualisation, and the Straussian approach as full-description (Fernandez, 2004). This difference relates to a disagreement regarding the unit of analysis. The Straussian approach emphasises word by word analysis (Strauss and Corbin, 1990), while the Glaserian approach deals with units of meaning at the line or sentence level (Glaser, 1978). Strauss and Corbin give an example (1990) of taking an hour to discuss what an individual meant by the word ‘once’.

Our purpose in undertaking this analysis is to discover what research is being reported in information systems journal papers. As described above, the purpose of such papers is the transmission of cognitive knowledge, a concomitant of which is for the papers to be written using represented objectivities that are, ideally, transparent. In this case then, there should be little need for extensive analysis of individual words and their possible meanings. Our aim is to abstract, rather than commentate. The Glaserian approach therefore appears to be more appropriate in this study.

Another significant difference in the two approaches to grounded theory, for our purposes, is the method of coding data. The Strauss and Corbin method requires that all data be coded against a single coding family — context, conditions, action/interactional strategies, intervening conditions and consequences. The Strauss and Corbin coding family is, however, clearly inappropriate for ontological analysis. This coding family was a variation of only one of eighteen coding families proposed as a significant part of the book Theoretical Sensitivity by Glaser (1978). Several of those coding families refer to ontological and mereological concepts (e.g. The Dimension Family, The Type Family, The Theoretical Family and The Conceptual Ordering Family — see Table 1). The list presented by Glaser is inclusive rather than exhaustive, and it is clearly intended that researchers using this method could derive their own coding families (Glaser, 1978) and this occurs in practice (Urquhart, 2001). Glaser expanded on his original list of coding families in two later books (Glaser, 1998, 2005).

We propose to use Ingarden’s ontological categories, defined in his framework, as the basis of a coding family for the grounded theory method coding technique. Further categories dealing with matters relating to publication, other than what is contained within the papers themselves (time to publication, reviewing status, intellectual property status) and which were suggested in Lamp (2002), may be added.

It should be noted that while this is not fully in accord with either the Glaserian or Straussian approaches, it draws on aspects of both. The unit of analysis and aim of conceptualisation tend towards the Glaserian approach. The use of a predetermined, rather than emergent, coding family is a Straussian feature, but the coding family is not that prescribed by Strauss. For the purposes of this research, we are seeking a technique that can be applied to discovering concepts within scientific works, and which is philosophically compatible with Ingarden’s framework. The coding technique employed in grounded theory would appear to offer this, but we do not claim that what we are undertaking is grounded theory since it is not in accord with either the Glaserian or Straussian approaches. We appropriate the coding technique from grounded theory and adapt it to the ontological analysis of text.

The process for coding where abstract conceptualisation is sought can be considered to consist of two phases: substantive coding and theoretical coding. Although described here sequentially, they are not executed in this way.

Substantive coding itself consists of two phases: open coding and selective coding. In open coding the analyst aims to ‘generate an emergent set of categories and their properties which fit, work and are relevant for integrating into a theory’ (Glaser, 1978). Units of meaning are examined and coded against as many categories as may fit. New categories emerge, and new units of meaning fit existing categories. In undertaking open coding the analyst considers three questions (Glaser, 1978):

The first question serves to remind the analyst that the data may not match any preconceptions held by the researcher. The second question serves to remind the analyst to consider codes already used. The third question serves to remind the analyst to consider whether a particular code might be a core category.

Selective coding occurs when the analyst identifies core categories and limits coding to ‘those variables that relate to the core variable in sufficiently significant ways to be used in a parsimonious theory’ (Glaser, 1978).

Theoretical coding uses the coding families, in our instance based on Ingarden’s ontological categories, to ensure that the analyst works at the conceptual level, writing about concepts and interrelations, rather than being bogged down in the data. The coding families assist in ensuring that the analyst has not overlooked the dimensions of a particular approach to understanding the data. Coding continues until the main concern of the research can be accounted for and further coding fails to add significant value in categories or properties. At this point theoretical saturation is deemed to have been achieved.

It should be noted that while the description has necessarily been presented as a sequential set of steps, it is by no means this simple. Glaser (1998 cit Fernandez, 2004) describes it as a method that ‘happens sequentially, subsequently, simultaneously, serendipitously and scheduled’. A model of the grounded theory method proposed by Fernandez (2004) is shown in Figure 2, and demonstrates this complexity. A detailed explanation of the Fernandez model is, however, beyond the scope of this paper.

Analysts applying this technique will use memos and other tools to aid the abstraction process and to provide a trail of decisions made while applying the technique. The output is expected to be a web of terms linked with high level categories. The terms will aid searching and be based on meaningful concepts that have a relationship with the texts making up the relevant body of work.

Two questions automatically arise. First, what is the group for whom the texts are meaningful? Second, how durable is the categorial scheme produced? Recall that we are interested in categorising scientific works based on the terms and concepts that exist in the work. Thus, we address these questions in that context.

Regarding the first question, within a readership there may be several distinct communities of practice each of which brings different perspectives to reading the same works. The major influences are the different unfulfilled schematised aspects held in readiness by the readers from these groups. To adequately support such a heterogeneous readership, an application of the technique is required for each group. This would provide a consistent analysis reflecting the different perspectives of the readers. It could begin to capture the diversity of perspectives that exist in information systems. This is significant because it is accepted that the nature and scope of the information systems domain are diverse; the approaches to researching information systems are diverse; the approaches to teaching information systems are diverse, and there is a lack of any single clear theoretical basis underlying the study of information systems (Lamp and Milton, 2004).

With the second question, one can say that as disciplines mature and research progresses, our needs and readings of scientific works will change. For example, early theories about the structure of atoms are now largely historic and no longer have the currency they had when first published in the late nineteenth century. Clearly, however, the work itself does not change. What changes is the context of research and research theory, and also the needs and goals of the readership. Thus, the technique will need to be periodically reapplied to maintain durability.

Nevertheless, we can say that the approach is theoretically feasible. We say this because, firstly, we have the basis of a coding family based on a sound ontological theory of research texts. Secondly, we have a methodical well-respected qualitative research method designed to meaningfully analyse text to produce conceptual categories. Together, we therefore have an approach designed to produce terms (categories) that are meaningful, emergent, and relate to high-level ontological categories for research texts.