Having outlined the key theoretical features of Q methodology, this section
examines Q as technique. Given Stephenson’s stance of Q being outside of existing methodological paradigms, Q theorists refer to Q as a technique rather than a method (for example, Brown, 1991; Watts & Stenner, 2012; Wolf, 2012).
Stephenson was a research assistant for the British psychologist Charles Spearman who (with Karl Pearson) developed the method of correlation used in regression analysis (R) and was greatly influenced by him (Stephenson, 1993/94). Factor analysis is a data reduction technique whereby “a number of tests (variables) are applied to a sample of persons” (Stephenson, 1953, p.15) to determine whether the variables are inter-related. The underlying relationships (which are not dependent on each other) are referred to as factors (Bryman, 2008). Stephenson (1953) reconceptualised Spearman’s and Pearson’s method to enable by-person factor analysis, that is, “an experiment (is designed) in terms of people . . . to assess qualities of performance with respect to each person in turn, and then to make correlations between people” (p.16). Stephenson used the letter Q to distinguish his approach.
Put simply, in R, variables are characteristics (test scores, traits) of a person and factor analysis looks for which characteristics go together. In Q, variables
are the people who participate in a Q study. Factor analysis is used to group participants together according to some underlying dimension of commonality in their viewpoints (Wolf, 2012). This enables shared meanings or viewpoints to be identified (referred to as orientations) and the extent of each participant’s association with a particular orientation. (This correlation is analogous to the step in R that correlates, for example, tests results on maths and music achievement) (Wolf, 2012).
Despite sharing the same statistical techniques, R and Q are based on very
different epistemological paradigms and have dissimilar purposes. Drawing on the literature, the key differences between R and Q are summarised in Table 6.1.
Table 6.1 Differences between R and Q
R Q methodology
Purpose The objective analysis of a topic of interest. Identifies the structure of opinion or attitudes in a
population of interest (Spearman, 1904).
“The purpose is to reduce and/ or eliminate the qualitative and subjective” (Stenner, 2011, p.198).
Reveals differences in points of view, attitudes, opinions about a topic of interest. The focus of Q is on “the constructions, rather than the constructors (participants)” (Stainton Rogers, 2005, p.180). “The purpose is to maximize the qualitative and subjective” (Stenner, 2011, p.198). Logic “Hypothetico-deductive” (Stephenson, 1953, p.17). “Postulatory-dependency” (Stephenson, 1953, p.17). Participant selection
A representative sample of the population of interest.
One or more individuals who have been selected using purposive sampling.
What is being collected
The degree to which a person has a certain trait/characteristic (assessed one at a time)
During a Q sort, the participants “put meaning upon and draw meaning from the statements” in the sort (Wolf, 2008/09, p.27). In this way, Q provides access to “the unrestricted viewpoint of its participants” (Watts, 2011, p.45).
How data is being collected
By use of a standardised data collection instrument, such as a test or survey with fixed categories.
By a multi-item comparison and ranking/scoping on a grid.
How the data are treated statistically
In R, the measuring units are “objectively scorable traits” (Brown, 1980, p.19). In R, correlation summarises the relationships among the traits, and factor analysis identifies the clusters of traits.
Unlike R, “there is no common unit of measurement other than the person’s self-referential viewpoint” (McKeown & Thomas, 2013, p.48). This means that Q involves the correlation and factoring of people - correlation summarises the views among the people, and factor analysis identifies the clusters of people with shared views (McKeown & Thomas, 2013).
Outputs The outputs of R describe the characteristics of the sample population that are statistically associated with the topic of interest.
The outputs of Q reveal factors or clusters of viewpoints about a topic of interest.
The rest of this section gives a brief overview about how a Q study is undertaken to provide a context for the following section describing how I conducted my Q research.
The first task for the Q researcher is to collect a large number of statements representing the range of discourses about the research topic (pictures, objects or sounds may be used as an alternative to written language). Using a matrix (or other framework) of themes/subthemes in the discourses, the researcher sorts the individual statements by theme/subtheme and then systematically selects statements which together provide “a representative miniature” of the larger concourse (Brown, 1991, p.6) (referred to as the Q set). Wolf (2012) notes that if the Q researcher is already very familiar with the research topic, they may create the matrix (or other framework) first and then collect statements that fit within it. Recruiting participants for a Q study is done by purposive sampling to provide a differentiated sample (Wolf, 2012). Participants (referred to as the P set) are asked to rank each of the items in the Q set according to a specific instruction (referred to as the condition of instruction) and an ordinal ranking scale (such as from - 4 to +4) (Watts & Stenner, 2012). Participants are further guided by the requirement to sort the statements according to a predetermined distribution (the degree of flatness or steepness of the distribution is referred to as the kurtosis) (Brown, 1980). This sorting and ranking of the Q set by an individual is referred to as a Q sort.
Through their ranking of the items, participants express their viewpoint on the topic (McKeown & Thomas, 2013), “tell a story” (Stainton Rogers, Stenner, Gleeson & Stainton Rogers,1999, p.249) or provide “a picture (of their) . . . conception of the way things stand” (Brown, 1980, p.6). A participant’s subjectivity is expressed in how the items are understood and how they are ranked (Brown, 1991).
The Q sorts undergo correlational and Q factor analysis (using Q software, in this case PQMethod) to identify statistically significant patterns of associations (referred to as factors) and the extent of each participant’s association with a particular factor. Stainton Rogers (2005) describes the distinctiveness of a factor: “Each factor represents a fully alternative understanding of the topic of interest” (p.191).6 The researcher’s task is then to interpret each of the factors expressed as a factor array. Wolf (2012) identifies two broad stances researchers may use to interpret patterns from Q factor analysis. Firstly in a person-centred Q study, the researcher enquires into the ways in which people view a matter from their perspective and the underlying predispositions that may influence a person’s response to the items in the Q sort. In a discourse-centred Q study, the researcher is interested in the discourses with which participants align. My approach is person-centred, examining how evaluative reasoning is understood by professionals undertaking public sector evaluation.