Q Methodology, also referred to as Q-Sort, is a research method used to investigate the ‘subjectivity’ of participant’s viewpoints on a specific topic.
This is achieved by having participants rank and sort a series of statements, known as a Q-Set.
Q Methodology was first developed by psychologist William Stephenson (1902-1989) and is used in clinical and research settings.
Q Methodology provides a means of exploring qualitative, subjective perspectives but using quantitative techniques to do so.
This is valuable for the researcher as it provides a level of transparency around the data collected which is not present when using most other qualitative research methods.
Some of the main reasons for selecting Q Methodology may include:
Before progressing through this guide to Q-Methodology it is advisable to become familiar with some of the key terms used within Q-Methodology. This is not an exhaustive list of key terms but instead a selection of terms which are frequently used in this article.
Over time Stephenson developed an interest in research methods and in turn developed an interest in psychology. These interests saw him undertaking further study at the University College London (UCL).
At UCL, Stephenson found himself under the teaching of Charles Spearman who was a pioneer of factor analysis. Stephenson received his second PhD, this time in psychology, in 1929.
Stephenson is best known for the development of Q-Methodology and the associated alternative form of factorial analysis.
He was concerned with the science of subjectivity and in 1935 Stephenson published his first paper on Q Methodology in the prestigious Journal, Nature.
As a result of his work on Q-Methodology and other prominent research, Stephenson was made Assistant Director of Oxford’s Institute of Experimental Psychology.
He was forced to take a break from his position during the Second World War when he joined the British Military, serving in India as Brigadier General.
After the war he returned to Oxford but left soon after in 1948, joining the University of Chicago.
It was during his period at the University of Chicago that he progressed his work on Q-Methodology and published: The Study of Behavior: Q-Technique and Its Methodology (1953). This is the work for which he is best known.
In 1955 Stephenson left Chicago, and after a short period working for an advertising company as Director of research, he returned to University life by joining the University of Missouri in 1958.
His final move was to the University of Iowa before retiring in 1977.
Even after his retirement, Stephenson continued to write and speak on Q-Methodology and the study of subjectivity.
There are seven stages of Q Methodology:
Below is a basic summary of the seven stages of Q Methodology but this will be followed by a more detailed exploration of each stage individually, providing an overview of the key considerations for the researcher.
When devising the statements that are to be used for the Q-Sort, it is essential that they cover the full range of viewpoints that the participants may have on the topic being explored.
This will ensure that every participant is able to express their own views on the topic, irrespective of what those view might be.
This is achieved through the development of a concourse.
The concourse was considered by Stephenson as being one the fundamentally essential aspects of conducting a meaningful q-sort.
In fact, Stephenson considered concourse, in general terms, to be an extremely important component of progressive thinking. This is reflected in the following quote.
The term ‘concourse’ refers to the collection of possible statements people may make about the topic being investigated.
The concourse should aim to include all perspectives and viewpoints that a potential participant may hold on the topic.
No statement should be considered irrelevant or not applicable and every participant should have their own personal viewpoint represented within the collection of statements.
The following quote, by Van Exel and De Graff (2005), summarises what a concourse is.
A concourses can be established using two different methods:
Naturalistic concourses are typically obtained from the participants in the form of verbal or written information.
This may come from communications with the participants or via questionnaires, interviews or focus groups.
In practice, the researcher would use structured feedback from the participants in order to develop a ‘naturalistic’ concourse for the topic.
Ready-made concourses are obtained from existing published information, such as books, newspapers, journals, websites or social media.
The researcher can use these sources to identify a suitable concourse which covers a broad range of viewpoints on the topic.
If is often the case, however, that the researcher will use a combination of naturalistic and ready-made sources of information to establish a concourse.
The Q-Set is a selection of statements which have been drawn from the concourse, and which should be representative of all viewpoints included within the concourse.
The Q-Set is of critical importance in allowing participants the ability to express their onw personal viewpoint on the topic. However, the creation of the Q-Set is entirely at the discretion of the researcher and is considered as much an art as it is a science.
The statements included in the Q-Set should be subjective opinion statements that have the potential to generate feelings, emotions and views regarding the topic.
It is expected that the researcher will be able to explain and justify their choice of statements.
The researcher should therefore consider the following guidelines proposed by Webler, Danielson and Tuler (2009) on what makes a good Q-Sort statement.
Due to the subjectivity of the researcher’s selection of statements there is a need for continuous revision and testing of the statements to refine the Q-Set and ensure it is fit for purpose. this will typically involve piloting the statements and obtaining feedback from the participants in order to make amendments.
When the researcher gets to the point where feedback from participants is adding little additional perspective on the topic confidence can be taken that the statements cover al viewpoints associated with the topic.
There is no rule for the total number of statements that should be used for a Q-Sort. It ultimately depends on the complexity of the topic and the type of participant who will be completing the Q-Sorts.
In their book, Doing Q Methodological Research: Theory, Method & Interpretation, Watts and Stenner (2012) propose that 40-80 statements would be expected. However, there are examples where over 100 statements have been used and other examples with much less.
Ultimately, the needs to be a sufficient number of statements to capture all associated viewpoints, so too few would not be sufficient, whilst also considering that an excessive number of unnecessary statements may reduce the motivation of the participants to maintain engagement throughout the entire Q-Sort.
As long as the statements fully cover the viewpoints of the topic, you can justify the statements you have selected, and you can justify the number of statements being used, you will have a sound Q-Set.
There are a number of alternatives to statements that can be used for a Q-Sort.
The researcher could use single words, images, sounds, smells, tastes, objects or anything else they feel is appropriate for the purpose of the study.
It is the ranking of the items that is of importance, not what those items are.
Overwhelmingly statements are used when conducting Q-Sorts, but using these alternatives can provide good options when dealing with participants who may find written statements challenging to deal with, such as when working with young children, images may prove to be a useful method.
Similarly, if the research is exploring participants preferred food combinations, rather than writing these combinations on paper the researcher could provide samples for participants to taste.
As with many aspects of Q-Methodology, the researcher has control over how they structure and manage what is to be sorted and ranked.
However, for the remainder of this article we will be referring to ‘statements’ as they are most commonly used in Q-Methodology. But when statements please be aware that the term statements could be replaced with any of the alternatives suggested above.
The third stage of Q-Methodology research involves the selection of the Q participants, collectively known as the P-Set (with P standing for Participants).
The selection of the P-Set is not achieved randomly, instead participants are deliberately selected to ensure the P-Set is as heterogeneous as possible. In other words, the researcher wants the P-Set to be diverse in viewpoints and character.
Exel and Graff (2005) explain that Q-Methodology promotes the use of a structured sample of respondents who are theoretically relevant to the problem under consideration. For example, this would include all relevant people who would have a clear and distinct viewpoint on the topic.
As the participants in Q-Methodology are the variables, and not the samples, it is not necessary for the total number of participants to be excessively large.
Brown (2003), student of Williams Stephenson, suggests that typically no more than 40 participants is necessary.
More recently, Webler, Danielson and Tuler (2009) proposed that the typical number of participants required for an effective Q-Methodological study is between one and three dozens (12-36).
So it is clear that even the experts of Q-Methodology appear to propose slightly differing and wide ranging approximate numbers of total participants.
So, although I cannot provide an exact number of total participants, it is considered appropriate to use a total participant number in the tens rather than the hundreds or thousands.
It is the researcher’s responsibility to include participants that cover the full breadth of the concourse being explored.
The participants who are included will therefore depend on the problem being explored.
For example, if the researcher is aiming to explore different people’s viewpoints on a political topic, it is important that the participants include people with different political backgrounds, from different areas and regions, ages, genders, etc.
Alternatively, if the researcher is aiming to explore the political views of recent College graduates then the breath of the participants may narrow slightly to meet this specific research need.
However, for this example, the researcher would still need to ensure the recent college graduates selected still offer a range of viewpoints and sufficiently varied enough to cover the concourse being investigated.
To summarise, the participants should enable the researcher to fully explore all viewpoints associated with the research topic being investigated.
There are three main elements that need to be considered when conducting a Q-Sort.
The Q-Grid can be considered as the ‘score sheet’ for the Q-Sort.
With in reason, the Q-Grid can take any shape the researchers believes will best suit their specific research study.
Traditionally, Q-Grids tended to have a quasi-normal distribution (a U-shaped normal distribution).
However, over time Q-Grids have become flatter in their structure.
The shape and structure of the Q-Grid is determined by the researchers and will be largely dependent on the types of responses they are looking for as well as the number of statements being ranked.
For example, if the researcher wants to find out which single statement is the most salient to the participants they would likely create a Q-Grid that has a single option at each extremity. Alternatively, if the researcher is less interested in specific single statement rankings and would prefer to explore a broader placement of statements then a flatter structure might be adopted.
Steven Brown, in his online workshop videos (found at the end of this article), explains that the structure of the Q-Grid has practically no impact on the statistical analysis and outcomes of the Q-Sort. It is therefore an aspect of Q-Sort that the research has freedom to manipulate to best meet the needs of their study.
Once the structure is in place, the researcher will then need to assign a score or value to each column contained within the Q-Grid.
This typically takes the form of the middle column being 0 and the columns to the left increasing as positive numbers and the columns to the right decreasing in negative numbers (as shown in the diagrams above).
As with the structure, the scoring of each column is determined by the researcher and will generally follow a natural increase/decrease in numbers until all columns have a value assigned.
The final consideration in this section is the condition of instruction.
This is used during the Q-Sort process and contains a short introduction to what the study is about, the question that the participants will be answering/exploring, as well as clear instructions on exactly how the participants can complete the Q-Sort.
Also included in the condition of instruction are the relative terms used by participants when completing the Q-Sort. In the examples given above, most relevant and least relevant are used, however this could be replaced by other terms and phrases to meet the need for the specific study being conducted, e.g. agree/disagree, like/dislike, most favourite/least favourite, etc.
It is important that all participants understand these aspects of the study to ensure they complete the Q-Sort in the correct way in order to generate meaningful results.
If the Q-Sort is being conducted using physical statements then the condition of instruction can be provided physically on a piece of paper, but if the Q-Sort is being conducted using an online software (more on this later) then the condition of instruction could be shared digitally with the participants prior to the start of the Q-Sort.
the researcher will provide the participants with the Q-Set (the full set of statements), the Q-Grid and the condition of instruction. The participants will then rank the statements based on their own personal viewpoints on the topic being investigated.
The process of ranking statements via the Q-Sort is referred to as a ‘forced distribution.’
It is common for some participants to find it challenging to differentiate between some of the statements. This is particularly common if the Q-Grid requires the participant to rank a single statement as most relevant and least relevant.
However, when questioned following the Q-Sort, participants are almost always able to explain and justify why they ended up ranking a specific statement as most or least relevant which is valuable information for the researcher.
Forced distribution is an important component of Q-Methodology and this can be explained by comparing the difference between using a Q-Sort and a Questionnaire to explore the same research question.
If a 20 statement questionnaire was used and participants were asked to mark if they agree or disagree with each statement using a standard Likert scale, the researcher may find that a participant agrees with all statements equally, or at least agrees with a large number of the statements equally.
Whereas, if the same 20 statements were used for a Q-Sort the participant is forced to rank the statements they agree with in a hierarchical manner. By doing so it forces the participant to think more critically about the topic and allows the researcher to dig a little deeper into the most salient views on the topic.
The Initial Sort
The first step of completing the Q-Sort is the initial sort. This initial sort can vary in accordance with the research study, and could be removed all together if a small Q-Set is being used. However, it will typically be carried out in the following way.
The participant is shown all the statements from the Q-Set and asked to place each statement into one of three buckets:
This process of initially sorting the statements into three buckets allows the participant to familiarise them self with the Q-Set and also make the next step of sorting the statements in line with the Q-Grid more manageable.
Now that the participant has the statements in three buckets they will find it much easier to rank the statements on the Q-Grid.
The statements which the participant had agreed with will tend to be placed at the positive end of the Q-Grid, the neutral statements will tend to be placed in the middle of the Q-Grid and the statements the participant had disagreed with tend to be placed at the negative end of the Q-Grid.
The participant has the ability to completely change any statement placement from the initial sort and also move statements around as they progressively complete the Q-Grid.
Once the participant is confident that their ranking of the statements on the Q-Grid accurately represents their views on the topic they will confirm they have completed the Q-Sort with the researchers.
This sorting process can either be completed using physical materials with statements on pieces of paper being placed on a large Q-Grid on a table in front of the participant, or alternatively there are some good software packages which will allow the researcher to set up the Q-Sort for participants to complete digitally. There are advantages and disadvantages to both approaches so it is up to the researcher to decide which best suits the research study being conducted.
Once all participants have completed the Q-Sort the researcher will have collected their data.
the post Q-Sort interview is placed at stage 5 of this Q-Methodology process, however it could be completed after or in conjunction with stage 6.
The researcher may wish to conduct the participant interviews prior to analysing the Q-Sort data or the researcher may prefer to analyse the data and use this information to inform which participants are selected for interview as well as what themes to explore and what questions to ask.
Like many elements of Q-Methodology, this decision is up to the researcher and is based on what they think will be most beneficial to the study.
However, if you feel this is achievable, or if you are using a relatively small number of participants, then there is no reason why you can’t interview all participants.
Alternatively, you could use focus groups instead of 1-2-1 interviews to engage with more participants in a more efficient manner.
Selection of participants is ultimately determined by the researcher.
Participants could be selected at random for interview, or if data analysis has already taken place (step 6), then participants could be purposefully selected to ensure all, or predominant, viewpoints are hear.
For 1-2-1 interviews, the researcher could select individuals who align closely with each of the factor profiles that have been produced following data analysis.
For focus groups, the researcher can create groups in any way they see fit. Examples may include, but are not limited to:
How the focus group is created will depend on what the researcher is aiming to achieve.
Ultimately, the researcher has the ability to select the participants that they feel will best enable them to explore and answer the research question they are pursuing.
The themes covered and the questions asked will vary depending on the researcher’s aims.
However, the researcher will typically ask the participant to justify their placement of statements (or items) on the Q-Grid, with particular interest in the statements they placed at each extreme end of the Q-Grid.
Below are some examples of typical questions the research may ask a participant during a 1-2-1 interview.
“Why did you select this statement as ‘most relevant’?”
“Why did you select this statement as ‘least relevant’?”
“Why is statement 12 more relevant than statement 4?”
“I can see that most of the statements you most agree with fall under the same theme, why do you think this might be?”
It is important to be responsive and flexible in your questioning to ensure the dialogue is progressive and allows for the participant and researcher to pursue different lines of thought as they arise.
The types of questions and themes discussed in a focus group will largely be the same as those discussed in 1-2-1 interviews.
You will want to ask the participants to justify theory placement of specific statements on the Q-Grid.
But when conducting focus groups some additional considerations need to be taken into account.
Focus groups should not be too large as some participants will not be hear in such cases. 3-6 participants is generally preferred.
It is helpful to set some ground rules at the start around allowing others to speak, not interrupting each other, timeframe, expectations, themes to be covered, etc. This helps to control the discussion and generate more meaningful input from all participants.
If the focus groups are made up of sub-groups within the participants, such as different age groups, genders, ethnicities or job roles, the views of these groups can be explored to identify any common areas of agreement or disagreement. This approach to focus groups may not always be appropriate but when relevant it can provide some useful findings.
If the focus group has been selected on account of all participants having a profile aligned with a specific factor, then all participants will be broadly in agreement. In this situation, common themes can be covered with each participant taking it in turns to share their own explanations and experiences.
If the focus group has been selected using individuals from different factors, then the participants will be mixed in their agreement and disagreement towards specific statements and themes. In this situation, the researcher can expect more debate which will need to be facilitated accordingly.
It is important to record any interviews or focus groups conducted to ensure you have evidence to return to should you wish to do so at a later date (you almost certainly will need to).
Ideally, videoing or recording the audio of the discussion would be the preferred option as this will offer you a richer record. It also leaves you free as the researcher to engage in discuss without having to worry about making written notes as this can be highly distracting.
the other alternative is taking written notes but when using this methods it is often impossible to capture everything that is said. You could use another person as the note taker but similar issues will usually remain present as the discussion flows naturally.
if you are conducting interviews or focus groups virtually, the software being used will normally have a recording option so this is an ideal approach to take in this situation.
Obviously if you are recording anything you will need to gain consent from the participants.
Data analysis is not a simple process if you do not have prior knowledge of statistics. It not possible to cover the full extent of data analysis requirements for Q-Methodology in this guide but instead some key points and terminology will be covered.
You are advised to explore the additional resources section at the end of this post for more detailed guides on this aspect of Q-Methodology.
Brown (1980) is considered a must read if you are serious about understanding the statistical knowledge for Q-Methodology. This is really where the methodology becomes more complex if you do not have a background in statistics.
However, there are some general rules to follow and consider which will support you in this process. I will do my best here to try to cover these without going too deep into the stats.
You can run Q-analyses using most data analysis software packages such as SPPS, R or Stata.
There are also a few free analysis tools online which may be of use to you. A good example of this is Ken-Q Analysis where you can upload your data via an excel spreadsheet and your data is analysed for you following your input on a few basic data calculation instructions.
Q-Methodology follows a process of factor analysis which have an infinite number of acceptable solutions. Therefore, it is best to be clear about your own analytic aims and objectives prior to starting your data analysis.
An appropriate solution to factor analysis will achieve the following:
The correlation matrix reflects the nature and extent of the relationships that pertain among all the Q-sorts in the data.
Variance refers to the full range of meaning and variability present in the data. Variance can be divided into three types:
Factors are derived from common variance and often called common factors.
Factor extraction involves the identification and removal of distinct portions of common variance from the correlation matrix.
Principal component analysis (PCA) is not classed as factor analysis. Instead ‘centroid factor analysis’ is the extraction method most commonly used and preferred when using Q-methodology.
Factor loading or factor saturation occurs when a factor is extracted from the correlation matrix and a measure is provided that shows the extent to which each individual Q-sort exemplifies. Prior to factor rotation, these are known as unrotated factor loadings.
When a factor is extracted from the correlation matrix, a sizeable portion of shared or common meaning is lost with it. The interrelationships of the Q-sorts and their intercorrelations change to reflect this lost influence/ The relationships that remain are captured by the residual correlations.
Communality is a good indicator of how communal a specific Q-sort is, for example, how much it has in common with the other q-sorts in the data set.
Factors have ‘eigenvalues (EV), which are calculated by summing the squared factor loadings of all Q-sorts on that factor. Combined with the factor variance, it provides clear indication of the strength and explanatory power of an extracted factor.
The Kaiser-Guttman criterion suggest that factors with EVs in excess of 1.00 should be extracted. This is generally accepted in the research community as an appropriate rule. However, this can lead to too many extractions of too may factors, particularly when analysing a large data set.
The question is often, how many factors should I extract? Brown’s (1980) original guidance was to use the researcher’s experience or alternatively the magic number 7. Extracting 7 factors is typically considered to be a good place to start.
In factor rotation, the unrotated factor loadings are used as coordinates and as a means of mapping the relative positions of all the Q-sorts in the data set.
Factor space: the mapping takes place in a space that is defined by the study factors. Each factor defines one of its dimensions.
Every position in the space represents a unique and meaningful viewpoint that might legitimately be adopted by an individual Q-sort.
Factor rotation is the system by which we ensure that each factor offers us the best possible vantage point from which to view our subject matter. This is achieved by rotating the factor axes through factor space.
By-hand rotation, the factors must be rotated manually and you will have to decide where each should be positioned. Varimax rotation conducts the process automatically using statistical criteria. the two systems are complementary and neither is preferable over the other. The method used will depend on the data set and nature of the research.
Factor rotation identifies Q-sorts whose position and viewpoint closely approximate that of a particular factor. these Q-sorts can be used to derive a sound and representative estimate of that factor’s viewpoint and the estimate can then be used to support a sound factor interpretation.
Factor rotation changes the position of the factors and their viewpoints, relative to the Q-sorts. However, the position of the Q-sorts relative to one another fixed by their unrotated loadings. Factor rotation shifts the researcher’s viewpoints, not the viewpoints captured in the data via Q-sorts.
A factor estimate is normally calculated via a weighted average of all the individual Q-sorts that load significantly on that factor, and that factor alone. Not all significantly loading Q-sorts need to be used but if you limit this number too much it can reduce the reliability of the estimates you produce.
Q-sorts processing a significant factor loading in relation to more than one factor are considered to be confounded. Normally, confounded Q-sorts will not be included in the construction of any of the factor estimates.
When calculating a factor estimate, the contribution of each Q-sort is weighted according to its factor loading. The higher the factor loading the greater the contribution made to the final estimate.
A total weighted score is calculated for each item in the factor estimate. The higher the total weighted score, the higher the value accorded to that item by the factor in question.
The total weighted score for each item is converted into a z score. This allows for cross-factor comparisons to be made. PQ Method includes multiple table and uses z scores in this way.
Bipolar factors are defined by both positively and negatively loading Q-sorts.
Before data interpretation, it is common to convert all z scores for each individual item into a single factor array. A factor array is a single Q-sort configured to represent the viewpoint of a particular factor.
A factor array is simply a best possible estimate of the factor’s viewpoint.
It is important to check the intercorrelation of your factor arrays carefully because a significant correlation may mean that the two factors in question might simply represent alternative manifestations of a single factor. in this situation it may be beneficial to reconsider your factor solution and reduce the number of factors.
Stephenson developed the Q-sort for the purpose of making a gestalt or holistic form of data collection easier.
The key word here is ‘interpret’. In keeping with Q-Methodology, to interpret a factor thoroughly, you should interpret. In other words, you should be able to explain or account for the entire item configuration captured in the relevant factor array. If this can be achieved your final interpretation will capture the factor’s full viewpoint.
Cross factor item comparison is not the primary concern in factor interpretation. Instead, the interpretation should be driven by the interrelationship of the many items within a particular factor array.
Generating a crib sheet for each factor is an effect means of ensuring that nothing obvious is overlooked. The system works effectively because it forces your engagement with every item in a factor array.
The crib sheet provides a structured way to consistently interpret each and every factor and that has the potential to produce holistic factor interpretations.
A crib sheet contains four categories: the highest and lowest ranked items in a factor array are listed, as are the items that the relevant factor has ranked higher or lower than any of the other study factors. The categories allow you to identify the most important issues.
It is also important to consider the items in the middle of the distribution. Sometimes these items can act as fulcrum for the viewpoint being expressed. Such items can prove pivotal to our understanding of a relevant factor.
Your attention should oscillate between the individual items and the whole viewpoint. Keep your eye on the whole story and focus on the individual items, considering it place in the overall picture.
Bipolar factors should be interpreted twice: once for the viewpoint of the positive pole and once for the viewpoint of the negative pole.
This article has provided a comprehensive introduction to Q-Methodology but if you would are interested in exploring Q-Methodology further, here are some recommended resources for you to explore.
Q Methodology – https://qmethod.org
William Stevenson Wiki Page – https://en.wikipedia.org/wiki/William_Stephenson_(psychologist)
Ken-Q Analysis – https://shawnbanasick.github.io/ken-q-analysis/
YouTube channel, qsiteadmin, have published a series of workshops conducted by Steven Brown who was a student of William Stephenson (founder of Q-Methodology). These videos are very detailed but do provide more depth in understanding for anyone who is looking to explore Q-Methodology in more detail. Just be warned, they are old videos so the sound and overall quality is poor. But much thanks to qsiteadmin for making these videos available for public viewing.
Video 1: Introduction
Video 2: Statement Generation
Video 3: The Q Sample
Video 4: Generating the Q Sort
Video 5: Q Method Software
Video 6a: Rotation (Part 1)
Video 6b: Rotation (Part 2)
Video 7: Factor Scores
Video 8: Final Q&A
When I first planned to use Q Methodology I went through the process of researching it and found it very difficult to find any examples of a straight-forward, accessible overview of what Q Methodology is.
I found this extremely frustrating and I spent a considerable amount of time trying to find and consume information on this research method.
The problem is that Q Methodology is not a simple topic to summarise so any attempt to simplify it will typically be lacking some essential information.
With this in mind, if you are serious about using Q Methodology for your research then this article will provide you with a fantastic starting point to build upon, but you will need to read beyond the content of this article eventually (some resources for further reading have been provided at the end of this article).
However, if you are just looking for a general understanding of what Q Methodology is and how it can be used then this article will provide you with a fantastic oversight.
To demonstrate the complex nature of Q Methodology, and the reason why many of those new to Q Methodology struggle to engage with it, take this definition of Q Methodology as an example.
For those lacking existing knowledge of research methodology, terms such as ‘operant subjectivity’ (eliciting stakeholder views) and ‘inherent epistemology’ (essential or natural understanding of knowledge) can be rather off-putting.
So, this article aims to break through the terminology to provide a more accessible overview of Q Methodology, without losing the core components and purpose of this research methodology.