I used to teach my students that they had to get the unit of analysis right and everything in their research design would flow from that. It is always important to avoid category errors but especially here. The example I used was if the unit of analysis was opinion. Institutions like universities do not have opinions, individuals do. Institutions have policies, a different category. If they want opinions, they then need to sample individuals. It appeared to be ridiculous to expect a university to have an opinion. A university can have a policy supporting students who have English as their second language, but cannot have an opinion on the Quad’s influence in the Pacific region. Lately however universities, corporations and government departments have tried to pretend that they can and do have opinions, usually ones that are popular and memeable.

Doing so means flatting the varied stances of the individuals who make up the institution which is why the opinions tend to be bland statements of The Nice. Separating out categories for analysis matters as not doing so causes a lot of confusion. To take one example, the current ‘We are more than a faceless corporation intent on shoving your hopes and dreams into the data grinder’ discourse seems keen on celebrating identity. Yet not everything is or can be an identity. Trans, gay, neuro-divergent, working class, British-Asian, female: all are very different in nature and your job if you are using any identity category is to work out the difference, what it might look like from the perspective of those concerned, and what elements of it are grounded in which material, physical and social structures. Is it other defined or self assigned would be the first question to ask. The ultimate aim is always working out what aspects of the world matter to people, why and how, and if we should or can do anything about them.

Let us take a look at the work you have to do in order that your research design gives the best account of reality possible. It should be grounded, sensitive to relevant distinctions, and carve reality at the right joints. The ‘carving reality at the joints’ metaphor is a good one because it infers you are only taking the right bits, and also because the history of carving humans at their joints should alert us to the blind alleys you can easily wander into. For example, for years the clitoris was thought to take one, very limited form, in part because of a failure to consider women’s bodies as anything other than weaker, inward turned versions of male bodies.

The clitoris has a wholly different structure than thought and is far more extensive than had been assumed. As an aside, I do not see how you can examine the anatomical history of women’s bodies without acknowledging that one version is more correct than another, and that knowledge has at times been confused or suppressed due to fear and loathing of women’s bodies. The conclusions drawn are not just arbitrary social constructions mixing up bio and social phenomena. That does not mean knowledge of women’s sexual and pleasure anatomy is complete, but we can say one understanding is better than another. Even better, it is the one that manages to recognise women as autonomous human beings – win! This brings us onto some crucial distinctions we need to get right: what is the unit of analysis and what is the unit of observation? Both these two need to link clearly.

The unit of analysis is the quality of the entity being reported on, and the unit of observation the unit-data type being aggregated. In the above case the unit of analysis could be women’s sexual function, and the unit of observation the clitoris. The reason for the multi-millenia ignorance about the clitoris might be assumed to be because the unit of analysis should have been women’s sexual pleasure but was not, and also because anatomy as a science had trouble with an organ that did not immediately appear as a clearly distinct whole. In fact, women’s sexual pleasure has been studied throughout medical history going back many centuries, usually as something mysterious and complicated: there just seems to have been a taboo around taking the clitoris seriously. Another study might take the unit of analysis to be women’s sexual satisfaction, and attempt to correlate variations in clitoral anatomy to degrees of reported sexual pleasure (hopefully there would be hypotheses driving these fictional studies rather than just ‘let’s see what happens’ but I know how you all think). The unit of observation would then be individual women or potentially women in different relationship types. Very different, but not unrelated, data types are needed. They connect because they are related questions.

I am going to stick with the anatomy metaphor because I like it. From these two elements, unit of analysis and unit of observation, we get the reality bag of tricks, your methodological dissection instruments with which you divide and reassemble your object of study. They are called ‘research methods’ because you need to be methodical with them. The ones you use and the way you use them are dependent on those two baseline assumptions, so it is worth taking the time to get them right.

Charlier, Philippe, Saudamini Deo, and Antonio Perciaccante. “A brief history of the clitoris.” Archives of Sexual Behavior49.1 (2020): 47-48.