Home Environment Research Methods in Anthropology: Qualitative and Quantitative Approaches

# Paired Comparisons

The method of paired comparisons is an alternative way to get rank orderings of a list of items in a domain. For any set of things, there are n(n — 1)/2 pairs of those things. Suppose you have a list of five colors: red, green, yellow, blue, and brown. Figure 10.1 shows the paired comparison test to find out an informant’s rank-ordered preference for these five colors. In this case, the question would be: ‘‘Look at each pair of colors and, for each pair, tell me which one you like more.’’

In each of the following pairs of colors, please circle the one you like best:

FIGURE 10.1.

A paired comparison test for rank-ordered data.

You might say: “Here are two animals. Which one is the more_where

the blank is filled in by ‘‘vicious,’’ or ‘‘wild,’’ or ‘‘smarter,’’ or some other descriptor.

You could ask informants to choose ‘‘the food in this pair that is better for you,’’ or ‘‘the crime in this pair that you’re most afraid of.’’

I’ve presented the pairs in figure 10.1 in such a way that you can easily see how the 10 of them exhausts the possibilities for five items. When you present a paired comparison test to an informant, be sure to scramble the order of the pairs to guard against order effects—that is, where something about the order of the items in a list influences the choices that informants make.

To find the rank order of the list for each informant, you simply count up how many times each item in a list ‘‘wins’’—that is, how many times it was circled. If you are studying illnesses and cancer is on the list, and if the question is ‘‘which of these pairs of illnesses is more life threatening,'' you expect to find it circled each time it is paired with another illness—except, perhaps, when it is paired with AIDS. Because this is so predictable, it’s not very interesting. It gets really interesting when you have illnesses like diabetes and high blood pressure in your list and you compare the average rank ordering among various ethnic groups.

The paired comparison technique has a lot going for it. People make one judgment at a time, so it’s much easier on them than asking them to rank order a list of items by staring at all the items at once. Also, you can use paired comparisons with nonliterate informants by reading the list of pairs to them, one at a time, and recording their answers.

Like triad tests, paired comparisons can only be used with a relatively short list of items in a domain, unless you apply balanced incomplete block designs. With 20 items in a paired comparison task, for example, informants have to make 190 judgments (Further Reading: rankings and paired comparisons).

There is one more method for studying the attributes of things: rating scales. This one is so important, it deserves a chapter of its own . . . next.

Kinship studies: Alexander (1976); Dousset (2008); Houseman and White (1998); Kronenfeld (2004); Leach (1945); Lehman (1992); Read (2001).

Methods of data collection for domain analysis: de Munck and Sobo (1998); Handwerker (2001);

J. C. Johnson and Weller (2002); Weller and Romney (1988).

Free lists: Ross and Medin (2005); Ryan et al. (2000); K. D. Smith et al. (2007); Thompson and Juan (2006); Verma et al. (2001).

Sentence frames: D’Andrade et al. (1972); Frake (1964); Hruschka et al. (2008); Metzger and Williams (1966).

Triad tasks: Durrenberger and Erem (2005); Furlow (2003); Nyamongo (2002); Ross et al. (2005). Pile sorts: Collins (2006); Longfield (2004); Roberts et al. (1986).

Rankings and paired comparisons: Chavez et al. (1995); Durrenberger (2003); Erickson (1997); Kozak et al. (2008); Thurstone (1927).

Found a mistake? Please highlight the word and press Shift + Enter

Subjects