Determination of Qualities and Quantities

Measures may be differentiated based on the assumptions implicitly made about how the magnitude of a phenomenon is assessed. For example, if behaviors are counted, there must be a clear definition of what those behaviors are.

Measures that are created by comparisons to a standard (e.g., a yardstick) yield magni­tudes (amounts) of a different sort. Time and distance are prototypical amounts, but many measures may be constructed by defining a nonmaterial yardstick and having respondents or coders make comparisons to it as in the following example of using a magnitude scale:

If 0 is not at all feeling hostile, and 100 is the level of hostility you feel when you are cut off in traffic, how hostile do you feel in this negotiation? There is no highest number.

An example of amount (or magnitude) scaling in studying war and conflict is found in Sulfaro and Crislip’s (1997) study on Americans’ perceptions of foreign policy threats. Participants rated their perceptions of 19 countries’ hostility toward the United States with two magnitude scales and one 7-point scale varying from “least hostile” to “most hostile.” Results indicated that the two magni­tude scales were almost identical in measuring hostility across the 19 countries with an “R² for [the logarithmically transformed variables] is.987” (p.

Through the use of ratios and differences, counts and amounts can be used to create derivative measures; examples include the ratio of the number of hostile words to the total number of words expressed (ratio of a count to a count), and the acceleration of aggressive­ness in interaction, as assessed by the change in the magnitude of expressed aggressiveness over time (a change in an amount divided by an amount). One study in which a ratio is derived and employed (counts over amounts) is Fuller, Murphy, Ridgley, and Ulack’s (2000) research on potential conflict in Southeast Asia in which ratios were used to determine “how many times greater (or less) the potential for conflict with one ethnic group is over another” (p. 321). Another example is the use of physi­ological data of amount-over-time ratios: D. M. Buss, Larsen, Westen, and Semmelroth (1992) measured the acceleration of a neg­ative emotion, jealousy, with physiological measurements such as electrodermal activity, pulse rate, and electromyographic activity.

The methods described here (counts, amounts, and their derivatives) allow greater precision, typically evidence high levels of reli­ability, and assist in determining the functional forms that relate variables of interest to each other when assessing hypotheses (Neuendorf, 2002; Woelfel & Fink, 1980). We realize that the typical investigator employs scales such as a unidimensional 1- to 7-point scale, with response alternatives bounded at both ends (i.e., one cannot go below 1 or above 7). Such scales are generally not examined for their many implicit assumptions: that the distance between adjacent pairs of scale units is equal for all pairs; that the number of response alternatives is adequate for the variable being scaled; that the boundedness of the scale does not cause scale distortions due to floor or ceil­ing effects; that the vectors emanating from the scale’s neutral point to the end points are separated by 180° (i.e., the end points are indeed polar opposites; see Torgerson, 1958, for a discussion of some of these issues). For example, Woelfel and Fink (1980, p. 235) showed that good and evil were not polar opposites, although a unidimensional scale would force good and evil to be polar oppo­sites by using them as scale anchors (i.e., “rate X on a 1 to 7 scale, where 1 = good and 7 = evil”). Thus, results based on the 1 to 7 scale would likely distort and thereby invalidate the respondents’ evaluations. Conflict research, like all social science research, would benefit from greater consideration of measurement options in terms of response scales and their assumptions.