Thurstone Scaling

Thurstone scaling is a covering term representing a series of three techniques developed by psychometrician Louis L. Thurstone during the 1920s. These scaling methods are commonly known as

• 1.
  The Method of Equal Appearing Intervals
• 2.
  The Method of Successive Intervals
• 3.
  The Method of Paired Comparisons

[Page 1126]Beneath all of these scaling techniques was a common perspective on the measurement of attitudes—the Law of Comparative Judgment (Thurstone, 1927).

Imagine that you wish to compare the weights of multiple objects and have no modern scale or set of standard weights. Rather than despair about an inability to judge weights precisely, one can estimate the weights of objects relatively—ordering them from lightest to heaviest. This could be done by comparing each object against the others one at a time, judging them in pairwise comparison. Place one object on either side of a balanced beam and identify the heavier object. A full set of comparisons should yield a set of objects ordered by weight. Thurstone was the first to realize and implement a system to do the same thing with psychological rather than physical objects.

A scaling methodology designed to assign numbers to the value or intensity of psychology objects must reflect the discriminations made by individuals judging multiple stimuli. The bottom line underlying the scaling of comparative judgments is that objects that are easily discriminated should be placed further apart than objects that are difficult to distinguish.

The Method of Paired Comparisons involves explicit judgment by all subjects of each combination of objects. This task becomes formidable as the number of objects increases, because the number of paired comparisons rises geometrically. That is, assuming a pairing of A and B is the same as a comparison of B and A, the number of comparisons required equals N(N − 1)/2. Four objects will be paired in 6 ways, 5 in 10 ways, 10 in 45 ways. If order matters, the number increases even more rapidly: N(N − 1). Four objects will generate 12 comparisons, 5 call for 20, and 10 require 90. The ordering of a large number of objects becomes unwieldy.

The Method of Equal Appearing Intervals is a technique developed to capture individual assessments of attitude objects without the necessity of capturing all paired comparisons. Judges are employed to calibrate a set of items, which is then administered to the study population.

Begin with a large set of statements reflecting diverse opinions toward an object. Present these statements to a panel of judges, each of whom determines the degree of support for (or opposition to) the object under study by placing the statement in one of 11 ordered piles ranging from “most favorable” to “neutral” to “least favorable.” These initial judgments provide the basis for the selection of items for the final scale.

Items are ranked in terms of their average support score (MEDIAN) and the variation in support scores (INTERQUARTILE RANGE). Those items that are inconsistently rated by the judges are dropped, and a final scale of 5 to 20 items is created by selecting items covering the range of support and approximately evenly spaced (as presented by the median assessment of the judges).

...