Historical Inferences from Guttman Scales: The Return of Age-Area Magic?

Abstract

Inferences about historical sequences are frequently drawn from Guttman scales, on the supposition that each item in the scale serves as a "functional prerequisite" for successive items. The validity of such inferences is here called into question on both theoretical and empirical grounds. Logically, such inferences are subject to the same limitations as inferences about the age of a trait based on its geographic distribution (the discredited "age-area" hypothesis). This is true even if specific scale items are seen as indices of broader evolutionary "stages," or if the unidimensional nature of the Guttman scale serves only as the basis for inferring some underlying unilinear evolutionary process. To test these notions empirically, 40 Mexican communities, varying widely in location, population, economic base, and social attributes, were investigated. Within each, data were gathered on the presence or absence of 24 traits typically selected for Guttman scaling, together with their actual historical sequence. A Guttman scale of high reproducibility was constructed from these trait lists, though some issues concerning the reliability of scaling were also raised. The developmental sequence inferred from this scale was then correlated with the historical sequences. Few correlations were negative, but the majority fell below .7, thus accounting for less than 50% of the variance. An alternative explanation for even this level of correlation was provided in the association found between population size and scale step (.83). An examination of the historical data also suggested that the actual processes occurring in Mexican communities resemble more closely a "multilinear" or a "systems" model of developmental growth than the unilinear model which Guttman scales are sometimes thought to imply.