A multidimensional approach to the classification of neoclassical combining forms

iThis paper aims to show the usefulness of multidimensional analysis techniques for linguistic classification, and to propose a solution to the much debated categorial status of combining forms providing a classification of such elements based on a detailed analysis of a representative corpus taken from the Italian language. Computational methods for statistical analysis of observed correlations allow for internally compact clusters that are well apart from each other, and also permit the establishment of descriptive variables in consideration of a given data set. This methodology satisfies the requirements of structural stability and flexible adaptability called for in linguistic prototype theory and at the same time it resolves both the problem of an optimal number of classes and predictivity of classification. This analysis gives both a structural description of the studied data set as a whole and a precise allocation of each analyzed item. The analysis, performed on a corpus of 563 Italian combining forms, is based on explicit criteria so that its linguistic interpretation is relatively straightforward. Solid empirical evidence is given to demonstrate that typical combining forms are bound lexemes (stems), consequentially their classification does not require any further linguistic category beyond those of affix and lexeme (the term “combing form” is a convenient descriptive tool for grasping together bound elements used to form morphological complex words not sharing all the characteristics of lexemes or affixes of a particular language). The results support the idea of a more widespread use of cluster analysis in linguistics, both from a methodological and empirical point of view.