The Agglutination Hypothesis: A belated empirical investigation

[Conference abstract]: Linguistic typology started out with (and for many decades remained largely confined to) the study of morphological macro-types such as fusional, agglutinating, isolating and incorporating languages. This taxonomy, as established by Humboldt and Schleicher, has become a classical model that no introductory textbook fails to mention. Although few linguists in the 20th century have systematically studied the typology of morphological techniques and many have criticized the classical taxonomy, it is used routinely for characterizing languages or explaining specific features.

      My talk is mostly concerned with the agglutinating/fusional distinction (taking for granted the validity of the parameter of degree of synthesis/isolation). Although some critics have claimed that this is a purely taxonomic distinction that makes no prediction about other aspects of the grammar (i.e. does not embody a typological correlation), the old idea that there are (predominantly) agglutinating and (prediminantly) fusional languages in fact makes two implicit predictions. First, agglutination/fusion is characteristic of whole languages rather than individual constructions; second, the various parameters of agglutination/fusion correlate. The (unstated, but widely assumed) Agglutination Hypothesis can thus be formulated as follows:

The Agglutination Hypothesis

      (i) First prediction: If a language is agglutinating/fusional in one area of its morphology (e.g. in

            nouns, or in the future tense), it shows the same type elsewhere.

      (ii) Second prediction: If a language is agglutinating/fusional with respect to one of the three

            agglutination parameters (a-c) (and perhaps others), it shows the same type with respect to

            the other two parameters:

                        (a) separation/cumulation

                        (b) morpheme invariance/morpheme variability

                        (c) affix uniformity/affix suppletion

      While from the start linguists have been careful to point out that individual languages may (and often do) represent mixed rather than pure types, everybody seems to be committed to the Agglutination Hypothesis as a statistically valid correlation. Without the first prediction, it would hardly make sense to speak of “agglutinating languages” (rather than, e.g., “agglutinating tenses”). Without the second prediction, we would have to single out “separating”, “morpheme-invariant” and “affix-uniform” languages rather than “agglutinating languages”.

      Amazingly, the Agglutination Hypothesis has never been subjected to empirical verification. This is not an easy matter, because the three parameters are not easily quantified and compared across languages. It presupposes, for instance, a clear-cut identification of inflectional categories, and an assessment of the importance of certain inflection classes and subclasses. However, if the Agglutination Hypothesis is correct, it should be possible to get valid results even if arbitrary decisions have to be made in unclear cases.

      In this paper, I will report on a study of the nominal and verbal inflectional morphology of a reasonably balanced world-wide sample of 40 languages. I will try out  a variety of measures for the agglutination parameters and determine whether they are cross-linguistically significant. Preliminary initial results suggest that the validity of the Agglutination Hypothesis is by no means a foregone conclusion – it may well be that it will turn out that “agglutination” is just one way of trying to capture the strangeness of non-Indo-European languages, which all look alike to Eurocentric eyes.

      (However, at this stage of the study, I remain totally agnostic as to the eventual outcome, which will hopefully be seen as relevant whether the results are negative or positive.)