Languages may be compared with respect to the degree of grammaticalization of certain cognitive or communicative categories. Such a comparison has two presuppositions:

1. A grammatical category (like pural) can be defined as an interlingual category such that it can be identified in different languages.
2. The criteria to measure the degree of grammaticalization are applicable in a unified way to formatives or constructions of different languages.

The comparison of two formatives or subsystems with respect to their degree of grammaticalization is methodologically safest when they belong to the same language. There are two possibilities here: either the compared elements may be diachronically identical members of two stages of the same language, as are Latin ad ‘to’ and French à dative; or they may be distinct members of one synchronic state of a language, as are the Latin preposition ad and the dative. In both cases there is a common basis that is kept constant; in the first case it is the diachronic identity of the compared signs, in the second case it is the system comprising the compared units. Therefore it is easy to prove that Latin ad and French à differ according to the grammaticalization parameters, and equally that ad and the dative in Latin differ by them, and consequently that they are grammaticalized to different degrees.

Interesting problems arise when we compare elements or subsystems of different languages, such as the case systems of Latin and Turkish, or unrelated elements/subsystems of two stages of a language, e.g. the Latin dative and French à. Such comparisons have, of course, always been made. Edward Sapir, in comparing the plural in Nootka and English, observes (1921:104) "that the [Nootka] plural concept is not as abstractly, as relationally, felt as in English," which implies that it is less grammaticalized. Similarly, the Latin dative is more grammaticalized than French à, and the Latin case system is more grammaticalized than the Turkish one. Plural marking on nouns and pronouns is done by the suffix -men in Mandarin and by the suffix -s (and its allomorphs) in English. It is more grammaticalized in English than in Mandarin.¹

The basis of comparison of these various subsystems is obviously a purely functional one. Given that the criteria of grammaticalization are purely formal in nature, they do not tell us which subsystems of different languages it makes sense to compare. However, once the common functional denominator that remains constant throughout a grammaticalization path is established, the criteria of grammaticalization tell us for any subsystem which stage of the path it has to be attributed to, and for any two subsystems, how similar or different they are in regard to grammaticalization.

We all apply concepts such as case suffix and postposition, allative and dative, personal pronoun and personal affix, on a cross-linguistic level. We subsume the German and the Turkish case suf- fixes under a common term, and similarly the Turkish and Cabecar postpositions; we decide that German zu ‘to’ is an allative preposition, but the cognate English to is a dative preposition; the preverbal pronominal elements of English are called personal pronouns, but in French they are pronominal clitics. For all these pairs of concepts, there is a common functional denominator and a difference in the degree of grammaticalization which we use as a criterion in applying the terms to categories of various languages. Given that differences in the degree of grammaticalization can by definition only be gradual, these analytical concepts must be prototypical concepts. We know certain focal instances of the concepts of case suffix and postposition, personal pronoun and personal affix and so forth, and we apply these terms to new phenomena depending on their similarity to our focal instances. Sometimes a new phenomenon lies somewhere in the middle between two focal instances, having an intermediate degree of grammaticalization for which our tradition provides no term. The Japanese case elements are a noteworthy example.

Several of them, as shown in , are desemanticized to a degree that we would expect them to be case suffixes; but tradition calls them postpositions, which may be more adequate if we have their (low) degree of cohesion in view. Therefore the vacillation in terminology may be justified. Given that we are dealing with scalar phenomena, terminological settlements in the space between the focal instances are necessarily somewhat arbitrary. We will have to define our notions of case suffix and postposition, of personal pronoun and personal affix more precisely by fixing the grammaticalization parameters for them, before we will be able to settle for a correct term for the Japanese case elements or for pronominal formatives of different degrees of grammaticalization.

¹ Iljic 2001: 74: “the ‘singular/plural’ distinction is not grammaticalized” in Chinese.