Guideline for Clusters of Change

For further reading: Facing interfaces: A clustering approach to grammaticalization and related changes by Norde & Beijering 2014.


Terminological problems in grammaticalization studies

Since there is no generally accepted definition of grammaticalization, it is a heterogenous concept which captures widely diverging developments. As a result, an instance of grammaticalization in view A does not necessarily qualify as an instance of grammaticalization from perspective B. Moreover, in the literature there is a tendency to pick out a subset of properties (such as Lehmann’s 1995 parameters or Hopper’s 1991 principles), on the basis of which one attributes the label 'grammaticalization' to a particular change. Not only do predefined lists of properties run the risk of being circular (e.g. grammaticalization has the properties A, B, and C; hence, property A+B+C is a case of grammaticalization), they may also lead to misconceptions and mismatches.

For these reasons, we propose to abandon the idea that all linguistic changes can be classified in terms of predefined categories such as 'grammaticalization'. We will therefore refrain from formulating yet another definition of grammaticalization, but focus on clusters of microchanges in grammaticalization, and the characteristics that it shares with other types of language change like lexicalization, pragmaticalization or degrammaticalization (in the following, we will use the cover term 'ization' for these phenomena). We put forward an analytic model in which we reduce grammaticalization and related changes to three levels of observation:

• Main mechanisms: formal reanalysis and semantic reinterpretation
Primitive changes: micro-changes on the levels of phonology, morphology, syntax, semantics, discourse
• Side effects: e.g. obligatorification or layering

Why primitive changes and not parameters?

As ‘izations’ are multi-dimensional, they are termed ‘composite changes’, consisting of basic mechanisms of change, several primitive changes, and their side effects. Mechanisms, primitive changes, and side effects are not themselves specific to one particular ization, but the way in which they cluster differs for each ization. While some of these clusters can be identified as a “prototypical” instance of a given ization, there are clusters which are hybrid cases at the interface of these prototypes.

The advantage of analyzing composite changes in terms of primitive changes is that primitive changes are far less controversial. Firstly, they only pertain to a single linguistic level, not to several levels simultaneously (as do izations). Secondly, they are parametric, meaning (in the case of primitive changes) that a linguistic property may undergo reduction, expansion, or neither, and which of these three possibilities is at stake is usually quite straightforward to make out. In Lehmann’s 'parameters of grammaticalization', on the other hand, each parameter is associated with one or more primitive changes. We argue that it is often difficult to classify izations on the basis of these parameters alone.


The Clustering Approach: Primitive Changes

Primitive changes are independent of the type of ization. They have the values expansion, reduction or zero (no change).

Morphological compositionality: concerns the degree of formal transparency of compositional items (discrete internal morpheme boundaries)
• more morphological compositionality: the better analyzable a compositional form is
• less morphological compositionality: less analyzable forms due to constituent internal reanalysis or univerbation

Morphosyntactic properties: concerns definiteness (combinability with determiners) and inflectional properties such as tense, case, number and subcategorization features. An item shifting from one word class to another gains the morphosyntactic properties of its new category and lose those of its original word class.
• more morphosyntactic properties: MISSING
• less morphosyntactic properties: MISSING

Phonological substance: concerns the loss or gain of phonological material (number of phonemes, segmental features like full vs. reduced vowels) and prosodic features (e. g. stress)
• more phonological substance: gain in phonological material, including gain of (main) stress (e. g. in the noun ism in contrast to the suffix -ism)
• less phonological substance: loss of segments or lenition/fortition of remaining segments (e.g. going to -> gonna) or loss of sandhi effect

Semantic compositionality: concerns the extent to which the meaning of an expression can be derived or constructed from its component parts as a sum of these (degree of semantic transparency or opacity of compositional forms)
• more semantic compositionality: MISSING
• less semantic compositionality: can but need not to correlate with loss of morphological compositionality (Dutch schatje ‘sweetie’ lost compositional meaning of little treasure, though it can still be morphologically analysed as diminutive)

Semantic substance: concerns the referential meaning, which makes relational meanings more salient when it fades
• more semantic substance: when new referential dimensions are gained (concretion, specialization)
• less semantic substance: more general, abstract meanings (bleaching, generalization)

Syntactic autonomy: concerns the degree of syntactic integration, cohesion and dependencies for a given item
• more syntactic autonomy: the more optional an item is, the more syntactic autonomy it has
• less syntactic autonomy: when items develop to obligatory elements of the core grammar (e. g. inflectional affixes)

Syntactic variability: concerns the degree of flexibility of a linguistic item; the number of syntactic slots that an element may occupy
• more syntactic variability: elements developing into adverbs or discourse markers gain in flexibility
• less syntactic variability: elements moving towards bound status

For information regarding the mechanisms of change and side effects, please read the original paper (linked above).