Note: I haven’t finished writing this file yet.
1. The Problems With The Traditional Scheme For Classifying Language Morphology
Wikipedia is still stuck using the outdated system, in large part because all the sources that get cited on Wikipedia tend to use the outdated morphology.
It wasn’t until I took my typology class during my third year of college that I learned that the morphology classification system I was taught in my introductory linguistics class is actually ambiguous and outdated.
Even languages that are described as “isolating” have morphology that ought to be considered agglutinating.
2. The Bickel & Nichols (2007) Model for Morphology
Bickel & Nichols (2007) have shown that this linear scale actually merges three different parameters, fusion, exponence and flexion, all of which are in themselves relevant for morphological typology and all of which may combine with each other. Also, any given language may employ a variety of combinations of these parameters, a fact which is obscured if we make whole-language typology statements, as already argued by Edward Sapir.
Fusion, Exponence, Flexivity
Fusion denotes the degree to which morphological markers (or formatives in Bickel & Nichols’ terminology) attach to a host stem. Following Bickel & Nichols (2007 and 2011b) I distinguish three types of fusion (and maybe a fourth type counting reduplication):
- A marker that stands alone as a free morpheme, that is, as an independent word, is isolating.
- Markers that are bound, i.e. that have to attach to a host, are concatenative.
- Markers that involve modifying the host in some way are non-linear:
- Root-and-pattern (semitic)
- Vowel Gradation
- Replacement / Substitution (replacing part of the stem)
- Suppletion (replacing the whole stem)
- Languages also differ as to how many grammatical categories may be expressed by one and the same morpheme.
- Separative (or monoexponential) morphemes encode only one single category
- Cumulative (or polyexponential, also called portmanteau) morphemes encode several things at the same time. This parameter may interact with fusion, so that we get six logical combinations: isolating, concatenative and non-linear separative markers plus isolating, concatenative and non-linear cumulative markers.
Automatic reduplication is when an affix obligatorily triggers reduplication but the reduplication itself does not add any meaning to the construction.
Languages also differ as to how many grammatical categories may be expressed by one and the same morpheme. Separative (or monoexponential) morphemes encode only one single category, while cumulative (or polyexponential, also called portmanteau, e.g. Booij 2005) morphemes encode several things at the same time. This parameter may interact with fusion, so that we get six logical combinations: isolating, concatenative and non-linear separative markers plus isolating, concatenative and non-linear cumulative markers.
Languages also differ in how much allomorphy they have, termed flexitivity in Bickel & Nichols (2007).
The Indo-European declension and conjugation classes are examples of flexitivity, where a set of inflectional affixes are chosen depending on which class the noun or verb belongs to.
If, on the other hand, a given grammatical marker is always the same, i.e. does not vary according to classes of verbs or nouns, it is nonflexive.
2.4. How The Terminology Of The Outdated, Non-Descriptive Model Translates To The Newer Model
- Isolating: High Isolating Fusion
- Analytic: Medium for all fusion scales, little to no exponence
- Fusional: High grammatical exponence, has at least some concatenating fusion, maybe some suprasegmental fusion, probably some flexivity
- Agglutinative: high fusion, low exponence
- Polysynthetic: Very high for all kinds of different fusion
3. Improving The Bickel & Nichols (2007) Morphology Model
In my opinion, the Bickel & Nichols morphology model is an improvement to the original, but it still conflates several aspects of morphology together, which is why I favor using six or seven parameters for categorizing morphology instead of just three.
Bickel & Nichols (2007) listed three types of fusion, with reduplication being the fourth missing type. It might just be better to divide these four types of fusion into separate categories for describing linguistic morphology. Thus, there would be a total of at least six main parameters for describing a language’s morphology.
Going further, it might be possible to improve this model by dividing the fusion scale into a linear fusion scale and a non-linear fusion scale (all non-linear fusion is bound).
Even further, there could be a distinction between bound linear fusion and isolating linear fusion.
Conclusion: There are six parameters for describing linguistic morphology: Bound Linear Fusion, Isolating Linear Fusion, Non-Linear Fusion, Reduplication (includes types of reduplication), Exponence (how many grammatical categories may be expressed by just one morpheme), and Flexivity (amount of allomorphy).
In my opinion, “exponence” should be further divided into ’grammatical exponence’ and ’lexical exponence’. For example, Chinese languages and English have roughly the same amount of grammatical exponence in their morphology, whereas English definitely has more lexical exponence in its morphology. Creoles probably have the lowest lexical exponence of all languages.
Shower Thought: Languages with higher exponence, but few terms for the semantic differences may be more vulnerable to the Sapir-Whorf Effect.