Hermit Crab: Morphological Characteristics
Unlike the situation with generative phonology, there has never been a consensus among generative linguists as to what morphology was (or even if it exists as a separate "module"). The following are some characteristics of the way Hermit Crab does morphology:
- Process-based Morphology: Hermit Crab implements process-based morphology. This means that all affixes are treated as morphological rules, rather than as 'things' that get concatenated onto the stem. (This is the original meaning of Item-and-Process morphology, as the term was first used by Hockett in his 1954 Word article "Two models of grammatical description." The term was later used to mean that affixes were attached to stems in their underlying form, then modified by morpho-phonological rules. Hermit Crab implements this latter idea of processual morphology, too, see here.) The use of morphological processes makes it fairly straightforward to implement nonconcatenative morphology, such as infixes or reduplication.
- Morphosyntactic Feature Percolation: Stems bear morphosyntactic features (with a serious limitation: they cannot be hierarchical, see "Limitations" below). An affix can specify that the stem must belong to a particular part of speech (category), and that its morphosyntactic features must unify with a certain set of features. In addition, the affix can bear morphosyntactic features of its own; when an affix is attached to a stem, the output bears all the morphosyntactic features of the affix plus any nonconflicting features of the stem. This is useful for modeling derivational affixes, and under some theories may be used to model inflectional affixes as well.
- Realizational Morphology: In addition, Hermit Crab supports "realizational morphology" (again, with the limitation to non-hierarchical features) Under the theory of realizational morphology, inflectional (but not derivational) affixes are treated as being triggered by a set of morphosyntactic features that are to be realized on a particular word. Intuitively, for each inflectable part of speech (category), there is a template consisting of a sequence of slots. (The order of the slots corresponds to the order in which the affixes are attached to the stem, which is not necessarily the same as their left-to-right order.) Within a slot, the order of affixes corresponds to the order in which they are tested for compatiblity with the features to be realized on the word, with the ordering being disjunctive. If the last affix in the slot has no features which it realizes, it is a default ('elsewhere') affix. If there is no default affix, and a word is to be generated whose realizational features do not match any of the affixes of the slot, the effect is that of a 'zero affix', that is, no affix from that slot is attached. Realizational morphology is discussed in Anderson's 1992 book A-morphous Morphology, and numerous other places.
With regard to morphology, Hermit Crab has the following limitations:
- Morphosyntactic features: Hermit Crab does not (yet) support hierarchical morphosyntactic features. This makes it difficult to deal with languages which have (say) agreement with both the subject and the object marked on the verb.
- Compounding: Hermit Crab does not (yet) support compounding or (productive) incorporation. (The specification calls for it, but it has not been implemented yet.)
- Cyclic Application: As currently implemented, Hermit Crab does not support cyclic rule application. (See Phonological Characteristics for more info.)
- Prosodic Morphology: Since Hermit Crab does not (yet) support metrical phonology, it does not support prosodic morphology. (That does not prevent you from analyzing reduplication by prefixing a bare syllable and copying--but not spreading--the features from the base onto the syllable; but it does mean that you have to do it using Cs and Vs, rather than a nice 'syllable' node.)