The reordering of morphemes
Several kinds of proposals have been put forth to explain the relative sequencing of morphemes we find in languages. Some are purely synchronic, such as Baker’s 1985 Mirror Principle that morphological derivations must directly reflect syntactic derivations. According to his analysis, for example, if a sentence in some language is analyzed as the result of first an applicative process (changing ‘The zebras handed the crowbar to the girl’ to ‘The zebras handed the girl the crowbar’) and then passivization (changing ‘The zebras handed the girl the crowbar’ to ‘The girl was handed the crowbar’), applicative and passive affixes will be added to the verb in that order. Other explanations of morpheme order, such as that proposed in Rice 1991, have focused on the relative semantic or grammatical scope of morphemes. Thus in a word like [[[beauti]ful]ly], the outer suffix-ly has scope over the full adjective stem beauti-ful to its left, deriving an adverb from the adjective. The inner suffix-ful has scope over just the noun root beauty to its left, deriving an adjective ‘full of beauty’. Bybee (1985:33–5) accounts for morpheme order in terms of ‘relevance’: ‘the semantic relevance of an affix to a stem is the extent to which the meaning of the affix directly affects the meaning of the stem’ (1985:4). More relevant affixes are predicted to occur closer to the root. Her prediction can be illustrated with markers of aspect and tense. A shift in aspect can significantly alter the nature of an event: sleeping is a different kind of activity than its inceptive counterpart falling asleep. Tense, by contrast, simply situates an event as a whole at some point in time, with little effect on its essence: sleeping today is much the same kind of activity as sleeping yesterday. Aspect affixes are accordingly predicted to occur inside of tense markers within verbal morphology.
Yet we know that morpheme order is not usually the result of spontaneous on-line decisions at each moment of speech. Morphological structures are built up over time, as independent words are gradually reduced to dependent affixes through processes of grammaticalization. Diachronic considerations were appropriately brought into the discussion of morpheme order with Givón’s provocative maxim that ‘today’s morphology is yesterday’s syntax’ (1971:413), further discussed by Comrie (1980). Bybee proposes that ‘the order of morphemes is in large part a result of the order of words in the verb phrase, and the frequency of occurrence of certain categories as verbal inflections is a reflex of their frequent occurrence contiguous to the main verb’ (1985:41). Of course a full sentence is not reduced to a morphologically complex word in a single operation. Morphological structures usually develop step by step, category by category. Morpheme order might thus be better seen as a reflection of the historical sequence of grammaticalization of affixes.
The various synchronic and diachronic accounts of morpheme order need not be at odds. Synchronic analyses of morpheme order cast in terms of derivational processes could, for example, reflect the historical order of the grammaticalization of these processes. These might in turn mirror their relative scope or relevance. Such relations might have been the basis underlying earlier word order. Comparative evidence within many language families indicates that morpheme order often does reflect the sequence of grammaticalization of affixes: those affixes closest to the root are indeed the oldest, and those on the periphery of words can be seen to be more recent additions. The picture is more complex, however. Once built up, morphological structures are not immune to further change. A variety of processes may alter morpheme order after attachment has occurred. (Discussions of some of these changes are in Haspelmath 1993, Mithun 1995, and Harris 1997). In what follows we will see a kind of reordering that can be stimulated by a shift in status of certain markers from derivational to inflectional.
