Types of morpheme order: layered versus templatic morphology
A distinction sometimes drawn in work on morpheme order is that between ‘layered’ or ‘hierarchical’ structures on the one hand, and ‘templatic’ or ‘slot and filler’ structures on the other. In layered structures, words are analyzed as having been built up step-by-step, affix-by-affix, with each morphological process creating a new stem that can serve as a base for another, as in [[beauti]ful]ly discussed above. Such structures are elaborately developed in the Eskimoan languages, illustrated here with material from Central Alaskan Yup’ik. Examples have been drawn from the speech of Elena Charles, Elizabeth Ali, and George Charles of Bethel, Alaska.
Both nouns and verbs in Yup’ik consist of an initial root, any number of derivational suffixes, and a final inflectional ending.
The inventory of suffixes is extremely rich; Jacobsen (1984) has identified over 450 derivational suffixes in Yup’ik and even more inflectional suffixes. A root may be inflected directly to form a word, or it may serve as the input to a derivational process which creates a new stem that can in turn be inflected as a word, and so forth. (Yup’ik examples are presented in the practical orthography.)
The Yup’ik suffixes do not always occur in the same order with respect to each other. Ordering differences can reflect differences in semantic or grammatical scope. Each added suffix qualifies the preceding stem as a whole, rather than just the root or the preceding suffix. In (2a), for example, the stem ‘giant’ is first derived (from ’person-big’), then qualified by ‘little’, while in (2b), the stem ‘midget’ is first derived (from ‘person-little’), then qualified as a unit by ‘big’.
In (3a), the suffix-yugnarqe- ‘probably’ modifies just the projected going, expressed by the material to its left. It is followed by a suffix-ni- ‘claim’ and a past tense marker-llru-. In (3b), the same suffix follows both-ni- and-llru and modifies the full derived stem: ‘claimed he would go’. In both, the past tense-llru- appears to the right or outside of the future tense-ciq-, conveying a past statement about a projected event.
Yet even within a system like Yup’ik, morpheme order is not fully free. Of course some orders occur more often simply because the scope reflected in one order corresponds to what speakers wish to say more often than the reverse. Certain recurring orders in Yup’ik are reflected in the development of what could be called compound suffixes: multi-suffix chunks that are processed as a single unit. One such unit is the negative desiderative seen in (1):-yuumiite-, from the combination-yug-umi-ite- DESIDERATIVE-STATIVE-NEGATIVE. There is no grammatical or semantic reason why a different sequence of suffixes could be used to mean ‘not want’, such as DESIDERATIVE- NEGATIVE: speakers simply use the ready-made unit-yuumiite- rather than building up a negative desiderative each time from possible components. Suffix clusters often show greater phonological fusion than other sequences of suffixes, and they are not necessarily equivalent in meaning to the sum of the meanings of their parts. Such semantic independence is well illustrated in the cluster-ngssaar- ‘secretively, unnoticeably, without letting oneself be seen’.
Jacobson (1984:517) derives this suffix complex from-ngssak- ‘small, non-vital …ing’ plus-arar- ‘to … a little at a time’. The link between ‘small’ and ‘secretive’ is not difficult to imagine. One often makes small movements in order to avoid attracting notice. The base atu-ngssaarak- can be used, for example, for someone humming under her breath. In the modern language, the movements no longer need be small: someone could be singing robustly out of the earshot of those from whom she wanted to conceal her actions. The verb in (4) was used to refer to the fact that traditional songs were kept alive through most of this century because ceremonies, forbidden by missionaries, were held in secret.
The routinization of frequently occurring morphological sequences con strains morpheme order to some extent. Yup’ik morpheme order is further constrained in another way. There is a clear distinction between the derivational and inflectional suffix zones. Every noun and verb must have one and only one inflectional ending. For nouns, the ending encodes number and case and, if the referent is possessed, the number and person of the possessor as well. For verbs, the inflectional ending consists of two parts, a mood marker (indicative, interrogative, optative, etc.) and a pronominal suffix specifying the core arguments of the clause. The mood and pronominal suffixes form a tight-knit unit: the choice of mood determines the choice of pronominal paradigm, and the mood and pronominal suffixes are fused.
Templatic systems, by contrast, show a fixed order among morphemes. The order does not necessarily reflect scope considerations. An example of templatic ordering can be seen in languages of the Iroquoian family. As in many polysynthetic langauges, morphological complexity is concentrated in the verb. Verb structure is schematized in Figure 2.
None of these morphemes ever appears outside of its regular position. It would be difficult to argue that the order consistently reflects semantic or grammatical scope. Two of the prepronominal prefixes, the cislocative and the translocative, appear to represent perfectly opposite distinctions. The cislocative typically indicates action directed toward a point of reference (‘come’, ‘bring’), and the translocative indicates motion away (‘go’, ‘take’). Yet the two prefixes occur in different positions in the template. The cislocative is an inner prepronominal prefix, usually appearing immediately before the pronominal prefixes; the translocative is an outer prepronominal prefix, occupying the first or second position in the word.
The difference in the positions of the two directional prefixes reflects their different histories: the cislocative prefix developed some time before the translocative prefix.
Any hope of describing Iroquoian affix order in terms of either synchronic derivation or scope relations is further undermined by the existence of dependen
cies among non-adjacent morphemes. Within the Northern Iroquoian languages, for example, a large number of verb stems require the presence of a prefix te- termed the dualic. The semantic contribution of the dualic is elusive; it some times implies two-ness of some kind, sometimes a change of state or position. (The Cherokee cognate of the Northern dualic prefix has the same shape te-, but it serves as a distributive, spreading events over various locations or absolutive participants.) In any case it is often impossible to predict the appearance of the dualic in the Northern languages. It simply always cooccurs with certain verb stems, a fact that must be learned with each stem. It occurs in the Mohawk verb ‘stand up’, for example, but not in the verb ‘sit down’, though both could be argued to involve a change of position.
The dualic is never adjacent to the verb root; it is always separated from it by at least the pronominal prefix and often many others. We thus have a discontinuous dependency.
Another characteristic of many templatic systems, noted by Rice (1991), is that different areas of the morphology may show different kinds of phonological boundaries. In Athabaskan languages, for example, the inner or ‘conjunct’ prefixes are more tightly fused phonologically than the outer or ‘disjunct’ prefixes. The Iroquoian languages show similar differences between inner and outer prefixes. Such differences are easily explained as a function of their longer history as affixes, subject to more internal sandhi processes.
The ordering properties of layered and templatic structures thus differ in significant ways, with layered structures showing more variability, more often reflecting scope relations among morphemes. In both layered and templatic systems, however, morpheme order is substantially more rigid than word order in sentences.
