Morphological analysis vs. etymology

In contrast with the definitions of the preceding section, the following discussion will be simplified: Attention will be restricted to the combination of a root with one other morpheme. The principles formulated for such problematic cases carry over to the segmentation of more complex stems.

Morphological analysis consists of segmentation and classification. Problems of segmentation are always problems of classification, too: segmenting a complex form into components A and B requires assigning A and B to a class.

In the examples accompanying the definitions mentioned, segmentation of the complex forms and classification of the components as particular types of forms was straightforward because these components were members of substitution classes which generate productive patterns of complex stem formation. Methodological problems arise if one or both of the components do not combine freely or if the formation is not compositional, i.e. there is no regular relation between expression structure and meaning structure.

Consider English aspect, prospect and respect as a first example. They are composed of a bound root -spect which recurs in them (and in stems of entirely different structure like spectator) and a set of prefixes which recur in other, more transparent English words like regain. However, while one or another of these prefixes may have a more or less well-defined function, the morpheme -spect cannot be assigned a meaning which would recur in the meanings of the three complex stems. Such cases are commonly the result of diachronic processes which isolate complexes which once had been formed by productive rules from their paradigmatic relations, so they fossilize. It is therefore tempting to reconstruct the original stage and analyse the complex as consisting of those elements which composed it by the rules of that stage. (In the present example, this would reveal the meaning ‘look’ for -spect.) This, however, would mean doing etymology instead of morphological analysis. While etymology may be a heuristic aid in morphological analysis, it is inappropriate to pass it off as synchronic morphological analysis. The question here is whether the synchronic description of English aspect, prospect and respect profits if we employ a morpheme -spect ‘look’ for it. The answer is probably ‘no’.

Roots only occurring with a set of derivational affixes

Roots which do not inflect by themselves but only occur in combination with one of a set of derivational affixes are well-known from English verbs with a Latin-Romance etymology. The two series appose, compose, depose, dispose, impose, repose and compel, dispel, impel, repel may stand for many others of this kind. These series differ from the one of the preceding section not only in being verbs, but also in being more regular. Assigning a generic semantic component ‘put’ to the root -pose and a generic semantic component ‘move by force’ to -pel, though not enabling an entirely compositional analysis of these stems, does grasp aspects of their meaning. Significata of linguistic signs can be very abstract ...

Unique morphemes

Basics

A unique morpheme is one which only occurs in combination with one other linguistic sign. Examples include -cant in English decant, cran- and rasp- in English cranberry and raspberry and Schorn- in German Schornstein (?:stone) ‘chimney’. The unique context may be a morphological one, as in these examples, or a syntactic one, as in the German idiom (jemandem) Paroli bieten (somebody:Dat paroli present:Inf) ‘to defy (somebody)’.

Since unique morphemes cannot, by definition, be used in different contexts, their meaning cannot be ascertained by methods of synchronic descriptive linguistics. It would, however, be mistaken to say that they are meaningless. If English cran- and rasp- were meaningless, then cranberry and raspberry would be synonymous. An unspecifiable meaning is not the same as no meaning. A unique morpheme is still a morpheme; it has a significans and a significatum. The latter is distinct, but cannot be specified. Spelling it out: Suppose German Stein as a linguistic sign has the following structure:

Then Schorn- has something like the following structure:

Combining these two morphemes in the compound Schornstein does not, of course, yield its meaning by a compositional procedure. The word remains idiomatic. This, however, is no different from certain compounds which are composed by completely transparent and productive stems like harvestman (kind of spider), which is neither a man nor has it anything to do with harvest.

Unique morphemes differ gradually from morphemes some of whose semantic features are recognizable without, however, sufficing to distinguish such a morpheme from all others. This is the case of the English roots -pose and -pel mentioned in the preceding section.

Roots only occurring with a particular derivational affix

Certain unique morphemes only cooccur with one particular derivational formative. If this has a clear-cut function, the case may seem tractable. German has a small set of rather productive verb suffixes, including -el Attenuative, as in zischen ‘hiss’ – zischeln ‘hiss lightly/intermittently’. The phonotactic shape of verb stems like schmunzel- ‘grin benignly’ and trippel- ‘patter’ leaves little doubt that they contain this formative, and their meaning embodies an attenuative component, too. However, the roots schmunz- and tripp- do not occur outside this combination. Assigning them a meaning by subtracting the attenuative feature from the meanings of schmunzel- and trippel- may be possible, but appears like an idle linguistic exercise. It is more prudent to recognize a unique morpheme in such cases.

Things are even more desperate if the meaning or function of the derivational operator cannot be pinned down. If we cannot say what de- means, -cant in English decant remains obscure. German vergess- ‘forget’ and verlier- ‘loose’ obviously share a derivational prefix which is otherwise productive, but has lots of different functions. Consequently, nothing can be made of -gess- and -lier-.

Roots only occurring in a particular compound

The same goes for roots which only occur in one particular compound. Beside the famous examples of the first morphemes of English cranberry and raspberry, their German counterparts Preiselbeere and Himbeere (as well as most other German berries) may be mentioned. If such nominal compounds AB are semantically compositional at least to the extent that they denote a kind of B determined by A, then a tentative meaning might be assigned to the unique morpheme A, viz. whatever it is that distinguishes a cranberry or a raspberry from other berries. This, however, is probably superfluous and also hazardous in the sense that there is little chance of matching some diachronic reality.

Completely idiomaticized complex stems

Some complex stems are straightforwardly segmentable on a formal basis, but their meaning bears no relation to the meanings of the components. This is the case of the harvestman mentioned before, but also of verbs of Germanic origin like understand and verbs of Romance origin like conceive, deceive, perceive, receive. This series differs from the series based on -pose and -pel mentioned above because even the fact that the root -ceive combines with a set of prefixes which otherwise occur productively in such verbs provides no clue to its meaning.

In all the cases of §§2-4, the structure of the expression leaves little or no doubt that it is composed of two elements which would otherwise pass as morphemes. The problem is, in all these cases, the semantics. These are, thus, cases where the form and the meaning of a complex unit do not match (completely). This, however, should not prompt the analyst to throw the baby out with the bathwater: The fact that the meaning is not composed in a way analogous to the composition of the form does not entail that the form is not composed.

Thematic affixes

Some roots require a thematic affix before they inflect. For instance, the Latin root laud- ‘praise’ only conjugates if provided with the thematic vowel -a to produce the verb stem lauda-. It is clear, however, that the root is laud- and not lauda- because the noun laus ‘praise, eulogy’ does decline on the basis of the root laud-. Likewise, nouns like German Eiche ‘oak’ end in a thematic vowel -e without which they do not decline. It is, however, not part of the root, because it disappears in compounds like Eichbaum ‘oak tree’.

Whenever there are stems derived into their distribution class by such a thematic suffix, this may be assigned the function of marking this class. Thus, Latin -a may mark a stem as a verb stem. (It would be homonymous, though, with the -a which marks a stem as a [feminine] noun stem.) In cases like the German nominal thematic -e, this works less well as there is no derived noun substantivized by it. Here we are approaching a certain type of submorphemic units, viz. elements which have a significans like a morpheme, but no significatum.

Such thematic formatives are not only presupposed by the inflection; they also commonly merge in phonologically more or less regular ways with the inflection desinences. As a consequence, the thematic formative is considered an inseparable part of the desinence. If the language has a set of such thematic formatives for stems of a given word class, this then generates inflection classes.

Segmentability and phonotaxis

Examples like the above illustrate the following situation: There is an inventory of derivational affixes and (free or bound) roots. The structure of many stems transparently involves a combination of an affix and a root. The transparency results from sufficient recurrence of the affixes, from the phonotactic shape of the derived stem and from its grammatical properties. However, there is no regular relationship between the structure of the significans and the structure of the significatum. Consequently, such stems cannot be derived by rules of word formation.

Lexicalization and subsequent semantic and/or phonological changes transform stems into roots. German has a derivational prefix ge- which was productive many centuries ago. Many of its products are so transparently related to an existent base that there is no doubt about their formation, like gedenken ‘commemorate’, derived from denken ‘think’.

Then there are many verbs like genießen ‘enjoy’, which display the same structure, but cannot be derived from anything, since what appears to be the root does not recur.1 Stems like genieß are halfway between complex stem and root.

Finally, Middle High German gelouben ‘believe’, originally derived in the same way, is glauben in contemporary German. Despite the existence of a few alternations of the type G(e)leise ‘rails’, decomposing glaub into a prefix g- and a root laub would be inappropriate in a synchronic analysis; glaub is now indisputably a root.

This example series shows how the phonotactics of verb roots depends on the morphological analysis: Because of glaub, we have to reckon with onset clusters of the structure CL (which, of course, we have to do for other reasons, too). But if genieß is a root, we have to reckon with (native) bisyllabic roots stressed on the second syllable, which complicates our phonotaxis of verb roots.

The decision in such matters is one of theoretical elegance: That analysis is the best which allows for the most general description of the facts. As long as we don't have the complete formal language description at our disposition which we could analyze for complexity or simplicity, it seems prudent to avoid the ‘rule-list fallacy’: The necessity to list the stem genieß in the inventory does not exclude the possibility of recognizing a formal structure consisting of two morphemes, even if one cannot pin down their function.

1 It is true that there is the noun Nießbrauch ‘usufruct’; but this is of little help.