Separationist theories aim at finding a principled solution to the frequently observed mismatches between form and function in morphology. One affix may often express a number of different grammatical or semantic functions, whereas one function is often expressed by a whole range of affixes including zero. Consider, for example, English plural nouns which can be marked by a number of different suffixes (e.g. -s as in dogs, -en as in oxen, -∅ as in sheep). At the same time, these suffixes are used elsewhere in the language to mark completely different categories. For instance, the suffix -s expresses also third person singular on verbs (as in works) and genitive case (as in Mary's), and the suffix -en is used for the purposes of verbal derivation (black - to blacken). To complicate matters further for sign-based theories, there appear to be affixes that are meaningless. Thus, the suffix -al in words such as syntactical, or dramatical can be omitted without change of meaning, which is an indication of the meaninglessness of -al in these derivatives. Similarly, in German compounds one often finds a linking element (so-called Fugenmorphem, as in Verwaltung-s-gebäude 'administration building') that does not carry any meaning (see e.g. Wiese 1996a:143-147 for some discussion).

Although traditional sign-based approaches to morphology (called LMH-theories by Beard, from 'Lexical Morpheme Hypothesis') acknowledge the existence of such mismatches, ancillary theories are needed to explain these violations of the assumed one-to-one correspondence of form and meaning. For example, in order to account for the versatility of -s in English, LMH-morphologists would have to posit three homophonous suffixes -s, each with its own function. Although this leads to a perhaps undesired proliferation of suffixes, there are independent arguments for such an approach, namely that there are striking differences in behavior between these three suffixes. One suffix -s attaches to verbs, the other attaches to nouns, and the third (i.e. the clitic) attaches to noun phrases (hence -s on verbs is never interpreted as plural marking, for example). In addition, the nominal suffix and the clitic differ in their allomorphy due to haplology effects, i.e. the clitic has a zero-allomorph which surfaces after plural -s (e.g. the boys'). In a sign-based approach such differences between suffixes are expected.

Separationist theories solve the problem in a more principled way by strictly separating meaning and form in morphology by postulating the independence of lexical derivation from its morphological spell-out. In this view, morphological asymmetry results from using several spelling operations to express a single derivation or from a single spelling operation used to express several derivations (Beard 1995:47-48). The central claim is that "the dissociation of derivation and morphology in word formation and inflection accounts for asymmetries of affixal sound and meaning such as polysemy, synonymy, zero and empty morphology, which plague other approaches" (Beard 1990a: 103).

Let us illustrate this with one of Beard's own examples, deadjectival nominalizations (Beard 1995:50-51). There are at least five suffixes which are used to mark such nouns in English, -th, -ce, -ity, -ness, and -∅:

(1) 

To account for these facts Beard assumes that there is only one rule of lexical derivation with five possible spell-out operations. The derivational rule does no more than transpose underlying qualitative adjectives into nouns. The crucial argument for the existence of such a unitary derivational rule is of course that the meaning of the different suffixes in (1) is identical. Note also that the restriction of the lexical derivational rule to non-relational adjectives holds for all suffixes. According to LMBM, the appropriate phonological spell-out of the derivational rule is done separately and is conditioned by different restrictions (such as the phonological make-up of the base, the semantics of the base, etc.). Hence Beard can state that "[a]ffixation [i.e. morphological spell-out, I.P.] truly operates on conditions that differ radically from those determining derivation (meaning)" (Beard 1995:51). No ancillary theories are needed to solve the mismatches: zero morphology is derivation without subsequent spell-out operations (see white_NOUN), empty morphemes are spell-out operations without derivation (see syntactic-al above), and morphological asymmetry in general results from the multiple mapping of derivations and spell-out possibilities. All mismatches predicted by this model occur and other types of mismatches do not occur.

Having outlined the basic ideas of separationist theories, let us see how separationist theories deal with verbal derivation in English.