Lexical Strata in English: Morphological Causes, Phonological Effects (review)

This monograph is organized into eight chapters and concludes with a very helpful apparatus—notes, references, a subject index, and an index of words, roots, and affixes. What Giegerich explores here is a theory of base-driven stratification. Such a theory ‘defines strata by reference not to affixes but to affixation bases’ (4). In this model, English has two strata (root and word), and German has three (root, stem, and word). On stratum 1, affixed forms are not rule-governed (affixes are free in principle to attach on more than one stratum), but their potential for linguistic behavior is listed. G also incorporates cyclicity, though with a constrained field of operation. Intriguingly, in his projection for the new model G suggests that well-known alternations such as totem/totemic, atom/atomic are not predictable: ‘such alternations can only be driven by orthographic information’ (5).

In Ch. 2, ‘Affix-driven stratification: The grand illusion’, G reviews key literature behind the English two-stratum model (Siegel 1974, Selkirk 1982, Kiparsky 1982). In the earliest incarnations of the two-stratum theory, English affixes were seen as subject to an affix ordering generalization (AOG) such that Class I affixes (eg. the -ic of atomic) are attached before Class II (atomless). And no Class II affix can occur inside a Class I formation (*atomlessity, cf. atomicity). Once these insights have been incorporated into the lexicon, what is then needed is a theory of bracketing that can express the distinction between kinds of affixes. Kiparsky 1982 supplied such a theory, wherein roots and words are encased in [ … ], but where prefixes are handled as [_prefix and suffixes as _suffix]: compare the bracketing of the compound [[word][count]] with that of [[word]less].

G notes further problems concerning the AOG: (i) some Class I affixes cannot stack nicely (*sensuousize), (ii) some affixes seem to belong to both postulated strata (-ism), and (iii) the neat two-stratum lexicon is interfered with by history: non-Germanic affixes cannot attach to Germanic bases (*bookic), while Germanic affixes are rather promiscuous, attaching to almost any base, whatever its pedigree (musicless, solemnly). Clearly, ‘[t]he significance of any observed effects of the AOG in English is … limited’ (13).

The remainder of Ch. 2 is devoted to an analysis of further problematic cases for the earlier lexicalist model(s), particularly those where affixes appear to attach on both strata. I provide a summary in Table 1 below, adapted from Ch. 2. For those unfamiliar with thinking in morphology, the middle column, Class I affixation, is where strange things happens (stress-shifting, affixation to bound roots, cranberry morphs), while the right-hand column is where the regular, productive, word-based (Class II) morphology works.

The uncontroversial idea of Ch. 3, ‘Principles of base-driven stratification’, is that the English lexicon occupies two strata. Stratum 1 is the root-level morphology. Roots include not only free forms such as [nation] or [keep], but also, in the new model, items such as [al] and [ise]. In addition, roots are recursive; that is, multiple stacking is possible, where each output of each stack is still a root. Thus stratum 1 in English contains entries such as [nation]_R, [[nation]_Ral]_R, [[[nation]_Ral]_Rise]_R, and so on, in which ‘R’ is just a label (‘I am and will behave like a root, but am devoid at this stage of categorial content’). At what point in a derivation do roots—which, note, somewhat counter-intuitively include what we all used to think of as affixes—become words? There are two possibilities: (i) they become words on the last cycle of stratum 1, or (ii) they become words as the first operation on stratum 2. G cites root-to-word conversion (RWC; see 78ff. and see also 106ff., where RWC is renamed as morphological default) as the last operation occurring on stratum 1. Once it has applied, proto-words...