The data

Which data form the basis for the discussion? While Fabb used Walker (1924) and his own collection of items to provide him with examples, I will considerably enlarge the data base with the help of Lehnert (1971) and the OED on CD.¹ Due to the nature of the OED, the use of examples from this source requires some caution and discussion. The problems concern the lemmata on the one hand, and their treatment by the lexicographers on the other.

The OED's near-comprehensive coverage of the English lexicon from the twelfth century onwards is extremely informative, especially with regard to historical studies, but proves to be problematic for synchronic studies like this one, which try to describe the morphological competence of today's speakers. Thus, many of the attested forms may be no longer in usage, although perhaps no information in the relevant entry indicates this. As a general policy, the OED labels these words as 'obsolete' or 'rare', but the decision to classify a word as obsolete is certainly not always easy and the treatment may not be consistent. To illustrate this, it is not immediately clear whether a word with its last citation in, say, 1890, should be considered no longer in usage.

Furthermore, the OED's aiming at comprehensiveness almost necessarily leads to the listing of what one reviewer called 'esoteric words', i.e. attested words that are nevertheless unfamiliar to most native speakers or otherwise somehow strange, and that, consequently, could be argued to constitute doubtful or even irrelevant examples of certain derivational types. While it seems clear that such words should not provide THE crucial evidence for or against certain analyses, it seems unwise to exclude them A PRIORI from one's data base, for the following reasons.

First of all, it is not clear where esotericness starts and where it ends. What is esoteric for one speaker, may be rather natural for another. But even if one would overcome this general difficulty (for example through careful experimental studies with native speakers), there is a second problem. Unfamiliarity or uncomfortableness with a certain word is not necessarily an indication of its morphological ill-formedness, but can have a number of causes, with a violation of morphological restrictions being only one of them. Thus, the rejection of esoteric words by a speaker may depend on pragmatic factors, or be the result of prescriptive rules a speaker applies. For example, linguists who work with native speaker informants often experience that words or sentences are first rejected by informants because the speakers fail to make sense of them, and not because the data violate morphological or syntactic rules of their language. Presented with an appropriate context which provides a possible interpretation, the same informants may readily accept the data presented to them. In essence, the claim that a putative word violates a morphological restriction should therefore be based not only on sound morphological arguments but also on the prior exclusion of other possibly intervening factors. Thirdly, and ideally, one would like to have a theory of morphological competence that can account for everyday words as well as for the esoteric ones. If a theory can handle both, it is to be preferred to theories that have to exclude esoteric words from the range of data they want to explain.

Following this line of reasoning we will not exclude esoteric words as evidence, but we will neither use them as primary evidence. In the vast majority of the cases presented below, I believe we can make a point on the basis of non-esoteric words, with esoteric words only providing additional evidence in some cases. In any case, the OED labels 'obsolete', 'rare', 'rare1' etc. will always be given.

In addition to the attested forms taken from Lehnert, the OED and the linguistic literature, we will also occasionally consider potential word forms that are not found in any of the available sources but are made up to test some of the proposed predictions. In such cases, it should be noted that the potential forms are again only used as ADDITIONAL illustration. Even if the reader does not accept these forms as well-formed, the main argument still holds.

¹ In addition, examples discussed in the pertinent linguistic literature (like Marchand 1969, Jespersen 1942, and many others) have also been included. With regard to Lehnert (1971) and the OED, one anonymous reviewer pointed out that there are even larger dictionaries or word-lists available (e.g. Brown 1963). These could of course provide even more counterexamples to Fabb's generalizations than those presented below, but they were not available to me.