The percentages of words formed in English and of those borrowed from other languages in the recent corpuses contrast strikingly with those in The Shorter OED, as reported by Thomas Finkenstaedt (1973, 118-56). In the following table, the SOED percentages represent the history of English over approximately 1,200 years, as recorded in that dictionary. The OED2 percentages are of the sample from volume 1 of the second edition of the OED (1989), representing a slightly longer chronological range but taken from only the first one-twentieth of the alphabet. The 'Recent' percentages are an average of the five corpuses used above that recorded new words from about a twenty-five-year period, 1963-88:

SOED OED2 Recent

Native formations 25.6 81.0 91.7

Loanwords 70.4 18.8 6.4

Unknown origin 4.0 0.3 1.9

These figures are certainly skewed. The extraordinarily high percentage of loanwords in the Finkenstaedt statistics for the SOED words is belied by both a sample from the OED2 and five recent collections, which differ among themselves comparatively little (their percentage of loanwords ranging from 4.3 to 7.5). Borrowing may well have declined in recent years as a source of new words in English, but a decline of the proportion suggested by the discrepancies between these figures is unbelievable.

A partial reason for the discrepancy is that the SOED data concerns only headwords, whereas the OED2 and the recent corpuses include new senses of old words. If the percentages are adjusted by omitting all shifts (semantic and grammatical), the native/loanword ratio becomes 75/24 for the OED2 and 90/8 for the recent collections. That is slightly closer but still far from the SOEDs 26/70.

Finkenstaedt (1973, 117) himself points to another possible cause of the problem. The SOED (and OED) etymologies by preference cite the earliest classical etyma. When such etymologies are reported as the sources of the English vocabulary, we have an incorrect account of the origins of English words since many words are composed in English of morphemes from classical languages. Every OED etymology has to be evaluated in terms of what we want to know about the origin of the form and what the OED editors were probably telling us. Uncritical statistical reports of OED etymologies are likely to be not just useless but badly misleading. The OED etymologies need to be reworked to clarify what they report.

It is probable that borrowing has declined in importance as a source of new words in English. That it has declined as radically as a comparison of the SOED figures with those of recent studies would suggest is very unlikely.

• Types of recent neologisms

Estimates of the relative productiveness of one or another type of word formation are subject to many variables and consequently uncertainties. Not least among those is establishing the correct etymology of a word. For example, unconscious 'that part of the mind not available to introspection, which nevertheless affects behaviour' might

• Greek and Latin

Greek and Latin formatives are highly productive sources for new technical terms coined in English. Consequently, very recent words in the scientific and technical registers that look like loans from the classical languages may actually have been formed within English from morphemes abstracted from loanwords that entered English long ago.

• Gauging changss in the si-Ze of the vocabulary

Given such ffuctuation in what we mean by the 'vocabulary' of English and the problems in counting it, any estimate of its increase in size since 1776 must be viewed sceptically. Yet it seems certain that the vocabulary has increased significandy. In a sample of words from the OED (the

• Wordfrequency

The frequency with which words are used has implications as a practical matter in stylistics, for example in setting an appropriate reading level for school books. The word frequencies in two standard corpuses of English, the Brown Corpus for American and the LOB Corpus for British, are reported by Hofland and

• Affixing

Affixation is also in some respects a lexis/grammar borderline phenomenon (Hirtle 1970; Hudson 1975; Ljung 1976). It is also an area of word formation particularly susceptible to vogues and oddities of use, such as the several senses of non - (Algeo 1971), interposing as in in-damn-defensible (McMillan 1980), and the