¹ Robert P. Stockwell and C. Westbrook Barritt, Some Old English graphemic-phonemic correspondences, (SIL Occasional Papers, No. 4; 1951); Fernand Mossé, Manuel de l'anglais du moyen âge (1945); M. Daunt, Old English sound-changes reconsidered in relation to scribal tradition and practice, Transactions of the Philological Society 1939.108-37 (1939).

I wish to express my thanks to Sherman M. Kuhn, who graciously read an earlier version of this paper and gave me the benefit of his broad scholarship in Old English. His help does not constitute endorsement of my conclusions, about which he has some serious doubts. As a result of his criticism I have been enabled to eliminate a number of obvious errors, and have tried to clarify certain points on which my earlier discussion impressed him as obscure.

² Sherman M. Kuhn and Randolph Quirk, Lg. 29.143-56 (1953); see now also the same authors in Lg. 31.390-401 (1955), replying to Stockwell and Barritt, Lg. 31.372-89 (1955).

³ See my Manual of phonology, esp. §244 (1955).

⁴ For Old English consonantism I find the arguments of William G. Moulton in Lg. 30.1-42 (1954) entirely convincing. In particular, his insistence on the phonetic (and possibly phonemic) distinction between syllable-initial [h] and postvocalic [x], in all the old Germanic languages and thus in reconstructible Proto-Germanic, is much to be preferred to Stockwell and Barritt's assumption (stated without evidence) that even in Old English h-was something like [x].

The consonant system of Midlands Old English by the time of the Vespasian document was then as follows: four voiceless stops /p t ċ k/, spelled p, t, c, c; four voiced stops /b d ġ g/ of limited distribution, written b, d, g, g; four spirants /f þ s x/, written f, ð ~ þ, s, h ~ g, voiced or voiceless depending on environment; one spirant /j/ (not a semivowel!), always voiced, also spelled g; and /m n r 1 w h/, spelled m, n, r, l, w, h.

⁵ J. Bosworth, The elements of Anglo-Saxon grammar 51 (London, 1823).

^5a Since I have not been able to undertake this task, the following outline may be partly wrong.

⁶ The restriction ‘alphabetical’ is essential: obviously one's assumptions must be different for a morphemic writing system such as that of Chinese, and doubtless in part different for a nonalphabetical but basically phonemic writing system such as some of those of the ancient Near East.

⁷ Rudolf Thurneysen, A grammar of Old Irish, translated from the German by D. A. Binchy and Osborn Bergin (Dublin, 1946). The passage is from §86, corresponding to §82 of the original: ‘Schliesst ein palataler Konsonant oder eine palatale Konsonantengruppe eine haupttonige Silbe, so tritt hinter jeden Vokal oder Diftong ausser ... ein Übergangsoder Gleitelaut i. Er muss sehr hörbar gewesen sein, da er in der Schrift selten unbezeichnet bleibt. Aber er war kein voller Vokal; er bildete mit einem vorhergehenden kurzen Vokal zusammen keine Länge (keinen eigentlichen Diftong), sondern die Silbe blieb kurz.‘ Thurneysen's Handbuch des Alt-Irischen is dated 1909; it is firmly in the prephonemic neogrammarian tradition.

⁸ Fide Norman A. McQuown.

⁹ For both the Vespasian records and the Glossaries, as already noted in the second paragraph of this paper, I have relied on Henry Sweet, Oldest English texts, EETS (1885).

¹⁰ It is highly unlikely that posttonic unstressed e had by these times been retracted to anything like []; but if it had been, the argument still applies, for we still find both writings for the stressed syllabic before identical vocalism in following syllables.

¹¹ That is, in Martinet's sense of ‘functional load’; see his Économie des changements phonetiques: Traité de phonologie diachronique (Berne, 1955). The contrast between /š/ and /ž/ in New English is not rendered nonphonemic by the existence of at most a handful of minimal pairs in any one person's speech; but this contrast does less of the work of keeping utterances apart than does, say, that between NE /p/ and /b/ or between /p/ and /t/.

¹² This is Sweet's view of the symbol ę; see his History of English sounds ² 102 (1888). Kuhn tells me that the symbol was borrowed from Latin writing habits and thus, for Old English, should probably not be described as a cursive form of æ. I believe that the difference does not bear integrally on the line of argument in the text.

¹³ The coalescence also preceded r-metathesis: The Vespasian glosses consistently show forms like orn ‘ran’, born ‘burned’. Later developments, all the way to New English, attest to the simultaneous existence of dialects—even in the Midlands—in which the coalescence did not occur (man, hand, lamb alongside long, strong). The commonest alternative hypothesis proposes an additional separate lower-mid back rounded vowel phoneme from earlier a before nasals. I see no justification for this alternative, but discuss it briefly in §2.4 below.

¹⁴ In sacerd the a may have been long: see Alois Pogatscher, Zur Lautlehre der griechichen, lateinischen, und romanischen Lehnworte im Altenglischen 31 ff., 42, 116-7 (1888).

^14a For these earlier forms from the Corpus glossary, I again rely on Sweet's OET, cited in fn. 9.

¹⁵ Sweet, History of English sounds ² 101.

¹⁶ Taken from the Latin forms in Sweet's reordering of the Glossaries, in his Oldest English texts.

¹⁷ Thurneysen, op. cit. 23.

¹⁸ Sweet, History of English sounds ² 97.

¹⁹ Einar Haugen, First grammatical treatise, Language monograph no. 25, Supplement to Lg. 26:4 (1950).

²⁰ Sweet, History of English sounds ², especially 108-9.