¹ Since we wrote the paper which appeared in Lg. 29.143-56 (1953), four further discussions of this subject have come to our attention: M. L. Samuels, The study of Old English phonology, Transactions of the Philological Society 1952 15-47 (London, 1953); M. Daunt, Some notes on Old English phonology, ibid. 48-54; K. Brunner, The Old English vowel phonemes, English studies 34.247-51 (1953); and C. E. Bazell, review in Litera 1.75-7 (1954). Further reference will be made to these studies wherever they contribute to the present discussion, but in order not to obscure the points at issue between Stockwell and Barritt and ourselves, we shall confine such contributions to the footnotes.

² Some Old English graphemic-phonemic correspondences—ae, ea, and a (Studies in linguistics: Occasional papers, No. 4; Washington, 1951). This study will be referred to hereafter by page, the rebuttal in this number of Language by section.

³ Language monograph No. 23 104-9 (Baltimore, 1948).

⁴ Stockwell and Barritt follow the traditional chronology in regarding i-umlaut as later than breaking. On this point we agree with them rather than with Samuels, who argues that breaking was contemporaneous with i-umlaut and continued after i-umlaut was complete (33-46). Although we are in substantial agreement with Samuels in many of his views and conclusions, the evidence that he has so far presented in favor of the lateness of breaking seems to us inadequate. The non-WS material (21, 36-7) consists of fourteen spellings collected from eight different texts written in three (possibly four) dialects at various times from about the 6th to the 9th century. These sporadic monophthongal forms should not be allowed to outweigh the abundant evidence of breaking that we find in the earliest MSS, especially since the fourteen spellings (and some others like them, which Samuels did not include) have already been explained by other scholars. We should like to point out, moreover, that the proper names in Bede's Ecclesiastical history, from which Samuels takes four of his forms, also include spellings like guruiorum (= gyrwa), -angli (=engle), strenaes halc (= strēones-), estranglorum (= ēastengla), and that the two Kentish charters used contain the forms hlothari (= hlōðhere), enfridi (= ēanfriðes), and bercuald (= beorht-). If sporadic Latinisms and scribal errors can be used to prove that breaking had not occurred when these texts were written, they can also be used to prove that i-umlaut was not yet complete in the 8th century, or that WGmc. au and eu were monophthongal in early Old English, or that the cluster htw lost its t in early Kentish. How much to infer from occasional spellings is a matter for personal judgment, but Samuels goes further than we are prepared to follow.

The WS material used to illustrate WS phonology before the time of King Alfred (43-5) consists of eight forms, which appear to be non-WS or local dialect, taken from two documents of the 9th century. It seems to us unsafe to treat these texts as though they contained typical pre-Alfredian West Saxon, first, because of the evidences of dialect mixture and, second, because of the dates at which they were written. The one mentions the date of Alfred's accession, and the other cannot have been composed earlier than the year preceding his birth. Both texts are available in H. Sweet, Oldest English texts, EETS, OS 83.179, 433-4 (1885).

For the Age of Alfred, Samuels cites ie-spellings (chiefly from Bodleian MS Hatton 20 of Alfred's translation of the Pastoral care) such as iernan and gewrietum, to show that breaking occurred after the period of i-umlaut in the first type of word, and that something akin to velar umlaut occurred in the second. But can forms like these be treated in isolation from other ie-spellings in the Hatton MS? It is difficult to assume a diphthong of any kind or of any origin in forms like the following: briengan, hieder, nieðemesð, siendon, sient, tieglan, ðieder, ðienga, ungeriesenlice, wieten (opt.). These examples of ie for historical /i/ are paralleled in the same MS by forms with ie for /i:/; e.g. adriefð, flietað, gegriepð, gestieganne, iedelnes, riece. We should have liked to see Samuels take account of such forms as these in postulating his revised chronology. He might have explained why they and the examples which he himself cites should no longer be regarded as reverse spellings, i.e. as an indication that early WS īe and ie were well on their way to becoming monophthongal by the close of the 9th century.

It would, therefore, seem necessary to have further evidence and discussion before we proceed with Samuels to assume a new sound-change in the form of a secondary i-umlaut (45) to account for the ie in WS wielle, fiellan. Caution seems all the more desirable when we note that forms like tellan and hærfest (adequately accounted for, along with wielle and fiellan, in the accepted chronology) are not satisfactorily explained by Samuels' hypothesis. Nor, by the same token and for the reasons previously stated, do we feel able to accept his arguments as regards the lateness of breaking in relation to i-umlaut.

⁵ Usage varies from MS to MS: in the Vespasian Psalter (British Museum MS Cotton Vespasian A. 1), long vowels and diphthongs are never distinguished by diacritics; in Beowulf and Judith (MS Cotton Vitellius A. 15), there is on the average about one acute in every twenty lines of poetry, and all appear over long vowels or long diphthongs; in the Caedmonian poems (Bodleian MS Junius 11), somewhere between half and two-thirds of the lines contain one or more acutes, but these appear over short as well as long vowel elements. The use of the acute over diphthongs is rare in all of the MSS, so rare that it might be termed sporadic. For detailed information on the usage in the poetic MSS, one may consult the introductions to the various volumes of the Anglo-Saxon poetic records (New York and London, 1931-53). Information concerning other MSS must be sought in scattered sources, too numerous to be listed here.

⁶ This diphthong is indispensable for the description of Northumbrian, Kentish, early Mercian, and early West Saxon.

⁷ When lengthened before a consonant cluster (as in Merc. āld), the vowel became rounded in some dialects of Middle English. But this development is too late to be included in a description of Old English, cf. K. Luick, Historische Grammatik der englischen Sprache 358-64 (Leipzig, 1921); R. Jordan, Handbuch der mittelenglischen Sprache 68-71 (Heidelberg, 1934); K. Brunner, Die englische Sprache 1.224-5 (Halle, 1950); F. Mossé, Manuel de l'anglais du moyen âge, II. Moyen anglais 1.41 (Paris, 1949); and other standard secondary sources.

⁸ For the generally accepted phonetic interpretation of OE h as [h] initially (except perhaps before /w/) and as [x] elsewhere, see K. Bülbring, Altenglisches Elementarbuch 188-9 (Heidelberg, 1902); J. Wright and E. M. Wright, Old English grammar 169-72 (Oxford, 1925); H. C. Wyld, A short history of English 61 (New York, 1927); etc.

⁹ We also anticipate some confusion in the phonemicization of pairs like bā, byge; hēa (pl. of hēah), hēaw (imperative); hygd, hād; s , sēaw; sēo, sēow (pt. of sāwan); Þē, Þēo (dat. sg. of Þēoh); w g, wēa; wē, weg.

¹⁰ Even if they are correct in supposing that OE ea was lengthened in all such cases, it is nevertheless true that they separate the Southern ME reflexes of WS ea from the reflexes of æ. We do not accept some of the theories of lengthening advanced. Although we believe that we understand what Stockwell and Barritt mean by ‘analogical’ lengthening (§7.13), the process is still too vague and undefined to be acceptable as an explanation of any large body of linguistic facts. Lengthening ‘by processes not now described in the traditional handbooks’ (§7.12, fn. 4) is so general and all-inclusive as to be meaningless.

¹¹ At the present time this area is so lacking in uniformity that modern dialectologists recognize several distinct dialects in Southern England. All of the major dialect areas of modern England contain smaller subareas, speech islands, etc. In OE times the major areas may have been more homogeneous than at present, but the evidence suggests that they were by no means uniform.

¹² As an example of the way in which a scribe might respell place names, we offer the following pairs of spellings from two different MS copies of a single document in Walter de Gray Birch, Cartularium Saxonicum 1.466-7 (London, 1885-93): Appincg lond, Apping Land; Fefresham, Febresham; Grafon aea, Grafonea. Four spellings of Worcester occur in two copies of a single document (ibid. 2.266-7): Wigra Ceastre—Weogerna ceastre, Wegerna ceaster—Wigerna cestre. Obviously somebody altered something, and probably not all of the spellings cited represent local usage.

¹³ The phoneme which we label /a/ is that which Stockwell and Barritt call /ɔ/, inaccurately we believe.

¹⁴ The shift in their argument is worth pointing out, since it suggests that the gap between their views and ours may be closing. They originally stated that OE ea and æ were identically treated in Middle English, further maintaining that ‘when secondary influences [which would include lengthening] operate, they operate to affect ae and ea in identical ways when all other conditions are identical’ (8). This is the sentence expunged by §2.11.

¹⁵ The spellings are not limited to the ME diphthongal types this time but include various types indicating that the reflex of ea was not identical with that of æ.

¹⁶ This feature of the orthography is most clearly seen in the Nthb. portions of the Rushworth Gospels (Bodleian MS Auct. D. 2.19), in which the long ēa (< Gmc. au) is frequently replaced by the spelling eo. The same spelling is also frequently substituted for the short ea.

¹⁷ The uses of the acute have already been mentioned, cf. our fn. 5.

¹⁸ We may be misinterpreting the vowel system set forth in §§5.1-2, but it seems to us that Stockwell and Barritt analyze WS ie, ā, ea, and ēa as orthographic representations of four phonemically different vowel nuclei. We suppose that they would also regard Merc. a, æ, , e, ē, ea, and ēa as representing more than one phoneme.

¹⁹ We note a similar suggestion by Samuels: ‘there is ... nothing to show that the fronting in cāse could not have been due simply to the combined influence of the initial palatal consonant and i-mutation’ (36). We hesitate to enter into a detailed discussion of this point, since it is presented only incidentally and may not represent Samuels' considered view. Nevertheless, it should be observed that he has suggested a new sound-change (a sort of combined palatal diphthongization, i-umlaut, and rounding), which must be assumed to have affected cāse alone. Since the change in question did not affect other words, there is no possibility of testing or verifying it.

²⁰ Samuels arrives at the conclusion (43) that the breaking diphthongs became phonemic in the 7th century. Although absolute datings for prehistoric developments are hazardous, we believe that we can accept this view as a working hypothesis.

²¹ We have in mind especially such contrasts as those between /æ/ and /ɔ/, /o/ and /u/, /o/ and /ö/, all of which Stockwell and Barritt seem to regard as phonemic (§5.1). Similar difficulties arise in the analysis of dialects (e.g. Kentish) in which the materials for study are very limited.

²² Cf. Brunner's remarks, Eng. studies 34.1.

²³ Most of §§12.1-2 sounds like old-line historical linguistics restated in contemporary structuralist terms.

²⁴ For some additional minimal and analogous pairs, see Samuels (22-3).

²⁵ Quirk, MLR 45.1-5 (1950).

²⁶ For a fuller discussion of this point, see Bazell, Litera 1.76.