¹ Bernard Bloch and George L. Trager, in Tables for a system of phonetic description (New Haven, 1940), used [s] for the dorsal variety and called it ‘laminal’, and presented a new symbol for the apical sibilant. But because in this paper we must avoid a prejudicial implication that one is the norm and the other an aberration, the two are here represented by an ad-hoc use of underlined and dotted [s] symbols, where the line may be thought of as a diagram of the tongue surface and the dot as diagramming the tip of the tongue. This makes plain [s] or /s/ available for use as a cover-symbol for either or both.

² The writer's [s] hiss is about a musical fourth higher in pitch when he produces the dorsal sibilant that is normal in his native English than when he produces the carefully drilled apical sibilant of his Castilian Spanish; from the latter down to the pitch of his [š] (also apical, as it is for most Americans but by no means nearly all) is a slightly greater musical interval, about a fifth.

³ For a keen ear or kinesthetic perception, there may be more discernible varieties than these two, filling in the gap between them; but these seem to be relatively rare, and usually an [s] can be unhesitatingly assigned to one of the two extreme articulations. About one-third of the writer's phonetics students use apical articulation for [s, z] in hiss and his; nearly two-thirds use dorsal articulation; about one in twenty is atypical. For Europeans our data are scanty; see fn. 23.

⁴ In any language that has only one /s/ phoneme, the difference between dorsal and apical [s] has by definition some entirely different status. It is never ‘heard’ in the sense that the word has here; and yet there are various possibilities, of which two will be mentioned by way of illustration. In Japanese only the apical variety occurs before /u/ and (if we write nothing where some descriptions have /u/) before consonant or pause, as in desu ga /desuŋa/ ‘though it is’, desu ka /deska/ 'is it?'. Elsewhere in Japanese, e.g. before /a/, the dorsal and apical [s] are both current and appear to be distributed pretty much as in English. In English the freedom is complete: there is no absolute limitation, no structural restriction; each English speaker does what he pleases, so to speak, in every context. And yet most of us use one of the two extreme varieties wherever it is possible, in a sense of 'possible' that is personal to each speaker, the other one either rarely or never. The writer, for example, has only the dorsal variety otherwise than before the apico-alveolar consonants; but before /l/ he has only the apical variety, and before /t, n/ he has either, in free and capricious variation; and although /t, n/ are otherwise always apical, in /st, sn/ they are dorsal whenever the sibilant is. Other speakers have quite different distributions, and theoretically any distribution at all is possible, but the one just described may well be the commonest. Descriptivists who use the term ‘allophone’ sparingly will then say, with Trager and Smith (An outline of English structure 34), ‘the phonemes /f, θ, s, š/ have one allophone each.‘

⁵ For phonologists like Roman Jakobson and John Lotz, however, this is no accident. They would point out that few languages have anywhere a distinctive feature that could serve to separate dorsal and apical [s] into two phonemes. At present such phonological arguments seem to be very strong if not absolutely decisive, and they will not be avoided in this paper. Being based on experience with many languages, they have the authority of good statistics; but in single instances they may be upset.

One language that has both these sibilant phonemes (besides /š/) is Basque. Jean Larrasquet, Le basque de la Basse-Soule orientale (Paris, 1939), clearly describes the dorsal [š] and the dorsal sibilant, but calls the apical sibilant ‘cacuminal’.

⁶ That one spirant out of many obstruents should have such a unique phonological status is at least highly improbable; see fn. 5.

⁷ Carl Karstien, Historische deutsche Grammatik (Heidelberg, 1939). Karstien is cited here as a responsible exponent of standard doctrine, differing from others principally in his relative willingness to speak out on difficult points. The spirant z is characterized without argument on page 125, the s on page 138 with a reference to Behaghel for the literature. But Behaghel's sources show only that s was employed for foreign [š] sounds in early times, when sch did not yet have the [š] value because it represented an earlier stage on the road from /sk/. The compromise identification of s with Polish ś appears to be Karstien's invention; it does not demonstrate a very firm grasp of Polish phonetics.

⁸ Another argument from general phonological theory; cf. fn. 5.

⁹ Where MHG s is not /s/ or /z/ today, it is /š/. That is as far as the evidence goes to show that it was always [ś] or always [š] in MHG. and that is not far enough.

¹⁰ Mildred K. Pope, From Latin to modern French 207 (Manchester, 1934). While this book presents hardly any theoretical innovations, it appears to overlook nothing whatever in its 600 very full pages, and is organized with extraordinary thoroughness and clarity.

¹¹ The symbol /ç/, borrowed from French orthography (not from any phonetic alphabet), and the cover-symbols /s/ and [s] of fn. 1, are used at this point to avoid prejudicing the argument by anticipating data introduced later.

¹² Except when palatalization supervened much later than the borrowing, as in spacious.

¹³ It is of course conceivable that buzzel had MHG /c/ [ts], which was occasionally spelled zz instead of tz; but the same would be true of any single MHG example, and it is only the accumulation of examples that establishes the normal correspondence between the two languages. This word was the name of a wine vessel, a cask or jug, taken over from the French original by the German paraphraser because (presumably) he was not sure of its meaning. For the vowel quantity see fn. 14. English bushel is another word, normal OF boissel, Anglo-Norman bousselle.

¹⁴ The MHG vowel had to be short before a geminate (and therefore necessarily fortis) obstruent between vowels, and single s between vowels was presumably already lenis in this stage of MHG, so that this küssen says nothing about the quantity of either ou or ss in OF, but only says that OF ss was fortis; it is agreed (on other grounds) that OF ss was not long, but was merely written double to distinguish it from voiced s between vowels. The unexpected umlaut, ü rather than u, like the modern i for this ü, says nothing about the OF ou either; vacillation between ü and u, or between ü and i, is common in German in words not firmly anchored in the derivational system, as is küssen ‘kissing’ by a regular pattern that relates it formally and semantically to Kuss ‘kiss’. The two MHG words küssen were available for punning, as in Walther's si hât ein küssen, daz ist rôt (54.7).

¹⁵ The area certainly did not extend into English, nor apparently outside OF and MHG in any other direction, being sealed off to the eastward especially by the quite different sibilant systems of the Slavic languages; the somewhat similar Castilian Spanish situation belongs to a later period.

¹⁶ There is a possibility that re-examination will find the MHG sibilant contrast surviving in Low Allemanic dialects where the three words isst, ist, and fischt have been somewhat unreliably reported to be all different in their spirants.

¹⁷ It is generally held that MHG single s was lenis between vowels already in the 13th century, perhaps even initially before a vowel; but note that there was no single fortis s for it to contrast with in either of these positions, and hence no separate /z/ phoneme. This could only be established later, through at least two shifts: merging of ss with zz, and shortening of all geminates. The modern standard opposes /z/ and /s/ only when a long vowel or diphthong precedes and a vowel follows, as in reisen and reissen—extremely light functional burdening.

¹⁸ This [t] is not underlined here, but was probably dorsal too; the difference is not worth inquiring into, since it could not have had any consequences.

¹⁹ The highest vowels /i, ü/ maintained their status as extreme vowels; and if they were ever retracted, then that divergence from /i, ü/ elsewhere was cancelled out before it had any consequences. It is practically certain that they were compensatorily lengthened when a following [s] vanished before vowel, along with the other vowels discussed below, and then also analogically lengthened before preserved ss like the other vowels; but because we have no evidence on their retracted qualities (if any), they can add nothing to the argument and are therefore omitted from the discussion. Fortunately, for the same reasons, they could not weaken the argument either.

²⁰ The approximate simultaneity of the two gradual phonetic drifts is not involved in the argument. It is hard to see how there could have been any significant connection between them, since /c/ did not occur before a consonant, the only position where /s/ changed so early. At most it might be suggested that the apical articulation of s was not ancient, and the contrast between the two sibilant articulations developed at this very time by way of maintaining at least some phonetic difference. But it is easier to suppose that the phonetic contrast between dorsal and apical was ancient—not that it matters.

There was also an early OF /č/ [tš], and there were voiced partners for /c/ and /č/, making altogether four affricates which now lost their stop component phonetically. Further, there was a /z/ beside the /s/; but there was no opposition between /z/ and /s/ before consonants, where the phonetic change occurred that we are now examining: one occurred before a voiced consonant and the other before a voiceless consonant. The /z/ before a consonant faded out, it seems, at the same time as /s/ before a consonant, and similarly by way of voiced h. Its very first weakening was to a [z] so weakly squeezed as to seem like a slack [d] (as when in Southern American English business becomes [bidnis]), so that it was often written d and gave modern English meddle beside French mêler from mesler. The effects of its disappearance upon the vowel system were identical with the effects of the loss of /s/, and there were the same analogies before preserved /z/, so that including the history and sequels of /z/ here would add nothing to the discussion. Similarly, what is here said of older /c/ is equally valid, mutatis mutandis, for its voiced counterpart. As for /č/ and its voiced counterpart, they are outside this picture, unless we suppose that they lost their stop component because there was room for a new /š/ and /ž/ after preserved /s, z/ had become usually dorsal at last.

²¹ We speak as if word-final sounds were followed by pause, because the sandhi data seem neither to add to the argument nor to detract from it.

²² Communicated in conversation by G. R. Shipman from his unpublished materials. Meigret was a spelling-reformer, and hence was seldom deceived by orthographic tradition.

²³ Wilhelm Viëtor, Elemente der Phonetik des Deutschen, Englischen und Französischen (Leipzig, 1915), quotes numerous authorities and concludes that the sibilants [s, z] are apical in English and usually in North German, but predominantly dorsal in South (High) German and French. The statement for English is certainly too categorical. Incidentally, Viëtor quotes numerous statements about older German and French sibilants, among which views similar to those in this paper are casually included (and rejected), for example that of Wilmanns for MHG.