1 In chronological order the important phonetic monographs are: Jón Ófeigsson, Træk af moderne Islandsk Lydlære, in S. Blöndal, Islandsk-dansk Ordbog xiv-xxvii (Reykjavík, 1920–24); Valtýr GuÐmundsson, Islandsk Grammatik (Copenhagen, 1922); Kemp Malone, The phonology of modern Icelandic (Menasha, Wis., 1923); Stefán Einarsson, Beiträge zur Phonetik der Isländischen Sprache (Oslo, 1927); Stefán Einarsson, A specimen of southern Icelandic speech (Oslo, 1931); Sveinbjörn Sveinbjörnsson, Icelandic phonetics (Aarhus, 1933); Bruno Kress, Die haute des modernen Isländischen (Berlin, 1937); Sveinn Bergsveinsson, Grundfragen der isländischen Satzphonetik (Copenhagen and Berlin, 1941); Stefán Einarsson, Icelandic (Baltimore, 1945); Björn GuÐfinnsson, Mállyzkur (Reykjavik, 1946); Björn GuÐfinnsson, Breytingar á framburÐi og stafsetningu (Reykjavik, 1947). Earlier studies by Storm, Sweet, and Goodwin (Svenska Landsmålen 1905–8) may also be consulted with profit. Important is Eli Fischer-Jørgensen's review of Bergsveinsson, Nordisk tidsskrift for tale og stemme 6.1–20 (1942).

2 The paper was published as The phonemes of modern Icelandic, Studies in honor of Albert Morey Sturtevant 5–21 (Lawrence, Kansas, 1952); the note was entitled Long and short in Icelandic phonemics, Lg. 29.61–2 (1953). Cf. the critique by Håkon Hamre in Maal og Minne 1955. 86.

3 The only previous considerations of Icelandic structures are, chronologically arranged, this writer's paper On the consonant pattern of modern Icelandic, Acta linguistica 2.98–107 (1940); Stefán Einarsson, Um kerfisbundnar hljóÐbreytinger í íslenzku (Studia Islandica 10; Copenhagen and Reykjavik, 1949); Árni BöÐvarsson, SitthvaÐ um lokhljóÐ, AfmæliskveÐja til Alexanders Jóhannessonar 1–8 (Reykjavik, 1953).

4 Informants include Icelandic students at the University of Wisconsin from 1940 to the present, and numerous Icelanders encountered on two visits to Iceland in 1955 (September) and 1956 (February to May).

4a [The author prefers to cite linguistic forms in roman type rather than in the italics customary in this journal.—Editor]

5 For valuable methodological statements see Zellig S. Harris, Methods in structural linguistics (Chicago, 1951), esp. 5 ff.; Charles F. Hockett, A manual of phonology 176 ff. (Baltimore, 1955).

6 One does not use a different name for the structure of one apple and for that of all apples. It does not, at least, affect the nature of one's description, say of vowels and consonants, whether one is thinking of a specific instance or of the language in general when one says that in pa the consonant /p/ precedes the vowel /a/. That there is no other difference in Hockett's use of ‘pattern’ and ‘structure’ becomes evident from the definitions he gives: pattern is ‘a stock of ultimate phonological constituents and a set of (hierarchic) arrangements in which they occur in one or another utterance’; structure is ‘the spatio-temporal srrangement of the parts of a specific historical event’. The application of a little discourse snalysis to these statements shows that they are equivalent except for the difference of general and particular. The dress and the dress pattern he uses as an analogy have identical not just ‘similar‘) structures; this identity is the pattern.

7 For the early Icelandic orthography, see my edition and translation of the First grammatical treatise (Language Monograph No. 25, 1950).

8 From Stefán Einarsson, Icelandic 2–3, with some minor changes of symbols.

9 Hockett, Manual 52 ff.

10 Initial /þ/ may begin with a closure which makes it sound like /t/, as I have often had occasion to observe, e.g. in þaÐ; the same phenomenon occurs in þú as pronounced by Bergsveinsson (2.126), cf. comment by Fischer-Jørgensen 4.

11 Hockett, Manual 165.

12 Beiträge 36; cf. also the discussion in Kress 169.

13 Beiträge 12. According to Halldór Sigurdsson, Icelandic children misspell geta as gjeta.

14 Cf. Árni BöÐvarsson, SitthvaÐ um lokhljóÐ 7; but his explanation of gæi as being deliberately kept apart from gægi 'window'· is unlikely. This is a common loanword phenomenon, with many parallels in the other Scandinavian languages; cf. the pronunciation with stop [k] and [g] in words like Norw. gear [gi:r] 'gear' and gis [gis:] ‘call to pigs’ for the usual spirants before front vowels.

15 Lg. 29.62. This is exactly what Bergsveinsson does in postulating length of vowels as phonemic, that of consonants as allophonic.

16 For a similar suggestion for Norwegian see Haugen, On the stressed vowel systems of Norwegian, Illinois studies in lang. and lit. 29.1.69 (1942); Norwegian language in America 419–20 (Philadelphia, 1953).

17 Einarsson, Beiträge 102; Bergsveinsson, Grundfragen 83.

18 For information on stress distribution see Einarsson, Beiträge 126–7.

19 Cf. the sentences given by Fischer-Jørgensen 18 ff., which clearly point to the same conclusion.

20 This appears to be what Bergsveinsson is trying to say in his passage on accent, Grundfragen 177 ff. He claims that ‘sentence accent’ is somehow different from ‘word accent’ because the former has ‘psychic’ Voraussetzungen. The difference, however, is rather one of distribution than of the phonemic entities themselves. His recorded material shows clearly the need of marking accent. His division into four degrees of stress corresponds to ours, if his two middle categories are combined into one (our secondary); there is no evidence that the perceptual difference between them which he has noted is anything but allophonic. (Fischer-Jørgensen, who listened to his recording, found his designation of these two quite uncertain; cf. her review 17.)

21 Einarsson also uses the half-long mark for the first consonant in a cluster, but from our point of view this is to be regarded as an allophone of full length, correlated to following consonants.

22 This treatment coincides exactly with the point of view expressed by Eli Fischer-Jørgensen in her review of Bergsveinsson 16: ‘Ved akcent forstaas fremhævelse; den erkendes subjektivt. De objektive midler til denne fremhævelse er intensitet, stigende melodi og varighed.‘ On the melody see below.

23 This also has the purely orthographic advantage of not confusing the Icelandic reader by placing accents over the vowels.

24 The transcriptions are those of Einarsson, Icelandic 10; note that [a] is phonetically central, with front and back allophones in [ai] and [au] respectively.

25 Blöndal xxvii; Björn GuÐfinnsson, Breytingar á framburÐi og stafsetningu 29 (Reykjavik, 1947).

26 Cf. Freystein Gunnarsson, Ritreglur 1 (Akureyri, 1942).

27 Icelandic 11.

28 KonraÐ Gíslason, Oldnordisk Formlære (Copenhagen, 1858); Valtýr GuÐmundsson, Islandsk Grammatik 3 (Copenhagen, 1922).

29 Cf. the following spellings in a letter of 1567 (Dipl. Isl. 15.1): mijnu (mínu), tijundun (tíundum), kuijttan (kvíttan), þijngvelli (þingvelli); ij was of course a common symbol for ii.

30 GuÐfinnsson, Breytingar 27–8.

31 For a more complex statement see Einarsson, Icelandic 9. Some writers have changed the orthography here to agree with their pronunciation, writing á ei í ó ú ý au for a e i o u y ö respectively, in words like langur : lángur etc.

32 GuÐfinnsson, Breytingar 29–30.

33 The system set up here is surprisingly similar to that which Einarsson assumes for pre-19th-century Icelandic in his essay Um kerfisbundnar (24), but not for the 20th century. I had not seen his essay when this system was worked out, entirely on a synchronic basis; my own system was originally set up exactly like his:

But further consideration led to the new formulation.

34 Cf. discussion in GuÐfinnsson, Mállyzkur 65, who finds them intermediate between long and short; this is not surprising in view of the absence of contrast. If there is a pronunciation of múgi with short [u], however, as suggested by Ófeigsson, who writes it [mui:ji], we would be forced to accept a complex nucleus with three members: /muwj'i/; the parallel word rúgi ‘rye’ (dat. sg.), however, he writes [ru:ji], i.e. /ruw'ji/. GuÐfmnsson claims that the usual pronunciation is /muw'ji/, and this is confirmed by my informants.

35 Ísl. MálfræÐi 7 (Reykjavík, 1923), here cited from Einarsson, Beiträge 21–2.

36 This is true in spite of the examples given by Malone (14) of short consonants in three-consonant clusters, e.g. vöxtur, which he transcribes as [vöxsdør], contrary to Ófeigsson's [vöx·sdør]; whether or not these are physically long, at least they bear the accent, which the following consonant does not.

37 William G. Moulton, Syllabic nuclei and final consonant clusters in German, For Roman Jakobson 372–81 (The Hague, 1956).

38 Beiträge 6.

39 Haugen, The consonant pattern (reference above). Note that Icelanders speaking English invariably preaspirate in words like met, got.

40 The nasal /n/ has a set of allophones before spirants; phoneticians have differed as to whether they are vocalic or consonantal. For the discussion see Einarsson, Nasal plus spirant or liquid in Icelandic, JEGP 39.462–4 (1940). There seems no reason to differ with his analysis of these as allophones of /n/; Eli Fischer-Jørgensen confirms it in the previously cited review of Bergsveinsson (8). There is also said to be sporadic voicing of /s/ after these nasals. A phonetic observation which I have not seen in the literature concerns the quality of the /s/: it is strongly apical, with the point approaching the gums at about the same point as the other dentals, so that its acoustic quality is reminiscent of East Norwegian retracted [ṣ] after r, e.g. in vers.

41 Harris calls this procedure ‘rephonemicization’, Methods 90–6.

42 GuÐfinnsson, Breytingar 19 ff.

43 Einarsson, Beiträge 35; Kress, Die Laute 105; GuÐfinnsson, Breytingar 19. Kaþólskur is most often pronounced with a /t/. After juncture the [þ] is expected: íþrótt ‘sport’, áþján ‘oppression’; pronouns beginning with [þ] get [Ð] as enclitics, e.g. sérÐu ‘do you see’, viÐ þau [viÐÐöy] ‘with them’.

44 See GuÐfinnsson, Mállyzkur 151–243 and Breytingar 16–9 for details on the regional distribution.

45 Sveinn Bergsveinsson, Grundfragen, and Einarsson, Beiträge, both have intonation curves of various Icelandic utterances.

46 The terms ‘fortis’ and ‘lenis’ are introduced to avoid a commitment on the voice correlation, which appears to be largely allophonic. Stops are predominantly voiceless, while continuants (except /s/) vary according to position. Fortis, when applied to the latter, refers to initial voicelessness, lenis to initial voicing.

47 Einarsson, Icelandic 26.

48 A few additional clusters listed in dictionaries may occur without assimilation when the morphemic pressure is great enough, esp. with final /s/, e.g. arfs, arms, bergs, efld (usually metathesized as /elvd/).

49 Cf. Einarsson, Icelandic 5–6. I have not included above the subordinate rule about lengthening before juncture and /h/ or vowel, since /h/ does not form a postnuclear cluster; also it often disappears in this position, and we have then only an intervocalic position, in which the consonant is short by definition. The exceptions to the rule, like kaupmaÐur /köip'ma^sÐur/, are simply explained as loss of juncture, as suggested by Einarsson in Um kerfisbundnar 9.

50 Two of these clusters do actually occur finally, but in rare forms: emj ‘cry’ and emr (in the expression emr-mál a kind of pig Latin).

51 Kress 7.

52 It should be noted, however, that Kress regards the preaspiration as phoneme /h/ and therefore places the division after it, writing e.g. vakna [vahgna]. An exception also is kensla, where he makes an epenthetic consonant [d] between [s] and [1] the syllabic boundary. He also places the boundary within the geminated consonants, in effect assigning the first to the stressed, the second to the unstressed syllable.

53 Cf. Freysteinn Gunnarsson, Ritreglur 67.

54 Manual 64.

55 Cf. my articles The syllable in linguistic description, For Roman Jakobson 213–21 (The Hague, 1956), and Syllabification in Kutenai, IJAL 22.196–201 (1956).

56 Cf. Bergsveinsson's evidence based on a recorded specimen of his own speech, in Grundfragen.

57 In Helgafell 2.63–9 (1943), Björn GuÐfinnsson illustrated a ‘spelling based on pronunciation’ which is not unlike a phonemic spelling. This was called to my attention by Stefán Einarsson, who kindly read and criticized this paper in an earlier version. While GuÐfinnsson's spelling is highly simplified, it does not carry the analysis of Icelandic as far as possible. He marks neither stress nor junctures, and fails to resolve the diphthongs, writing for instance á and ei as in the traditional orthography. As I have done, he resolves the palatals into clusters; but he uses different symbols for allophonic voiced and voiceless spirants (q–x, Жþ), and inconsistently writes h for preaspiration but not for postaspiration. On the other hand, he uses h for voiceless spirants, even where the h is redundant, as before fortis consonants.