CC BY
0
DOI: 10.30842/ielcp2306901528119
Alexander Mankov
Institute of Linguistics, RAS / St. Tikhon's University for the Humanities,
Moscow, Russia. [email protected]
ON THICK L1
The article describes the distribution of two allophones of the phoneme /l/ in Swedish dialects of Estonia and, in the first hand, in the dialect of Gammalsvenskby, where I carried out fieldwork and interviewed speakers of the dialect. It is shown that these allophones appeared in the Proto-Scandinavian period, before the syncope of the unstressed vowels, though the present-day quality of thick l may be due to a later development. I suggest that the emergence of two allophones of /l/ is related to the emergence of similar allophones of /r/, which, in turn, was caused by PScand. *R having become an allophone of /r/.
Keywords: retrofex flap, "thick" l, historical phonology, Swedish dialects of Estonia.
А. Е. Маньков
Институт языкознания РАН / ПСТГУ, Москва, Россия. [email protected] О «толстом» l
В статье рассматриваются правила распределения двух аллофонов фонемы /1/ в шведских диалектах Эстонии и прежде всего в диалекте села Старошведское (Херсонская обл.), где автор проводил полевые исследования. Показывается, что эти аллофоны возникли в прасканди-навскую эпоху, до синкопы безударных гласных, хотя современное качество «толстого» l могло возникнуть позже. Выдвигается предположение, что возникновение двух аллофонов /1/ связано с возникновением двух аллофонов у фонемы /r/ в праскандинавском, которое, в свою очередь, обусловлено превращением прасканд. *R (< *z) в аллофон /r/.
Ключевые слова: ретрофлексный флэп, «толстое» l, диахроническая фонология, шведские диалекты Эстонии.
Thick l (Sw. »tjockt 1») is a retroflex flap. It is pronounced with the tip of the tongue raised towards the top of the hard palate2 and then making a single touch of the area just above the protuberance
1 I am grateful to Elis J. Ahlberg for his meticulous comments on this article.
2 Hence the Swedish term kakuminalt l used alongside tjockt l, to Lat. cacumen 'peak'.
behind the upper teeth. Auditively, it is a very short, one-touch consonant which may resemble a lisped r. However, in the dialect speech that I heard, it is clearly distinct from both r (trilled) and l (dental). In the IPA it is designated as t, in the Swedish Dialect Alphabet § or — in the 1880s — i (Eriksson 1961: 77). Below it is designated as /. Standard Swedish has lost it and generalised the dental l in all positions, but it is preserved in dialects. M.I. Steblin-Kamenskij (1965: 21) correctly described the general tendency for l and / to be in complementary distribution. In this article I will study the distribution of l and / in Swedish dialects of Estonia, primarily in the dialect of Gammalsvenskby; combinations of r and dental consonants are also taken into account. The examples are taken from this dialect unless otherwise specified. The source of data is interviews with its fluent speakers recorded during my trips to the village. I have devised the following spelling system for the dialect of Gammalsvenskby (Mankov 2019).
Vowels: a [a], a [a:], e [e], e [e:i, e:],3 i [i], T [i:], o [o], o [o:], u [u], u [u:], y [y], a [e], & [e:], o [re], 5 [re:], u [s],4 U [u:], ai [ei, ei], ou [re:u, reu].5
Consonants: b [b], d [d], d [d] (as in Sw. bord), f [f], g [g], h [h], j [j], k [k], l [l] (as in Sw. lara, mal), / [t], m [m], n [n] ([g] before k), n [o], y [g], p [p], r [r] (trilled), * [s], 5 [J],61 [t], t [t], v [v], x [x], z [z], ? [3].
Some forms from previous descriptions of Swedish dialects of Estonia may sometimes raise doubts as to the accuracy of their representation in published works, i.e. it is not always clear if they were recorded correctly. However, in this article I do not deal with this problem and cite the forms as they are given in the sources. Previous scholars used a variety of spelling systems, sometimes I provide a transliteration using the spelling that I devised for the dialect of Gammalsvenskby.
3 [e:i] is usually pronounced within morphemes, [e:] at the end of morphemes.
4 Central rounded open vowel; lower than Standard Sw. [e], lips are rounded and protruding.
5 The diphthongs usually shorten before long consonants or groups of consonants. In this case the second element is pronounced extrashort, e.g. g[ei]g/ar 'gills'.
Note that it is alveopalatal and sounds identical to Russ. m or Germ. sch; it is not postalveolar as Standard Sw. r*.
1. Distribution of I and I in synchrony
(for more examples see Mankov 2019: 78).
l occurs in the following positions, cf. Eriksson (1961: 77-78): 1) word-initially or, to be more precise, at the beginning of stressed morphemes,7 e.g. lias 'obey', lagge 'put'; 2) when doubled: bill 'uncle; old man', vallar 'herdsman'; if ll shortens, it remains l and does not give /: valiy 'soup', cf. Sw. valling; 3) before d, t within morphemes (see section 2.1); 4) after d, t: fraidlendar 'cheerful', bittle 'early'; 5) after 5, x: slo 'to strike', gnixl 'to whine'; 6) before 5 + consonant: alsk 'to want'; 7) after i, i, ai:8 fila 'to wag tail', fil 'file' (tool), dail 'to divide'; historically, here also belongs aild 'fire' (see section 3). The fact that it is only l rather than / that occurs after these vowels is due the phonetic dissimilarity between / and i, i, ai: / is pronounced with the tongue being backward whereas i, i, ai are front vowels. Note also the combination lld, where d is an epenthetic consonant which appears on a morpheme boundary: falldar pres. sg. offall 'fall'.
/ occurs: 1) intervocalically and word-finally after vowels: fa/an 'beautiful', fU/ 'bird'; 2) before k, g, p, m, v, f: fo/k 'people', fo/ge 'to follow', ho/p 'to help', bo/m 'to stir', ha/m 'straw', go/v 'flour', ho/ft preterite of holp; earlier, / occurred intramorphemically before 5 and n, but the combinations /s, /n assimilated and gave s, n (see section 2.1); 3) after k, g, p, b, v, f: tikk/e 'often', g/em 'forget', p/ant 'plant', b/u 'blood', f/U 'to fly', gnav/ar pres. sg. of gnavo/ 'to gnaw'; 4) in the suffix -/iy: arm/iy 'orphan'.
Loanwords usually have l: alesk 'dumpling', blud 'saucer' etc.
2. Combinations of l, r with dental consonants (d, t, s, n); the combination rl.
2.1. Combinations l, r + d, t, s, n within morphemes.
When the combinations ld, lt in East Swedish dialects are originally intramorphemic, they are not assimilated and remain ld, lt (Jansson 1942: 51; Wessen 1965: § 53). Examples from
7 / never occurs at the beginning of stressed syllables. This explains why the suffix -laik has [l], e.g. brailaik 'breadth'. In all likelihood, it used to carry the secondary stress, due to which forms with this suffix were prosodically equalled with compounds. Cf. the suffix -/iy, which was always unstressed and is therefore pronounced with /.
8 In the dialect of Runo, / is possible after i, see section 3.2.
Gammalsvenskby include farald 'butterfly', hold 'to hold', mtild 'dry earth; dust', sold 'sieve', galt 'hog', gilt 'sow', malt 'spleen', here also belongs It in old borrowings: filt 'blanket', balt 'belt'. Note that PGerm. *lp gives ll: gull 'gold' (Tamm 1887: 78).
The preservation of ld, lt makes a contrast to rd, rt, which are assimilated within morphemes and give postalveolar d, t. This is because in the combinations ld, lt the consonants have different manners of articulation (lateral vs. plosive), but the same place of articulation (apico-dental), so they have fewer prerequisites to assimilate. In this case, the assimilation is only possible by the manner of articulation and not by the place, because consonants which have the same place of articulation obviously cannot assimilate one another by the place of articulation. In Swedish dialects of Estonia, the assimilation by the manner did not take place in ld, lt, but in Standard Swedish it did (in ld), hence halla 'to hold' etc. (Tamm 1887: 79; Wessen 1965: § 53). In the combinations rd, rt the consonants differ both in regard to the place (alveolar vs. dental) and to the manner (sonorant vs. plosive), i.e. they allow assimilation by both these parameters, which actually happened: rt > rt (progressive assimilation by the place) > tt (regressive assimilation by the manner), e.g. sjatta 'heart'. This also explains why PGerm. *lp (unlike lt) gave ll: it is the same type of assimilation as in rt, rd. It should be noted that the places of articulation of consonants in the combinations rt, rd (and lp) are not so distant as to prevent the assimilation. In such combinations as Ik the distance is considerable, which explains the absence of assimilation.
The combination rt does not cause lengthening of the preceding vowel9, due to which the intramorphemic tt is long. As for rd (< OSw. rd), the preceding vowel was lengthened already in Old Swedish (Wessen 1965: § 37), e.g. god 'yard'. The development was therefore rd > rd > d, rather than *dd, because otherwise the overlength * Vdd would have appeared, which is not allowed in the present-day prosody of the dialect.
The combinations ls, rs within morphemes are always assimilated, which is due to the same reason: the consonants have different places and manners of articulation. In the contemporary
9 However, cf. ia>t 'cloudberry' (Sw. hjortron) in the dialect of Nucko (Danell 1951: 160).
dialect of Gammalsvenskby the results of the assimilation of ls and rs are the same in terms of articulation, cf. has 'neck' (< *hals; note that a was lengthened before ls) and foss 'rapid stream' (Sw. fors, the combination rs has not lengthened the preceding vowel). However, the earlier dialect did distinguish *ls and *rs, the former combination gave a retroflex s (designated as • in the Swedish Dialect Alphabet), the latter a postalveolar s (s), cf. has,, fop (Karlgren 1953: §§ 21, 33).
The intramorphemic combination rn in the dialect of Gammalsvenskby is invariably reflected as an ordinary dental n: bjun 'bear'.
2.2. Combinations l, r + d, t, s, n on morpheme boundaries.
On synchronically distinguishable morpheme boundaries in inflexion, the combinations ld, lt etc. are not preserved. The following assimilations have taken place: /+t > tt (postalveolar t), e.g. fatt neuter of fa/an 'beautiful'; /+d > d: skjdd preterite of skjo/e 'rinse'; /+s > s: skjds (< *skjd/s) pres. sg. of skjo/es 'be washed off; /+n > n and n: spdn def. sg. of spS/ 'mirror', bUn 'be inflamed' (Sw. bulna). The consonant n regularly occurs only in monosyllabic stems when the definite article -n < -en is added. In di- and trisyllabic stems n often gives n, e.g. kvarvan / kvarven def. sg. of kvarvo/ 'door bolt'. There is also fluctuation in the quantity of t, cf. snat/snatt 'soon' (originally the neuter of the otherwise unattested adjective corresponding to Sw. snar 'quick'), bot/bott neut. of bo/an 'blunt', hat/hatt, i.e. [ha:t], [hat:], and also [ha t ], neut. of ha/an 'slippery'. The fluctuation is due to the fact that either the long tt is generalised or the root vowel preserves its length. The overlength, i.e. the preservation of both the long vowel and the long consonant, is not allowed in the contemporary dialect. It should also be noted that the phonetically regular preterite forms of such verbs as skjo/e 'rinse', sjo/e 'sell' (with the initially short root vowel) would be *sjodd, *skjodd, i.e. with V/d > V/d > Vdd. However, such forms never occurred in my interviews, the vowel is always lengthened, the consonant is short: sjdd, skjdd. This is because the long dd is generally not common in the dialect due to such cases as god (see section 2.1). Analogy with preterites of verbs where the root vowel was always long is also possible, e.g. hed, id, lQd, mid, red to her 'hear', ir 'whirl' (of snow), ldr 'learn', mir 'build from stones', rer 'mix'. Moreover, /g+d also gives d, the preceding vowel lengthens:
fo/ge 'follow' — f5d, svo/ge 'swallow' — sv5d, tjo/ge 'carve' — tj5d. A long d has only occurred in the preterite offer 'lead' which was pronounced [fed:], but also [fe: d].
The same assimilations take place in combinations with r, i.e. r+t > t/tt: k/att/k/at neut. of k/aran 'clear'; r+d > d (see hed above), r+s > ss: r[e]ss pres. sg. of reras 'move', r+n > n in monosyllabic stems: tjUn def. sg. of tjUr 'bull'. In disyllabic stems rn often gives n: biggjan/ biggjan def. sg. of biggjar 'builder' (Mankov 2019: 80).
On morpheme boundaries in word formation, namely in compounds, combinations of / with the following dentals can be both assimilated and preserved: bumo/-takke 'cotton blanket', ha/-stikke 'hailstone', ko/-strunk 'cabbage head', may(y)o/-stukk 'rolling pin', alongside juttrti 'fir tree', may/-stukk 'rolling pin'. The same concerns r + dentals in compounds: er-snipa 'ear lobe', finstas-k/e 'curtain', fo-skokk alongside for-skokk 'flock of sheep', or-tTtas and o-tTtas 'for years' (Sw. artiotals).
As for the phonological status of t, d, s, I propose the following solution: within morphemes they are phonemes, on morpheme boundaries they are realisations of /rt/, /rd/, /rs/, /ld/, /lt/. The postalveolar [o] on morpheme boundaries is a realisation of /rn/ and /ln/, in such forms as godn def. sg. of god, it is an allophone of /n/.
2.3. The combination rl.
It was shown above that in combinations with r, first a progressive assimilation by the place of articulation, then a regressive assimilation by the manner of articulation took place. The combination rl should therefore have given *r/, then *//, where / is a postalveolar (rather than retroflex) l. This would have created an opposition l ~ r ~ I. However, it would be assymetrical and redundant. Such sequence as [t] ~ [|] (retroflex t) ~ [t] (postalveolar t), which in all likelihood existed in the dialect (see 2.1), was a realisation of /t/ ~ /lt/ ~ /rt/. The sequence [l] ~ [t] ~ [[] would be a realisation of /l/ ~ /rl/ (because [l] and [t] are allophones of /l/); this would create assymetry. For this reason, in rl there was no progressive assimilation rl > r{, and only the regressive assimilation rl > ll > l took place. The examples are kaliy 'woman' (OSw. karling), hole 'very'. The latter form goes back to *holet, which originally is the neuter of the non-attested adjective *holendar <
*horlendar < *hordlendar, cf. Sw. hardlig (Danell 1905-34: 183); as for the loss of d, cf. /gd > d in svo/ge — svdd in 2.1.
On synchronic morpheme boundaries occur rl alongside l, but never a postalveolar l: Hon jtir lat [-rl-] 'She's lazy'. Compare also farlitse alongside falitse 'to hope'.
3. The heteromorphemic combination Id vs. d.
As was mentioned, the original intramorphemic ld is preserved in Swedish dialects of Estonia, while ld with a morpheme boundary between l and d gives d. There are however forms in which l and d are divided by a morpheme boundary, but in one case the outcome is ld (as if the form were originally intramorphemic), in the other d. Such forms in the dialect of Gammalsvenskby are buld 'furuncle' and kjdd 'cold' (noun). Before we study these two forms, it should be noted that the Standard Swedish words that are relevant in this context are bold 'id.', kold 'id.', bolde 'furuncle', dald 'small valley', mald 'grist', stold 'theft', vald 'unfairness' (Tamm 1887: 78-81). The question is why they preserve ld and not assimilate it to ll (as in halla, kall etc.). Tamm explained it by the influence of those forms where the suffix -d- was preserved, e.g. tyngd (ibid.: 79). According to A. Kock (1893: 259), all these forms had thick l, which, unlike the dental l, did not cause the assimilation ld > ll. It was also proposed that the thick l in such forms as OSw. byld, kyld could also be due to the influence of bulen, kulen where / was phonetically regular; thus e.g. H. Pipping (1914: 27-28). I think that we should take into account the fact that these forms are bimorphemic (unlike hall-, kall), and the morpheme boundary is clearly distinguishable. All these forms preserve a link to the related roots with l: bulen 'swollen', kall 'cold', dal 'valley', mala 'grind', stjala 'steal', vilja 'want'. It was the clearly visible morpheme boundary between l and d rather than the supposedly retroflex character of l that has prevented ld from being merged into ll in Standard Swedish.
The word eld 'fire' probably never had the thick l: it goes back to PScand. *eilda- < PGerm. *aili-da- (on its derivation see Kroonen (2013: 11)), and / was not possible after ai < ei (see section 1). This word lost its link with the primary root, which triggered the assimilation, hence the form ell attested in the 16th century (SAOB: E 405). A similar case is probably Sw. kvall 'evening' (cf. Olcel.
kveld). If the etymology that relates it to OE cwelan 'to die' is correct, this noun was derived with a suffix that contained PScand. *d; its relation to the primary root became obscure a long time ago, which made this form monomorphemic, having eliminated the obstacle to the assimilation ld > ll.
Now we need to look into why Swedish dialects of Estonia sometimes have ld and sometimes d on a morpheme boundary. I suppose that the reason for this was the distribution of the varieties of l in the older period. Basically, this distribution seems to have been the same as in contemporary dialects. Immediately before d, the phoneme /l/ was realised as a dental l, both intra- and heteromorphemically. In the intervocalic position, /l/ had a different quality and ultimately gave the thick l. If a vowel that followed such l later syncopated, l and the following d assimilated, i.e. *VlVd > *V/d > Vd. This means that l had two different allophones before the syncope, i.e. as early as in Proto-Scandinavian. However, we do not know what the original quality of the second allophone was; its retroflex character could be a later development. For the dating of / see Ronge (1993: 200; 15th century in Sweden), Andersson (2016: 21; first half of the 14th c.), Fridell (2017: 167; late 14th c. — 16th cc.); cf. Kuzmenko (1995: 125).
These regularities may explain the words for 'furuncle' and 'cold' in Swedish dialects of Estonia. If a suffix that contained a dental consonant (namely, *d) was added directly to the root, the combination has remained ld; if there was a vowel between l and *d (this vowel later syncopated), we get d. One more circumstance should be taken in to account. There was a whole range of suffixes with the consonant *d, some of them contained i before d: PScand. *-di-, *-idu, *-da- (PIE *-ti-, *-ita, *-to-), *-dan-, *-idan-. Some of them were synonymous, and, consequently, derivatives with these suffixes could contaminate, which obscured the initial distribution of ld (< *-l-dV-) and /d (< *-l-idV-).
3.1. The noun bUld 'furuncle'.
It has two varieties in Swedish dialects of Estonia, with ld and with d. The form with ld occurs in the contemporary dialect of Gammalsvenskby: bUld, def. sg. -n, pl. -ar, def. pl. -a(na) m. It was recorded by Karlgren (1953: 17): buld, -n, -ar, -a m., and in the dialect of Nucko: bmld m. (Danell 1951: 60). I suppose that it is an
old formation, going back to PGerm. *bl-da- m., PIE *bhel- 'to grow, spread, swell' (IEW: 120-121). The suffix is PIE *-t-o-, which in this case was added directly to a zero-grade root, the stress is on the thematic vowel. This suffix was used to make participial adjectives and corresponding nouns (Brugmann 1906: 394ff.; Olson 1916: 546; Kluge 1926: § 117). The development is PScand. *bulda- > *buld- > bUld; the form without a-umlaut has been generalised, cf. Wessman (1936-38: I, 115ff.); ld usually lengthens the preceding vowel (cf. hold 'to hold'). However, it cannot be excluded that the suffix was *-dan-: *bul-dan- m. with the following transition from the weak declension to the strong: buld is strong; the weak form would be *buld-an in the def. sg. form, but it was never heard in my interviews. In this case, this form would identical to Sw. bolde, which either goes back to PScand. *bul-d-an- or to *bul-i-d-an- m. The suffix *-(i)p-an- was used in nouns for ailments, e.g. OE brunepa 'itching', spiwepa 'vomiting', wl&tta 'nausea' and many others, including the word for death, OIcel. daudi (Kluge 1926: § 118; Krahe, Meid 1967: § 119); this corresponds well with the sense of the Swedish word.
The form with d has the following phonetic variants.
With u+d. Vendell recorded it in the dialects of Dago, Gammalsvenskby, Runo, Nucko: bud (= bud in my transliteration), pl. -ar, feminine in Gammalsvenskby, masculine on Dago (Freudenthal, Vendell 1886: 36); buD m. on Runo (Vendell 188287: 84); bud (u = u here) m. on Nucko (Vendell 1904-1907: 96). It probably goes back to PScand. *bul-i-d-an- m. or *bul-id-o f.; the intervocalic l gives / with the following assimilation, i.e. *bu/du > *bu/d > bud).
With e + d in the dialect of Nucko: beD n. (Danell 1951: 68) and d + d in the dialect of Ragoarna: bod, -er f. (Freudenthal, Vendell 1886: 39); o is [®:]. I explain it as a contamination of *bul-di- f. and *bul-id-o f. The form *bul-di- gave PScand. *byld- (with a dental l); the vowel e in the dialect of Nucko is due to the delabialisation of o (lengthened before ld) < OSw. y (Danell 1905-34: 98; Hultman 1939: 162). The consonant d is taken from bud < *bul-id-. The neuter gender of beD instead of the original feminine (cf. OSw. byld f.) should not necessarily be old; it could be caused by the fluctuation between f. and n., typical of the dialect of Nucko; for its explanation, see (Danell 1905-34: 226). Note that in Standard
Swedish, *bul-di- gave bold (Tamm 1877: 28); judging by the available materials, this form has no direct cognates in Swedish dialects of Estonia.
The discussed forms are summarised as follows.
PScand. *bul-da- m.: bUld Gsvby, bwld Nucko;
PScand. *bul-d-an- m.: OSw. bulde, bolde, boldhe; Sw. bolde; ?
bUld m. Gsvby, Nucko;
PScand. *bul-id-an- m.: bud m. Dago, Nucko, bu^ Runo;
PScand. *bul-di- f.: OSw. byld, bold f.; Sw. bold;
PScand. *bul-id-u f.: bud f. Gsvby;
Contamination of *bul-di- f. and *bul-id-o f.: beD n. Nucko, bod f. Ragoarna.
3.2. The noun kj5d 'cold'.
Similar developments are seen in the word for 'cold'. In my interviews it occurred as both fem. (ute slTk kj5d 'in such a cold') and masc. (e slikar kj5d). In Swedish dialects of Estonia, this noun is probably a contamination of PScand. *kul-id-an- m. (Hultman 1908: 196) and *kul-di- f.; d (< */d) comes from the former, the i-umlauted vowel from the latter. Both these forms are attested in Old Swedish: kulde m., kyld f. The vowel y before d lowered to 5 and developed a palatal element j (Hultman 1939: § 88). An identical form was recorded by Vendell in the dialects of Gammalsvenskby, Ragoarna, Vippal: kiod f. (Freudenthal, Vendell 1886: 103); Tiberg (1940: 329) recorded ki544 on Ragoarna, the same form was recorded by R. Pohl (1944: 21): ki®D. Other dialects differ in terms of the development of the vowel.
The dialect of Nucko has ked f. (Freudenthal, Vendell 1886: 101), keD f. (Danell 1905-34: 98; 1951: 236; Hultman 1939: 162).
As for the dialect of Runo, Vendell (1882-87: 105) recorded kiu^ f. This dialect shows a specific development of long i and y,
which give U or jU before /, e.g. hul 'to rest', stur 'rudder, steering wheel', piul 'willow' (Icel. hvila, styri, pill), see Vendell (1882-87: 20), Hultman (1894: § 67), Franzen (1959: 51). Before the thick l the short vowel y gives o > a in this dialect, cf. falt 'to follow' (Hultman 1894: § 1.11). But in front of d the vowel y lengthened
before the change y o a started to operate and took part in the change y > ju, hence kjud (Hultman 1894: § 1.28, footnote 3)10.
The dialect of Ormso has i: kid f. (Freudenthal, Vendell 1886: 103). Here the original vowel y also lengthened before d prior to the change y o, and then delabialised.
4. The origin of I.
I suppose that the emergence of / or, to be more precise, of two allophones of /l/ is related to the emergence of R in Proto-Scandinavian. As a result of the operation of Verner's law, PGerm. /s/ split into two allophones, [s] and [z]. After the stress established on the first syllable, s and z became two different phonemes. While /s/ occurred in all positions, the occurrence of /z/ was limited to the intervocalic and word-final positions. The limited distribution of /z/ has led to this phoneme being used intervocalically and word-finally as an optional allophone of /s/. The optional character and, consequently, phonological weakness of [z] resulted in the disappearance of the allophonic variation, with [s] having been generalised. This is seen in Gothic. In West-Germanic and Scandinavian languages the development was different because here /z/ was subject to rhotacism. The result was the emergence of a new phoneme /R/. It had the same restrictions in the distribution as /z/, i.e. it was possible intervocalically, word-finally, and before d, n, whereas the old phoneme /r/ occurred in all positions. The articulatory proximity between /r/ and /R/ accompanied by the limited distribution of /R/ led to /r/ and /R/ becoming allophones, with /R/ repeating the fate of /z/, because at a certain stage it came to be an optional allophone of /r/. This allophonic variation was also eliminated. However, different languages generalised different varieties of the articulation of /r/. For example, Faroese and English
10 An interesting case in connection with this is the word kiolmt f. 'measure for dry goods, e.g. grain' in the dialect of Runo (Vendell 1882-87: 105). It is borrowed from Est. kulimit, dial. kulimet (Lagman 1971: 64; cf. Saxén 1895-1897: 159). As Elis Ahlberg suggests (personal communication), it was probably borrowed after the change y > ô before the thick l and therefore did not take part in this change. However, it took part in the change y/î > jû, but with a slightly different outcome (the difference is due to the fact that the vowel was originally short), hence the supposed development kulimit > *kilmit > *kjolmit > kjolmt.
have generalised the pronunciation as an approximant which may go back to /R/, while Icelandic, Norwegian and Swedish have the trilled articulation which was characteristic of /r/. The emergence of two allophones of /r/, viz. [r] and [R], had one more consequence for the system of consonants. Under the influence of [r] ~ [R] the phoneme /l/ also split into two allophones, i.e. the one occurring word-initially, in gemination and in combination with t, d, and the one occurring intervocalically and word-finally. The first allophone was in all likelihood a dental [l]. We do not know how the second allophone was initially pronounced, but eventually it gave the retroflex flap in Scandinavian dialects. The velarised l, typical of English, may also go back to the second allophone of the old /l/. Its original quality could therefore be due to the backward position of the tongue.
References
Andersson, T. 2016: Dateringen av supradentalerna rn, rs, rt i Sverige.
Namn och bygd, 5-49. Brugmann, K. 1906: Grundriss der vergleichenden Grammatik der
indogermanischen Sprachen, II. 1. 2 ed. Strassburg. Danell, G. 1905-34: Nuckömälet. Stockholm. Danell, G. 1951: Ordbok över Nuckömälet. Uppsala; Copenhagen. Eriksson, M. 1961: Svensk ljudskrift 1878-1960. Svenska landsmäl och
svenskt folkliv. Franzen, G. 1959: Runö ortnamn. Uppsala; Copenhagen. Freudenthal, A., Vendell, H. 1886: Ordbok öfver estländsk-svenska
dialekterna. Helsinki. Fridell, S. 2017: Supradentalernas datering. In: Östman, J.-O. et al. (eds).
Ideologi, identitet, intervention. Helsinki, 161-169. Hultman, O. 1894: De östsvenska dialekterna. Helsinki. Hultman, O. 1908: Hälsingelagen och Upplandslagen Ärfdabalk i Cod. Ups.
B49. Spräkhistorisk undersökning jämte tvänne exkurser. Helsinki. Hultman, O. 1939: Efterlämnade skrifter, II. Helsinki. IEW — Pokorny, J. 1959: Indogermanisches Etymologisches Wörterbuch. Bern; Munich.
Jansson, V. 1942: Om de östsvenska dialekternas ställning. Svenska
landsmäl och svenskt folkliv, 47-71. Karlgren, A. 1953: Gammalsvenskby. Uttal o ch böjning i
Gammalsvenskbymälet. Stockholm. Kluge, F. 1926: Nominale Stammbildungslehre der altgermanischen Dialekte. 3 ed. Halle.
Kock, A. 1893: Behandlingen av fornsvenskt kort y-ljud och supradentalers invärkan pä vokalisationen. Arkiv för nordisk filologi, 50-85, 235268.
Krahe, H., Meid, W. 1967: Germanische Sprachwissenschaft, III. Wortbildungslehre. Berlin. Kroonen G. 2013: Etymological Dictionary of Proto-Germanic. Leiden; Boston.
Kuzmenko, J. 1995: Det tjocka l-ets historia i nordiska spräk. In: Äström,
P. (ed.) Studier i svensk spräkhistoria, 4. Stockholm, 119-127. Lagman, H. 1971: Svensk-estnisk spräkkontakt. Stockholm. Mankov, A. 2019: The Dialect of Gammalsvenskby: an Outline of its Phonology. Svenska landsmäl och svensktfolkliv. Vol. 142, 67-105. Olson, E. 1916: De appellativa substantivens bildning i fornsvenskan. Lund. Pipping, H. 1914: Till frägan om l- och n-ljudens kvalitet i de nordiska
spräken. Studier in nordisk filologi. Vol. 7. Pöhl, 1944: Typordför undersökning av de svenska folkmälen (2:a uppl.). Estland, Lilla Rägö. Undersökningen gjord är 1944 av Rain Pöhl. Manuscript kept at the Institute for Language and Folklore (ISOF), Uppsala, Sweden, folder 17102. Ronge, H. 1993: Kakuminalt l — älder och ursprung. In: Wollin, L. (ed.).
Stu dier i svensk spräkhistoria, 3. Uppsala, 193-201. SAOB — Svenska Akademiens ordbok över svenska spräket, I-. Lund,
1898- (https://www.saob.se/). Saxen, R. 1895, 1897: Finska länord i östsvenska dialekter. Svenska
landsmäl och svenskt folkliv. Steblin-Kamenskij, M. I. 1965: Om alveolarer og kakuminaler i norsk og
svensk. Norsk tidskrift for sprogvidenskap. Vol. XX, 18-27. Tamm, F. 1877: Om fornnordiska feminina afledda pä ti och pä i^a. Uppsala.
Tamm, F. 1887: Fonetiska kännetecken pä länord i nysvenska riksspräket. Uppsala.
Tiberg, N. 1940: Rägösvenskan. In: Söderbäck, P. (ed.). Rägöborna.
Stockholm, 327-356. Vendell, H. 1882-87: R nömälet. Stockholm.
Vendell, H. 1904-1907: Ordbok över de östsvenska dialekterna. Helsinki. Wessen, E. 1965: Svensk spräkhistoria, I. Stockholm. Wessman, V. E. V. 1936: De finlandssvenska dialekternas labiala vokaler, I. Helsinki.
