## Quantity and Syllable Structure

The basic facts about vowel length in Modern Icelandic can be informally stated as follows:

Stressed vowels are long if no more than one consonant follows

The exception to this simple rule is that stressed vowels are also long before two consonants if the first one is a member of the set /p, t, k, s/ and the second of /j, v, r/. Hence the stressed (first) vowels in (a) in the following list are all long and so are the stressed vowels in (b) whereas the stressed vowels in (c) are short, as indicated:

a bua ['pu:a] 'live', tala ['tha:la] 'talk', lesa ['leisa] 'read' b nepja ['ne:phja] 'coldness', katra ['k^yit^a] 'happy (gen. pi.)', flysja ['fliisja] 'peel'

c elda ['elta] 'cook', andi ['anti] 'spirit', belja ['pelja] 'bellow', inni ['in:i] 'inside'

It is assumed here that long consonants are in fact geminates, or at least equivalent to double consonants phonologically, and that consonant length is basic or underlying in Icelandic and vowel length derived. It should also be noted here that the only vowels that occur in completely unstressed syllables in native Icelandic words are /i, a, u/. (Note that this does not hold for syllables that carry secondary stress, such as in words like 'asnajegur (see the section on stress above, pp. 148-9).)

The consonant sets mentioned in the exception to the vowel quantity rule stated above (and exemplified in the (b) (vs (c)) examples in the list) suggest that syllable boundaries may play a role in vowel quantity since the members of the first and second sets are probably at the opposite ends of the sonority hierarchy for Icelandic consonants. The question is how to build this into the quantity rule.

It is a well-known fact for many languages that have positionally determined vowel length that vowels tend to be long in open syllables, i.e.

syllables that are not closed by consonants. Hence it would seem natural to assume that the syllable boundary in the words in the list is as follows: (a) bu.ci, ta.la, le.sa\(b) ne.pja, ka.tra,fly.sja\ (c) el.da, an.di, bel.ja, in.ni. As the reader may have noticed we have in fact been assuming a syllabification along similar lines above. This would mean that one intervocalic consonant always forms part of the second syllable, and given two intervocalic consonants the boundary varies depending on the sonority of the consonants (this would have to be spelled out in more detail). This would give open syllables in (a) and (b) but closed in (c). If this were correct, we could simply say that stressed vowels are long in open syllables in Icelandic.

The problem with this is the quantity in monosyllabic words that end in one consonant and hence would seem be closed syllables. There the vowel is long too. Examples include words like tal ['^ail] 'speech', les ['le:s] '(I) read\fit ['fi:th] 'web'. There are various ways to solve this problem. One is to say that there is something special going on in monosyllables. Another is to say that the syllabification we have been assuming is wrong and should be like this: (a) bu.a, tal. a, les .a; (b) nep.ja, kat.ra,flys.ja; (c) eld. a, and.i, belj.a, inn A. This could be called 'the final-maximalistic' syllabification, meaning that you let 'as many consonants as you can' follow the preceding vowel. Then you could say that the vowel in stressed syllables is long if at most one consonant follows.

The main motivation for this last analysis is the fact that it seems to allow us to have one rule for vowel quantity in monosyllables and polysyllables. That is desirable, of course. Unfortunately, it is not obvious that this works, however. The test case would be monosyllables that end in consonant clusters of the sort /p, t, k, s/ + /v, j, r/. These are very rare in the language but the few that can be formed certainly contain long vowels. In that respect the words in (a) differ from the ones in (b): (a) snupr ['s^Yipf] 'scolding', flysj ['fins?] 'peeling', pukr ['pVikj] 'secretiveness', sotr ['sceitf] 'slurping'; (b) kumr ['khYmr] 'bleating', emj ['emj] 'wailing', bolv ['pcelv] 'cursing'. So either we need a more sophisticated theory of syllables, namely one that does not consider final consonants and certain final consonant clusters part of the preceding syllable in some sense, or the length of stressed vowels in Modern Icelandic does not depend on syllable boundaries.

### Some Consonantal Processes

Whereas aspirated stops are very common in the world's languages, pre-aspirated ones seem to be rather rare, although they occur in some Scandinavian dialects. Icelandic pre-aspiration is illustrated in the following examples: (a) tappi [thahpi] 'cork', katt [khayht] 'happy (n.)', pakkar [phahkar] 'parcels'; (b) epli [ehpli] 'apple', rytmi [nhtmi] 'rhythm', vakna [vahkna] 'wake up'. The stops /p, t, k/ are aspirated in initial position, for instance. Double (or geminate) consonants are normally long in Icelandic, as explained in the preceding section, but where we would expect long /pp, tt, kk/ on historical or synchronic grounds we get pre-aspirated stops instead.

This is illustrated in the (a) examples above (the example kdtt involves synchronic alternation since it is the neuter form of the adjective kdtur where the stem is kat- but the neuter is formed as usual, by adding a #-f#, see p. 155). In addition, /p, t, k/ are pre-aspirated when they precede /l, m, n/. This is illustrated in the (b) examples.

Devoicing of sonorants is also not very common in the world's languages, but it occurs in Icelandic (and in certain Scandinavian dialects too). The sonorants are not all equally susceptible to devoicing and there are some dialectal differences. In short, /r/ is devoiced before /p, t, k, s/ and in the most common dialect /l, m, n/ are also devoiced before /p, t, k/. This can be illustrated by the following examples: (a) nom. far [fair] 'fare', gen. fars [fajs]; (b) f. ful [furl] 'sour', n.fult [fujt]; f.fim [film] 'nimble', n.fimt [ftijit]; I fin [fi:n] 'fine', n. fint [figt]. Sonorants are also devoiced word-finally (or rather phrase-finally) after voiceless consonants (and optionally after voiced segments in phrase-final position): vatn [vahtg] 'water', rusl [rystj] 'garbage'. In addition, most speakers of the devoicing dialect also devoice /fl/ before /k/ (it does not occur before /p, t/). Note also that devoicing of sonorants before /p, t, k/ leads to de-aspiration of the stops. In general, Icelandic stops are not aspirated after voiceless consonants (see Table 6.5).

There are various types of alternations between stops and fricatives in Icelandic. Thus we have fricativization of /p, k/ between a vowel and N as in f. adj. txp [t^aixp11], n. txpt [thaift] 'uncertain'; f. adj. rik [ri:kh], n. rikt [rixt] 'rich'. Similarly, /p, t, k/ are sometimes realized as their homorganic fricatives between vowels and /s/, but this does not hold for all words and is usually only optional when it can apply: nom. skip [sci:ph], gen. skips [serfs] 'ship'; acc. bat [pay:th], gen. bats [pays:] 'boat'; nom. pak [0a:kh], gen. paks [Gaxs] 'roof. On the other hand, the fricatives /v, y/ show up as [p, k] before /l, n/. This 'stopping' occurs for instance when the appropriate environment is created by an ellipsis of unstressed vowels (actually, intervocalic [v] could either be analysed as If I or /v/ since there is no contrast between the two in that position): fem. sg. grafin [kraivm], pi. grafhar [krapnar] 'buried'; nom. sg. saga [saiya], gen. pi. sagna [sakna] 'saga'; acc. sg. hefil [heivil], nom. pi. heflar [heplar] 'grader'; f. sg. pogul [OceiyYl], pi. poglar [Goeklar] 'silent'. This process does not apply to /5/ before /l, n/, however.

Homorganic (dental or alveolar) unaspirated stops are inserted between /rl/, /rn/, /si/, /sn/. The proper environment can again be created by ellipsis of unstressed vowels: f. sg.farin [fa:rm], pi. farnar [fartnar] 'gone'; f. sg. lasin [laism], pi. lasnar [lastnar] 'sick'; acc. sg. feril [feinl], dat. sg. ferli [fertli] 'career'; acc. sg. drysil [triisil], dat. sg. drysli [tristli] 'devil'.

0 0