Íslenskt mál og almenn málfræði

						Ritdómar                                          305
ones are normally aspirated. There is significant dialectal differentiation in word medial
position, which Árnason mentions (p.9) but this does not explain why, for example,
heslur 'horse' is transcribed [hesdYr] on pp. 15 and [hes-tYr] on p. 36 and 57 (the half-
long consonant is discussed in the text). If no difference is intended, it is only confusing
to use both notations. If thcre is a difference, it should be explained. Likewise we
are not told why the labial hard plosive is aspirated in nepja [nE:phja] 'cold weather'
but unaspirated in tepra [de:pra] 'sadness' (p. 22, examples in (6)). On gets the im-
pression that there exists, within one and the same dialect, a three-way distinction,
say [p — ph — b], clearly a mistaken picture. The voice contrast is also confusing
with the spirant [v] ; with consonantal clusters vowels are long before any of /p t k
s/ followed by any of /v r j/, but the examples illustrating this transcribe the tv of
uppgötva 'discover' and the kv of skrökva 'lie' as [tf] and [kf] respectively (for the
northern dialect???). Since [f] and [v] are claimed to be distinctive (p. 10), either the
examples do not fit the rule, or the rule has to be modified to allow /f/ to condition
the lengthening, or one must be told in what sense the [f[ in the above examples is
an instance of /v/. Finally, there is the question of the hard (voiceless) palatal velar
plosive which Árnason transcribes as [c], although he does not include it in his inventory
of plosives (p. 9). If it is indeed a single segment, then it is not clear at all how forms
such as sækja [sai:cha] 'fetch' are examples of lengthening before a consonantal cluster
(Jk+jf). It would seem that the long diphthong is due to lengthening before a single
consonant, just as in lœkur [lai:khYr] 'stream'. The treatment of the palatal velar as
a single consonant sometimes produces results which are mildly amusing: on p. 151
Árnason transcribes vekja 'wake' as [ve:cha] and syllabifies it as vek$ja. One should
hope that syllabification principles refer primarily to the phonetic or phonological form
rather than to spelling. But there is a more serious issue involved here: Árnason must
be assuming some kind of a phonological process /kj/ -» [c] — i. c. the palatal
velar may wcll be a single segmcnt phonctically but phonologically it should be treated
in some contexts as a cluster/k(+)j/. The lengthening of vowels could be easily dismissed
as suggested above: there is a single consonant rather than a cluster, hence there is
no problem; derivational morphology, however, cannot be so easily dismissed: the de-
rivation of veikja [vei:cha] 'weaken' from veikur [vei:khYr] would be hopelessly complica-
ted, if not impossible, unless we assume /j/ as a verbalising suffix. Hence there must
be a stage in the (phonological) derivation where /k+j/ corresponds to the phonetic
[ch].
Now for the length rule. Árnason never really makes it clear why he is unhappy
with the traditional formulation specifying the condition when vowels are long (e. g.
Einarsson 1967:4). In brief, stressed vowels are long if followed by at most one conso-
nant (hence also word-finally, e. g.: nú [nu:] 'now' and prevocalically, e. g. búa [bu:a]
'live') or a cluster of any of /p t k s/ plus any of /v j r/, e. g.: nepja [ne:phja] 'cold
weather', titra [thi:thra] 'shiver', skrökva [sgræ:kfa] 'lie' etc. It is clear that the reason
behind the drive for a reinterpretation is precisely to make sense of the consonantal
clusters which blur an otherwise neat rule.
Also the traditional claim that stressed vowels are short if two or more consonants
follow (with the specified exceptions) may be regarded as purely observational, divorced
from any structural patterns of the language. The re-analysis which Arnason develops
íslenskt mál IV 20