IV: VOICE
Classified according to vocal technique, the sixteen informants
at the center of the study fall roughly into two groups with Ebenezer,
Gunnar, Jón, Karl, Kristján, Margrét, Pétur,
Thorgils, and Thórur in the first and the rest in the second.
The voice quality of the former is narrow and tense and usually
quivering on long held syllables; the general term for it is kvæardd
, i.e., kvæda -voice. The latter group is characterized
by a more relaxed voice without quivering, resembling more or
less the performer's ordinary speaking voice. This manner of performance
is expressed by the verb raula (n. raul ), which refers generally
to artless or modest delivery, as pointed out earlier (p.2). Some
informants of the oldest generations in the west think that raul
is incompatible with kvæaskapur, whereas others tend to
disregard the vocal style, and a few consider almost any performance
of rímur or stanzas in rímur metres to be kvæaskapur
. The stricter view limits the concept of kvæaskapur greatly
and is consistent with the widespread opinion in the west that
kvæaskapur is a difficult art mastered by few: the kvæamenn.
Most would only raula and did not deserve the name.
When the text is performed with a kvæa -voice, the vowels
tend to be distorted, so that often a word becomes incomprehensible
or the meaning is altered. Fig. 2 shows the results of a spectral
analysis of a number of vowel samples from the field recordings.
The mean values for the frequencies of formant 2 are plotted against
the values for formant 1, and the vowel triangles thus obtained
for Kristján's speech and kvæaskapur (a) and Thórur's
singing and kvæaskapur (b) compared.1 Fig. 2c shows the
non-averaged composite formant distribution for the kvæa
-vowels of Kristján, Thórur, and Gunnar. It is easily
seen from Fig. 2 that the vowels of the kvæa - voice differ
drastically from those of speech and singing. On the other hand,
there are only minor differences between the formant values of
Thórur's spoken and sung vowels. The first formant (F1)
of all vowels except Kristján's and Thórur's o is
much higher in kvæaskapur than speech and the second formant
of the the front vowels lower. In this regard the values of í,
i, ú approach the values of i, e, and o as spoken, but
e has either the same values as i or those typical of the cardinal
vowel [æ] normally not occurring in Icelandic speech. This
confusion between e and i is found in the speech neither of Kristján
not Thórur; the same applies to u and , which in their
kvæaskapur have the same or similar formant values.
Fig. 2
Frequency of the first formant (F1) on the x axis and the second
formant (F2) on the y axis in kilo Herz (1 kHz = 1000 cycles per
second). Broken lines indicate spoken and sung vowels and solid
lines vowels of kvæa -voice.
Besides, the shift of F1 and F2 the most striking feature of
the kvæa spectrograms is the great concentration of spectral
energy in the region between 3 and 4 kHz (for all vowels except
ú and in some cases o) characteristic of the vocal "twang,"
according to Vennard.2
The performers are at least partly aware of different articulations
is singing and kvæaskapur, and the adjective andthröger
has been used to explain why to kvæa is more strenuous than
to sing. Andthrönger literally means "narrow-breathed"
and probably indicates in this connection a constricted pharynx.
At the same time no tension of face and neck muscles can be observed,
and the opening of the jaw is not greater than in speech, but
occasionally the erect and slightly backward posture of the head
is exaggerated so that the jaw is pulled back into the throat.
The spectrograms of an experiment with two subjects imitating
this gesture suggested that pressing the head rather far backwards,
keeping everything else equal, is sufficient to change the first
two formats of the vowels in a similar way as in Fig. 2, but the
strong high partials are not present, and the voice does not sound
exactly like a true kvæa -voice. Vennard states that twang
is produced by stretching the palatopharyngeus muscles (the back
pillars extending from the soft palate into the back wall of the
pharynx) and that nasality may be present as a by-product.3 Nasal
resonance does not seem to be of any importance in kvæa-voice,
and at least in the case of Ebenezer it is completely out of the
question since his nose is clogged, but nevertheless his voice
rings in a typical manner. If the head is held backwards only
a slight tension of the pillars is needed in order to get the
right sound, but on the contrary if it is held forward a relatively
great tension is required. Thus no more than a moderate stretching
of the pillars and possibly of other muscles connected with them
might be expected with the usual erect or slightly backward position
of the head. Such posture would not be too straining for the organs
-- an important point, since a good kvæamaur could perform
for hours without showing any fatigue.
There will be no attempt here to explain how the articulatory
gestures discussed above constrict the pharynx, and whether they
can produce kvæa-voice by themselves with formant frequencies
similar to those observed. Other factors could contribute to this
effect too, especially tongue shape and position of the larynx,
and since the jaw is most often relatively high and rigid, a high
position of the larynx is to be expected.
Spectrograms of samples from performances of Sigurur, Gisli, and
Gumundur show formant frequencies much closer to those of Kristján's
speech and Thórur's speech and singing than their kvæaskapur
, and i and e, u and are always distinguished clearly. There can
be no doubt that a text performed with kvæa-voice is less
intelligable that if performed with the other voice productions,
but the different formant values seem not to be brought to consciousness
as gross changes of vowels; in fact only those who have tried
to transcribe texts performed with kvæa- voice seem to be
aware of them. Apparently, therefore, they are normally heard
as part of the voice quality rather than as distortions of meaningful
units of the language, and only exclusive concentration on this
aspect will bring them to light.
Generally speaking the spectral changes occasioned by singing
and raul are small and rather unpredictable, whereas those characteristic
of kvæa-voice are large and follow a definite pattern. Possibly
every cultivated singing style has its own pattern of deviations
from speech, producing the "professional" sound of the
voice. In Western artistic singing, for instance, there is a spectrum
envelope peak near 3kHz for all vowels, often called the "singing
formant," as it represents the most important spectral difference
between spoken and professionally sung vowels. According to Sundberg,
lowering the larynx is "the major articulatory gesture associated
with the production of a singing formant," but at the same
time "a low larynx position makes the tissues between the
thyroid and cricoid cartilages more lax. The shifts in the relative
position of these two cartilages, required for the big pitch variations
in singing, could be made with less muscular activity, if the
tissues joining them are lax."4 A high larynx, on the other
hand, would not favour the production of wide intervals and smooth
transitions between registers. Actually, intervals of less than
a minor third predominate in kvæaskapur, and intervals larger
than a fourth are exceptional. The melodic range is less than
an octave, always staying within the same register. Male performers
use high chest register but women low; thus the seimur of Margrét
and of the three men Thórur, Gunnar, and Ebenezer is usually
found near "a" (220 Hz). The others are not much lower
("d" to "g"), so most of kvæaskapur
performed with a kvæa-voice falls within the range of "d-g1".These
limitations should probably be regarded as essential features
of kvæa-voice no less than of melodic motion, which could
indeed be regarded as conditioned by it.
The ornament usually called dillandi or dill is connected exclusively
with kvæa-voice. It is either a narrow trill (Fig. 3b) or
what seems to be a true vocal tremolo, i.e., a pulsation of the
same pitch rather than oscillations between two adjacent pitches.
Fig. 3 (a and c) shows the characteristic tremolo pattern as it
appears on melograms. Of the nine performers with unimpaired kvæa-voice
five used tremolo only or mainly, three (Margrét, Karl,
and Kristján) use trill and tremolo or a mixture of both,
and Pétur alone mainly trill. The lowest rate is Pétur's
5 cycle per second and the highest is Karl's 7 cycles. A similar
range of frequencies has been reported for vibrato in artistic
singing.5 As can be seen from Fig. 3a and b, Karl's trill and
tremolo have only a rate in common; not only is the waveform different,
but also the intensity fluctuations accompanying the trill tend
to be twice as fast as the pitch oscillations; notwithstanding
special attention is needed in order to hear the differences.
On the other had melograms of the trills often look like melograms
of vibrato in singing, but the perceived quality is clearly different,
the characteristic timbre of kvæa-voice probably playing
an important role in creating the aural impression. Unlike vibrato
dillandi is never continuous, and usually it is only applied to
long pitches, especially the seimur .
Fig.3
Melogram a-d shows the tremolo on the last syllable of the second line and the trill on the last syllable of the third line of Karl's No. 30b, and melogram c the tremolo on the last syllagble (seimur ) of Thórur's No. 11, 6. The higher curve presents pitch and the lower amplitude with pitches of the tempered scale (marked at the top) and decibels (marked at the bottom) as references.
The shorter pitches in kvæaskapur corresponding to a
syllable of normal length or half a syllable are for the most
part unsteady: rising, falling, or rising and falling rather than
level. Unless exaggerated and relatively long, these inflections
are not heard as pitch glides, but if tremolo is added to them
a succession of different pitches could result, as illustrated
by the following stylized design:
The sustained pitch (a) yields to a short dillandi but the
inflected (b) a related ornament with the same rate. Rapid ornaments,
other than dillandi, are generally hnykkir (sing.hnykkur ) or
rykkir, whether produced by a tremolo or not. It is hard to draw
a clear boundary between dillandi and hnykkir produced by a tremolo,
and the same is true of the boundaries between inflected pitches,
short hnykkur produced without tremolo, and melismas, as they
form a continuum stretching from the domain of sound quality to
that of melodic motion.
Fig. 4
Fig. 4a, a melograph of the first measure of No. 36, shows
considerable pitch movement within all syllables, but it only
leads to a hnykkur in the second. The glide at the beginning of
the third syllable (-ding) is difficult to detect, and the whole
syllable is heard as a melisma of two pitches of the same length
(transcribed as "g" and "f"), the second falling
on the nasal n. Actually the first pitch is twice as long as the
second, and the n occupies two-thirds of the syllable on the melogram.
In Fig. 4b, melogram of the first measure of No. 35, tremolo is
added to the the level pitch of the first metrical syllable (á),
the downward glide of the split second (ég a), and the
first pitch of the third (hald-), where the tremolo stops on the
second pitch falling exactly on the 1. At the beginning of the
second syllable there is a pitch inflection of a similar nature
as the hnykkur in the preceding example, but following tremolo
is neither heard as dillandi not hnykkur, probably because of
the shifting sounds [jeya], all voiced and each occupying approximately
one cycle of the tremolo. On the whole the interplay between sounds
of the language and pitch movement seems to be varied and complex,
language interfering with melody and vice versa.
When performances are highly ornamented, as in No. 14, the voiceless
consonants tend to be short, whereas in plain performances they
occupy much more space; voiced consonants, on the other hand,
are for the most part rather prominent in both cases, especially
the nasals. No. 26 shows Pétur's kvæaskapur at its
most ornate. It is almost continuous voicing from beginning to
end, the only interruptions being a very short t in the first
measure, k in the fourth, and t in the fifth. As a result the
text is incomprehensible until after repeated hearings, when the
missing words can be guessed with the help of alliterations and
rhymes. Pétur's speech is always clear, and he had repeatedly
stressed the importance of good delivery of the text in kvæaskapur,
so he does not seem to be aware of the lack of clarity in his
performance, which is usually considerable, even if it rarely
reaches the extremes of the example cited. But such performances
should be seen in a different context. As mentioned earlier (p.
8) it was a merit to be able to perform long phrases and even
whole stanzas without a pause, or as Gunnar states, "If it
could not be heard that he needed to take breath in a stanza he
was considered a good kvæamaur." It is clear too that
the performers tend to keep the volume constant until they reach
the seimur, where it drops. What all this adds up to is a constant
flow of well sustained sound which the performer strives for even
at the cost of verbal clarity. According to the old kvæamenn
in Breiafjörur, strength is needed in order to support it,
and they all give the same reason why their performances cannot
stand comparison with their earlier achievements: lack of strength.
And when Pétur says "if great power is put into it
[i.e., the performance]...it will be braver," it seems that,
to him at least, sustained kvæa-voice is the voice of courage.
FOOTNOTES: Voice
1 The Icelandic vowels are usually transcribed phonetically as follows: í=[i], i=[I], e=[e], a=[a], o=[o], ú=[u], ö=[ö], u=[Y]. Stressed short and long vowels differ systematically with regard to formant values, as Magnus Pétursson (1969, 135-40) has pointed out. The formants of kvæa vowels and the sung vowels on the other hand do not vary either according to length or stress in any systematic way. Therefore, no distinction is made between long and short vowels in Fig. 2. The spectrograms were made at the Royal Institute of Technology (KTH), Stockholm.
2 Vennard (1967), 115-116, 218.
3 Vennard (1967), 115.
4 Sundberg (1972), 52; and (1970), 45.
5 See Vernard (1967), 195-196.