See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/342788595

A Phonetic Illustration of the Sound System of Icelandic

Article · July 2020

CITATIONS READS
0 386

2 authors:

Stefon Flego Kelly Berkson

Indiana University Bloomington Indiana University Bloomington
13 PUBLICATIONS   2 CITATIONS    62 PUBLICATIONS   111 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

3D Ultrasound imaging of tongue morphology during speech production View project

"Canadian" Raising in Fort Wayne, Indiana View project

All content following this page was uploaded by Stefon Flego on 09 July 2020.

The user has requested enhancement of the downloaded file.

 A Phonetic Illustration of the Sound System of Icelandic
Stefon Flego, Indiana University, [email protected]
Kelly Harper Berkson, Indiana University, [email protected]
1 Introduction
Icelandic, a member of the North Germanic branch of the Indo-European language family,
is spoken by over 320,000 people worldwide, with 300,000 in Iceland (Eberhard et al. 2019). A
distinction between two major dialect areas is often made: a northern area (harðmæli, or "hard"
varieties) and a southern area (linmæli, or "soft" varieties), although as Hansson (2003:50) points
out, these labels are somewhat misleading, as harðmæli is spoken mainly in the northeast and
linmæli in the rest of the country.
A substantial portion of the literature on Icelandic focuses on its unusual laryngeal
phonology, and description of the dialect groups is no exception: while harðmæli exhibits plain vs.
aspirated contrasts in word-internal onsets (in addition to word-initial onsets), linmæli has such
contrasts only in word-initial position (Hansson 2003, Pétursson 1976, Þráinsson & Árnason 1992).
Furthermore, speakers of linmæli exhibit a phonation contrast in both word-initial and word-
internal sonorants, while this contrast is limited to word-initial position in harðmæli. A
comprehensive account of the major phonetic differences between these dialect areas, including
detailed description of vowel realizations, is given in Þráinsson & Árnason (1992)’s seminal paper
on dialectal variation in modern Icelandic.
The focus in this description, however, is on the southern variety (linmæli), which is widely
spoken and often considered to be the "standard" (Árnason 2011:104). The data presented in this
study are from Guðbjartur Hákonarson, a speaker from Reykjavík. At the time of recording, he
was in his early twenties and had been living in Bloomington, Indiana for approximately three
years while pursuing a four-year undergraduate degree at Indiana University.
The characteristics of the Icelandic sound system that are most often noted include the
following. Icelandic has a contrast between plain and aspirated series of voiceless stops. The
aspirated series is realized as post-aspirated word-initially and pre-aspirated word-internally. This
pattern is extended into sonorants, resulting in a typologically uncommon phonation-type contrast
between modal and voiceless sonorants. There is also a large vowel inventory, with a vowel length
contrast found in both monophthongs and diphthongs. The following sections deal separately with
consonants, vowels, and prosody (stress and intonation).

2 Consonants
An inventory of Icelandic consonants is given in Table 1. The inventory may look
deceptively large at first because it includes all commonly transcribed positional variants. However,
the set of sounds which can surface word-initially and word-internally are quite distinct, so the
phonotactic restrictions for each of these contexts will be treated separately in subsequent
subsections. Stops and nasals are found at four major places of articulation, and fricatives are found
at five major (supralaryngeal) places of articulation. Icelandic’s consonant inventory also includes
voiceless sonorants, and typologically rare pre-aspirated stops (Silverman 2003), in which voice
offset time precedes stop closure (i.e. [hp ht hc hk]).

1
Table 1 Consonants
Bilabial Labio- Dental Alveolar Palatal Velar Glottal
dental
Plosive p(ː) ph h
p t(ː) th ht c(ː) ch h
c k(ː) kh h
k ʔ
Nasal m(ː) m̥ n(ː) n̥ ɲ ɲ̊ ŋ ŋ̊
Trill r(ː) r̥
Fricative f(ː) θ s(ː) ç x h
ʋ ð j ɣ
Approximant
Lateral l l̥

Word-internal length contrasts are possible for certain consonants, marked in Table 1 with
a parenthetical length diacritic. The supralaryngeal non-strident fricatives come in voiced and
voiceless pairs, with the voiced counterpart appearing on the chart midway between the fricative
and approximant rows. These sounds can be realized with or without frication—that is to say, more
approximant-like and more fricative-like articulations are acceptable and freely vary (Árnason
2011:108, Helgason 1993). Although coronal stops and nasals are produced with a dental
articulation (i.e. [ t̪ n̪ ]), keeping with traditional Icelandic transcription, the dental diacritic will
not be included.
Further description is found in the following subsections, where consonantal contrasts are
divided into initial position (that is to say, before the stressed vowel, which is typically the first
syllable of the word) and internal position (following the stressed vowel). Major allophonic
alternations, additional phonotactic restrictions, and connected speech processes are described in
a third subsection.

2.1 Initial Contrasts

There is a laryngeal contrast at most places of articulation and for most manners in word-
initial position. This is realized as an aspiration contrast in stops and a voicing contrast in fricatives
and sonorants. Unaspirated and voiced sounds are given in the left column of Table 2, and aspirated
and voiceless sounds in the right column. The glottal stop only occurs before vowel-initial stressed
syllables, and is generally not considered a phoneme of Icelandic (Árnason 2011:100).
Concerning the word-initial aspiration contrast in stops, it has been found in lab speech
that the voice onset time (VOT) of the aspirated series co-varies with speaking rate, while VOT of
the unaspirated member is relatively stable across differing speech rates (Pind 1995). Results from
glottographic experiments have demonstrated that peak glottal opening occurs close to (Löfqvist
& Pétursson 1976) or at (Löfqvist & Yoshioka 1981) the release of the stop in the aspirated series,
but nearer to stop closure in the unaspirated series. In addition, it was found that peak glottal
opening area was larger for the aspirated series than the unaspirated series.
In word-initial position, voiceless sonorants are realized with both longer duration and
greater H1-H2 differences than their voiced counterparts (Bombien 2006). Some of their
productions could therefore be characterized as having lax phonation (i.e. breathy rather than
voiceless). According to Bombien, this is especially true of nasals and somewhat true of laterals,
but rhotics are most often truly voiceless.

2
Table 2 Initial Contrasts1
[p] bötum [ˈpœːtʏm] recovery.DAT.PL [ph] pötum [ˈphœːtʏm] motion.1PL
[m] mötum [ˈmœːtʏm] feed.1PL
[ʋ] vöðum [ˈʋœːðʏm] wade.1PL [f] fötum [ˈfœːtʏm] vat.DAT.PL
[t] dökkum [ˈtœhkʏm] dark.DAT.PL [th] tökkum [ˈthœhkʏm] button.DAT.PL
[n] nökkva [ˈnœhkʋa] dinghy.ACC.SG [n̥] hnökkum [ˈn̥œhkʏm] nape.DAT.PL
[r] rökkva [ˈrœhkʋa] get dark.INF [r̥] hrökkva [ˈr̥œhkʋa] recoil.INF
[θ] þökkum [ˈθœhkʏm] thank.1PL
[s] sökkva [ˈsœhkʋa] sink.INF
[l] lökkum [ˈlœhkʏm] enamel.DAT.PL [l̥ ] hlökkum [ˈl̥ œhkʏm] anticipate.1PL
[c] gjöldum [ˈcœltʏm] pay.1PL [ch] kjöltum [ˈchœl̥ tʏm] lap.DAT.PL
[k] göltum [ˈkœl̥ tʏm] stack.DAT.PL [kh] köldum [ˈkhœltʏm] cold.DAT.SG
[j] jörðum [ˈjœrðʏm] earth.DAT.PL [ç] hjörðum [ˈçœrðʏm] herd.DAT.PL
[ʔ] öldum [ˈʔœltʏm] century.DAT.PL [h] höldum [ˈhœltʏm] hold.1PL

2.2 Internal Contrasts

Stressed syllables in Icelandic exhibit a word-internal quantity contrast: ˈVːC vs. ˈVCː,
examples of which are given in Table 3. Note that not all consonants which can occur after the
stressed vowel have a corresponding short or long counterpart (i.e. [ˈVfː], but *[ˈVːf], and [ˈVːl],
but *[ˈVlː]). For the purposes of the quantity contrast, pre-aspirated stops behave like geminate
consonants in that they are always preceded by a short vowel.2

Table 3 Internal Quantity & Manner Contrasts

[p] krapa [ˈkhraːpa] slime.GEN.PL [hp] krappa [ˈkhrahpa] narrow.F.ACC.SG
[ʋ] krafa [ˈkhraːʋa] claim.NOM.SG [pː] krabba [ˈkhrapːa] crab.ACC.SG
[fː] kaffi [ˈkhafːɪ] coffee.NOM.SG
[m] kremur [ˈkhrɛːmʏr]3 crush.2SG [mː] klemmu [ˈkhlɛmːʏ] clip.ACC.SG
[t] skata [ˈskaːta] stingray.NOM.SG [ht] skatta [ˈskahta] tax.ACC.PL
[ð] skaða [ˈskaːða] damage.ACC.SG [tː] skadda [ˈskatːa] harm.INF
[s] masa [ˈmaːsa] chat.INF [sː] massa [ˈmasːa] mass.ACC.SG
[n] mana [ˈmaːna] dare.INF [nː] manna [ˈmanːa] man.GEN.PL
[r] mara [ˈmaːra] sea.GEN.PL [rː] marra [ˈmarːa] creak.INF
[l] mala [ˈmaːla] grind.INF
[c] þekja [ˈθɛːca] roof.NOM.SG [hc] þekkja [ˈθɛhca] recognize.INF
[k] saka [ˈsaːka] accuse.INF [hk] sakka [ˈsahka] plumb.NOM.SG
[j] segja [ˈseiːja] say.INF [cː] seggja [ˈsɛcːa] man.GEN.PL
[ɣ] saga [ˈsaːɣa] story.NOM.SG [kː] sagga [ˈsakːa] dampness.ACC.SG

1
Note than some older speakers may also have word-initial [x] in words like [xʋaːlʏr] hvalur “whale,” realized by
younger speakers as [khʋaːlʏr] (Þráinsson & Árnason 1992).
2
Þráinsson (1978) analyzes intervocalic preaspirates and geminates as identical to a series of two consonants on the
prosodic tier.
3
The sonorants /r/ and /l/ are very often devoiced in phrase-final position, although this is considered optional
(Árnason 2011:237). We have chosen not to transcribe this phenomenon in our examples since it is context-specific.
Thus, words such as kremur are transcribed herein without the optional devoicing, i.e. as [ˈkhrɛːmʏr].

3
 Spectrograms exemplifying the VːC vs. VCː quantity contrast are shown in Figure 1.
Corresponding durational measures for the segments of interest appear in Table 4. Regarding
production of [r], the tokens presented here exemplify the behavior of our consultant: short [r] was
typically produced with one or two occlusions and long [rː] with three or four occlusions.

m aː s a m aː r a m aː n a

m a sː a m a rː a m a nː a
Figure 1 Words minimally contrastive in quantity: the VːC member is shown in the top row, VCː in the bottom row.
Spectrograms produced using Praat speech analysis software (Boersma & Weenink 2019).

Table 4 Duration for vowels and consonants illustrated in Figures 1 and 2.

Length of V Length of C
(in ms) (in ms)
[ˈmaːsa] 180 156
[ˈmasːa] 87 250
[ˈmaːra] 243 71
[ˈmarːa] 122 158
[ˈmaːna] 179 119
[ˈmanːa] 83 188
[ˈk raːpa]
h
187 149
[ˈk rapːa]
h
89 213

Both Garnes (1976) and Pétursson (1976) reported that in sets of words minimally
contrastive in quantity (such as those in Table 3), consistent differences in duration were
maintained between short and long vowels, but that durational differences in certain classes of
consonant were not as robust. Vowels would accordingly seem to be the primary cue to the quantity
contrast. However, the speakers in Pind (1995)’s study systematically manipulated the duration of
the long member of the pair (i.e. VːC and VCː) at different speech rates, while the duration of the
short member remained relatively stable across rates.
Figure 2 shows the three-way contrast between [khraːpa] ~ [khrahpa] ~ [khrapːa]. The top
and bottom panels show the tokens from which the last two rows of duration measures in Table 4
([aːp] vs. [apː]) were taken.

4
 k h
r aː p a

h h
k r a p a

k h
r a pː a
Figure 2 Illustration of a plain voiceless singleton stop (top), a pre-aspirated voiceless stop (middle), and a geminate
voiceless stop (bottom).

Despite extensive discussion in the literature on Icelandic phonetics and phonology, there
is no consensus on whether pre-aspirated stops should be thought of and transcribed as a single
segment, e.g. [hp], or as a sequence of [h] followed by a stop, e.g. [hp] (Árnason 2011:219-227,
Hoole & Bombien 2010, Pétursson 1972, Ringen 1999, and Þráinsson 1978, among others). On
the one hand, post-vocalic aspiration is found only before stop closures, so word-internal [h] would
be subject to strict positional requirements if it were to be considered an autonomous segment.
This close association between voice-offset time and stop closure suggests that the two phases be
considered part of a single segment (i.e. [hp]). However, as can be seen in the middle panel of

5
Figure 3, word-internal aspiration is quite long in duration, and that pre-aspirated stops are always
preceded by a short vowel suggests that they be considered phonologically long, or consisting of
two prosodic units for the purposes of the quantity contrast. This behavior would perhaps be better
captured by an analysis in which pre-aspirated stops are treated as two equal units (i.e. [hp]). As
the orthographic representation of these sounds suggests, the three-way contrast between modern
Icelandic -p [p] ~ -pp [hp] ~ -bb [pː], to use the labial place of articulation as an example, is thought
to go back to -p [(h)p] ~ -pp [(h)pː] ~ -bb [bː] in Proto-Norse (Helgason 2002). It then seems that the
obstruent system has evolved in such a way that voicing during stop closure was eliminated, yet a
durational contrast between -p and -pp and a laryngeal contrast between -pp and -bb were both
retained.
Word-internal contrastive voicing in fricatives and sonorants is found only in the context
of consonant clusters, as seen in Table 5. In addition, it can be seen that the dorsal nasals [ɲ ɲ̊ ŋ ŋ̊]
only surface before a word-internal homorganic stop.

Table 5 Internal Laryngeal Contrasts

[m] lambi [ˈlampɪ] lamb.DAT.SG [m̥] lampi [ˈlam̥pɪ] lamp.NOM.SG
[n] vanda [ˈʋanta] do carefully.INF [n̥] vanta [ˈʋan̥ta] need.INF
[ɲ] hangi [ˈhauɲcɪ] hang.3SG.PRS.SJV [ɲ̊] hanki [ˈhauɲ̊cɪ] loop.NOM.SG
[ŋ] hanga [ˈhauŋka] hang.INF [ŋ̊] hanka [ˈhauŋ̊ka] loop.ACC.SG
[r] hjarða [ˈçarða] herd.GEN.PL [r̥] hjarta [ˈçar̥ta] heart.NOM.SG
[l] hálfa [ˈhaulʋa] half.F.ACC.SG [l̥ ] hjálpa [ˈçaul̥ pa] help.INF
[ʋ] höfðum [ˈhœʋðʏm] head.DAT.PL [f] höftum [ˈhœftʏm] constriction.DAT.PL
[ɣ] sigðum [ˈsɪɣðʏm] sickle.DAT.PL [x] sigtum [ˈsɪxtʏm] strainer.DAT.PL

According to Bombien (2006), there are a host of cues which contribute to the voicing
contrast in sonorants, including duration, frication, and phonation type, and the relative strength
of these cues varies by position and by sonorant. For example, similarly to word-initial position,
word-medial voiceless sonorants are often longer than their voiced counterparts (visible in Figure
3 below). They also usually involve some degree of frication. Frication is greater in medial position
than in initial position, and is greatest in rhotics, followed by laterals, and least in nasals (Bombien
2006). 4
As shown with the sample pair [ˈlampɪ] and [ˈlam̥pɪ] in Figure 3, voiceless sonorants tend
to be produced with noticeable frication and poor or absent formant structure. Hoole & Bombien
(2010) reported that pre-aspirated stops and word-medial voiceless nasals have a very similar

4
The tendency for voiceless laterals and rhotics to be produced with some degree of frication is not unique to Icelandic.
Production of /l̥ / as [ɬ] and /r̥/ as [ʂ] or [ʃ] is also attested in, for example, Tibeto-Burman languages (Lotven et al.
2020). In work focused specifically on laterals, Maddieson and Emmorey (1984) point out that the tendency for
voiceless approximants to be produced with frication is so strong that some scholars have argued that voiceless laterals
must be produced as fricatives. The authors present acoustic data from five languages (Burmese, Navajo, Taishan
Chinese, Tibetan, Zulu) which clearly establishes that while there are measurable differences between voiceless lateral
approximants and fricatives, the differences are subtle. A typological survey of phonotactic patterns in approximately
60 additional languages, however, reveals that voiceless approximants and fricatives pattern differently in terms of
positional restrictions, attested allophonic variants, and so forth. The distinction is therefore both acoustically subtle
and phonologically important. For our purposes, while a more thorough investigation of these sounds is beyond the
scope of the present work, but is certainly a topic to be addressed in the future.

6
gesture of glottal abduction, both in terms of duration and the timing of its onset with respect to
the preceding vowel.

l a m p ɪ

l a m̥ p ɪ
Figure 3 Illustration of a voiced bilabial nasal (top) and a voiceless bilabial nasal (bottom).

2.3 Processes Affecting Consonants

There are a number of additional observations that can be made about the patterning of
consonants in Icelandic. Obstruents exhibit regular alternations directly preceding [l] and [n]. For
example, intervocalic plain stops [p t c/k] alternate with pre-aspirated stops [hp ht hk] before [l] and
[n], and voiced approximants [ʋ ɣ] with plain stops [p k].5 These alternations often arise from
concatenation of the GEN.PL -na suffix (e.g. [ˈkaːta] street.NOM.SG ~ [ˈkahtna] GEN.PL, [ˈsaːɣa]
story.NOM.SG ~ [ˈsakna] GEN.PL), or when word-medial syncope obtains in disyllabic roots ending
in [l] or [n] (e.g. [ˈʔɔːpɪn] open.F.NOM.SG ~ [ˈʔɔhpnar] F.NOM.PL, [ˈphaiːcɪl] brine.ACC.SG ~
[ˈphaihklɪ] DAT.SG, [ˈtjœːʋʏl] devil.ACC.SG ~ [ˈtjœplɪ] DAT.SG). In addition, the voiced
approximants [ʋ ð ɣ] and [l r] are often devoiced to [f θ x] and [l̥ r̥] when they appear in pre-pausal
and utterance final position (Árnason 1980, Dehé 2014).
Voiced fricatives and [r] are frequently reduced or deleted in connected speech (Árnason
1980), and regularly, or even obligatorily, in many consonant clusters (Côté 2004, Helgason 1993,
Rögnvaldsson 1989). In triconsonantal clusters (C1C2C3), which are often encountered at

5
The contrast between palatals and velars is only maintained prevocalically. Before consonants or word-finally they
are neutralized to velars.

7
morpheme boundaries, C3 is always preserved, so if the cluster is to be simplified, C1 or C2 are
omitted (Côté 2004, Rögnvaldsson 1989). Côté (2004) argues that consonantal deletion is driven
by perceptual factors and sonority requirements. For example, deletion mainly targets stops and
non-strident fricatives, the segments with the weakest internal cues. Côté also identifies the
following post-sonorant hierarchy of deletion: C2 obstruents are least likely to be deleted (and
never obligatorily so) when C1 is [r], are likely to be deleted (and even obligatorily so for some
speakers) when C1 is [l], and are obligatorily deleted when C1 is a nasal.
There exists a similar sonorant-specific hierarchy with respect to the patterning of word-
internal voiced fricatives. In native monomorphemic words, all three of the voiced fricatives [ʋ ð
ɣ] surface post-vocalically, only [ʋ ð] surface after [r], only [ʋ] surfaces after [l], and none surface
after nasals (Flego 2017). This r > l > n hierarchy was also identified in Bombien (2006) for degree
of frication and degree of voicelessness in non-modal sonorants, and subsets of this hierarchy also
seem to pattern together in other areas of the sound system (e.g. devoicing of utterance final [l]
and [r] but not [n], manner/laryngeal alternations among obstruents before [l] and [n], but not
before [r]). These sonorant-specific implicational relationships intersect with many phonetic and
phonological phenomena in Icelandic, from static phonotactics inherited in the native vocabulary,
down to low-level phonetic tendencies in the modern language.

3 Vowels
The inventory of contrastive vowel types in Icelandic is traditionally thought to consist of
eight monophthongs and five diphthongs (Garnes, 1974, Pétursson 1976, Þráinsson & Árnason
1992, among others). In the vowel quadrilaterals in Figure 4, approximate positions of
monophthongs and diphthongs are given in the left and right panels, respectively.

i● ●u

ɪ● ●ʏ

ɛ● ●œ ●ɔ
ei øi ou

a ai au
●
Figure 4 Icelandic Vowel Charts.

Vowel length participates in the quantity contrast discussed above. Briefly, all vowels
(monophthongs and diphthongs) are short if followed by a geminate, pre-aspirated stop, or
consonant cluster, and surface as long otherwise.6 The addition of inflectional suffixes can alter
word-internal syllabification, which often induces vowel length alternations (e.g. [ˈfjœːðʏr]
feather.NOM.SG ~ [ˈfjœðrʏm] DAT.PL, [ˈheiːl] whole.F.NOM.SG ~ [ˈheil̥ t] N.NOM.SG, [ˈplauː]

6
There is one particular group of clusters which are not preceded by a short vowel: if C 1 is a stop or [s] and C2 is
one of three segments ([ʋ r j]), the preceding vowel is long. Gouskova (2004) analyzes this phenomenon by
appealing to syllable contact, in which low sonority C1 and high sonority C2 are heterosyllabified as a post-tonic
onset cluster.

8
blue.F.NOM.SG ~ [ˈplauht] N.NOM.SG). Interestingly, Pétursson (1976) reported a slight difference
in intrinsic length between monophthongs and diphthongs, with both long and short diphthongs
being 12-20 ms longer than long and short monophthongs. Garnes (1974) notes that short
diphthongs are still diphthongal, but the overall distance across the vowel space is quite reduced.
A near minimal set of the eight monophthongs and five diphthongs is presented in Table 6.
Due to the post-vocalic context shared among the example words, all stressed vowels are long.

Table 6 Near-minimal set for vowel contrasts

[i] bíður [ˈpiːðʏr] wait.2SG.PRS.IND
[ɪ] biður [ˈpɪːðʏr] ask.2SG.PRS.IND
[ei] beiðu [ˈpeiːðʏ] mantis.ACC.SG
[ɛ] beðin [ˈpɛːðɪn] wait.PST.PTCP.N.NOM.PL
[ai] bæði [ˈpaiːðɪ] both.N.NOM.PL
[ʏ] buðum [ˈpʏːðʏm] invite.1PL.PST.IND
[øi] bauð [ˈpøiːð] invite.1SG.PST.IND
[œ] böðum [ˈpœːðʏm] bathroom.DAT.PL
[u] búðir [ˈpuːðɪr] shop.NOM.PL
[ou] bjóðir [ˈpjouːðɪr] invite.2SG.PRS.SBJV
[ɔ] boði [ˈpɔːðɪ] invitation.DAT.SG
[au] báðir [ˈpauːðɪr] both.M.NOM.PL
[a] baði [ˈpaːðɪ] bathroom.DAT.SG

Formant measurements were taken for each of the long monophthongs in the example
words in Table 6. These are plotted in the left panel of Figure 5, where arrow start-point represents
the formant value taken at 20% of the vowel duration and arrow endpoint indicates the
measurement taken at 80%. It can be seen that the mid vowels exhibit considerable spectral
movement. Garnes (1974) and Pétursson (1976) have previously noted that the short and long
variants of /ɛ œ ɔ/ have very different spectral characteristics. Garnes (1974) reported that the long
allophones of the three mid vowels /ɛ ɔ œ/ are diphthongal, opening and centralizing over the
course of the vowel, and that the short allophones are considerably lower and centralized. This
description is consistent with our informant’s behavior, which is represented in the right panel of
Figure 5.

Figure 5 Vowel plots for our informant. Left: measurements taken at 20% and 80% of each of the eight
monophthongs in the example words given in Table 6. Right: measurements taken at 20% and 80% of short and
long allophones of /ɛ œ ɔ/, averaged across all example words in this paper not beginning with a palatal consonant.

9
 For one of the informants in Pétursson (1976)’s study, durational differences between the
canonically long and short allophones of mid vowels were nearly merged, yet distinct differences
in vowel quality were preserved. Pind (1999) hypothesized that speakers may allow for more
durational overlap between [ɛ] and [ɛː] than between [a] and [aː], since the former contrast includes
the additional disambiguating cue of vowel quality. However, the output of a simple neural
network classifier showed that there was no significant difference in misclassification between
words with [a]/[aː] and those with [ɛ]/[ɛː] when only given durational information.
We note two phonotactic restrictions concerning vowel quality. First, word-initial post-
consonantal [j] precedes only six of the 13 vowel types ([a ɛ œ u au ou]). Given this restriction,
and that post-consonantal [j] participates in the numerous grammatically conditioned vowel quality
alternations found in the language (e.g. boði ~ bjóðir ~ buðum ~ bauð in Table 6, all semantically
related to “invite”), it is unclear whether it should be considered the final consonant in an onset
cluster, or as part of the following vowel (for a discussion of sub-syllabic constituency of onglides
in American English, see Davis & Hammond 1995). This raises the question of whether Icelandic
should be analyzed as having 19 contrastive vowel types instead of the traditionally posited 13.
The second positional restriction involves the dorsal nasals [ɲ ɲ̊ ŋ ŋ̊]. Only a subset of vowel
qualities ([i u ai ou ei øi au]) may precede them, although older speakers may be more likely to
retain monophthongal realizations of the last three ([ɛ œ a]) in this context (Þráinsson & Árnason
1992).
Length does not alternate for vowels in less prominent syllables.7 These weakly stressed
vowels are often subject to deletion in medial syllables or when word-final and followed by a
vowel-initial word (Árnason 1980, Dehé 2008). According to Dehé (2008), likelihood of word-
final vowel deletion is affected by a number of factors, including prosodic boundaries (in turn
affected by syntactic structure and focus structure) and eurhythmy (stress clashes and lapses). For
example, in the illustrative passage below, fara úr kápunni is realized by our informant as
[ˌfaːruˈkhauː…], avoiding a stress lapse between the two prominent syllables [ˌfaː] and [ˈkhauː].
When not deleted, however, unstressed vowels may still show significant spectral reduction.
Helgason (1993) found that unstressed vowels were extremely centralized and much higher in the
vowel space than their short stressed counterparts.

4 Stress and Intonation

In disyllabic words, the first syllable of a word generally receives main stress. In trisyllabic
words (and longer), the first syllable is assigned main stress, but odd-numbered syllables have a
secondary stress and even-numbered syllables are weak (Árnason 1985). Alternations in stress
patterns can be induced by certain prefixes (e.g. ó- “un-”), and some derivational suffixes attract
stress, even if they occupy an even-numbered syllable (Árnason 1987, Bergsveinsson 1963).
Icelandic does not have contrastive lexical intonation like mainland Scandinavian
languages (i.e. Norwegian and Swedish). In general, the strongest syllable of the rightmost
constituent in the utterance bears a pitch accent. However, this seems to interact with the hierarchy
nouns > verbs > prepositions > personal pronouns, where constituents to the right on the
hierarchy are less likely to bear the pitch accent than constituents to their left, even when they
occupy the rightmost position in the sentence (Árnason 1998). In addition, contrastive focus can

7
While the most common vowel qualities in syllables of lesser stress are [ɪ], [a], and [ʏ] (corresponding historically
to the vowel system of inflectional and derivational endings), many proper nouns and loanwords exhibit greater
variation in vowel quality in unstressed syllables (see Árnason 2011:67 for discussion and examples).

10
assign phrasal stress to nearly any part of the utterance when needed (Árnason 1985). In neutral
declarative sentences, the syllable bearing phrasal stress typically has a H*L pitch accent if long,
or H, followed by L on the next syllable, if short (Árnason 1998). This is contrasted with a L*H
contour on the prominent syllable in yes/no questions. In both of these clause types, though, a low
boundary tone (L%) is usually associated with the end of the utterance, and thus with finality. A
high boundary tone (H%) at the end of a domain signals non-finality (i.e. continuation). The H*L
pitch accent and low boundary tone are clearly present on sterkara and veginum, the final words
of the first two sentences in the illustrative passage below. In addition, a high boundary tone
signaling continuation is evident on blés, the final word of the first phrase in a “the more…, the
more…” construction.
In addition to the two bitonal pitch accents (H*L and L*H) advanced by Árnason (1998),
Dehé (2009)’s findings suggest the presence of two monotonal pitch accents (H* and L*), although
the two sets may be in complementary distribution. Dehé makes no claim as to whether there is a
phonological contrast between bitonal and monotonal pitch accents, and leaves this question to
future research. Finally, Icelandic prenuclear accents are characterized by a late rise from a low
accented syllable (L*H), while final nuclear accents are characterized by an early rise and
immediate fall (Dehé 2010).

5 Illustrative passage
Icelandic is a highly inflected fusional language. To save space, its inflectional suffixes are
glossed using one abbreviation each for person, plurality, tense, mood, gender, case, and
definiteness, without periods, in the order they appear here:

Verbal Inflection:
1ST PERSON (1) SINGULAR (S) PRESENT (PS) INDICATIVE (I)
2ND PERSON (2) PLURAL (P) PRETERITE (PT) SUBJUNCTIVE (S)
3RD PERSON (3)

Nominal Inflection (nouns):8

NOMINATIVE (N) SINGULAR (S) DEFINITE (D)
ACCUSATIVE (A) PLURAL (P) INDEFINITE (I)
DATIVE (D)
GENITIVE (G)
Adjectival Inflection (adjectives, pronouns, numerals, participles):
MASCULINE (M) NOMINATIVE (N) SINGULAR (S)
FEMININE (F) ACCUSATIVE (A) PLURAL (P)
NEUTER (N) DATIVE (D)
GENITIVE (G)
Einu sinni deildu norðanvindurinn og sólin um,
ˈʔeiːn-ʏ ˈsɪnː-ɪ teil-tʏ ˈnɔrðanʋɪnt-ʏrɪn ɔ ˈsouːl-ɪn ˈʔʏmː
one-NDS time-DSI dispute-3PPTI north.wind-NSD and sun-NSD about
"One time the north wind and sun were quarreling over"

8
Icelandic nouns do not inflect for gender in the way adjectives do—rather, grammatical gender is an intrinsic,
immutable property of a noun in Icelandic. As such, in an effort to streamline interlinear glosses, we opted not
to include gender for nouns.

11
hvort þeirra væri sterkara. Þau sáu þá
k ʋɔr̥-t
h
θeira ʋair-ɪ ˈstɛr̥k-ar-a θøi ˈsauː-ʏ θau
which.of.two-NNS NGP be\PTS-3SPTS strong-CM-NNS NNP see\PTI-3PPTI then
"which of the two of them was stronger. Then they saw"

mann í hlýrri kápu á ferð á veginum.

ˈmanː-Ø i ˈl̥ i-rːɪ ˈkhauːp-ʏ au ˈfɛrθ-Ø au ˈʋeiːj-ɪnʏm
man-ASI in warm-FDS coat-DSI on route-DSI on way-DSD
"a man in a warm coat coming along."

þeim kom þá saman um að það þeirra skyldi

θeim ˈkhɔːm-Ø θau ˈsaːman ˈʔʏmː a ˈθaːθ ˈθeirːa scɪlt-ɪ
NDP agree\PTI-3SPTI then together agree CONJ NNS NGP FUT.PTS-3SPTS
"They agreed that the one who could force the man”

teljast sterkara sem gæti neytt ferðamanninn

t ɛl-ja-st
h
ˈstɛr̥k-ar-a sɛm cait-ɪ ˈnei-ht ˈfɛrðaman-ɪn
count-INF-RFL strong-CM-NNS REL can\PTS-3SPTS force-PTCP.NNS traveler-ASD
"to take his coat off would be considered”

til þess að fara úr kápunni. Norðanvindurinn tók þá til að

t ɪl θɛs a
h
far-a ur ˈk auːp-ʏnɪ
h
ˈnɔrðanʋɪnt-ʏrɪn ˈt ouk-Ø
h
θau ˈthɪ:l a
to go-INF out.of coat-DSD north.wind-NSD start\PTI-3SPTI then to
"the stronger one. The north wind then began”

blása af öllum mætti, en því meira sem hann

ˈplauːs-a aʋ ˈʔœtl-ʏm ˈmaiht-ɪ ɛn θi ˈmeiːr-a sɛm han
blow-INF with all-MDS might-DSI but CM much.CM-NS REL MNS
"to blow with all his might, but the harder he”

blés, því þéttara vafði ferðamaðurinn kápunni að

ˈpljɛːs-Ø θi ˈθjɛ t-ar-a
h
ˈʋaʋ-ðɪ fɛrðamað-ʏrɪn ˈkhauːp-ʏnɪ ˈʔaːð
blow\PT-3SPTI CM tight-CM-NS wrap\PT-3SPTI traveler-NSD coat-DSD around
"blew, the more tightly the traveler wrapped his coat around”

sér; og að lokum gafst norðanvindurinn upp.

sjɛr ɔ a ˈlɔːk-ʏm ˈkaf-st nɔrðanʋɪnt-ʏrɪn ˈʔʏhp
RFL.3SG.DAT and at end-DPI give.up\SPTI-RFL north.wind-NSD give.up
himself; and finally the north wind gave up."

Svo fór sólin að skína og það varð

ˈsʋɔː four-Ø ˈsouːl-ɪn a ˈsciːn-a ɔ θað ʋarð-∅
so start\PTI-3SPTI sun-NSD to shine-INF and NNS become\SPTI-3SPTI
"So the sun began to shine and it became”

12
hlýtt. Þá fór ferðamaðurinn undir eins úr kápunni.
ˈl̥ i- t
h
ˈθauː four-Ø ˈfɛrðamað-ʏrɪn ˈʔʏntɪr eins ur ˈkhauːp-ʏnɪ
warm-NNS then go\PTI-3SPTI traveler-NSD immediately out.of coat-DSD
"warm. Then the traveler took off his coat at once.”

Norðanvindurinn áttaði sig þá á því að sólin

ˈnɔrðanʋɪnt-ʏrɪn ˈʔauht-aðɪ sɪx θau ˈʔauː θʋi a ˈsouːl-ɪn
north.wind-NSD realize-3SPTI RFL.3SG.ACC then CONJ sun-NSD
"The North wind realized then that the sun”

væri sterkari en hann.

ʋair-ɪ ˈstɛr̥k-ar-ɪ ɛn han
be\PTS-3SPTS strong-CM-MNS than MNS
"was stronger than him.”

Acknowledgements
We would like to thank our language consultant, Guðbjartur Hákonarson, for his knowledge,
time, and patience. We would also like to thank four anonymous reviewers for their insight and
thoughtful suggestions.

References
Árnason, K. (1980). Some processes in Icelandic connected speech. The Nordic Languages and
Modern Linguistics, 4, 212-222.
Árnason, K. (1985). Icelandic word stress and metrical phonology. Studia Linguistica, 39(2), 93-
129.
Árnason, K. (1987). The stress of prefixes and suffixes in Icelandic. In Nordic Prosody IV:
Papers from a Symposium (pp. 137-146). Odense University Press.
Árnason, K. (1998). Toward an analysis of Icelandic intonation. Nordic Prosody, 7, 49-62.
Árnason, K. (2011). The Phonology of Icelandic and Faroese. Oxford University Press.
Bergsveinsson, S. (1963). Die Wortbetonung im Isländischen. Phonetica, 10(3-4), 213-221.
Boersma, P. & Weenink, D. (2019). Praat: doing phonetics by computer [Computer program].
Version 6.0.54, retrieved 6 June 2019 from http://www.praat.org
Bombien, L. (2006). Voicing alterations in Icelandic sonorants– A photoglottographic and
acoustic analysis. Arbeitsberichte des Instituts für Phonetik und digitale
Sprachverarbeitung der Universität Kiel (AIPUK), 37, 63-82.
Côté, M. H. (2004). On perception and sonority in cluster simplification: Consonant deletion in
Icelandic preterites. Journal of Germanic Linguistics, 16(3), 203-243.
Davis, S., & Hammond, M. (1995). On the status of onglides in American English. Phonology,
12(2), 159-182.
Dehé, N. (2008). To delete or not to delete: The contexts of Icelandic final vowel deletion.
Lingua, 118(5), 732-753.
Dehé, N. (2009). An intonational grammar for Icelandic. Nordic Journal of Linguistics, 32(1), 5-
34.
Dehé, N. (2010). The nature and use of Icelandic prenuclear and nuclear pitch accents: Evidence
from F0 alignment and syllable/segment duration. Nordic Journal of Linguistics, 33(1),
31-65.
13
 Dehé, N. (2014). Final devoicing of /l/ in Reykjavík Icelandic. In Speech Prosody, 7, 757-761.
Eberhard, D. M., Simons, G.F., & Fennig, C. D. (eds.). (2019). Ethnologue: Languages of the
World. Twenty-second edition. Dallas, Texas: SIL International. Online version:
http://www.ethnologue.com
Flego, S. (2017, September). Evidence for markedness hierarchies from continuancy
neutralizations in Icelandic. Paper presented at the 22nd meeting of the Mid-Continental
Phonetics & Phonology Conference, Columbus, OH.
Garnes, S. (1974). Suprasegmental aspects of Icelandic vowel quality. OSU Working Papers in
Linguistics, 17, 144-159.
Garnes, S. (1976). Quantity in Icelandic: Production and Perception (1. Aufl ed., Hamburger
Phonetische Beiträge, Bd. 18). Hamburg: Buske.
Gouskova, M. (2004). Relational hierarchies in OT: The case of syllable contact. Phonology
21(2), 201-250.
Hansson, G. Ó. (2003). Laryngeal licensing and laryngeal neutralization in Faroese and
Icelandic. Nordic Journal of Linguistics, 26(1), 45-79.
Helgason, P. (1993). On coarticulation and connected speech processes in Icelandic.
Málvísindastofnun Háskóla Íslands.
Helgason, P. (2002). Preaspiration in the Nordic languages: Synchronic and diachronic aspects
(Doctoral dissertation, Institutionen för Lingvistik).
Hoole, P., & Bombien, L. (2010). Velar and glottal activity in Icelandic. Between the Regular
and the Particular in Speech and Language, 171-204.
Lotven, S., Berkson, K., Wamsley, J. C., Danaher, J., Van Bik, K., & Davis, S. (2020). The
syllable in Kuki-Chin. Journal of South Asian Languages and Linguistics, 6(2), 281-308.
Löfqvist, A., & Pétursson, M. (1976). Swedish and Icelandic stops-A glottographic investigation.
The Nordic Languages and Modern Linguistics, 3, 454-461.
Löfqvist, A., & Yoshioka, H. (1981). Laryngeal activity in Icelandic obstruent production.
Nordic Journal of Linguistics, 4(1), 1-18.
Maddieson, I., & Emmorey, K. (1984). Is there a valid distinction between voiceless lateral
approximants and fricatives? Phonetica, 41(4), 181-190.
Pétursson, M. (1972). La préaspiration en islandais moderne. Examen de sa réalisation
phonétique chez deux sujets. Studia Linguistica, 26(2), 61-80.
Pétursson, M. (1976). Quantität und vokalische Klangfarbe im modernen Isländischen. STUF-
Language Typology and Universals, 29(1-6), 49-63.
Pind, J. (1995). Speaking rate, voice-onset time, and quantity: The search for higher-order
invariants for two Icelandic speech cues. Perception & Psychophysics, 57(3), 291-304.
Pind, J. (1999). Speech segment durations and quantity in Icelandic. The Journal of the
Acoustical Society of America, 106(2), 1045-1053.
Ringen, C. O. (1999). Aspiration, preaspiration, deaspiration, sonorant devoicing and
spirantization in Icelandic. Nordic Journal of Linguistics, 22(2), 137-156.
Rögnvaldsson, E. (1989). Íslensk hljóðfræði. Málvísindastofnun Háskóla Íslands, Reykjavík,
Iceland.
Silverman, D. (2003). On the rarity of pre-aspirated stops. Journal of Linguistics, 39(3), 575-598.
Þráinsson, H. (1978). On the phonology of Icelandic preaspiration. Nordic Journal of
Linguistics, 1(1), 3-54.
Þráinsson, H., & Árnason, K. (1992). Phonological Variation in 20th Century Icelandic.

14

View publication stats