ESUKA – JEFUL 2018, 9–1: 209–244
ESTONIAN WORD PROSODY ON THE PROCRUSTEAN BED
OF MORAE
Natalia Kuznetsova
Institute for Linguistic Studies, Russian Academy of Sciences,
and the University of Turin
Abstract. The paper analyses existing moraic conceptions of Estonian quantity. Main
features of functional, generative and phonetically-instructed moraic accounts of Estonian are considered. In most generative accounts, morae simultaneously represent
several layers of functionally and structurally diverse information. This brings along
a considerable increase in formal analytical machinery and internal controversies. In a
structural functional framework, morae can be used to formalise the prosodic contrast
of long and short stressed syllables in Estonian. Its relevance is seen in actual functioning of the prosodic system. This contrast is built upon the segmental contrast of short
and long phonemes and, in turn, serves as a basis for the contrast of two distinctive foot
accents, light and heavy. As an example, a formal morphonological algorithm of calculating Estonian foot accents, which also shows the place of the syllable weight contrast, is
proposed.
Keywords: Estonian, structural functional phonology, autosegmental phonology, word
prosody, mora, quantity
DOI: https://doi.org/10.12697/jeful.2018.9.1.09
1. Introduction
Estonian phonology has been a subject of extensive research since
the 1930s, which has progressed within three main frameworks: structuralist, generative, and the studies by phoneticians who have sought to
formalise the results of phonetic experiments.
Structural functional analysis of Estonian during the 1960s and
1970s laid the basis for its current phonological understanding (Asu
et al. 2016: 238), but now this paradigm is underrepresented in Estonian phonology. Most of research is conducted either within generative paradigms or in phonetics. However, its potential has not yet been
fully exhausted. On the contrary, a functional and cognitive view on
the language structure, when the latter is seen as a function of general
210 Natalia Kuznetsova
processes of human communication and cognition, seems to be gaining
popularity again, also in phonology. For example, paradigms represented
by Mompéan-González (2014), Hall et al. (2018), and Kapatsinki (2018)
are in the same line with those proposed in earlier functionalist works
(e.g. Jakobson 1960, Kasevič 1983). Phonology in a structural functional paradigm is seen as one of the auxiliary levels helping to transmit
a message and achieve efficient communication. Phonological structure
and its dynamics are results of language use by human beings and therefore follow general cognitive mechanisms. The present paper aims at
developing the functional-structural approach to Estonian phonology
further, as well as at its critical comparison with other frameworks.
The main object of the present study is the mora in Estonian. It is an
important notion for many accounts of Estonian phonology, discussed
below, and is typically referred to as “one of what heavy syllables have
two of” (Prince 1980: 525–526). A circular definition allows a very loose
usage of the term. At the same time, the mora lies at the very core of
formal representations, so it is extremely important to understand its
real content in various accounts. The typological concept of the mora
is addressed in Section 2, and existing moraic accounts of Estonian in
Section 3. Section 4, contributes to the grounding of the functional-structural analysis of the Estonian morae, and Section 5 contains conclusions.
Any moraic account in Estonian relies on a certain conception of
stress (both primary and secondary). There is still not a final consensus on
stress placement rules among Estonian phonologists and phoneticians (cf.
latest remarks in Hint 2001: 253–257, Pajusalu et al. 2005: 100, Asu and
Lippus 2018). Estonian stress requires further phonetic and phonological
investigation, so the very facts which form the basis for most moraic
accounts of Estonian are still to be verified. This remark made, Estonian
stress will be left aside here.
2. The mora in linguistic typology
2.1. Functional structural approach of the Prague school
The term mora (‘duration of time, delay’ in Greek) originally referred
to the smallest rhythmic unit that permitted a description of stress rules
in the poetic metrics of Ancient Greek and Latin; similar units were used
in Sanskrit (mātrā) and Japanese (on).
Jakobson (1931/1962, 1937/1962) and Trubetzkoy (1935/1968,
1939/1969) introduced the term into linguistic typology. The former
Morae in Estonian 211
applied it to a distinction of pitch-accent types in certain languages.
For Trubetzkoy, mono- vs. bimoraicity was a way to describe general
prosodic homogeneity vs. heterogeneity of the syllable nucleus. Short
syllable nuclei behave as a single prosodic domain, while the two
parts of long syllable nuclei are “treated differently prosodically, the
treatment of the difference being distinctive” (Trubetzkoy 1939/1969:
175). Several points of this conception (ibid.: 173–181) which have
entailed major inconsistencies in later theoretic advances on the mora
are summarised below.
(1) Trubetzkoy gives no clear list of relevant prosodic properties
that might allow morae to be detected. He mentions stress and stød
placement rules, contrasts of tones and pitch accents, and also prosodic
equality between long vowels and polyphonematic diphthongs or
combinations of a vowel and a consonant.
(2) The maximal number of morae in a syllable remains unclear. The
contrast between long and short syllable nuclei is generally considered
as logically privative, but three- and four-mora nuclei are assumed
possible for tonal languages.
(3) The mora is an abstract phonological notion, but Trubetzkoy
refers to its possible phonetic correlates, such as a longer physical duration or a presence of stød in long nuclei.
(4) The very nature of the mora raises questions. It is an analytical
tool helping Trubetzkoy to describe the dual nature of syllable. The latter
is a segmental phonotactic unit which consists of phonemes, but also
a prosodic domain which carries word-prosodic units (Hyman 1985:
1). Trubetzkoy observed various possible correlations between the two
functions of the syllable, and the mora notion allowed him to typologise
those. However, the two roles of the syllable are not clearly distinguished by him. On the one hand, mono- vs. bimoraicity is claimed to
be an abstract prosodic property of syllable nuclei; on the other hand,
the mora is referred to as the smallest prosodic unit and a constituent of
long syllable nuclei (Trubetzkoy 1939/1969: 170, 177, 182).
Trubetzkoy did not specify any exact relations between moraic,
syllabic and segmental boundaries. He considered Estonian to be a
mora-counting language, obviously on the same grounds as Saami
(Lapp), where “long vowels occur only in the same environments as
clearly biphonematic diphthongs”. For such multiple length systems, he
still admitted only a binary phonological contrast of mono- and bimoraic syllable nuclei (ibid: 180–181).
212 Natalia Kuznetsova
2.2. Formalisation of the mora: metrical and autosegmental
approaches
Subsequent years saw an acute interest in syllable structure and a
rise in research on African tonal languages (cf. Pike and Pike 1947,
Kuryłowicz 1948). The term ‘syllable weight’ was coined by Allen
(1965) and brought into general use by Newman (1972), who defined
syllable weight as an intrinsic structural rather than a positional property
of syllables.
Syllable weight was further formalised in two main directions, both
of which used morae (viz. Fox 2000: 79–85). A metrical approach originated from studies on stress and an autosegmental one from those on
tone. In both, weight became primarily a positional property of syllables. The mora, the weight unit, kept its dual nature as either a concrete
part of the syllable or a way to measure its prosodic properties. Attempts
to establish a formal prosodic hierarchy and clear correspondences
between morae and segments aggravated the original opacity of the term
rather than resolved it.
Within the metrical approach, Prince declared the mora an “immediate constituent of the [syllable] rime”, notably in his study on Estonian quantity (Prince 1980: 526; see Fig. 1a). His metrical grid structure
presumes that lower level units are constituents of higher level units.
However, the statement that “long vowels have two moras, short vowels
one” (Hayes 1989: 256, cf. Fig. 1c) implies that morae can be also
constituents of vowels. In this case, the moraic level should be logically
placed below the phonemic one. On the other hand, morae also represent
the syllable’s positional prosodic properties, because syllable weight is
detected according to the word stress rules, etc. Morae formalise a link
between syllables and higher hierarchic levels (foot/word). In this case,
it would be logical to place the moraic tier over the syllabic one. Prince’s
metrical grid makes neither choice, as it places morae between syllables
and segments. The same basic way of representation was maintained by
Hyman (1985: 17, Fig. 1b, ‘x’ stands for mora) and Hayes (1989: 254,
Fig. 1c; 1995) and became standard in moraic phonology.
Morae in Estonian 213
a.
b.
c.
σ
σ
σ
σ
Rime
X
Onset
Mora
Mora
C C
V
VC
t
X
a
m
t
μ
μ
a
t
μ
t
μ
a
Figure 1. Morae within a metrical stress grid by (a) Prince,
(b) Hyman, (c) Hayes
An alternative representation of morae is their autosegmentalisation.
This approach was first developed for tonal languages by Hyman, who
claimed it “incompatible with the branching syllable structure” (Hyman
1985: 3). He followed Goldsmith (1976: 27–28), who distinguished
between two independent layers of information realised simultaneously
in the phonetic signal: sequences of segmental and suprasegmental
units. None of these levels is subordinated to the other, they are just
associated. Hyman (1985: 16; Fig. 2a) introduced the level of morae
(X-tier) between segments and suprasegmentals for tone to formalise
the rules of their association.
His moraic representation of stress, however, is not autosegmental,
as it follows the metrical grid structure (Fig. 1b). A real autosegmentalisation for stress morae could look like that provided by Odden (1997:
175, Fig. 2b), where an X-tier shows segmental slots, and morae are
placed above the syllable. The dual nature of the syllable as a domain for
segmental grouping and a bearer of prosodic units can be clearly seen,
if segmental and suprasegmental information is given separately. The
necessity of autosegmentalising tone in formal representations became
apparent to researchers early enough, but this is still not universally
admitted for other kinds of lexical prosody, including suprasegmental
quantity in Estonian (see below).
Hayes (1995: 299, Fig. 2c) also proposed a dual distinction of weight
for languages in which CVC syllables can behave as light or heavy
depending on the prosodic process they are involved in. However, both
moraic tiers are still placed below the syllabic level, so the syllable
weight is not autonomous from segmental composition. The relationship
between the two moraic tiers is one of dependency rather than association. Fox (2000: 110, Fig. 2d) suggested a more autosegmental version
214 Natalia Kuznetsova
of the dual distinction of weight, with the “foot quantity” level above the
syllabic one. Seeing syllabic and segmental quantity as properties rather
than constituents, he eventually placed the two outside a strictly-layered
hierarchy of “true” units, as dependent tiers associated with the latter
(ibid.: 111, Fig. 2e).
a.
H
X
t
X
r
c.
σ
μ
Q Q Q
XX X X X
μ
μ
t a
t
p
i
e.
F
μ
X
a
d.
σ
μ
L
X
e
b.
a
k
σ
σ
μ
μ
F
Q tier
n ε v ə
σ
μ tier
Figure 2. Autosegmentalisation of morae for (a) tone by Hyman
and (b) stress by Odden; dual distinction of weight by (c) Hayes
and (d, e) Fox
2.3. Phoneticisation of the mora: mora-splitting
The search for phonetic correlates of the mora became popular in the
1990s, with a rapid growth in experimental phonetic studies. Moraic analysis was picked up by phoneticians after a mora-splitting procedure was
introduced by Maddieson (1993: 14; Fig. 3) and promoted in Hubbard
(1994), Broselow et al. (1997), and others. Maddieson (1993: 9) claimed
that “moraic structure and duration show a general relationship”. As
surface durational contrasts often show a more complex structure than
just a binary distinction, mora-splitting/sharing between segments was
proposed to account for this.
σ
σ
σ
σ
σ
σ
μ
μ μ
μ
μ
μ μ
μ
a t e m b a
a t e m b a
Figure 3. Mora splitting (Maddieson): /atemba/ > [ateˑmˑba]
Morae in Estonian 215
Notably, even the split morae are not able to cover all observed durational microvariability language-internally and cross-linguistically. It is
admitted that morae do not need to reflect exact durations (e.g. Lehiste
1960: 51–52, Maddieson 1993: 9). On the other hand, on the road of
mora-splitting there is not a natural border between cases where the
moraic structure should still reflect this variability and where it should
stop doing so. Therefore, some have even proposed to separate phonological weight from phonetic duration entirely (Erwin 1996, Gordon
2004).
Putting phonological morae in direct correspondence with duration
is similar to an attempt to depict fine qualitative allophonic variability
of segments in phonological transcription. The ineffectiveness of this
idea for quantity might seem less obvious than for quality, because the
former is often seen as an entirely suprasegmental feature. It is a paradox
that a typical metrical grid at the same time de facto links quantity to
segments. Any kind of quantity linearisation at just one hierarchical
level fails to distinguish between various layers of linguistic information
transmitted by the speech signal:
Length is not just a matter for segments; whatever higher units of prosodic structure we postulate, they have extent in time, and the temporal
structure of utterances is a reflection of the timing relations present at
the different levels within this structure... Furthermore, since timing
relations at every level are manifested along a single dimension, time,
the actual lengths of individual segments and syllables are the complex
result of combining factors of different kinds (Fox 2000: 107–108).
Especially in such complex quantity systems like Estonian, a linearisation of an essentially non-linear structure brings along an everlasting
battle with formalism, with always more sophisticated tricks used but
results still far from satisfactory.
3. Existing moraic analyses of Estonian
3.1. Main facts about Estonian stress and quantity
Basic facts about Estonian stress, syllabic structure, segmental and
suprasegmental quantity are given below. The phonological interpretation mainly follows structural analyses by Viitso (1978, 1979, 2003),
Eek (1986, 1990), and Hint (1973, 2001, 2015a, b). I have discussed it
in detail elsewhere (Kuznetsova 2018) and here give a brief summary.
216 Natalia Kuznetsova
The distinction of three quantities is phonetically realised within the
disyllabic foot nucleus (a prosodically active sequence from the first
syllable vowel throughout the second syllable vowel)1. The Estonian
foot can contain from one to three syllables (cf. Lippus et al. 2013,
Pajusalu 2015 for the latest overviews). In a monosyllabic foot, the
foot nucleus is truncated, and there is only a binary segmental length
contrast. The following quantitative proportions between the three key
segments in the foot nucleus seem the most important factors in the
production and perception of quantities (Eek and Meister 2004: 271):
Q1: σ1 nucleus
<
σ2 nucleus
Q2: σ1 nucleus
≥
σ1 coda
≤
σ2 nucleus
Q3: σ1 nucleus
<
σ1 coda
>
σ2 nucleus
Authors’ abbreviations: ‘σ1 nucleus’ – a short vowel, the first part of a long vowel/
diphthong; ‘σ1 coda’ – the last element of the 1st syllable (i.e. the second part of a
long vowel/diphthong, the first consonant in a cluster, the first part of a geminate);
‘σ2 nucleus’ – a 2nd syllable vowel.
Figure 4. Proportions between three key segments in the foot
nucleus (Eek and Meister 2004: 271)
The surface ternary contrast contains two binary phonological
contrasts. First, there is a segmental contrast of long vs. short vowels and
fortis vs. lenis consonants. Second, there is a suprasegmental contrast of
two foot accents: light /´/ and heavy /`/. The foot accent is a lexicalised
(morphologised) prosodic pattern of realisation of the foot nucleus.
Di- and trisyllabic feet in Q1 and Q2 phonologically differ only in
their segmental content. The former starts with a short first syllable
(open with a short vowel), and the latter with a long first syllable (closed
and/or containing a long vowel or a diphthong), cf. types 2a vs. 2b–f in
1
A more conventional term for the foot nucleus would be e.g. minimal foot in Eek and
Meister (1997) or just foot in most generative accounts (Prince 1980 etc.). However,
the foot is understood here as “the formal vehicle for stress” (Odden 1997: 178), not as a
structural unit. It starts with a stressed syllable and ends before the following stressed
syllable. Unparsed syllables are permitted only at the beginning of a morphological
word, if the first lexicalised stress falls on its second syllable (Viitso 2008: 183–184).
The accent-bearer can be, however, shorter than the foot itself (it does not contain the
first syllable onset, the coda of the second syllable and the whole third syllable). This
prosodically active part of the foot is called the foot nucleus in the same way as syllable
nucleus defines the prosodically active part of a syllable.
Morae in Estonian 217
Table 2. The feet in Q3, in turn, can have an identical phonemic structure to those in Q2 and contrast to the latter only prosodically, i.e. in
accent, cf. types 2b–f vs. 3a–e in Table 2. All monosyllabic feet carry
heavy accent (Table 2, 1a–e).
Estonian syllabic structure is important for a discussion of moraicity.
Non-initial syllables will not be analysed in detail. There are in total five
quantitative types of the first syllable nucleus in monosyllabic feet and
six types in di- and trisyllabic feet (cf. Hint 2001, Viitso 2003, Eek and
Meister 2004). Five of the latter six types can occur in feet under both
accents, and one only under a light accent (see structures and examples
for mono- and disyllabic feet in Tables 1 and 2).
Table 1. First syllable quantitative structure in mono-, di- and trisyllabic Estonian feet
Monosyllabic foot
Heavy accent /`/
Di- and trisyllabic feet
Light accent /´/
Heavy accent /`/
´V
`VV
(C(C))
´V
`VV
´VC
`VC
´VCC
`VCC
`VVC
´VVC
`VVC
`VVCC
´VVCC
`VVCC
`VCC
(C(C))
S2(S3)
`VVCCC
Abbreviations: VV – a long vowel/diphthong; CC – a fortis consonant or a cluster;
(C(C)) – optional syllabic onsets; S2(S3) – presence of the 2nd and optionally the 3rd
syllable in the foot.
Estonian stress contains two main distinctions important for
phonology. First, the foot stress can be morphologically unbound or
bound. In other words, there is a default foot-rhythmic stress and a lexicalised foot stress (see examples in Viitso 2003: 17). The distinction
between rhythmic and lexicalised stresses was first made for Estonian
by Hint (1973) and discussed at length in Viitso (1979), Eek (1986).
Moreover, it finds robust grounding in modern typological works on
stress and accent (van der Hulst 2010, 2014). Second, within a multifoot
word, there can be primary and secondary foot stresses.
No S1 nucleus
structure
1
Disyllabic foot
Heavy accent /`/
Light accent /´/
V
2
VV
3
VC
4
Monosyllabic foot
VCC
(1a) öö
/`ȫ/
(1b) sakk
/`sak̄ /
[øːˑ]
2
(2a) ude
/´ute/
[ˈud̥ eˑ]
(2b) uude
/´ūte/
/´ut̄ e/
[ˈuːd̥ e(ˑ)]
(3a)
uude
[ˈutːe(ˑ)]
(3b)
utte
(2c) ute
[sakːˑ]
[saːˑg̊ ]
(2d) nitro
(2e) uute
/´nit̄ ro/
/´ūt̄ e/
[saːˑkː]
(2f)
/´sūrt̄ e/
5
VVC
(1c) saag
6
VVCC
(1d) saak
/`sāk/
/`sāk̄ /
7
VVCCC
(1e) poort
/`pōrt̄ / [poːˑrtː]
Heavy accent /`/
suurte
/`ūte/
/`ut̄ e/
[ˈuːˑd̥ e]
[ˈutːˑe]
uute
/`at̄ ra/
/`ūt̄ e/
[ˈuːˑtːˑe]
poorti
/`pōrt̄ i/
[ˈpoːˑrtːˑi]
[ˈnitːro]
(3c)
atra
[ˈuːtːe]
(3d)
[ˈsuːrtːe]
(3e)
[ˈɑtːˑrɑ]
Glosses: (1a) ‘night’; (1b) ‘jag’; (1c) ‘saw’; (1d) ‘prey, yield’; (1e) ‘border of a fabric/rug‘; (2a) ‘fuzz’; (2b) ‘innovation’; (2c) ‘ewe.GEN’; (2d) ‘nitro-’;
(2e) ‘new:GEN.PL’; (2f) ‘big:GEN.PL’; (3a) d ‘new:ILL.’; (3b) ‘ewe.PART’; (3c) ‘plough.PART’; (3d) ‘innovation.GEN’; (3e) ‘border of a fabric/rug.PART’
‘border of a fabric/rug.PART’ (examples from Viitso 2003: 13–16, Hint 1997: 130).
2
The phonetic transcription does not reflect minor phonetic effects of foot isochrony, when neighboring sound durations influence each other (e.g.
adjacent long vowels and fortis consonants), as they are not directly relevant for the present discussion. Vowels and consonants bearing the quantitative peak of Q3 are marked as [aːˑ]. Lenis stops and s are marked as [d̥ ], gemination of fortis consonants at the syllable boundary is not reflected.
Second syllable vowel in Q2 is reflected as [e(ˑ)] in some structures, as its duration in Standard Estonian can vary from short to long (Eek and
Meister 2004: 350).
218 Natalia Kuznetsova
Table 2. Examples of quantitative structures in mono- and disyllabic Estonian feet 2
Morae in Estonian 219
Rhythmic stress is always secondary and exhibits only the durational patterns Q1 and Q2, so no accentual contrasts. Lexicalised stress
is linked to a given position in certain morphemes and cannot shift to
another syllable. Besides, it is distinctive in a similar way to tone in
tonal languages: it distinguishes between the light and the heavy foot
accent. Only lexicalised stresses (i.e. foot accents) are relevant for
phonology, while the rhythmic stress is a phonetic phenomenon.
3.2. Morae in the formalisations of phonetic results
Estonian phoneticians understand the mora as a “minimal abstract
unit of temporal regulation” (Eek and Meister 1997: 87) or as an
“abstract isochronous unit of timing” (Lehiste 1990: 290). Eek and
Meister (1997, 2004) preserved the maximal bimoraicity of the syllable
and so had to use the procedure of mora-splitting/sharing with all its shortcomings: crossing association lines, generation of unattested structures,
the need to treat long vowels as sequences of two identical phonemes, etc.
(see Section 2.3 and Hint 2001: 170, Ehala 2003: 57, Prillop 2013: 9–10).
Plüschke (2011, 2013) used morae just to visualise her hypothesis that
in each quantity degree the pitch peak falls after the first mora, which
was actually not fully confirmed by her results. She adopted trimoraic
syllables and a varying duration of mora in each quantity degree, while
not entering into theoretical discussions about the term itself.
A more explicitly theoretical stance on the mora is found in earlier
accounts by Wiik (1982, 1985, 1991) and Lehiste (1960, 1990, 1997a, b).
Wiik applied morae to formalise the phonetic foot isochrony rules. He
counted morae in syllables rather than segments and left aside the question of overlength. This resulted is a very simple analysis: short syllables are monomoraic, long ones bimoraic. The first short vowel of a
syllable counts as one mora and all kinds of following “tails” as the
second mora (Wiik 1991: 298). The foot isochrony tendency requires the
second vowel of a disyllabic foot nucleus to lengthen after a monomoraic syllable and shorten after a bimoraic one (see further in Section 4).
Lehiste (1960: 51) originally treated mora as “the phonemic unit of
segmental quantity” and assigned one to three morae to each segment.
This resulted in syllables containing up to six morae (in type 3e, Table 2)
or with the same number of morae but belonging to two different quantities, and also cases where a Q2 syllable contained more morae than a Q3
syllable (Lehiste 1997b: 13). Lehiste admitted that the mora is a relative
220 Natalia Kuznetsova
perceptual unit of duration, but still searched for its physical correlate in
the “peaks of energy” inside sounds (1960: 51; cf. with Gordon 2004).
Lehiste (1990, 1997a) also explored mora-counting in whole syllables.
Apart from the problem of moraic ambiguity of unstressed syllables,
her results showed that, unlike in Japanese, “neither the syllable, nor the
mora constitute the basic unit of timing” in Estonian (Lehiste 1990: 290).
Her general conclusion is relatively drastic:
If it is considered desirable that phonological constructs have phonetic
backing, the theories that use mora-counting as supportive evidence
need to be revised; alternatively, the term “mora” should be redefined
in terms other than duration (ibid.).
Lehiste rejected morae due to the theoretical unacceptability of
either mora-sharing or the varying physical length of the mora. Even
her disregard for the maximal bimoraicity of the syllable was not able to
solve all the descriptive problems which arise with an attempt to explain
all Estonian durational phenomena through morae.
3.3. Earlier formal moraic accounts of Estonian
There exist several formal phonological moraic accounts of Estonian
both within generative (Prince 1980, Kager 1995, Hayes 1989, 1995,
Bye 1997, Odden 1997, Ehala 1999, 2003) and functional (Hint 1978,
1980, 2001) frameworks. The latest generative advances by Prillop are
considered in Section 3.4.
Hayes (1989) understood the mora in a dual way, as a unit of both
segmental quantity and positional syllable weight, and accepted trimoraic
syllables for Estonian. Prince (1980), who treated the moraic contrast
as binary, recognised the non-linear nature of Estonian quantity. He
saw it as “the product of multiplying a segmento-syllabic distinction
(heavy/light) by a prosodic distinction (foot/nonfoot)” (ibid.: 559) and
suggested that a Q3 syllable always constituted a foot on its own. This
view was followed by Kager (1995), but also extensively criticised
(Viitso 1982, Bye 1997, Odden 1997, Hint 2001, Ehala 2003).
First, it still implies trimoraic syllables in types like (3c), see Fig. 5b.
Second, as a consequence, the recursive foot (consisting of a smaller
foot plus unparsed syllables), had to be accepted, cf. extreme cases of
a double recursion in Fig. 5a, b (Prince 1980: 530, 549). The recursive
foot contradicts the classic version of Prosodic Hierarchy “word > foot
Morae in Estonian 221
> syllable > mora” (Selkirk 1980, 1984). This hierarchy is still widely
accepted in generative linguistics and prohibits feet from comprising
of feet (in non-generative accounts, however, its universal validity is
put under doubt, cf. Schiering et al. 2010). The recursive foot implies a
combination of at least two homonymic notions of the foot with different
content: “Fstress” and “Fquantity” (Odden 1997: 180). The stress foot
includes various kinds of structural “quantity feet” at lower levels. In the
present paper, the foot is understood strictly as “Fstress” (cf. Footnote 1),
and so no recursive feet are allowed.
Also, other problematic issues of Prince’s conception have been
discussed:
– unattested stress patterns generated by his theory, as well as some
unconfirmed data on Estonian stress patterns that he used;
– inability to account for word-initial unstressed (extrametrical)
syllables;
– a need to accept either a sequence of two unparsed syllables between
the two ‘minimal feet’ (as in Fig. 5a) or a so-called ‘weak layering’, in
which elements are directly linked to the higher levels of the prosodic
hierarchy, bypassing the lower ones (see Section 3.4);
– unnatural prosodic treatment of word-final geminates, cf. Fig. 5c
(Prince 1980: 532);
– inability to account for prosodic alternations in certain declinational
types.
Finally, the treatment of a Q3 syllable as a separate foot does not
correspond to existing perceptual data on Estonian: “accents ...can be
identified only when information about V2 has also been delivered to
listeners” (Eek and Meister 1997: 91).
a.
b.
F
c.
F3
M
s
F
s
F2
F1
s
w
F
F
F
sw ww
kau:kele
kaugele /`kaukele/
‘far:ALL’
vem p la
vembla /`vempla/
‘cudgel.GEN’
s
a
w
a:s:
sw
tat t (σ́ σ́)
aastat /`āstat̄ /
‘year:PART’
Figure 5. Some Estonian structures in Prince’s interpretation
222 Natalia Kuznetsova
Subsequent researchers tried to adhere to the maximal bimoraicity of
the syllable even more strictly, which provoked progressive loosening of
other principles of the moraic theory.
Bye (1997: 51, Fig. 6a) proposed representing Q3 syllables as formally
disyllabic, with either a “degenerative” syllable or a freestanding mora.
Eek and Meister (2004: 255–256), who used degenerative feet, noticed
that their prosodic structure would be completely different from regular
isochronic disyllabic feet. Bye accepted a ternary foot (kau-u-ge)-le in
kaugele ‘far:ALL’ to avoid a sequence of two unparsed syllables. His
framework would still imply either syllables unparsed in feet (such as
-le in Fig. 6a) and trimoraic syllables in structures like (2e–f) and (3d–e)
in Table 2 (Ehala 2003: 56). Even if an idea to distinguish between
different prosodic levels in the phonology of quantity is fruitful in itself,
the existing representational framework did not allow Prince and Bye to
depict it properly.
Hint (1978, 2001) and Ehala (2003), in turn, proposed to autosegmentalise all quantity by disconnecting it from segments and transferring it
to the level of syllables. Their syllable weight distinction corresponds to
the prosodic contrast of light and heavy accents in the present account.
Stressed syllables with Q1 and Q2 are ascribed one mora and those with
Q3 two morae, respectively. Hint (2001: 257–258) thoroughly discussed
the theoretical basis of his essentially structuralist account (see Section
4.2), but did not try to express it with a formal apparatus of the moraic
theory. Such an attempt was made by Ehala and turned out an impossible
endeavour. Representations of Q2 and Q3 as in Fig. 6b (Ehala 2003:
58), break several important conventions of this theory, e.g. bimoraicity
of long vowels and a ban on sequences of identical phonemes (Prillop
2013: 7–8). One could also wonder if a mora would be shared between
four segments in structures like poorti (‘border of a fabric/rug.PART’ (3e)).
Ehala (2003: 77) explicitly pointed out that the syllable weight, autosegmental in nature, is treated by the moraic theory “as totally dependent
on segmental structure”. His representations, which also depict morae
below syllables, are even more controversial than the standard moraic
formalism. He is not able to reflect the autosegmental nature of morae,
but is obliged by the existing representational convention to depict the
link between morae and segments, actually negated by him. Besides,
this convention does not allow a distinction between prosodically active
(stressed) and inert (unstressed) syllables, which is a very important
contrast in Hint and Ehala’s framework. Ehala tried other representational
alternatives, such as the dual distinction of weight by Hayes (see Section
2.2), but the latter, as said, also does not truly reflect mora autosegmentalisation.
Morae in Estonian 223
a.
b.
Q2
F
σ
σ
σ
σ
Q3
F
σ
σ
σ
σ
μ
μ
μ μ
μ
k a u u k e l e
s a a t a (t)
s a a : t u (s)
kaugele /`kaukele/
‘far:ALL’
saadad /´sātat/
‘send.2SG’
saadus /`sātus/
‘product’
Figure 6. Some Estonian structures in (a) Bye’s and (b) Ehala’s
interpretation
Odden (1997) gave an extremely cautious assessment of the moraic
theory validity for Estonian. He analysed more than a dozen of its problematic points raised by the Estonian case, making an inquiry into the
very nature of the mora and the foot. Without any resolute answers, he
still concluded that a foot-based account of Estonian quantity “accords
Q3 no direct status in the phonology” (Odden 1997: 190). If phonology
is represented as in current moraic theory, there seem to be indeed no
adequate formal tools to account for prosodic overlength.
3.4. The latest moraic accounts of Estonian by Prillop
Recent advances in generative Estonian phonology are represented by
Prillop (2011, 2013, 2015, 2018a, 2018b) and a Government phonology
account by Pöchtrager (2006, 2015). The latter uses Estonian to ground a
radical proposal to give up segmental phonology altogether, not addressed
here, as the mora notion is not exploited in this work.
Research by Prillop, in turn, develops the moraic theory further. This
account of Estonian is even more formally sophisticated than any of the
previous ones. The latest works (Prillop 2018a, b) also reveal important
changes in her conception. For example, the OT framework and the
distinction of strong and weak morae, heavily exploited earlier, were
abandoned.
Prillop (2011: 13, 2018a: 345–348) sees the mora as a smaller
building block of the syllable. However, the description of the mora’s
function “to distinguish between light and heavy syllables”, which is
224 Natalia Kuznetsova
important for the stress rules (Prillop 2018a: 349), does not match the
definition perfectly. It emphasises syllable’s prosodic behaviour as a
whole in regards to higher level prosodic units, not its composition of
smaller units.
Morae as units require exact correspondences to individual segments.
Phonological quantity linearised in a single moraic tier recalls the old
segmental phonological conception of Estonian quantity, extensively
criticised (cf. the latest remarks in Hint 2015a, b, Pajusalu 2015).
Prillop (2015: 185) explicitly admits that “in a certain sense we deal
with segmental quantity degrees: μ-, μμ- and μμ+-vowels, as well as
moraless, μ- and μ+-consonants are contrasted”. The whole four quantity degrees of vowels are identified here.
Prillop (2013: 3, 2015) sees the “consistency with regard to the
phonetic facts” as one of the main advantages of her description over
the previous ones. She agrees, however, that morae cannot correspond
to exact milliseconds (Prillop 2011: 181–182, 2018a: 348). As discussed
in Section 2.3, there is no general theory on the acceptable level of
inconsistency between the two. Prillop, too, does not formulate any
clear restrictions on this. Moreover, her mora has a variable duration
(Prillop 2018a: 349). If no clear duration of the mora and no rules of
the correspondence between duration and phonological length are set
in the theory, such phonetic argumentation is not falsifiable and can be
considered only as secondary in a discussion on phonological length.
Several sensitive points of her own account with regards to some
relatively prominent phonetic facts can be noted regardless. Prillop
(2011, 2013, 2015) introduced so-called prominent (strong) morae (see
below) and suggested that their phonetic correlate is a longer and/or
more intensive pronunciation (Prillop 2015: 183). This brings us back
to the peaks of energy as the mora’s correlate suggested by Lehiste (see
Section 3.2). However, no actual results from phonetic measurements
supporting this view are cited. One could wonder how a strong mora
in an unstressed syllable of Q2 (see Fig. 7a, c) and Q1 (2a ude ‘fuzz’,
Table 2) should be realised. One might expect to find an intensity peak
in the second syllable of such words. Existing research on Estonian
intensity (Liiv 1985, Eek 1986, Eek and Meister 1997, Asu and Lippus
2018) has attested to no such peaks in a second syllable of Q1/Q2.
Prillop (2015: 176) also claims that pitch in Estonian always falls
after the second mora. This would imply that in Q1 (cf. ude in 2a, Table
2) the pitch would fall inside the second syllable vowel. Ample experimental research since the 1930s has shown that pitch in Q1 falls more
or less at the same place as in Q2, i.e. at the border between the first and
Morae in Estonian 225
the second syllable (see Lippus et al. 2009, 2013 for the latest results).
Moreover, a simulated shift of the peak from the first vowel to the
second in Q1 changed the native perception of primary stress patterns:
´kanata [ˈg̊ ana(ˑ)tˑa] ‘hen.ABE’ was perceived as ka `natta [g̊ a ˈnatːˑa]
‘also fishpot.PART’ (Eek and Meister 2003: 910; transcription is mine,
N.K.). Prillop (2018a: 356) already placed the pitch fall after the first
mora, the same way as Plüschke (2013: 33) did.
The phonetic reduction of V2 in Q3 disyllables (cf. `uude ‘new.ILL’
and other types in 3a–e, Table 2) was used as an argument supporting
Prince’s idea that a Q3 syllable exhausts the foot (Prillop 2015: 182).
This V2 is claimed to be reduced exactly because it does not belong to
the foot. One could wonder if the same claim should then hold for an
unstressed V3 in trisyllables. Prillop (2018a: 351) sees the foot as maximally disyllabic. Phonetic results by Lehiste (1997b: 150–151) showed
that in trisyllabic Q3 words V2 is also reduced, but not V3. The length
of V3 is generally comparable to the length of a full short V1, irrespective of the quantity degree of the foot. Exactly for this reason, Eek and
Meister (1997: 95) phonologically treat V3 as not belonging to their
“minimal foot”, while V2 is included. If both V2 and V3 do not belong
to the foot, as Prillop’s conception implies, vowel reduction cannot be
considered as a sign of syllable’s extrametricality.
An aspiration for defining the relatively fine phonetic details of each
sound in a phonological representation already makes the task of Estonian quantity formalisation challenging enough. The restrictions of the
moraic theory which Prillop chose to follow made it even more complicated. She adopts Prince’s old idea that a Q3 syllable always makes
a foot on its own (see Section 3.3) and adheres to the following main
principles (Prillop 2018a: 351):
(1) the syllable is maximally bimoraic
(2) the foot is minimally bimoraic and maximally disyllabic;
(3) short vowels are monomoraic, long vowels bimoraic;
(4) short consonants are without mora, long consonants monomoraic;
(5) dependency lines in the prosodic tree do not cross.
In order to maintain these five principles, two admissions had to be
made:
(6) morae can be shared/split between several sounds;
(7) weak layering (projecting elements and even their parts directly
to higher levels of the hierarchy, bypassing the lower levels) is
allowed.
226 Natalia Kuznetsova
Weak layering, which helps to avoid the recursive foot (see Section
3.3) actually also violates the classic Prosodic Hierarchy, because the
latter prohibits skipping levels. Currently there are no strict criteria in
the theory about the acceptability of different variants of weak layering.
This allows various ad hoc solutions: extrasyllabic consonants, unparsed
syllables, and morae which are “unaffiliated with syllables and linked
to higher-level prosodic constituents. It is, however, unclear which
constituents can remain without a link to next level constituents and
how many levels can be skipped in this way” (Prillop 2013: 11). For
example, Kager and Martínez-Paricio (2018: 182) seem to allow even
whole syllables unaffiliated with any higher level constituents. Prillop
(2018a: 351) did not allow bypassing of more than two levels, and later
added an additional restriction that levels can be skipped only at units’
boundaries. However, the reasons for these particular choices and their
subsequent changes were not explicitly addressed.
Strong and weak (prominent and non-prominent) morae also require
a comment. This contrast was taken over from Kager (1992). The
latter used Hayes’ term “moraic trochee” to describe quantity sensitive
stress systems where the stress falls on either all heavy syllables or on
alternating light syllables. Kager, however, treats Estonian as having
syllabic rather than moraic trochee (ibid.: 304), and his theory does
not allow strong morae to occur outside the stressed syllables anyway.
Prillop, in turn, claimed that Estonian combines two different types of
rhythm: a trochaic one on the syllabic level and an iambic one on the
moraic level. Moreover, the two do not necessarily coincide, as strong
morae can occur in unstressed syllables (cf. Fig. 7a, c). This account is
incompatible with the metrical grid theory, as it generates uncontinuous
columns (Prillop 2013: 18), and actually constitutes a new amendment
to a general moraic theory.
The introduction of strong and weak morae also means that one more
informational layer is included in the moraic tier. In addition to quantity,
now also a part of the data on prominence (which was previously calculated from the metrical grid nodes) is linearised in morae. Prillop (2013:
18, 21), therefore, had to claim that mora strength is a lexical property,
even if it cannot create lexical length contrasts. No wonder that an
extremely sophisticated machinery was needed when so much information had to be included in morae. Notably, strong and weak morae do
not appear anymore in Prillop (2018a, b), for unclarified reasons.
As an example, let us examine how Q2-Q3 disyllabic feet with
the most complex first syllable structures (VVCC and VVCCC)
Morae in Estonian 227
are analysed. The most optimal (in an OT sense) representations for
´uute – `uute (types 2e – 3d in Table 2, Fig. 7a–b) follow Prillop (2013:
23, 2015: 184), while in Prillop (2018a: 355) the mora strength is no
longer depicted. Earlier variants of ´kaarte – `kaarte (types 2f – 3e in
Table 2) structural types are represented in Fig. 7c–d (Prillop 2015: 24).
In Prillop (2018a: 360), a prosodic tree for the `VVCCC type underwent
some significant modifications, which are represented in Fig. 7e (the
same type as in Fig. 7c).
a.
b.
ω
ω
F+
c.
d.
ω
F+
F
σ
σ+
σ
σ
σ
μ μ
μ+
μ μ+
μ
μ μ
μ+
t
e
u
t
e
F
F
σ+
u
e.
k a r t e
σ
σ
μ μ μ+ μ
k a r t
e s
σ
σ
μ μ
μ
uː
r tː e
uute /´ūt̄ e/
uute /`ūt̄ e/
kaarte /´kārte/ kaarte /`kārte/ suurte /´sūrte/
‘new:GEN.PL’ ‘innovation.GEN’ ‘arc:GEN.PL’ ‘map.PART.PL’
‘big:GEN.PL’
Figure 7. Complex Estonian structures in Prillop’s earlier (a–d)
and later (e) interpretation
Maximal bimoraicity of the syllable and the long vowels are
preserved, but at a high cost. Several problematic points for a classical
moraic theory and even Prillop’s own basic postulates are listed below.
(1) As said, the recursive foot is avoided due to the acceptance of
weak layering, which accepts direct projections both from segments
to syllables and from morae to feet (cf. Fig. 7c–d). Projections of one
segment’s parts to two different levels (moraic and syllabic, see geminate t in Fig. 7a–d) are also permitted, while no restrictions on this are
formulated.
(2) The mora can be shared across syllabic and even foot boundaries
(Fig. 7b). In the earlier version, it could also split between three segments
(Fig. 7c). In the later version, this issue is resolved by projecting both
parts of the geminate directly to a syllabic level (Fig. 7e). However,
such a representation violates one of the basic principles outlined
above: a long (fortis) consonant should have one mora (in Fig. 7e, it has
none). Notably in a monosyllabic form `suurt ‘big:PART’ the final fortis
228 Natalia Kuznetsova
consonant has a mora (Prillop 2018a: 360). Unfortunately, clear criteria
to distinguish between the cases of fortis consonants with and without
a mora were missing.
(3) Not all syllables are parsed into feet and not even all morae are
parsed into syllables (cf. with Bye’s freestanding morae in Section 3.3).
(4) While long vowels are always treated as bimoraic, first syllable
diphthongs are treated as either monomoraic (sharing a mora) or bimoraic, depending on the quantity degree (Prillop 2015: 177). However,
initial syllable diphthongs and long vowels manifest the same prosodic
behavior, so it remains unclear why the two should be represented
differently. Besides, Estonian diphthongs are combinations of two
phonemes rather than single phonemes (Viitso 2003: 2). Mora sharing
in a Q2 diphthong therefore violates one of Prillop’s main postulates
that a short vowel should always have one mora.
(5) As discussed before, an amendment on the combination of
syllabic and moraic trochees in the same language, especially when
strong morae can occur outside of strong (stressed) syllables, would be
very difficult to incorporate into a standard moraic theory.
3.5. Validity of Prillop’s analysis for Soikkola Ingrian
Prillop (2015) applied her moraic analysis also to the Soikkola
dialect of Ingrian. She uses this Finnic variety with a ternary quantity
contrast of consonants, similar to Estonian, as support for the existence
of strong and weak morae. A brief comment on her analysis will be
made.
One of its sensitive points is an incomplete dataset: proposed generalisations do not embrace the entire language system. For example, a
rule that stops and s were geminated if the next syllable contained a
diphthong or long vowel (Prillop 2015: 189) is not a general one. In
trisyllabic feet, those consonants could be geminated also before short
vowels, which created four-way contrastive patterns (4a–d in Table 3,
Kuznecova 2015: 207–208).
Prillop (2015: 190) claimed that Soikkola Ingrian “tends to preserve
original durational contrasts of words, syllables and segments”.
However, strong reduction characteristic of at least the modern state
of the dialect leads to typological changes in the durational contrast
of non-initial vowels. An original short vs. long vowel contrast turns
into a reduced vs. short vowel contrast (Kuznecova 2009, Markus 2011,
Kuznetsova 2016).
Morae in Estonian 229
According to our recent phonetic research (Kuznetsova and Brodskaya, in prep.), the original contrast of non-initial long and short vowels
is nowadays lost from trisyllabic feet. Phonetically long second syllable
vowels occur only in structures with the shortest foot nuclei (types 4e–f
in Table 3), while lost from longer structural types (e.g. 4b, d in Table 3).
Therefore, the statement that “original non-initial long vowels are linked
to strong morae and short vowels to weak morae” (Prillop 2015: 190)
does not hold in both of its parts. On the contrary, the only vowels which
are nowadays phonetically long in the second syllable of trisyllables are
exactly those which were short in Proto-Finnic (types 4e–f in Table 3),
while all original long vowels in this position shortened.
Table 3. Changes in trisyllabic feet of Soikkola Ingrian
ProtoFinnic
(C+V2)
(4) a. *CV
b. *CV̄
Soikkola Ingrian
older system
Gloss
newer
system
> [ˈvuːtˑava]
= [ˈvuːtˑava]
‘leak.PTCP.PRS.ACT’
> [ˈsuːtˑiːma]
> [ˈsuːtˑima]
‘judge.1PL’
c. *C̄ V
d. *C̄ V̄
> [ˈuːtːele]
= [ˈuːtːele]
‘wait.IMP’
> [ˈmuːtːiːma]
> [ˈmuːtːima]
‘change_oneself.1PL’
e. *CV
> [ˈoDaːma]
= [ˈoDaːma]
‘take.1PL’
f. *CV
> [ˈmatˑaˑła]
= [ˈmatˑaˑła]
‘low’
4. Estonian syllable weight in the structural functional perspective
4.1. Morae in the structure of Estonian quantity
Functional structuralism treats the language as a set of hierarchically organised and mutually associated layers of information, each of
which has its own function in enabling effective communication. In
fact, there are several hierarchies which are associated to each other (cf.
“double articulation” in Martinet 1949/1965, also Fox 2000: 333–344).
Length, among other linguistic phenomena, is also “a property of units
of different levels” (Fox 2000: 110). Physical duration of sounds in
concrete utterances is influenced by multiple phonological and phonetic
factors, each of which belongs to a concrete level of a certain hierarchy.
Additionally, various pragmatic and even metalinguistic components,
230 Natalia Kuznetsova
such as age, gender, social status, mood, health of a speaker, can affect
these concrete durations.
Let us analyse structural composition of Estonian quantity up to the
level of the foot. Postlexical phenomena are outside the scope of this
paper and involving the prosodic word level would require an additional
discussion on a non-trivial question of primary and secondary stress in
Estonian (viz. Kuznetsova 2018). Our main questions will be the structure of Estonian foot accents and the place of morae in it.
Kasevič et al. (1990: 20) typologically described any word-prosodic
unit through its association to three basic structural units: (1) a domain
(segmental unit associated with it); (2) a bearer (minimal segmental
structure which has to be present to make its realisation possible);
(3) a functional basis (morphological unit associated with it). Estonian
accents have the foot as their domain, the foot nucleus as a bearer, and
the morpheme as a functional basis. Two relevant prosodic levels are
distinguished in Table 4: the syllable and the foot. They both are associated with the morphemic level in a parallel grammatical hierarchy.
Besides, both hierarchies are associated with the segmental level,
which is subprosodic and subgrammatical. Quantitative phenomena
are divided into two types: phonological and phonetic. The latter are
outlined tentatively and can include, for example, intrinsic duration of
segments of different quality on the segmental level. On the prosodic
levels, phonetic information on quantity would define, for example, the
exact distribution of durational peaks and reduction among segments
and syllables in various types of feet.
Table 4. Structure of Estonian quantity up to the foot level
Main quantitative phenomena
(Levels of Level of Nr
grammatical prosodic
Phonology
Phonetics
units
units)
(a)
(b)
e.g. phonetic rules for a
Foot
3 light vs.
heavy accent distribution of lengthening
(=lexicalised and shortening between
foot stress)
syllables in a foot (incl. foot
isochrony tendency)
(Morpheme)
Syllable 2 long vs. short e.g. phonetic rules for a
syllables
distribution of lengthening
(mono- vs.
and shortening between
bimoraic)
segments in a syllable
Segment
1 long vs. short e.g. intrinsic duration of
phonemes
segments
Morae in Estonian 231
Morae in the generative accounts of Estonian discussed above can
represent in total five different quantity components simultaneously:
1–3(a) and 2–3(b), marked bold in Table 4. In order to represent all the
desired contrasts without violating some basic theoretical principles,
progressively more sophisticated analytical mechanisms have been
invented. Yet, no account has yet managed to preserve internal theoretical consistency entirely.
In a functional structural framework, morae are reserved just for
a component 2a, which is put in a bold frame in Table 4. Morae are
considered not as syllable constituents, but as an analytical measure of
the syllable’s prosodic properties (cf. Section 2.1). The main prosodically relevant syllabic contrast in Estonian, as outlined in Section 3.1,
is of short and long syllables (or mono- vs. bimoraic, or light vs. heavy
syllables). It serves as the basis for the accent distinction at а higher
foot level. Feet with a stressed long syllable can carry two types of
accent, while feet with a stressed short syllable carry only one. Accents
are distinctive word-prosodic units, such as tones, and should rather be
represented autosegmentally.
Estonian morae characterise the weight just of syllables, not
of feet, and therefore do not account for overlength. The Estonian
syllable weight contrast is, in a sense, auxiliary and can be automatically detected from the segmental composition of syllables. Only the
contrasts of segmental quantity and of accents need to be depicted in the
phonological transcription.
4.2. Structural moraic analysis by Hint
The moraic account of Estonian by Hint (1978, 2001: 257–258)
is also essentially structuralist, and naturally arrived at a complete
autosegmentalisation of morae from segments. A brief comment on its
differences from the present conception will be given.
Long vowels and consonants are considered as iterations of two short
segments. The question on mono- vs. biphonemic character of Estonian
long sounds has been extensively discussed in literature (see e.g. Viitso
2008: 184–189), and no final consensus has been yet reached. Long
sounds do not contain any internal phonetic boundaries, and for a number
of functional reasons they are treated here as monophonemic (as in Erelt
et al. 1995: 102–103).
Hint makes a distinction between phonetic and phonological rules
for quantity, similar to the one proposed in Section 4.1. He states that
232 Natalia Kuznetsova
a phonetic description of the syllable structure should determine the
way Q3 is realised inside the syllable, while its phonological description
should determine the quantitative peak placement. However, as explicated in Table 4, an exact description of peak placement in a syllable
and foot also belongs to the phonetic description. The place of peak can
be automatically detected if one knows the syllable and foot structures
and the foot accent type. Therefore, it is enough for a phonological transcription just to mark the syllable serving as a morphological anchor of
accents.
Most importantly, overlength is regarded as a syllabic feature by
Hint. That is why he used morae to describe the contrast between the
presence and absence of overlength in stressed syllables: “the weight
of one mora is obtained from the segmental composition of the syllable,
and of the second mora from the prosodic syllabic extra-quantity (heavy
accent, heavy stress, extra stress)” (Hint 2001: 257–258). He saw stress
and overlength as two separate prosodic phonemes added to segmental
syllabic structures. Syllabic extra-quantity can occur only in long stressed
syllables, while Q1 and Q2 syllables have stress only as a prosodic
marker. Therefore, the latter, as well as all unstressed syllables are monomoraic.
Hint’s wording does not allow lexicalised and rhythmic stresses to
be distinguished in the case of Q1 and Q2. Only lexicalised stress (light
accent) will be a “prosodic phoneme” to be depicted in a phonological
description. Rhythmic stress, which has the same phonetic cues as the
light accent, is a phonetic phenomenon.
In the present paper, overlength is seen as a property of lexicalised
stress, i.e. the accent (like in Viitso 1979, 2008 and Eek 1986), rather
than a separate feature. The accent has the foot as its domain and the
disyllabic foot nucleus as its bearer (therefore, no accentual contrast in
monosyllabic feet is present), even if it is morphologically anchored in
a certain syllable of a morpheme. The morae, as syllabic features, do
not, therefore, constitute an appropriate tool for formalising overlength.
4.3. Functional evidence for the prosodic contrast of short
and long syllables
Let us now examine further phonological and phonetic evidence
that only the contrast of long and short syllables is indeed functionally
relevant at the syllabic level.
Phonologists have long discussed whether Estonian is a mora- or
a syllable-counting language. Already Ojamaa (1976: 45) noticed that
Morae in Estonian 233
Estonian could be considered mora-counting on the basis of a Q1 vs.
Q2 distinction, and syllable-counting on the basis of a Q3 vs. non-Q3
distinction. Hint (1978) analysed the dynamics of changes in Estonian
nominal declension and concluded that Estonian is turning from a moracounting into a syllable-counting language. “The distinction between
inflectional classes that was formerly based on the quantity of the first
syllable is on the way out” (Lehiste 1983: 26).
Viitso (1982), however, re-analysed Hint’s data and offered a simpler
and more general algorithm which calculates the rules for the choice
of affixes both at an earlier and at a later stage of Estonian morphonological development. This algorithm is based on the contrasts of short
vs. long stressed syllables and of two accents in a long syllable. The
morphonological change is explained by the loss of a light accent in
certain syllables. Tendencies for the secondary stress loss from noninitial syllables and the primary stress shift to initial syllables are
general widespread processes in Estonian (Viitso 1979, 1982, 2003,
Pajusalu 2009). They signify a drift towards a replacement of marked
stress patterns by unmarked ones. Viitso concluded that mora-counting
is not needed to explain the processes described by Hint and that Estonian was stress-counting already at the initial stage of those processes.
Hyman (1985: 93), Kager (1992), and Alber (1997) also assumed that
overlength is not counted in the rules of stress placement.
Studies on runic poetry (Ross and Lehiste 2001: 57–58) showed
that Estonian folklore metre indeed does not distinguish between long
and overlong syllables. A contrast of long and short lexically stressed
syllables is, in turn, of utmost importance. Long stressed syllables have
to fall in an ictus position, while short stressed syllables have to avoid
it (in the latter case, the second syllable half-long vowel gets the ictus).
In modern poetry and modern readings of traditional poetry, this system
is being replaced with a trivial correspondence between ictuses and
stressed syllables, so the syllable weight sensitivity is being lost (Ross
and Lehiste 2001, Lehiste 1990, 1997a). Still, it is exactly the contrast
of short and long stressed syllables that is indicated in the old system as
being prosodically relevant.
A prosodic phonetic tendency to foot isochrony provides further
proof of this. Wiik (1991) formulated an elegant rule: a short vowel of
the second syllable is phonetically lengthened after a monomoraic (short)
syllable and shortened after a bimoraic (long) one (see 3.2). In the history
of Estonian and many other Finnic varieties, both reduction and apocope
after a long syllable and phonetic lengthening after a short syllable are
extremely typical processes attested to various degrees (viz. Kuznetsova
234 Natalia Kuznetsova
2016). For example, in Soikkola Ingrian, this lengthened short vowel
after a monomoraic syllable is phonetically much longer than the original
long vowel after a bimoraic syllable (see the research cited in Section
3.5).
4.4. Possible formalisations of the functional-structural
account of Estonian quantity
If a generative moraic theory cannot provide adequate tools for the
representation of Estonian quantity within a functional-structural framework, other possibilities could be explored. For example, we can try to
model a morphonological process of a synthesis of Estonian accents
(cf. ´suurte in Fig. 8). Both phonological and morphonological information is needed to calculate the accent, so two parallel processes will
be considered. In phonology, first, the phonemic length of segmental
phonemes has to be detected (this corresponds to the CV-tier in early
versions of moraic theory). On the next step, a quantitative structure of
syllables is determined. Next, the syllabic structure tells us the syllable
weight, which will play a role in higher-level prosodic processes. Finally,
on the foot level, we learn about the structure and the boundaries of the
whole foot and the place of the foot stress.
In morphophonology, first, the segmental structure of each allomorph
is established. Then we obtain information about the type of allomorph.
Main types of morphemes in Estonian are roots, derivative suffixes and
segmental and suprasegmental inflectional morphemes (cf. Viks 1992,
Erelt et al. 1995). Roots can have at maximum three prosodic types of
allomorphs. First, it is a monosyllabic stem (Rmon) with a heavy accent.
Second, it is a multisyllabic “vocalic” stem which carries light accent by
default for most words (R, or its abridged variant R-), apart from the class
of so-called contracted roots. The latter have the default vocalic stem
with a heavy accent (Rs) and also an additional lightly-accented stem
(Rw). The default stem for most nouns can be obtained from a genitive
singular form, Rw from a partitive singular form. Next, we establish the
prosodic properties of morphemes in regards to the accent. Following
the principle outlined for Proto-Slavonic by Zalizn’ak (1985), Estonian
morphemes could be classified into those attracting and non-attracting
accent (m+ and m in Fig. 8). Accent-attracting morphemes (e.g. all roots,
apart for certain unaccented personal pronouns) require an accent either
on their first syllable or on the last syllable of a preceding morpheme
(like all suprasegmental morphemes and a derivative suffix -nna for
female actors). The phonological domain of the accent is the foot, and
Morae in Estonian 235
its morphological domain is an accent-attracting morpheme, optionally
followed by unaccented morphemes. Finally, the type of accent is defined
for each morphological domain.
In the end, phonological and morphological information is synthesised to obtain accents in an actual grammatical word. The word suurte
contains one disyllabic foot with the first heavy syllable, on the one hand,
and a sequence of an accent-attracting and unaccented morphemes, which
would require a default accent for a vocalic stem, i.e. light.
lexicalised (morphologised) foot accent
/´sūrtē/ ‘big:GEN.PL’
morphonological
phenomena
phonological
phenomena
morphologically
required accent
LI
F
foot stress
accent attraction by
morphemes
m+
m
μ syllable weight
σ
σ
morphonological
types of morphemes
R-
morphemic structure
CVVC CCV CVVCC CV syllable structure
s ū r t̄ e s ū r t̄ e phonemic string
phonemic string
pl.gen
μμ
syllables
Figure 8. Example of calculating a foot accent in ´suurte
Further examples illustrate some other important phenomena in the
accent synthesis. A structural difference between the light accent and
the rhythmic stress is illustrated in two multifoot words given in Fig. 9:
a compound with two accent-attracting morphemes (Fig. 9a) and a
simplex word with one accent-attracting morpheme (Fig. 9b). The latter
will have two foot stresses, but only one lexicalised accent.
a.
b.
/´kana´muna/ ‘egg’ (‘hen+egg’)
/´kavalama/ ‘cunning:CPR.GEN’
LI
LI
F
F
LI
F
F
m+
m+
μ
μ
μ
μ
m+
m.m
μ
μ
μ
μ
R
R
σ
σ σ
σ
R
cmp.gen
σ
σ σ
σ
CVCV CVCV
CV CV CV CV CVCVCV CV
CV CV CV CV
kana muna ka na muna kavala ma
ka va lama
Figure 9. Calculating foot accents in a complex vs. simplex multifoot word
236 Natalia Kuznetsova
Fig. 10a shows a suprasegmental morpheme of partitive, which
changes the accent of the whole foot into heavy. An example of a word
containing the contracted stem (Rs) and also a derivative suffix ´-line
with prosodic properties similar to nouns (such morphemes are marked
as ADJ, ADJ-, ADJmon) is given in Fig. 10b.
a.
b.
/`sūri/ ‘big.PART.PL’
/`pȫrde´line/ ‘turning’ (‘turn:ADJ’)
LI
.HE
F
m+
m.m+
μμ
R-
pl.part
σ
HE
LI
F
F
μ
m+
m+
μμ
μ
μ
μ
σ
Rs
ADJ
σ
σ
σ
σ
CVVC V
CVV CV
CVVCCV CVCV
CVVC CV CV CV
s ū r i
s ū r i
p ö: r t e l i n e
p ö: r t e l i n e
Figure 10. Calculating foot accents in `suuri and `pöörde´line
Some prosodic differences between adjective suffixes which are
prosodically close to nouns (ADJ) and proper nominal stems (R) could
be observed in pairs like those in Fig. 11 (cf. Viitso 1979: 144, 2008:
182). A trisyllabic word can be composed of two feet, but this is a
marked stress model and there is a tendency to lose the stress of a final
monosyllabic foot. This, however, does not seem to happen if such a foot
contains a root morpheme (10a), while it is more common in adjective
derivative suffixes (10b). As a result, derivatives like kasulik ‘useful’ in
Fig. 11b face a prosodic conflict between phonology and morphonology.
On the one hand, the foot structure can manifest a lack of the second
foot stress. On the other hand, the system of prosodic morphonological
alternations imposes heavy accent on the last syllable. Eventually, to
resolve this conflict such words start changing their declension type,
as described by Hint (1978) and Viitso (1982). A functional-structural
description of accents, which takes into account both phonological and
morphonological information, can predict such changes in the system,
as it clearly traces this prosodic conflict.
Morae in Estonian 237
a.
̄
/´tuli`tik/ ‘match’ (‘fire+stick’)
LI
HE
F
F
m+
m+
μ
μ
R
Rmon
σ
σ σ
CVCV CVCC
t u l i t i k̄
μμ
CV CV CVCC
t u l i t i k̄
b.
̄
/´kasu(`)lik/ ‘useful’ (‘use:ADJ’)
LI
HE
F
(F)
m+
m+
μ
μ
R
ADJmon
σ
σ σ
CVCV CVCC
k a s u l i k̄
μμ
CV CV CVCC
k a s u l i k̄
Figure 11. Calculating foot accents in compounds vs. derivatives
5. Conclusion
To sum up, the mora notion might be of use in a structural functional
description of Estonian phonology to formalise the prosodic distinction of short (light) and long (heavy) syllables, which can be used for
example in larger cross-linguistic comparisons on syllable weight. The
relevance of this prosodic distinction of syllables can be traced in actual
functioning of Estonian prosody. Other quantity contrasts, relevant for
Estonian phonology, belong either to the segmental, or to the foot level.
The mora is understood in structural phonology as an abstract property rather than a constituent of the syllable. Therefore, the formal
apparatus of current generative moraic theory, which places morae
inside a dependency grid, is not suitable for a structuralist description.
At present, moraic theory has not yet found a way to describe such a
complex multilevel quantity system as Estonian without serious internal
contradictions or ad hoc theoretical solutions. As argued, this is a consequence of linearising a hierarchically organised quantity system into a
single prosodic tier of morae, with just a binary contrast of mono- vs.
bimoraic syllables.
Autosegmental representation of morae, showing associations
between hierarchies and their levels rather than dependencies, reflects
a structural functional conception of Estonian quantity much better.
Possible use of morae in structural formalism was demonstrated in a
morphonological algorithm calculating Estonian accent placement.
238 Natalia Kuznetsova
Acknowledgements
I am grateful to Pavel Iosad, Cormac Anderson, Larry Hyman, Brittany Blankinship, an anonymous reviewer, and the editors of the present
volume for their valuable comments.
Address
Natalia Kuznetsova
Institute for Linguistic Studies
Russian Academy of Sciences
Tuchkov per. 9
199053, St. Petersburg, Russia
E-mail: nkuzn@yandex.ru
References
Alber, Birgit (1997) “Quantity sensitivity as the result of constraint interaction”. In Geert
Booij, Jeroen van de Weijer, eds. Phonology in progress – progress in phonology,
1–45. (HIL Phonology Papers, 3.) The Hague: Holland Academic Press.
Allen, W. Sidney (1965) Vox Latina. Cambridge: Cambridge University Press.
Asu, Eva Liina and Pärtel Lippus (2018) “Acoustic correlates of secondary stress in Estonian”. In Katarzyna Klessa, Jolanta Bachan, Agnieszka Wagner, Maciej Karpiński,
and Daniel Śledziński, eds. 9th International Conference on Speech Prosody 2018,
13–16 June 2018, Poznań, Poland, 602–606. Poznań: Adam Mickiewicz University.
Asu, Eva Liina, Pärtel Lippus, Karl Pajusalu, and Pire Teras (2016) Eesti keele hääldus.
(Eesti keele varamu, 2.) Tartu: Tartu Ülikooli Kirjastus.
Broselow, Ellen, Su-I Chen, and Marie Huffman (1997) “Syllable weight: convergence
of phonology and phonetics”. Phonology 14, 47–82.
Bye, Patrik (1997) “A generative perspective on ‘overlength’ in Estonian and Saami”. In
Estonian prosody, 36–68. Tallinn.
Eek, Arvo (1986) “Problems of Estonian word prosody”. Estonian Papers in Phonetics
1984–1985, 14–66.
Eek, Arvo (1990) “Units of temporal organisation and word accents in Estonian”. Linguistica Uralica 26, 4, 251–263.
Eek, Arvo and Einar Meister (1997) “Simple perception experiments on Estonian word
prosody: foot structure vs. segmental quantity”. In Estonian prosody, 71–99. Tallinn.
Eek, Arvo, and Einar Meister (2003) “Foneetilisi katseid ja arutlusi kvantiteedi alalt (I):
Häälikukestusi muutvad kontekstid ja välde”. Keel ja Kirjandus 46, 11, 815–837;
12, 904–918.
Morae in Estonian 239
Eek, Arvo and Einar Meister (2004) “Foneetilisi katseid ja arutlusi kvantiteedi alalt (II).
Takt, silp ja välde”. Keel ja Kirjandus 47, 4, 251–271; 5, 337–357.
Ehala, Martin (1999) “Eesti väldete probleemi üks lahendusi”. Keel ja Kirjandus 6,
378–385; 7, 453–465.
Ehala, Martin (2003) “Estonian quantity: implications for Moraic Theory”. In Diane
Nelson and Satu Manninen, eds. Generative approaches to Finnic and Saami linguistics, 51–80. Stanford: CSLI Publications.
Erelt, Mati, Reet Kasik, Helle Metslang, Henno Rajandi, Kristiina Ross, Henn Saari, Kaja
Tael, and Silvi Vare (1995) Eesti keele grammatika I. Morfoloogia. Sõnamoodustus.
Tallinn: Eesti Keele Instituut.
Erwin, Sean (1996) “Quantity and moras: amicable separation”. In Matthew Pearson and
Ileana Paul, eds. The structure of Malagasy. Vol. 1, 2–30. (UCLA Occasional Papers
in Linguistics, 17.) Los Angeles: University of California, Los Angeles.
Estonian prosody = Lehiste, Ilse, and Jaan Ross, eds. (1997) Estonian prosody: papers
from a symposium. Tallinn: Institute of Estonian Language.
Fox, Anthony (2000) Prosodic features and prosodic structure: The phonology of
‘suprasegmentals’. Oxford: Oxford University Press.
Goldsmith, John A. (1976) Autosegmental phonology. Unpublished Doctoral Thesis.
Swarthmore College.
Gordon, Matthew (2004) “Syllable weight”. In Bruce Hayes, Robert Kirchner and Donka
Steriade, eds. Phonetically based phonology, 277–312. Cambridge: Cambridge
University Press.
Hall, Kathleen C., Elizabeth Hume, T. Florian Jaeger, and Andrew Wedel (2018) )
“The role of predictability in shaping phonological patterns”. Linguistics Vanguard,
20170027.
Hayes, Bruce (1989) “Compensatory lengthening in moraic phonology”. Linguistic
Inquiry 20, 2, 253–306.
Hayes, Bruce (1995) Metrical stress theory: principles and case studies. Chicago and
London: The University of Chicago Press.
Hint, Mati (1973) Eesti keele sõnafonoloogia I. Rõhusüsteemifonoloogia ja morfofonoloogia põhiprobleemid. Tallinn: Eesti NSV Teaduste Akadeemia.
Hint, Mati (1978) “Changes in the prosodical system of contemporary Estonian”. Estonian Papers in Phonetics 1978, 39–43.
Hint, Mati (1980) “Minevikuline ja tulevikuline aines keelesüsteemis. Prosoodiatüübi
nihked ja selle tagajärjed”. Keel ja Kirjandus 23, 215–223, 271–278, 349–355.
Hint, Mati (1997) “The Estonian quantity degrees in prosody and morphophonology”. In
Estonian prosody, 125–135. Tallinn.
Hint, Mati (2001) “Prosoodiaväitluste läbimurdeta”. Keel ja Kirjandus 2, 164–172; 3,
252–258; 4, 324–339.
Hint, Mati (2015a) “Kolme häälikuvälte eksituste mehaanika”. Keel ja Kirjandus 3,
163–177.
Hint, Mati (2015b) “Eesti väldete uurimine ei ole veel ajaloo lõpus”. Keel ja Kirjandus
4, 268–269.
240 Natalia Kuznetsova
Hubbard, Kathleen (1994) Duration in moraic theory. Unpublished Doctoral Thesis.
University of California, Berkeley.
Hyman, Larry (1985) The theory of phonological weight. (Publications in Language
Sciences. 19.) Dordrecht, Holland and Cinnanminson, NJ: Foris.
Jakobson, Roman (1931/1962) “Die Betonung und ihre Rolle in der Wort- und Syntagmaphonologie”. In Roman Jakobson. Selected writings. Vol. 1, 117–136. The Hague:
Mouton.
Jakobson, Roman (1937/1962) “Über die Beschaffenheit der prosodischen Gegensätze”.
In Roman Jakobson, Selected Writings. Vol. 1, 254–261. The Hague: Mouton.
Jose Antonio Mompéan-González (2014) “Cognitive linguistics and phonology”. In John
Taylor and Jeannette Littemore, eds. The Bloomsbury companion to cognitive linguistics, 253–276. London and New York: Bloomsbury Academic.
Kager, René (1992) “Shapes of the generalized trochee”. In Jonathan Mead, ed. The
Proceedings of the Eleventh West Coast Conference on Formal Linguistics, 298–312.
Chicago: University of Chicago Press.
Kager, René (1995) “Ternary rhythm in alignment theory”. Utrecht University: Rutgers
Optimality archive. Available online at <http://roa.rutgers.edu/files/35-1094/35-1094KAGER-0-0.PDF>. Accessed on 30.01.2018.
Kager, René, and Violeta Martínez-Paricio (2018) “Mora and syllable accentuation:
typology and representation”. In Rob Goedemans, Jeff Heinz, and Harry van der
Hulst, eds. The study of word stress and accent: theories, methods and data, 147–186.
Cambridge: Cambridge University Press.
Kapatsinski, Vsevolod (2018) Changing minds changing tools: from learning theory to
language acquisition to language change. Cambridge: MIT Press.
Kasevič, Vadim B. (1983) Fonologičeskie problemy obščego i vostočnogo jazykoznanija.
Moskva: Nauka.
Kasevič, Vadim B., Evgenija M. Šabel’nikova, and Viktor V. Rybin (1990) Udarenie i ton
v jazyke i rečevoj dejatel’nosti. Leningrad: Izd-vo LGU.
Kuryłovič, Jerzy (1948) “Contribution à la théorie de la syllabe”. Bulletin de la Société
Polonaise de Linguistique 8, 80–114.
Kuznecova, Natalia V. (2009) “Suprasegmentnaja fonologija sojkinskogo dialekta
ižorskogo jazyka v tipologičeskom aspekte”. Voprosy jazykoznanija 5, 18–47.
Kuznecova, Natalia V. (2015) “Dve fonologičeskie redkosti v dialektah ižorskogo
jazyka”. Acta linguistica Petropolitana 11, 2, 201–257.
Kuznetsova, Natalia (2016) “Evolution of the non-initial vocalic length contrast across
Finnic varieties of Ingria and the adjacent areas”. Linguistica Uralica 52, 1, 1–25.
Kuznetsova, Natalia (2018) “What Danish and Estonian can tell to modern word-prosodic
typology”. In Rob Goedemans, Jeff Heinz, and Harry van der Hulst, eds. The study of
word stress and accent: theories, methods and data, 102–143. Cambridge: Cambridge
University Press.
Kuznetsova, Natalia and Irina Brodskaya (in prep.). “Temporary structure of the trisyllabic foot in Soikkola Ingrian”.
Morae in Estonian 241
Lehiste, Ilse (1960) “Segmental and syllabic quantity in Estonian”. In American studies
in Uralic linguistics, 21–82. (Indiana University Publications. Uralic and Altaic
Series 1.) Bloomington: Indiana University Press.
Lehiste, Ilse (1983) “Prosodic change in progress: evidence from Estonian”. In Irmengard Rauch and Gerald F. Carr, eds. Language change, 10–27. Bloomington: Indiana
University Press.
Lehiste, Ilse (1990) “Moraic analysis of Finnish and Estonian”. In Congressus Septimus
Internationalis Fenno-Ugristarum, Sessiones sectionum 3B (Section “Dissertationes:
Linguistica”), 286–291. Debrecen: Kossuth Lajos Tudományegyetem.
Lehiste, Ilse (1997a) “The phonetic realization of the haiku form in Estonian poetry,
compared to Japanese”. In Kiritani Shigeru, Hajime Hirose, and Hiroya Fujisaki, eds.
Speech Production and language: In Honor of Osamu Fujimura, 241–249. (Speech
Research, 13.) Berlin: Mouton de Gruyter.
Lehiste, Ilse (1997b) “The structure of trisyllabic words”. In Estonian prosody, 149–164.
Tallinn.
Liiv, Georg (1985) “Akustičeskije korrel’aty ėstonskogo slovesnogo udarenija v
sootnošenii s differencial’noj dolgotoj”. Sovetskoe finno-ugrovedenije 21, 1, 1–13.
Lippus, Pärtel, Eva Liina Asu, Pire Teras, and Tuuli Tuisk (2013) “Quantity-related
variation of duration, pitch and vowel quality in spontaneous Estonian”. Journal of
Phonetics 41,1, 17–28.
Lippus, Pärtel, Karl Pajusalu, and Jüri Allik (2009) “The tonal component of Estonian
quantity in native and non-native perception”. Journal of Phonetics 37, 4, 388–396.
Maddieson, Ian (1993) “Splitting the mora”. UCLA Working papers in Phonetics 93,
9–18.
Markus, Elena (2011) “The phonetics and phonology of a disyllabic foot in Soikkola
Ingrian”. Linguistica Uralica 47, 2, 103–119.
Martinet, André (1949/1965) “La double articulation linguistique”. In André Martinet. La
linguistique synchronique, 1–41. Paris: Presses Universitaires de France.
Newman, Paul (1972) “Syllable weight as a phonological variable: the nature and function of the contrast between “heavy” and “light” syllables”. Studies in African linguistics 3, 3, 301–323.
Odden, David (1997) “Some theoretical issues in Estonian prosody”. In Estonian Prosody,
165–195. Tallinn.
Ojamaa, Koit (1976) Temporal aspects of phonological quantity in Estonian. Unpublished
Doctoral Thesis. University of Connecticut.
Pajusalu, Karl (2009) “Dynamics of Estonian phonology”. STUF — Language Typology
and Universals, Sprachtypologie und Universalienforschung 62, 109–121.
Pajusalu, Karl (2015) “Eesti vältesüsteemi olemusest”. Keel ja Kirjandus 4, 262–267.
Pajusalu, Karl, Toomas Help, Pärtel Lippus, Ellen Niit, Pire Teras, and Tiit-Rein Viitso
(2005) “On the temporal structure of Estonian secondary-stressed feet”. Linguistica
Uralica 41, 2, 98–106.
Pike, Kenneth L. and Eunice Viktoria Pike (1947) “Immediate constituents of Mazateco
syllables”. International Journal of American Linguistics 13, 2, 78–91.
242 Natalia Kuznetsova
Plüschke, Mareike (2011) “Peak alignment in falling accents in Estonian”. In Proceedings
of the 17th International Congress of Phonetic Sciences (ICPhS XVII), Hong Kong,
China, August 17–21, 2011, 1614–1617. Hong Kong: City University of Hong Kong.
Plüschke, Mareike (2013) Peak alignment in Estonian. Unpublished Doctoral Thesis.
München: Ludwig-Maximilians-Universität München.
Pöchtrager, Markus (2006) The structure of length. Unpublished Doctoral Thesis. Wien:
Universität Wien.
Pöchtrager, Markus (2015) “Beyond the segment”. In Eric Raimy and Charles E. Cairns,
eds. The segment in phonetics and phonology, 44–64. Wiley Blackwell.
Prillop, Külli (2011) Optimaalsusteoreetiline käsitlus eesti keele fonoloogilisest kujunemisest. (Diss. Phil. Estonicae Universitatis Tartuensis, 28.) Tartu: Tartu Ülikooli
Kirjastus.
Prillop, Külli (2013) “Feet, syllables, moras and the Estonian quantity system”. Linguistica Uralica 49, 1, 1–29.
Prillop, Külli (2015) “Hääliku-, silbi- ning jalavälde: ühe nähtuse mitu tähku”. Journal of
Estonian and Finno-Ugric Linguistics 6, 3, 169–195.
Prillop, Külli (2018a) “Mida teeb moora eesti keeles”. Keel ja Kirjandus 5, 345–364.
Prillop, Külli (2018b) “Varjatud muutused eesti keele prosoodilises struktuuris”. Keel ja
Kirjandus 6, 433–452.
Prince, Alan (1980) “A metrical theory for Estonian quantity”. Linguistic Inquiry 11,
511–562.
Ross, Jaan and Ilse Lehiste (2001) The temporal structure of Estonian runic songs: a
study of the relationship between language and music. Berlin: Mouton de Gruyter.
Schiering, René, Balthasar Bickel, and Kristine A. Hildebrandt (2010) “The prosodic
word is not universal, but emergent”. Journal of Linguistics 46, 3, 657–709.
Selkirk, Elisabeth O. (1980) “The role of prosodic categories in English word stress”.
Linguistic Inquiry 11, 563–605.
Selkirk, Elisabeth O. (1984) Phonology and syntax: the relation between sound and structure. Cambridge, MA, and London, England: MIT Press.
Trubetzkoy, Nikolai S. (1935/1968) Introduction to the principles of phonological
descriptions. The Hague: Martinus Nijhoff.
Trubetzkoy, Nikolai S. (1939/1969) Principles of phonology. Berkeley and Los Angeles:
University of California Press.
van der Hulst, Harry (2010) “Representing accent”. Phonological studies 13, 117–128.
van der Hulst, Harry (2014) “Representing rhythm”. In Harry van der Hulst, ed. Word
stress: theoretical and typological issues, 3–55. New York: Cambridge University
Press.
Viitso, Tiit-Rein (1978) “Accents in Estonian”. Estonian Papers in Phonetics 1978,
94–98.
Viitso, Tiit-Rein (1979) “Problemy udarenija v ėstonskom jazyke”. Sovetskoe finno-ugrovedenie 15, 3, 136–152.
Viitso, Tiit-Rein (1982) “Morosčitajuščij li jazyk ėstonskij?”. Sovetskoe finno-ugrovedenie 8, 1, 8–20.
Morae in Estonian 243
Viitso, Tiit-Rein (2003) “Structure of the Estonian language: phonology, morphology and
word formation”. In Mati Erelt, ed. Estonian language, 9–92. (Linguistica Uralica
Supplementary Series,1.) Tallinn: Estonian Academy Publishers.
Viitso, Tiit-Rein (2008) Liivi keel ja läänemeresoome keelemaastikud. Tartu and Tallinn:
Eesti Keele Instituut.
Viks, Ülle (1992) A concise morphological dictionary of Estonian. Vol. 1: Introduction
and grammar. Tallinn: Estonian Academy of Sciences.
Wiik, Kalevi (1982) “Viron äännekestojen sääntösynteesi”. Symposiumi 1982. Suomalais-neuvostoliittolainen itämerensuomalaisen filologian symposiumi 30.8–2.9.1982,
Jyväskylä: Esitelmien refereaatit, 76–77. Jyväskylä: Jyväskylän yliopisto.
Wiik, Kalevi (1985) “Regelsynthese zur Lautquantität im Estnischen”. In Heikki Leskinen,
ed. Ostseefinnische Untersuchungen. Ergebnisse eines Finnisch-Sowjetischen Symposions, 129–137. Helsinki: Suomalaisen kirjallisuuden seura.
Wiik, Kalevi (1991) “On a third time of speech rhythm: foot timing”. In Proceedings of
the XIIth International Congress of Phonetic Sciences. Vol. 3, 298–301. Université
de Provence.
Zalizn’ak, Andrej A. (1985). Ot praslav’anskoj akcentuacii k russkoj. Мoskva: Nauka.
Аннотация. Наталья Кузнецова: Эстонская словесная просодия в
Прокрустовом ложе мор. Статья посвящена анализу существующих
морных подходов к анализу количества в эстонском языке. Обсуждается
основные особенности морных концепций эстонского количества в рамках
функциональной и генеративной лингвистики, а также понимание моры
фонетистами. Генеративные концепции совмещают на одном иерархическом уровне репрезентации (морном) нескольких разных с функциональной и структурной точки зрения пластов информации, что приводит к
значительному усложнению формального описания и внутренним противоречиям. В рамках структурного функционального понимания эстонского
количества моры могут служить вспомогательным средством для формализации просодической оппозиции краткого (легкого) и долгого (тяжелого)
ударного слога. Ее релевантность проявляется в функционировании просодической системы языка. Этот контраст надстраивается над сегментной
оппозицией кратких и долгих фонем и, в свою очередь, формирует базис
для оппозиции лексикализованных стопических акцентов, легкого и тяжелого. В качестве примера в статье приводится формальный морфонологический алгоритм присваивания стопических акцентов в словоформе и
показано место в нем слогового веса.
Ключевые слова: эстонский, структурная функциональная фонология,
автосегментная фонология, словесная просодия, мора, количество
244 Natalia Kuznetsova
Abstrakt. Natalia Kuznetsova: Eesti sõnaprosoodia moorade Prokrustese sängis. Artiklis analüüsitakse eesti keele vältekontseptsioone, mis rakendavad mooralist analüüsi. Hinnatakse peamisi funktsionaalseid, generatiivseid
ja foneetilise moorasid arvestavaid seletusi eesti keele kohta. Enamasti esindavad generatiivsetes seletustes moorad samaaegselt mitut funktsionaalselt ja
struktuuriliselt hajusa informatsiooni tasandit. See toob kaasa märkimisväärse
formaalsete analüüsitehnikate keerustumise ja sisemised vastuolud. Strukturaal-funktsionaalses raamistikus saab moorasid kasutada eesti keeles pika ja
lühikese rõhulise silbi prosoodilise kontrasti esitamiseks. Selle sobivust jälgitakse prosoodilise süsteemi tegeliku toimimise põhjal. Kontrast põhineb pika
ja lühikese foneemi segmentaalsel kontrastil ja on omakorda aluseks kahele
erinevale kõnetakti aktsendile, kergele ja raskele. Artiklis on näitena välja
pakutud morfofonoloogiline algoritm eesti keele taktiaktsentide arvutamiseks;
see näitab ka silbikaalu kontrasti kohta.
Märksõnad: eesti keel, strukturaal-funktsionaalne fonoloogia, autosegmentaalne fonoloogia, sõnaprosoodia, moora, välted