Iambicity without stress in Kera
Mary Pearce
University College London
The Chadic language Kera (Ebert 1979; Pearce 1998, 2003) exhibits
iambicity without stress. This paper argues that feet are constructed
over a combination of light and heavy syllables, and that deletion and
lengthening of vowels takes place as necessary to form the iambic feet
(–) and ( –). We will consider inputs of the form /CVCV/, where in
phrase-medial position, the final vowel is deleted, giving a
monosyllabic, heavy foot (CVC). In phrase-final position, the second
vowel undergoes iambic lengthening, giving (CVCV:). As well as
lengthening and deletion, other clues to the iambic nature of Kera are
found in the system of vowel allophony and the domains of vowel
harmony and tonal/voice spreading. We will conclude that the foot
structure is a central part of Kera phonology, and that Kera adds to the
list of languages with foot structure in the absence of overt stress.
1 INTRODUCTION
Kera is a Chadic language, spoken in southern Chad and parts of Cameroon, with
50,000 speakers. Previous literature on the language (Ebert 1979; Pearce 1998, 2003)
includes nothing on the foot structure of Kera. Although previous work on Chadic refers
to quantity sensitivity, there are few references to any metrical structure (Newman 1972;
Roberts 2001; Wolff 2001; Jagger and Wolff 2002).
Kera is one of a set of iambic languages with only two types of feet: light-heavy
( ) and heavy (). Strategies are in place to avoid light-light ( ) feet. Iambs can
force not only lengthening but also vowel deletion.
(1)
For an input of /CVCV/:
Phrase medial  final vowel deletion: (CVC)
Phrase final  2nd vowel lengthened: (CVCV:)
With definite article, -ŋ  no change: (CVCVŋ)
In disyllabic feet, the contrast between the vowels is maximized, as is common in iambs,
so non-head vowels are shortened, while head vowels are lengthened. Other clues to
iambicity include vowel allophony, vowel harmony domains, and tonal spreading. Kera
is a tone language and as is the case in many tone languages, there is no overt stress. For
this reason, the evidence for the parsing of feet is not based on stress placement.
In the following section, we will see in more detail how the strategies of deletion
and lengthening are employed in Kera to avoid the ( ) foot. But to understand the
Kera foot, we must first consider the allowable structures in Kera syllables.
(2)
Kera syllable structure
Light syllables (L):
(C)V
Heavy syllables (H): (C)V: (vowel lengthened)
(C)VV (lexically long vowel)
(C)VC
 2005 by Mary Pearce
CamLing 2004: 100-107.
[baaŋa]
/cawa/[cәwa:]
[baaŋa]
[kan]
'elephant'
'fire'
'elephant'
'water'
2
Mary Pearce
2 KERA IAMBIC FEET
Kera has iambic feet which are constructed over light and heavy syllables: ( ) or ().
A foot which is constructed over two light syllables is inadmissible in Kera: *( ).
Combining these facts with the acceptable syllable structure, we get the following:
(3)
Possible foot structures
monosyllabic foot
disyllabic foot
(CVV)
(Cv.CVV)
(CVC)
(Cv.CVC)
Before looking at examples of Kera deletion and lengthening, we will look briefly at the
strategies employed by other languages which wish to avoid the ( ) foot.
Choctaw (Lombardi and McCarthy 1991), Chickasaw (Hayes1995), Hixkaryana
(Derbyshire 1979) and Yupic (Baković 1996) are all iambic languages that avoid ( )
feet by lengthening the second vowel to form a ( ) foot. On the other hand, Palauan
(Gouskova 2005), avoids ( ) feet by deleting the final syllable.
Several trochaic languages employ a similar strategy (Gouskova 2005). Panare
and Tonkawa avoid ( ) feet by deletion of the final vowel. Itelmen prefers consonant
gemination, so that /CVCV/  CVC.CV. Ilokano, Icelandic, Italian and Tagalog
lengthen the first syllable vowel as do Woleaian and Kiribati, but the related languages
Ponapean and Mokilese delete the final vowel (Spaelti 2005).
Now turning to Kera, the strategies are as follows:
(4)
/LL/ input undergoes deletion or lengthening to make a well-formed foot:
/bg/
phrase-medial:
[bg nuutu] animal his
'his animals'
phrase-final:
[beg:]
animal
'animals'
definite article:
[begŋ]
animal-DEF 'the animals'
The input for /bg/ must be /CVCV/ because /CVC/ would not need to be lengthened
and /CVCV:/ would not need to be shortened with vowel deletion. Only /CVCV/
explains both the shortening and lengthening processes.
There is also a third strategy where light syllables attach to the following word:
(5)
Syllable attaches to following word to avoid (LL) feet
(H)L# (H)  (H)(LH)
/saama nmti/
rope yours 'your rope'
word level: (saa)ma (nm)ti
phrase level: (saa)(mәnm)ti
These lengthening and deletion strategies support the case for the Kera iambic foot.
3 OTHER EVIDENCE FOR IAMBS
3.1 Vowel Allophones
Kera has 6 vowel phonemes and the three non-high vowels have allophones as follows:
(6)
Allophones in Kera vowels
phonemes:
/i/
/ /
Head (-ATR)
[i]
[]
Non-head (+ATR)
/u/
[u]
/ /
[]
[e]
/a/
[a]
[ә]
/ /
[]
[o]
Iambicity without stress in Kera 3
Lexical monosyllabic words always select vowels from the top set: [i] [ ] [u] [] [a] [].
As monosyllabic words must be feet, it follows that they are also heads. This implies
that these vowels are head vowels. On the other hand, the vowels [e] [ә] and [o] are nonhead vowels. They are of short duration and never appear in heads. Allophony and
duration both indicate the difference between a head and a non-head vowel.
(7)
(8)
Comparison of head and non-head vowels
[e] [ә] [o], always of short duration (approx. 30ms), found in non-head position.
e.g.
[pep:] 'god', [gol:] 'to look' (phrase-final)
[] [a] [], always of longer duration (greater than 50ms), found in head position.
e.g.
[pp] 'god' [gl] 'to look' (not phrase-final)
[pep:] 'god', [golŋ] 'looked' (phrase-final)
Possible Kera structures with vowel quality differences in head and non-head
Input
Output
Heads are underlined
LH
/ka.baŋ/
(LH)
(kәbaŋ)
'tree'
LL
(H)
'god'
/p.p/
(pp)
(LH) (phrase-final)
(pep:)
HH
(H)(H)
'hyena'
/gr.ny/
(gr)(ny)
/k.mm/
(k)(mm) 'rat'
H#LH /tn falaŋ/
(H)(LH)
(tn)(fәlaŋ) 'I found'
H#LL /tn gl/
(H)(H)
'I look'
(tn)(gl)
(H)(LH) (phrase-final)
(tn)(gol:)
HL]
/baa.ŋa/
(H)L (phrase-final)
(baa)ŋa
'elephant'
/kum.na/
(kum)na
'chief'
Kera is not the only language with vowel quality differences linked to a difference in
duration. English has phonemic length differences with +ATR vowels almost double the
duration of -ATR vowels (Wells 1962). There are also smaller phonetic differences
between vowel heights, high vowels being shorter. Cantonese (Yip 1996) has length
differences between +/- ATR vowels. Similarly, the Chadic languages Sokoro (Gordon
Martin p.c.) and Gœmai (Birgit Hellwig p.c.) show length differences between e/ and
o/. Like in Kera, [e] is shorter than []. Hausa also has a difference in vowel quality
corresponding to length, but unlike Kera, a short vowel is found in both CV and CVC
syllables (Carnochan 1988). In Kera, the shorter vowels are restricted to non-head
position.
In sum, the length and quality distribution of the vowels is exactly what is
expected if the word is parsed into iambic feet, and the head/non-head position controls
length and ATR.
3.2 Vowel harmony domains
Other evidence for the foot is found in vowel harmony domains. Kera has total harmony
on roots and height harmony between roots and suffixes, but it also has fronting and
rounding on central vowels. Front suffixes cause central vowels in the same foot to front
(9a). But between feet, the fronting does not take place (9b). If Kera does not have a
foot structure, these harmony facts are hard to explain.
4
Mary Pearce
(9)
Fronting within the foot1
(a) bal-(bel) love
(b) baad-(baa)d wash
bŋ-(biŋi) open
vg-(v)gi empty
s-(isi) sit down
sk-(s)ki hear
3.3 Tone spreading
A third piece of evidence for the foot comes from the tone spreading facts. In three
syllable words with two tones, the domain of the first tone is a foot.
(10)
Low High
High Low
High Mid
(gәdaa)(m:)2 horse, (dak)(tәlaw) bird
(kәsaa)(b:) locust, (man)(dәhaŋ) sack
(kuur)(si:) coal, (saa)(tәraw) cat
*(gәdaa)(m)
*(kәzaa)(b)
*(kuur)(si)
Kera is not the only language with tone and foot interaction. Creek (Zec 1999), Seneca
(Zec 1999) and Winnebago (Hayes1995) have alternating stress and tone on the head of
the foot. Shanghai (Duanmu 1993) allows tone spread from the head. In Llasa Tibetan
(Duanmu 1992) the head keeps its tone, but the tone on non-heads is lost. Three Chadic
languages, Bole, Kanakuru and Hausa (Newman 1972), show interaction between the
tone and syllable weight, possibly involving foot structure. The tone patterns in Kera are
similar to those seen in these languages and give supporting evidence for the Kera foot.
3.4 Slow speech pause breaks
A final piece of evidence comes from pauses in slow speech. A Kera speaker was asked
to say certain phrases slowly and instead of lengthening each segment, he introduced
pauses between each foot. This shows he is treating each foot as a minimal word and
that the foot is psychologically real to the Kera speaker.
(11)
/Tn asaŋ kul(i) nuutuŋ/ 'I saw his house'
(a) (slow): (Tn)..... (asaŋ)......(kul)..... (nuu)....(tuŋ)
(b) (slow): (Tn)..... (asaŋ)......(kul).......(li:)..... (nuu)....(tuŋ)
(c) (slow): (Tn)..... (asaŋ)......(ku:).......(li:)...... (nuu)....(tuŋ)
Having established the foot structure, we now consider the evidence for the absence of
( ) feet.
4 EVIDENCE THAT THERE ARE NO CVCV FEET
4.1 Words with underlying form /CVCV/
The phonetic lengthening and shortening provide strong evidence for the foot. Syllables
in the order CV.CVC shorten and lengthen respectively because they form one iambic
foot, while syllables in the order CVC.CV have no change in the expected duration. If
1
2
These examples also contain height harmony --i, which is not restricted by foot structure.
Bisyllabic feet with a single high tone are actually realized with the high tone on the non-head first syllable lowered
to mid. This implies that high tone does not spread.
Iambicity without stress in Kera 5
there was no iambic foot, we would not expect a contrast in these two patterns. We will
look first at /CVCV/ feet. The following facts are repeated from Section 1.
(12)
Changes in /CVCV/ input
(a) In non-phrase-final position, /CVCV/ becomes (CVC).
(b) In phrase-final position, /CVCV/ becomes (CVCV:)
(c) The first vowel is shortened as non-head vowel
(d) The second vowel is lengthened as head vowel.
The graph below shows these facts in the vowels of the /CVCV/ words in two folk tales.
First column is V1
Second column is
V2 in each case.
100
duration (ms)
80
60
40
20
0
/CVCV/, not phrase final
/CVCV/, right edge phon. phrase
Fig 1
First and second vowel duration in 55 /CVCV/ words from 2 folk tales
number of tokens
The ratio of the first syllable to the second in /CVCV/ words at the right edge of the
phonological phrase is on average 1 to 2.7. This is similar to other languages such as
Ciyao with a ratio of 1:2.2 and Luganda 1:3.2 (Hubbard 1995a and 1995b). The graph
below shows the ratio for phrase final /CVCV/ words. Although the spread of results is
quite large, the graph shows that very few words have a ratio of less than 1:2.
16
12
8
4
0
0
1
2
3
4
5
6
7
8
ratio duration V2/V1
Fig 2
Ratio of duration of vowels in /CVCV/ at right edge of phrase
There are no /CVCV/ words which give an output close to being [CVCV]. But in order
to discard the possibility of a (CVCV) foot, we must also consider longer words which
might contain a (CVCV) foot.
4.2 Words with underlying form /CVC.CVC/
/CVC.CVC/ words include a transition between the two middle consonants. If the
transition were shown to be a vowel, this would imply that a word-initial (CVCV) foot
must exist. For this reason, the transitions in 85 /CVC.CVC/ words were measured.
Transitions play no part in the phonological structure of the word (Levin 1986,
1987). The duration of the transition is dependent on the nature of the surrounding
consonants (Byrd 1996). In English, transitions can be up to 30 ms long (Hoole et al
2003) and transitions generally follow [-son] consonants (Matteson 1965). Dell and
6
Mary Pearce
Elmedlaoui (2002) note that transitions in Berber are always consonantal releases. These
facts are also true for Kera, as seen below in a comparison of the vowel or transition
with the following vowel. Transitions are measured in CVC CVC words. Vowels are
measured in (CVCVC) feet in the words: (CVCVC)CV or (CVC)(CVCVC).
(13)
Vowel ratio = vowel length
final vowel length
(Range: 0.4-0.8)
Transition ration = transition length
final vowel length
(Range: 0-0.4)
a. Sonorant ___C
b. Obstruent ___ Sonorant
Number of tokens
Number of tokens
20
Transition
15
10
Vowel
5
0
0
0.2
0.4
0.6
0.8
Length Ratio, ms
8
6
4
Transition
2
0
0
1
Vowel
0.2
0.4
0.6
0.8
1
Length Ratio, ms
Number of tokens
c. Obstruent___Obstruent
10
8
6
All of these results
are highly
significant with
results of p<0.001
Vowel
Transition
4
2
0
0
0.2
0.4
0.6
0.8
1
Length Ratio, ms
Fig 3
Comparison of transitions and vowels (using ratio compared with following vowel)
These measurements show that /CVC.CVC/ words have transitions and not vowels
between the two middle consonants. In fact, there are no [CVCVCVC] words in Kera.
4.3 Relative vowel lengths
In Kera there are various categories of vowels and also a transition. Each of these is
characterized by a different duration. Below, the tables show average durations for each
and the ratio between them. Firstly we compare transitions with short and long vowels.
(14) Vowel durations
vowel
transition
...C-C...
short vowel
(CVC)
long vowel
(CVV)
duration (ms)
9
57
86
number of tokens
(62)
(56)
(20)
ratio
1
6
9
The three categories above are clearly distinct from each other. In addition, short vowels
differ considerably depending on whether they are heads and the nature of the foot. The
Iambicity without stress in Kera 7
changes in duration take place in disyllabic feet as the non-head vowel shortens and the
head vowel lengthens to maximize the contrast between them.
(15)
vowel
duration (ms) number of tokens ratio
non-head, open syllable
(CVCVC) 28
(32)
2
head, closed syllable
(CVC)
57
(56)
4
head, disyllabic foot, closed (CVCVC) 70
(34)
5
head, long vowel
(CVCV:)
86
(20)
7
In none of the examples above have we found any evidence for a /CVCV/ foot. It is in
fact the case that no such foot exists in Kera. This explains the need for the strategies of
deletion and lengthening as described in the first section of this paper.
5 CONCLUSION
A more theoretical explanation for the presence of ( ) and () feet, and absence of
( ), can be found in the Iambic/Trochaic Law (Hayes 1985) with further
developments from Kager (1993, 1995), Mellander (2004), Revithiadou and van de
Vijver (1997) and van de Vijver (1998). For reasons of space, the reader is referred to
the literature.
For many languages, foot structure is assigned with reference to stress patterns.
But we have seen that Kera like Hausa (Leben 1997), Hee (Vaibra p.c.; Roberts p.c.),
Creek, Shingazidja (Kim 1999), Kirundi (Hayes 1995), and Choctaw (Lombardi and
McCarthy 1991) has foot structure in the absence of overt stress. Kera shows that iambs
can be the domain of tone and vowel harmony. It can also have a bearing on the
allophony of vowels. Kera is one of a number of iambic languages that allow only ( )
and () iambs with strategies in place to avoid the ( ) foot. Iambs can force
lengthening or vowel deletion in many languages, but in Kera, both strategies are
employed.
REFERENCES
Baković, E. (1996) ‘Foot Harmony and Quantitative Adjustments’. Rutgers Optimality Archive 168.
http://roa.rutgers.edu .
Byrd, D. (1996) ‘Influences on articulatory timing in consonant sequences’. Journal of Phonetics 24:2,
pp.209-244.
Carnochan, J. (1998) ‘The vowels of Hausa’. In Studies in Hausa, Language and Linguistics: In honour of
F.W.Parsons. G. Furness and P Jaggar (eds.). Kegan Paul, London.
Dell, F. and M. Elmedlaoui (2002) Syllables in Tashlhiyt Berber and in Moroccan Arabic. Kluwer
Academic Publishers.
Derbyshire, D. (1979) ‘Hixkaryana’. Lingua Descriptive Studies, no. 1. Amsterdam: North-Holland.
Duanmu, S. (1992) ‘An autosegmental analysis of tone in four Tibetan languages’. In Linguistics of the
Tibeto-Burman Area. 15.1: 65-91.
Duanmu, S. (1993) ‘Rime, length, stress and association domains’. Journal of East Asian Linguistics 2.1:
1-44.
Ebert, K. (1979) Sprache und Tradition der Kera (Tschad). Teil III: Grammatik. Berlin: Reimer.
Hayes, B. (1985) ‘Iambic and Trochaic Rhythm in Stress Rules’. In M. Niepokuy et al. (eds.),
Proceedings of NELS 17. pp.274-289.
Hayes, B. (1995) Metrical Stress Theory: Principles and Case Studies. University of Chicago Press.
Hubbard, K. (1995a) ‘ “Prenasalised Consonants” and syllable timing: evidence from Runyambo and
Luganda’. Phonology 12: 235-256.
Hubbard, K. (1995b) ‘Towards a theory of phonological and phonetic timing’. In B.A. Connell and A.
Arvaniti (eds.) Phonology and Phonetic Evidence: Papers in Laboratory Phonology IV. CUP.
Cambridge. pp. 168-187.
8
Mary Pearce
Gouskova, M. (2005) Syncope: Prosody or *STRUC? ms. University of Massachusetts, Amherst.
Jagger, P and H.E.Wolff (eds.) (2002). Chadic and Hausa Linguistics; Selected papers of Paul Newman
with commentaries. Rüdiger Köppe Verlog: Köln.
Hoole, P, S. Fuchs, and K Dahlmeier (2003) ‘Interarticulator Timing in Initial Consonant Clusters’. ms.
given at Proceedings of the 6th International Seminar on Speech Production, Sydney, December
7-10, 2003.
Kager, R. (1993) ‘Alternatives to the Iambic-Trochaic Law’, NLLT 11:3.
Kager, R. (1995) ‘Review Article on Metrical Stress Theory: Principles and Case Studies’, in Phonology
12: 437- 464.
Kim, Sun-Hoi (1999) The metrical computation in tone assignment. Doctoral dissertation. University of
Delaware.
Leben, W.R. (1997) ‘Tonal Feet and the adaptation of English borrowings into Hausa’. Studies in African
Linguistics 25: 139-154.
Levin, J. (1986). Between Epenthetic and Excrescent Vowels (or What Happens After Redundancy Rules).
ms. University of Texas at Austin.
Levin, J. (1987) ‘Between epenthesis and excrescence: what happens after redundancy rules’. In M.
Crowhurst (ed.) Proceedings of the West Coast Conference on Formal Linguistics 6: 187201. Stanford: The Stanford Linguistic Association.
Lombardi, L. and J. McCarthy. (1991) ‘Prosodic circumscription in Choctaw morphology’. Phonology 8:
37-71.
Matteson, E. (1965) The Piro (Arawakan) language. Berkeley: University of California Press.
Mellander, E. (2004) Quantitative Processes Without an Iambic/Trochaic Law. ms. Website: Mellander,
McGill University and Universität Leipsig.
Newman, P. (1972) ‘Syllable weight as a phonological variable’. Studies in African Linguistics 3: 301-23.
Reprinted in Jagger and Wolff (2002).
Pearce, M. (1998) ‘Consonants and Tone in Kera (Chadic)’. Journal of West African Languages 27.1:
33-70.
Pearce, M. (2003) Vowel Harmony in Kera (Chadic). ms. MA dissertation, UCL, London.
Revithiadou, A. and R. van de Vijver (1997). ‘Durational contrasts and the Iambic/Trochaic Law’.
Samiian, V (ed.) Proceedings of WECOL. Department of Linguistics, California State
University, Fresno, CA pp. 229-242.
Roberts, J. (2001) ‘Phonology of Central Chadic’. In N. Mutaka and S. Chumbow (eds.) Research Mate in
African Linguistics: Focus on Cameroon. Rüdiger Köppe Verlag: Köln. pp. 93-118.
Spaelti, P. (2005) ‘Weak Edges and Final Geminates in Swiss German’. Rutgers Optimality Archive 18.
http://roa.rutgers.edu .
van de Vijver, R. (1998) ‘The iambic issue: iambs as the result of constraint interaction’. Doctoral
dissertation, HIL dissertation series 37. Holland Academic Graphics, The Hague.
Wells, J. (1962) A study of the formants of the pure vowels of British English ms. Thesis for MA,
University of London. Available on John Wells website, UCL.
Wolff, E (ed) (2001) ‘Topics in Chadic Linguistics’. Papers from the 1st Biennial International
Colloquium on the Chadic Languages, Leipzig, July 5-8, 2001.
Verlag. hn.
Yip, M. (1996) ‘Lexicon Optimization in Languages without Alternations’. In J. Durand and B. Laks.
(eds.) Current Trends in Phonology: Models and Methods. CNRS, Paris-X and University of
Salford: University of Salford Publications. pp. 757-788.
Zec. D. (1999) ‘Footed tones and tonal feet: rhythmic constituency in a pitch-accent language’.
Phonology 16: 225-264.
Mary Pearce
Phonetics and Linguistics Department
University College London
Gower Street
London
WC1E 6BT
United Kingdom
[email protected]