Phoneme and Allophone - Speech Resource Pages
Phoneme and Allophone - Speech Resource Pages
Phoneme and Allophone - Speech Resource Pages
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Phoneme</strong> <strong>and</strong> <strong>Allophone</strong><br />
The nexus between<br />
phonetics <strong>and</strong> phonology<br />
Robert Mannell
Discriminative Elements (1)<br />
� Trubetzkoy (1939) wrote:-<br />
"It is the task of phonology to study which<br />
differences in sound are related to<br />
differences in meaning in a given language,<br />
in which way the discriminative elements ...<br />
are related to each other, <strong>and</strong> the rules<br />
according to which they may be combined<br />
into words <strong>and</strong> sentences."
Discriminative Elements (2)<br />
� Linguistic units which cannot be substituted<br />
for each other without a change in meaning<br />
can be referred to as linguistically contrastive<br />
or significant units.<br />
� Such units may be phonological,<br />
morphological, syntactic, semantic etc.
Discriminative Elements (3)<br />
Logically, this takes the form:-<br />
IF unit X in context A GIVES meaning 1<br />
AND IF unit Y in context A GIVES meaning 2<br />
THEN unit X AND unit Y belong to separate linguistic units<br />
e.g.<br />
IF sound [k] in context [ _ æt] GIVES meaning “cat”<br />
AND IF sound [m] in context [ _ æt] GIVES meaning “mat”<br />
THEN sound [k] <strong>and</strong> sound [m] belong to separate linguistic units
What is a phoneme? (1)<br />
� Its not true to say that a phoneme is a sound,<br />
or even that it’s a class of sounds<br />
� <strong>Phoneme</strong>s exist in human brains<br />
� They are abstract cognitive (linguistic) entities<br />
� They are conventions shared by a speech<br />
community but vary, sometimes very<br />
significantly, between speech communities
What is a phoneme? (2)<br />
� You might well ask … but surely they have<br />
something to do with speech sounds?<br />
� Well, yes, but indirectly …<br />
� When we speak we intend our listeners to<br />
underst<strong>and</strong> what words we have uttered.<br />
� A word in the brain is represented as a<br />
sequence of phonemes (or graphemes for the<br />
written word)
What is a phoneme? (3)<br />
� To communicate a sequence of words we must<br />
utter a sequence of sounds (or write the words<br />
or use sign language …)<br />
� A spoken word results from the production of a<br />
sequence of vocal tract gestures<br />
� These gestures result in a sequence of sounds.<br />
� We interpret this sequence of sounds as a<br />
sequence of phonemes
What is a phoneme? (4)<br />
� When we learn a language we learn to<br />
associate sequences of sounds in the<br />
physical world with sequences of phonemes<br />
(<strong>and</strong> therefore words) in the mental world.<br />
� The same sound may belong to a different<br />
phoneme in a different speech community or<br />
even in a different phonetic context.
What is a phoneme? (5)<br />
� OK … doesn’t that mean that a phoneme is<br />
realised as a class (or set) of physical sounds?<br />
� Yes, well almost … as some sounds can, in<br />
different contexts, belong to (or represent)<br />
different phonemes<br />
� This means that the phonemes are represented<br />
in the physical world by potentially overlapping<br />
sets of sounds
What is a phoneme? (6)<br />
� The set of sounds (in the external acoustic<br />
world) that represent a phoneme can be<br />
referred to as a set of allophones.<br />
� “allo-” indicates “difference, alternation or<br />
divergence” (Macquarie Dictionary)
What is a phoneme? (7)<br />
� An allophone is a sound that can represent a<br />
particular phoneme.<br />
� A phoneme can be said to have a huge<br />
number of slightly different allophones.<br />
� This is too complex to work with so we break<br />
this down into a much smaller set of discrete<br />
sounds that we can transcribe phonetically or<br />
measure in some other way.
What is a phoneme? (8)<br />
� One of the reasons for a phoneme having<br />
different allophones is coarticulation.<br />
That is, in different contexts the effect of<br />
adjacent phonemes can affect a phoneme’s<br />
physical realisation.<br />
� Another reason is convention. A speech<br />
community has an implicit (unconscious)<br />
agreement that certain allophones be used in<br />
certain contexts.
What is a phoneme? (9)<br />
Classical phonology took a simple view of the<br />
relationship between phonemes <strong>and</strong> allophones:-<br />
� phonemes are contrastive, allophones are not<br />
� an allophone belongs to a single phoneme<br />
� allophones of the same phoneme are in<br />
complementary distribution<br />
� allophones are phonetically similar<br />
We will examine the last three points in more detail<br />
later, but first we will examine phonemic analysis.
Phonemic Analysis<br />
� Phonemic analysis uses a narrow transcription<br />
of the speech of a language to determine what<br />
are the phonemes (the their allophones) for<br />
that language.<br />
� The phonemic analysis relies on the<br />
assumption that (a) the transcribed words have<br />
different meanings <strong>and</strong> (b) the transcription<br />
reliably captures the language’s sound system.
Minimal pairs (1)<br />
� <strong>Phoneme</strong>s are the linguistically contrastive or<br />
significant sounds (or sets of sounds) of a<br />
language.<br />
� Such a contrast is usually demonstrated by<br />
the existence of minimal pairs or contrast in<br />
identical environment (CIE).<br />
� The search for minimal pairs is the most<br />
important strategy in phonemic analysis.
Minimal pairs (2)<br />
� Minimal pairs are pairs of words which vary<br />
only by the identity of the segment 1 at a single<br />
location in the word (eg. [mæt] <strong>and</strong> [kæt]).<br />
� If two segments contrast in identical<br />
environment then they must belong to different<br />
phonemes. That is, if we change one sound to<br />
another <strong>and</strong> it changes the meaning then the<br />
sounds belong to different phonemes.<br />
(1) “segment” is another word for “single speech sound”
Minimal pairs (3)<br />
� A paradigm of minimal phonological<br />
contrasts is a set of words differing only by<br />
one speech sound.<br />
� In most languages it is rare to find a<br />
paradigm that contrasts a complete class of<br />
phonemes (eg. all vowels OR all consonants<br />
OR all stops, etc.).
Minimal pairs (4)<br />
e.g. English oral stops<br />
� the English oral stop consonants could be defined by the<br />
following set (paradigm) of minimally contrasting words:i)<br />
/pIn/ vs /bIn/ vs /tIn/ vs /dIn/ vs /kIn/<br />
� Only /ɡ/ does not occur in this paradigm <strong>and</strong> at least one<br />
minimal pair must be found with each of the other 5 stops to<br />
prove conclusively that it is not a variant form of one of them.<br />
ii) /ɡɐn/ vs /pɐn/ vs /bɐn/ vs /tɐn/ vs /dɐn/<br />
� Again, only five stops belong to this paradigm. A single minimal<br />
pair contrasting /ɡ/ <strong>and</strong> /k/ is required now to fully demonstrate<br />
the set of English stop consonants.<br />
iii) /ɡ{In/ vs /k{In/
Contrast in Analogous Environment (1)<br />
� Sometimes it is not possible to find a minimal<br />
pair which would support the contrastiveness<br />
of two phonemes <strong>and</strong> it is necessary to resort<br />
to examples of contrast in analogous<br />
environment (CAE).<br />
� CAE is almost a minimal pair, however the<br />
pair of words differs by more than just the<br />
pair of sounds in question.
Contrast in Analogous Environment (2)<br />
� Preferably, in CAE, the other points of<br />
variation in the pair of words are as remote<br />
as possible (i.e. not adjacent <strong>and</strong> preferably<br />
not in the same syllable) from the pair of<br />
sounds being tested.<br />
� The further away the other contrast is, the<br />
more unlikely it is to have any conditioning<br />
effect on the selection of pair phones of<br />
interest.
Contrast in Analogous Environment (3)<br />
� eg. /ʃ/ vs /ʒ/ in English are usually supported<br />
by examples of CAE pairs such as<br />
"pressure" [preʃə] vs "treasure" [treʒə].<br />
� The only true minimal pairs for these two<br />
sounds in English involve at least one word<br />
(often a proper noun) that has been borrowed<br />
from another language (eg. "Confucian"<br />
[kənfjʉːʃən] vs "confusion" [kənfjʉːʒən], <strong>and</strong><br />
"Aleutian" [əlʉːʃən] vs "allusion" [əlʉːʒən]).
Minimal pairs versus CAE<br />
� Even one example of a minimal pair might be<br />
considered good evidence that two sounds<br />
are allophones of different phonemes.<br />
� Minimal pairs are reliable evidence in<br />
phonemic analysis.<br />
� CAE is poorer evidence in phonemic analysis<br />
<strong>and</strong> ideally requires other supporting<br />
evidence (eg. similar patterns for similar<br />
pairs of sounds confirmed to contrast by CIE)
Syntagmatic analysis (1)<br />
� A syntagmatic analysis of a speech sound<br />
identifies all of the locations or contexts within<br />
the words of a particular language where the<br />
sound can be found.<br />
� Note that in the following examples (next<br />
page), "#" is used to represent a word or<br />
syllable boundary, "V" represents any vowel,<br />
<strong>and</strong> "C" represents any other consonant.
Syntagmatic analysis (2)<br />
For example, English [n] <strong>and</strong> [ŋ] :-<br />
� a syntagm of the phone [n] in English could be<br />
in the form:-<br />
( #CnV..., #nV..., ...Vn#, ...VnC#, ...VnV..., etc.)<br />
� whilst [ŋ] in English would be:-<br />
(...Vŋ#, ...VŋC#, ...VŋV..., etc)<br />
but would not include the word initial forms.
Syntagmatic analysis (3)<br />
� For example, sequences of the type<br />
"#CnV..." would include "snow" [snəʉ],<br />
"snort" [sno:t] <strong>and</strong> "snooker" [snʉ:kə].<br />
� In this case, the only consonant (for English)<br />
that can occupy the initial "C" slot is the<br />
phoneme /s/, <strong>and</strong> so the generalised pattern<br />
could be rewritten as "#snV...".
Syntagmatic analysis (4)<br />
� A syntagmatic analysis can provide information<br />
about different restricted distributions of two<br />
allophones of the same phoneme (always found in<br />
different locations).<br />
� A syntagmatic analysis might help to strengthen a<br />
case based on a CAE analysis. For example, we<br />
might find CAE for [p, b] <strong>and</strong> we find that all other<br />
oral stop pairs (e.g. [t, d] <strong>and</strong> [k, g]) have minimal<br />
pairs in the same syllable location. So this would<br />
strengthen the case for [p, b] being separate<br />
phonemes.
Complementary Distribution (1)<br />
� A phoneme may be realised by more than<br />
one speech sound <strong>and</strong> the selection of each<br />
variant is usually conditioned by the phonetic<br />
environment of the phoneme.<br />
� This is known as mutually exclusive or<br />
complementary distribution (CD)
Complementary Distribution (2)<br />
� The CD of two phonemes means that the two<br />
phonemes can never be found in the same<br />
environment (ie. the same environment in the<br />
senses of position in the word <strong>and</strong> the<br />
identity of adjacent phonemes).<br />
� If two sounds are phonetically similar <strong>and</strong><br />
they are in CD then they can be assumed to<br />
be allophones of the same phoneme.
Complementary Distribution (3)<br />
� eg. in many languages voiced <strong>and</strong> voiceless stops<br />
with the same place of articulation do not contrast<br />
linguistically but are rather two phonetic realisations<br />
of a single phoneme (ie. /p/=[p,b], /t/=[t,d], <strong>and</strong><br />
/k/=[k,ɡ]).<br />
� Whether the voiced or voiceless allophone is chosen<br />
depends upon syntagmatic distribution (e.g. where in<br />
the word) or phonetic context (e.g. whether the<br />
adjacent sounds are voiced or voiceless)
Complementary Distribution (4)<br />
� In some Australian Aboriginal languages word medial<br />
oral stops are voiced if both adjacent phonemes are<br />
voiced (e.g. between two vowels) <strong>and</strong> are voiceless if<br />
at least one of the adjacent sounds is voiceless.<br />
� For initial stops the patterns varies from language to<br />
language <strong>and</strong> even between dialects within a single<br />
language. In some dialects of a language the<br />
voiceless allophone is preferred, in others the voiced<br />
allophone is preferred, <strong>and</strong> in others the choice of<br />
allophone is a matter of individual choice.
Contrastive Distribution<br />
� Contrastive distribution is the opposite of<br />
complementary distribution.<br />
� Sounds in contrastive distribution can occur<br />
in the same location <strong>and</strong> when exchanged<br />
change the meaning of the word.<br />
� Sounds in contrastive distribution belong to<br />
different phonemes.
Free Variation<br />
� Occasionally speakers of a language are free<br />
to choose whether they use one or another of<br />
two possible allophones. The choice may be<br />
word-specific (often a sign of language change<br />
in progress). The choice may be pragmatic<br />
(discourse context) or sociolinguistic<br />
(e.g. Some French speakers choose to use the<br />
alveolar trill [r] when in the village <strong>and</strong> the more<br />
prestigious uvular trill [ʀ] when in Paris.)
Phonetic similarity (1)<br />
� <strong>Allophone</strong>s must be phonetically similar to<br />
each other.<br />
� In analysis, this means you can assume that<br />
highly dissimilar sounds are separate<br />
phonemes (even if they are in complementary<br />
distribution).<br />
� For this reason no attempt is made to find<br />
minimal pairs which contrast vowels with<br />
consonants. BUT…
Phonetic similarity (2)<br />
� Even the distinction between vowel <strong>and</strong><br />
consonants isn’t totally unproblematic.<br />
� For example, a sequence in one language<br />
which is perceived as a diphthong might in<br />
another language be perceived as a semivowel<br />
(approximant) <strong>and</strong> a vowel<br />
e.g. /ja/ ~ /ia/ OR /aj/ ~ /aI/<br />
� What happens when the speakers of one of<br />
these languages learns the other language?<br />
What is the interlanguage phonology?
Phonetic similarity (3)<br />
� Exactly what can be considered phonetically<br />
similar may vary somewhat from language to<br />
language <strong>and</strong> so the notion of phonetic<br />
similarity can seem to be quite unclear at times.<br />
� Sounds can be phonetically similar from both<br />
articulatory <strong>and</strong> auditory points of view. One<br />
finds pairs of sounds that vary greatly in their<br />
place of articulation but are sufficiently similar<br />
auditorily to be considered phonetically similar.
Phonetic similarity (4)<br />
� According to Hockett (1942), "...if a <strong>and</strong> b are<br />
members of one phoneme, they share one or<br />
more features".<br />
� Phonetic similarity is therefore based on the<br />
notion of shared features.<br />
� Such judgments of similarity will vary from<br />
language to language <strong>and</strong> there are no<br />
universal criteria for similarity.
Phonetic similarity (5)<br />
� Example 1. Glottal [h] <strong>and</strong> palatal [ç] are<br />
voiceless fricatives which are distant in terms<br />
of places of articulation, but they share<br />
features <strong>and</strong> are sufficiently similar auditorily<br />
(both weak sounding, voiceless, non-tonguetip,<br />
non-labial, fricatives) to be allophones of<br />
a single phoneme in some languages such<br />
as Japanese.
Phonetic similarity (6)<br />
� Example 2. In English, /h/ <strong>and</strong> /ŋ/ are in<br />
complementary distribution. /h/ only ever occurs<br />
at the beginning of a syllable (head, heart,<br />
enhance, perhaps) whilst /ŋ/ only ever occurs at<br />
the end of a syllable (sing, singer, finger).<br />
� The differ in place, manner <strong>and</strong> voicing <strong>and</strong> in<br />
tongue body, velum <strong>and</strong> laryngeal gestures.<br />
They are so different that no one regards them<br />
as allophones of the one phoneme.
Phonetic similarity (7)<br />
� So, if phonetic similarity is so difficult to pin<br />
down, then is it a useful tool in phonemic<br />
analysis.<br />
� Yes, it is. What we need to be aware of are<br />
the common patterns of phonetic similarity<br />
across languages. What follows are some of<br />
the common (but in no case universal)<br />
patterns…
Phonetic similarity (8)<br />
� i) two sounds differing only in voicing:<br />
[pb] [td] [kɡ] [ɸβ] [θð] [sz] [ʃʒ] [xɣ] etc...<br />
� ii) two sounds differing in manner of<br />
articulation only as oral stop vs fricative. The<br />
sibilant or grooved fricatives [s,z,ʃ,ʒ] are<br />
excluded from this category as they are quite<br />
different auditorily from the other fricatives.<br />
[pɸ] [kx] [bβ] [ɡɣ] etc...
Phonetic similarity (9)<br />
� iii) Any pairs of consonants close in place of<br />
articulation <strong>and</strong> differing in no other<br />
contrastive feature:<br />
[sʃ] [zʒ] [nɲŋ] [lɭ] [lʎ] [mɱ], etc...<br />
� iv) Any other pairs of consonants which are<br />
close in articulation <strong>and</strong> differ by one other<br />
feature but are nevertheless frequently<br />
members of the same phoneme<br />
[lɹ] [cɡ] [tθ] [dð]
Phonetic similarity (10)<br />
� In languages where voicing is noncontrastive,<br />
two phones differing in voicing<br />
<strong>and</strong> only slightly in place of articulation might<br />
be considered similar e.g. [cɡ] etc.<br />
� Further, for the purposes of this type of<br />
analysis, the place of articulation of the<br />
apicodental fricatives [θ,ð] is considered to<br />
be close enough to that of the alveolar stops<br />
[t,d] to be considered phonetically similar.
Phonetic similarity (11)<br />
� v) Any two vowels differing in only one<br />
feature or articulated with adjacent tongue<br />
positions<br />
[æ ɐ] [i I] [ɐ: ɐ] [i y] [ɑ ɑ̃]
Phonetic similarity (12)<br />
� Although it is implied above that the notion of<br />
"phonetic similarity" is in some way less<br />
linguistically abstract (more phonetic?) than<br />
the notion of complementary distribution, it is,<br />
nevertheless, a quite abstract concept.<br />
� The are no obvious <strong>and</strong> consistent acoustic,<br />
auditory or articulatory criteria for phonetic<br />
similarity.
Phonetic similarity (13)<br />
� There are many examples of very similar phones<br />
which are perceived by native speakers to belong to<br />
separate phonemes.<br />
� In English, for example, a word terminal voiceless stop<br />
may be either released <strong>and</strong> aspirated or unreleased.<br />
The homorganic (same place of articulation) voiced<br />
stop may also be released or unreleased.<br />
� Often the unreleased voiced <strong>and</strong> voiceless stops may<br />
actually be identical in every way except that the<br />
preceding vowel is lengthened before the<br />
phonologically voiced stop.
Phonetic similarity (14)<br />
� In terms of phonetic similarity, the two<br />
unreleased stops may actually be identical<br />
<strong>and</strong> yet be perceived by native speakers to<br />
belong to different phonemes because of<br />
their different effect on the preceding vowel.<br />
� /kɐp/→[kɐpʰ] ... [kɐp̚ ]<br />
/kɐb/→[kɐˑb] ... [kɐˑb̚ ] ... [kɐˑp̚ ]<br />
(nb. " ̚ " means unreleased stop <strong>and</strong> " ˑ "<br />
means partially lengthened vowel)
Phonetic similarity (15)<br />
� Conversely, phones which are very dissimilar<br />
(at least from certain perspectives) may be<br />
felt by native speakers to belong to a single<br />
phoneme.<br />
� eg. Japanese (1)<br />
/h/ → [ɸ] before /u/ eg.[ɸuku] "luck“<br />
/h/ → [ç] before /i/ eg.[çito] "man“<br />
/h/ → [h] before /e,a,o/ eg.[hana]<br />
(1) Japanese in the mid 20 th century. This pattern has undergone recent change.
Phonemic Pattern<br />
� A pair of phones in complementary distribution may<br />
sometimes be classified into separate phonemes on<br />
the basis of phonemic pattern.<br />
� In other words, is there a group of phonemes which<br />
exhibit a similar pattern of distribution (eg. clustering<br />
behaviour, morphology, etc.) to one of the phones<br />
being examined?<br />
� In the case of the pair [h], [ŋ] there are some<br />
similarities in patterning between [h] <strong>and</strong> certain<br />
fricatives, <strong>and</strong> between [ŋ] <strong>and</strong> the other nasal stops.
Phonological Space (1)<br />
� The greater the distance between a<br />
phoneme <strong>and</strong> its nearest neighbours, the<br />
greater the scope for allophonic variation.<br />
� In other words, the larger the number of<br />
redundant features (ie. features which when<br />
changed will not create another phoneme)<br />
the greater the number of allophones which<br />
can actually occur.
Phonological Space (2)<br />
� eg. English<br />
/p/ = [-voice] [+bilabial] [+stop] [-nasal] [+/-aspirated]<br />
(nb. + present, - absent, +/- optional)<br />
� Changing the feature [-voice] to [+voice] will create /b/,<br />
changing the feature [bilabial] may create /t,k/,<br />
changing the feature [stop] may create /w,f/,<br />
changing the feature [nasal] will create /m/.<br />
� The only feature with complete freedom of movement<br />
is aspiration, <strong>and</strong> variation of this feature does create<br />
the main pair of allophones of this phoneme in English.
Phonological Space (3)<br />
� eg. English<br />
� /r/ → [ɹ] alveolar approximant<br />
� /r/ → [R&] voiceless alveolar approximant (e.g. after /t/)<br />
� /r/ → [ɻ ] retroflex approximant (West Engl<strong>and</strong>)<br />
� /r/ → [ɾ] alveolar flap (Scottish) eg. [ɡɾIn]<br />
� /r/ → [ʁ] uvular fricative (Tyneside, UK)<br />
� The possible varieties of /r/ seem to include variations<br />
of manner, place <strong>and</strong> voicing. The only restrictions are<br />
that its allophones may not overlap with those of /l/<br />
<strong>and</strong> /w/.
<strong>Allophone</strong>s & Narrow Transcription (1)<br />
� It should now be obvious that broad<br />
(phonemic) transcription is only possible<br />
following phonemic analysis. We can’t<br />
transcribe the phonemes until we know what<br />
they are.<br />
� To what extent do narrow transcriptions<br />
match the result of the identification of<br />
allophones in phonemic analysis?
<strong>Allophone</strong>s & Narrow Transcription (2)<br />
� No narrow transcription captures all<br />
allophones of each phoneme. As we have<br />
already seen, each phoneme has an<br />
extremely large number of slightly varying<br />
allophones.<br />
� In this course we greatly simplify narrow<br />
transcription <strong>and</strong> ignore a lot of significant<br />
Australian English allophones.
<strong>Allophone</strong>s & Narrow Transcription (3)<br />
� What might be a maximal narrow transcription<br />
of a language (e.g. Australian English)?<br />
� The maximal set of possible distinctions that<br />
we can transcribe is limited by the availability<br />
of IPA symbols, including the diacritics <strong>and</strong> by<br />
our ability to hear (or otherwise measure)<br />
these differences.
<strong>Allophone</strong>s & Narrow Transcription (4)<br />
� For example we can use, not only the IPA<br />
vowel symbols but also the raising, lowering,<br />
fronting <strong>and</strong> backing diacritics to indicate a<br />
large variety of vowel qualities.<br />
� We can also provide diacritics for vowel<br />
nasalisation, breathy voice <strong>and</strong> creaky voice<br />
as well as voiceless (for whispered vowels).<br />
� For Arabic vowels we could add a velarised<br />
diacritic (for velarised vowels).
<strong>Allophone</strong>s & Narrow Transcription (5)<br />
� Traditionally, phonemic analysis was limited<br />
by the accuracy <strong>and</strong> degree of detail of a<br />
narrow transcription. If a feature was missed<br />
in a narrow transcription it wasn’t part of the<br />
phonemic analysis.<br />
� Today we can add detail from acoustic <strong>and</strong><br />
physiological measurement when judging<br />
allophonic variation.
<strong>Phoneme</strong> <strong>and</strong> <strong>Allophone</strong><br />
� This is a complex topic. You will not become<br />
sufficiently familiar with it without practice in actual<br />
phonemic analysis.<br />
� You should review the unit web site. The web site<br />
has some additional material not covered here<br />
(especially “The Premises of Phonemic Analysis”).<br />
� Read <strong>and</strong> try all of the phonemic analysis tutorial<br />
exercises, including the complex ones. Solutions will<br />
be available after the relevant tutorial.
Readings<br />
� The unit web site:http://www.ling.mq.edu.au/speech/phonetics/phonology/phoneme/index.html<br />
� Clark, J., Yallop, C., & Fletcher, J. (2007), An introduction to<br />
phonetics <strong>and</strong> phonology, Blackwell (3rd edition).
References<br />
Referred to in preparation of these notes, but not required reading<br />
� Hockett, C.F. (1942) “A System of Descriptive Phonology”, Language,<br />
18(1), 3-21<br />
� Pike, K.L. (1947) Phonemics, U.Michigan<br />
� Trubetzkoy, N.S. (1939) Grundzüge der Phonologie. Travaux du<br />
Cercle Linguistique de Prague 7, Reprinted 1958, Göttingen:<br />
V<strong>and</strong>enhoek & Ruprecht. Translated into English by C.A.M.Baltaxe<br />
1969 as Principles of Phonology, Berkeley: University of California<br />
Press.