Phoneme and Allophone

Phoneme and Allophone

Robert Mannell, Macquarie University, 2008

Topics

The course material for this topic is divided into the following subtopics:-

1) Introduction

2) The Premises of Practical Phonemics

Phoneme and Allophone: Introduction

Trubetzkoy (1939) wrote

"It is the task of phonology to study which differences in sound are

related to differences in meaning in a given language, in which way the

discriminative elements ... are related to each other, and the rules

according to which they may be combined into words and sentences."

Linguistic units which cannot be substituted for each other without a

change in meaning can be referred to as linguistically contrastive or

significant units. Such units may be phonological, morphological,

syntactic, semantic etc.

Logically, this takes the form:-

IF unit X in context A GIVES meaning 1

AND IF unit Y in context A GIVES meaning 2

THEN unit X AND unit Y belong to separate linguistic

units

eg. IF sound [k] in context [_æt] GIVES meaning "cat"

AND IF sound [m] in context [_æt] GIVES meaning "mat"

THEN sound [k] and sound [m] belong to separate linguistic

units

Phonemes

Phonemes are the linguistically contrastive or significant sounds (or

sets of sounds) of a language. Such a contrast is usually demonstrated

by the existence of minimal pairs or contrast in identical

environment (C.I.E.). Minimal pairs are pairs of words which vary

only by the identity of the segment (another word for a single speech

sound) at a single location in the word (eg. [mæt] and [kæt]). If two

segments contrast in identical environment then they must belong to

different phonemes. A paradigm of minimal phonological contrasts is a

set of words differing only by one speech sound. In most languages it is

rare to find a paradigm that contrasts a complete class of phonemes

(eg. all vowels, all consonants, all stops etc.).

eg. the English stop consonants could be defined by the following set of

minimally contrasting words:-

i) /pɪn/ vs /bɪn/ vs /tɪn/ vs /dɪn/ vs /kɪn/

Only /ɡ/ does not occur in this paradigm and at least one minimal pair

must be found with each of the other 5 stops to prove conclusively that

it is not a variant form of one of them.

ii) /ɡɐn/ vs /pɐn/ vs /bɐn/ vs /tɐn/ vs /dɐn/

Again, only five stops belong to this paradigm. A single minimal pair

contrasting /ɡ/ and /k/ is required now to fully demonstrate the set of

English stop consonants.

iii) /ɡæɪn/ vs /kæɪn/

Sometimes it is not possible to find a minimal pair which would support

the contrastiveness of two phonemes and it is necessary to resort to

examples of contrast in analogous environment (C.A.E.). C.A.E. is

almost a minimal pair, however the pair of words differs by more than

just the pair of sounds in question. Preferably, the other points of

variation in the pair of words are as remote as possible (and certainly

never adjacent and preferably not in the same syllable) from the

environment of the pairs of sounds being tested. eg. /ʃ/ vs /ʒ/ in

English are usually supported by examples of pairs such as "pressure"

[preʃə] vs "treasure" [treʒə], where only the initial consonants differ

and are sufficiently remote from the opposition being examined to be

considered unlikely to have any conditioning effect on the selection of

phones. The only true minimal pairs for these two sounds in English

involve at least one word (often a proper noun) that has been borrowed

from another language (eg. "Confucian" [kənfjʉːʃən] vs "confusion"

[kənfjʉːʒən], and "Aleutian" [əlʉːʃən] vs "allusion" [əlʉːʒən]).

A syntagmatic analysis of a speech sound, on the other hand,

identifies a unit's identity within a language. In other words, it indicates

all of the locations or contexts within the words of a particular language

where the sound can be found.

For example, a syntagm of the phone [n] in English could be in the

form:-

( #CnV..., #nV..., ...Vn#, ...VnC#, ...VnV..., etc.)

whilst [ŋ] in English would be:-

(...Vŋ#, ...VŋC#, ...VŋV..., etc)

but would not include the word initial forms of the kind described for

[n].

Note that in the above examples, "#" is used to represent a word or

syllable boundary, "V" represents any vowel, and "C" represents

another consonant.

For example, examples of the type "#CnV..." would include "snow"

[snəʉ], "snort" [snoːt] and "snooker" [snʉkə]. In this case, the only

consonant (for English) that can occupy the initial "C" slot is the

phoneme /s/, and so the generalised pattern could be rewritten as

"#snV...".

Allophones

Allophones are the linguistically non-significant variants of each

phoneme. In other words a phoneme may be realised by more than

one speech sound and the selection of each variant is usually

conditioned by the phonetic environment of the phoneme. Occasionally

allophone selection is not conditioned but may vary form person to

person and occasion to occasion (ie. free variation).

A phoneme is a set of allophones or individual non-contrastive speech

segments. Allophones are sounds, whilst a phoneme is a set of such

sounds.

Allophones are usually relatively similar sounds which are in mutually

exclusive or complementary distribution (C.D.). The C.D. of two

phonemes means that the two phonemes can never be found in the

same environment (ie. the same environment in the senses of position

in the word and the identity of adjacent phonemes). If two sounds are

phonetically similar and they are in C.D. then they can be assumed to

be allophones of the same phoneme.

eg. in many languages voiced and voiceless stops with the same place

of articulation do not contrast linguistically but are rather two phonetic

realisations of a single phoneme (ie. /p/=[p,b],/t/=[t,d], and

/k/=[k,ɡ]). In other words, voicing is not contrastive (at least for

stops) and the selection of the appropriate allophone is in some

contexts fully conditioned by phonetic context (eg. word medially and

depending upon the voicing of adjacent consonants), and is in some

contexts either partially conditioned or even completely unconditioned

(eg. word initially, where in some dialects of a language the voiceless

allophone is preferred, in others the voiced allophone is preferred, and

in others the choice of allophone is a matter of individual choice).

eg. Some French speakers choose to use the alveolar trill [r] when in

the village and the more prestigious uvular trill [ʀ] when in Paris. Such

a choice is made for sociological reasons.

Phonetic similarity

Allophones must be phonetically similar to each other. In analysis, this

means you can assume that highly dissimilar sounds are separate

phonemes (even if they are in complementary distribution). For this

reason no attempt is made to find minimal pairs which contrast vowels

with consonants. Exactly what can be considered phonetically similar

may vary somewhat from language family to language family and so

the notion of phonetic similarity can seem to be quite unclear at times.

Sounds can be phonetically similar from both articulatory and auditory

points of view and for this reason one often finds a pair of sounds that

vary greatly in their place of articulation but are sufficiently similar

auditorily to be considered phonetically similar (eg. [h] and [ç] are

voiceless fricatives which are distant in terms of glottal and palatal

places of articulation, but which nevertheless are sufficiently similar

auditorily to be allophones of a single phoneme in some languages such

as Japanese).

eg. In English, /h/ and /ŋ/ are in complementary distribution. /h/ only

ever occurs at the beginning of a syllable (head, heart, enhance,

perhaps) whilst /ŋ/ only ever occurs at the end of a syllable (sing,

singer, finger). They are, however, so dissimilar that no one regards

them as allophones of the one phoneme. They vary in place and

manner of articulation, as well as voicing. Further the places of

articulation (velar vs glottal) are quite remote from each other and /h/

is oral whilst /ŋ/ is nasal.

According to Hockett (1942), "...if a and b are members of one

phoneme, they share one or more features". Phonetic similarity is

therefore based on the notion of shared features. Such judgments of

similarity will vary from language to language and there are no

universal criteria of similarity.

The following pairs of sounds might be considered to be similar.

i) two sounds differing only in voicing:

[pb] [td] [kɡ] [ɸβ] [θð] [sz] [ʃʒ] [xɣ] etc...

ii) two sounds differing in manner of articulation only as plosive vs

fricative. The sibilant or grooved fricatives [s,z,ʃ,ʒ] are excluded from

this category as they are quite different auditorily from the other

("central") fricatives.

[pɸ] [kx] [bβ] [ɡɣ] etc...

iii) Any pairs of consonants close in place of articulation and differing in

no other contrastive feature:

[sʃ] [zʒ] [nɲŋ] [lɭ] [lʎ] [mɱ], etc...

iv) Any other pairs of consonants which are close in articulation and

differ by one other feature but are nevertheless frequently members of

the same phoneme

[lɹ] [cɡ] [tθ] [dð]

In languages where voicing is non-contrastive, two phones differing in

voicing and only slightly in place of articulation might be considered

similar eg. [cɡ] etc.)

Further, for the purposes of this type of analysis, the place of

articulation of the apicodental fricatives [θ,ð] is considered to be close

enough to that of the alveolar stops [t,d] to be considered phonetically

similar.

v) Any two vowels differing in only one feature or articulated with

adjacent tongue positions

[æ ɐ] [i ɪ] [ɐː ɐ] [i y] [ɑ ɑ̃]

Although it is implied above that the notion of "phonetic similarity" is in

some way less linguistically abstract (more phonetic?) than the notion

of complementary distribution, it is, nevertheless, a quite abstract

concept. The are no obvious and consistent acoustic, auditory or

articulatory criteria for phonetic similarity. Further, since a notion of

similarity implies a continuum the following question must be asked of

two phones in complementary distribution. How similar must they be

before they are to be considered members of the same phoneme?

There are many examples of very similar phones which are perceived

by native speakers to belong to separate phonemes. In English, for

example, a word terminal voiceless stop may be either released and

aspirated or unreleased. The homorganic (1) voiced stop may also be

released or unreleased. Often the unreleased voiced and voiceless

stops may actually be identical in every way except that the preceding

vowel is lengthened before the phonologically voiced stop. In terms of

phonetic similarity, the two unreleased stops may actually be identical

and yet be perceived by native speakers to belong to different

phonemes.

For example:-

/kɐp/→[kɐpʰ] ... [kɐp̚]

/kɐb/→[kɐˑb] ... [kɐˑb̚] ... [kɐˑp̚]

(nb. " ̚ " means unreleased stop and " ˑ " means partially lengthened

vowel)

Conversely, phones which are very dissimilar (at least from certain

perspectives) may be felt by native speakers to belong to a single

phoneme.

eg. Japanese (2) /h/ → [ɸ] before /u/ eg.[ɸuku] "luck"

→ [ç] before /i/ eg.[çito] "man"

→ [h] before /e,a,o/ eg.[hana]

From an articulatory perspective, these phones seem very dissimilar

(bilabial, palatal, and glottal) being produced at the extreme ends of

the vocal tract. They are, however, relatively similar acoustically and

auditorily (they are all relatively weak voiceless fricatives). This kind of

phonetic similarity is listener orientated rather than speaker orientated.

eg. English /t/ → [ʔ] medially and finally in some dialects

eg. Cockney - "butter", "wait"

→ [t] initially

nb. /k/ → [k,ʔ] does not occur although they are articulatorily closer

Phonemic Pattern

A pair of phones in complementary distribution may sometimes be

classified into separate phonemes on the basis of phonemic pattern. In

other words, is there a group of phonemes which exhibit a similar

pattern of distribution (eg. clustering behaviour, morphology, etc.) to

one of the phones being examined? In the case of the pair [h], [ŋ]

there are some similarities in patterning between [h] and certain

fricatives, and between [ŋ] and the nasals.

For example, there is a suffix which when placed before a word

commencing with a stop has the effect of negating the original

meaning. The suffix has the form /ɪ/ plus the nasal homorganic with

the stop.

ie. "impossible" [ɪmp...]

"intolerable" [ɪnt...]

"incalculable" [ɪŋk...] or [ɪnk...]

(free variation in citation form, but homorganic predominating in

rapid speech)

Clearly, this pattern suggests that [ŋ] behaves in some instances with

the same phonological pattern as the other nasals. It does in fact raise

the question of [ŋ] being an allophone of /n/. This was indeed the case

until the 1600's, but now there are quite a few minimal pairs which

have since crept into the language. ("sin"/"sing", "run"/"rung").

Phonological Space

The greater the distance between a phoneme and its nearest

neighbours, the greater the scope for allophonic variation. In other

words, the larger the number of redundant features (ie. features which

when changed will not create another phoneme) the greater the

number of allophones which can actually occur.

eg. English /p/ → [-voice]

[+bilabial]

[+stop]

[+/-aspirated]

(nb. + indicates that a feature is present, - indicates that a feature is

absent, +/- indicates that a feature is optional)

Changing the feature [-voice] to [+voice] will create /b/, changing the

feature [bilabial] will create /t,k/ (or potential allophones of them) and

changing the feature [stop] will create /w,f,m/. The only feature with

complete freedom of movement is aspiration, and variation of this

feature does indeed create the main pair of allophones of this phoneme

in English.

eg. English /r/ → [ɹ] alveolar approximant

[ɹ̥] voiceless alveolar approximant (after voiceless

sounds)

[ɻ] retroflex approximant (West England)

[ɾ] alveolar flap (Scottish) eg. [ɡɾɪn]

[ʁ] uvular fricative (Tyneside)

The possible varieties of /r/ seem to include variations of manner,

place and voicing. The only restrictions are that its allophones may not

overlap with those of /l/ and /w/.

The Premises of Practical Phonemics

(This section is after Pike (1947) (chapter 4, pp 57-66), all text below

in quotes has been taken from this source)

This section examines some of the basic assumptions behind phonemic

analysis. The first four premises are particularly important to remember

during the process of phonemic analysis.

"Phonemic analysis cannot be made with phonetic data alone; it must

be made with phonetic data plus a series of phonemic premises and

procedures".(p65)

"Phonemic procedures... must be founded upon premises concerning

the underlying universal characteristics of languages of the world... ."

(p57)

1. "Sounds tend to be modified by their environments" (coarticulation,

producing allophones)

The actual details of these processes vary from language to language.

2. "Sound systems have a tendency towards phonetic symmetry"

eg. IF unequivocal evidence that [p] vs [b] and [k] vs [ɡ] are

separate phonemes then it is likely that [t] vs [d] are separate

phonemes

3. "Sounds tend to fluctuate"

Free variation of allophones, eg. sometimes /tas/ = [tas] and

sometimes /tas/ = [das]

4. "Characteristic sequences of sounds exert structural pressure on the

phonemic interpretation of suspicious segments or suspicious

sequences of segments"

For example, in the interpretation of syllable structure:-

eg1. [ma] "cat"

[bo] "to run"

[su] "sky"

[sa] "leaf"

[ia] "moon"

[tsa] "ten"

If in all non-suspicious words the syllable structure was found to be CV

then

[ia] = /ja/

[tsa] = / t͡sa/

which would agree with the CV structure.

eg2. [maba] "dog"

[nasaɡ] "elephant"

[saplam] "egg"

[pasak] "to eat"

All clear syllable initials are found at the start of the words and are

always $CV... . All clear syllable finals are found at the end of words

and are either ..VC$ or ..V$. There are no unambiguous examples of

CC clusters at the start or end of a syllable therefore the most likely

analysis would be to place the syllable boundary in [saplam] thus

/sap$lam/. In the cases of [maba],[nasaɡ] and [pasak] the most

satisfactory syllabification would be to place the medial consonant in

the second syllable (placing at the end of the first syllable would require

an additional syllable initial $V... which is not unambiguously attested

(ie. no words begin with a vowel)).

Some extra premises (Pike lists more but these are the most

important)

1. "Every language has consonants and vowels"

2. "Certain kinds of segments may be vowels in one language but

consonants in another."

eg. [ia] = /ia/ in language 1 (L1) but [ia] = /ja/ in language 2 (L2)

3. "The dichotomy between vowel and consonant is not strictly an

articulatory one but is in part based on distributional characteristics."

4."A long vowel or consonant may in some languages constitute two

phonemes."

eg. [aː] = /a/ in L1 and /aa/ in L2

5. "A sequence of two segments may in some languages constitute a

single phonetically complex phoneme."

eg. [atsa] = /at$sa/ in L1 and /atsa/ in L2 (nb. $ = syllable

boundary)

It may be that L2 only allows open syllables (V and CV) and so the L1

form would be illegal.

6. "Some segments may be non-significant transition sounds"

eg. in English /eɡ/ may be [ʔeɡ], where the glottal stop is

phonemically non-significant.

7. "If two segments are sub-members of a single phoneme, the NORM

of the phoneme is that sub-member [allophone] which is least limited

in its distribution and least modified by its environments."

eg. /n/ → [ŋ] /__ {k/ɡ} and [n] elsewhere (here, [ŋ] is clearly an

environmental modification)

8. "In order to be considered sub-members of a single phoneme, two

segments must be (a) phonetically similar and (b) mutually exclusive

as to the environments in which they occur."

9. "When two phonemic conclusions each appear to be justifiable by the

other premises, and each seems to account for all the available facts of

all types, that conclusion is assumed to be correct (a) which is the least

complex, and (b) which gives to suspicious data an analysis parallel

with analogous non-suspicious data, and (c) which appears most

plausible in terms of alleged [coarticulations in] specific environments.

Reference

The following books/papers were referred to but aren't required

reading.

Hockett, C.F. (1942) "A System of Descriptive Phonology", Language,

18(1), 3-21

Pike, K.L. (1947) Phonemics, U.Michigan

Trubetzkoy, N.S. (1939) "Grundzüge der Phonologie". Travaux du

Cercle Linguistique de Prague 7, Reprinted 1958, Göttingen:

Vandenhoek & Ruprecht. Translated into English by C.A.M.Baltaxe 1969

as Principles of Phonology, Berkeley: University of California Press.

Additional Reading

Students should also read the following:-

Clark, Yallop & Fletcher (2007) (Chapter 4)

Footnotes

1. Homorganic sounds are two or more sounds that have the same

place of articulation but which differ in manner of articulation.

2. This pattern reflects Japanese in the mid 20th century. This pattern

has undergone recent change.

Share this post