Phonemes - Minimal Pairs
After having spent quite some time on phonetics and the different branches of it, we will now turn our attention to its more theoretical counterpart, phonology. During one of our first joint sessions, you have already briefly come across the two terms in opposition to each other in connection with the defintion of consonants and vowels.
First task in developing a phonological description of a particular language:
- determine which sounds can convey a difference in meaning (same thing a child has to do when beginning to learn a language and realize that, for example, there is a difference between the words white and right)
- when two sounds can be used to differentiate words, they belong to different phonemes
- however: there may be small shades of sounds that cannot be used to distinguish words, e.g. differences between the consonants at the beginning and end of the word pop (puff of air vs. no puff of air; opening of lips vs. no opening of lips) -- both belong to the same phoneme
- NOTE: phoneme not a single sound, but usually a group of sounds
- phonemic transcription (or broad transcription) = records only those sound variations that cause a difference in meaning (vs. allophonic or narrow transcription)
Speech does not simply consist of a string of target articulations linked by simple movement between them. In fact, articulation of individual sound segments is almost always influenced by the articulation of neighboring segments, often to the point of considerable overlapping of articulatory activities. Phonetic variability is due not just to differences among individual speakers, but very often also to the phonetic context (context sensitivity). However, those variations usually do not pose any difficulty to a listener - in fact, variations can be decoded with apparently unconscious ease.
Examples of context-sensitive variation:
- nasalization of oral vowels if preceding a nasal consonant (as in sand, can't, bend)
- palatalization of [s] when preceding a [j] -- turns into [ʃ] (as in this year, tissue)
- peripheral vowels may become centralized, esp. in rapid speech if unstressed (vowel reduction towards [ə])
There are three types of assimilation:
- assimilation of place (as in ratbag or oatmeal -- [t] often realized as [p])
- assimilation of manner (as in Indian pronounced as [ɪnʤən] -- stop [d] and approximant [j] merge to form an affricate [ʤ])
- assimilation of voicing (as in have to -- [v] often realized as [f], assimilating to unvoiced [t])
Yet another special case: Elision - instance of complete sound deletion, e.g.
- in consonant clusters, such as facts (deletion of [t]) or fifths (deletion of [θ]) -- to ease the articulation process
- when unstressed, the word and often loses its [d]
- entire unstressed syllables are often elided from longer words, such as February and library
Contrastive systems range in complexity from languages with less than 20 distinctive consonants and vowels to languages with 60 or more. English, depending on the particular dialect, has up to 24 consonants and up to about 20 vowel sounds (Warlpiri (=Australian Aboriginee language): only 3 distinctive vowel sounds -- /a/, /i/, and /u/).
- phoneme = contrastive/distinctive sound within a particular language (notation: /…/)
- allophone (also variant) = sound which counts as an alternative way of saying a phoneme in a particular language (notation: […])
- English /r/ may be realized as [r], [ɹ], etc. (different realizations of /r/ do not cause a change in meaning, contrary to, e.g., Spanish (e.g. pero (= but) vs. perro (= dog)))
- Warlpiri /a/ may be realized as [ɒ], [æ], etc. (in Warlpiri, different realizations of /a/ do not cause a change in meaning, contrary to, e.g., English)
English /n/ and its allophones:
- [n̪] - dental by assimilation before a dental fricative, e.g. tenth, month
- [n:] - lengthened before a voiced obstruent in the same syllable such as [d], [z], or [ʤ], e.g. tend, tens, plunge
- [n] - normal quality elsewhere, e.g. net, ten, tent
- NOTE: [ŋ] not relevant here because this sound exists as a distinctive phoneme in the English sound system, e.g. in sin vs. sing, ban vs. bang)
In sum - Two views of the phoneme:
- functional: focus on differences in pronunciation which have an effect on the meaning of a word; phonemes = sounds that serve to differentiate words from each other, cf. as in minimal pairs* such as red vs. led, real vs. zeal
- phonetic: focus on actual pronunciation of phonemes (demands narrow phonetic description) and phonetic variability within a single phoneme; phonemes = set of related sounds (phones) -- if a phoneme has more than one variant: phoneme consists of a set of allophones standing in complementary distribution
* minimal pair = word pairs whose sound structures are identical except one minimal difference, a single sound segment that occurs in the same place in the string -- the substitution of one for the other makes a different word, e.g. crick and creek (all the possible variations - crick, creek, crook, croak, crake, crack and crock - constitute a minimal set)
Decide whether the following pairs of words are minimal pairs or not and give reasons!
- Oma : Opa
- Rand : Rat
- Rad : Rat
- bitten : bieten
- Rosen (pronounced with an alveolar trill) : Rosen (pronounced with an uvular trill)
- Buch : Bücher
- dir : Tier
- Rasen : rasen
- Sache : Sachen
- Milch : mild
- blau : Bau
- Weg : Steg
- chunk : hunk
In general: allophones = conditioned variants of a phoneme; generated by phonological conditioning(= a matter of language-specific 'rules of pronunciation')
Examples of allophones:
- [ã] before a nasal consonant (Engl. can't (RP))
- [a] elsewhere
- [g] between two voiced sounds (in languages where there is no difference between voiced and voiceless sounds, e.g. many Australian Aboriginal languages)
- [k] elsewhere
- [ŋ] before a velar consonant (Span. banca, mango)
- [n] elsewhere
- [ð] between two vowels (Span. Toledo; see also Span. realizations of /b/ and /g/ as in Cuba and Diego -- weakening from plosive to fricative manner)
- [d] elsewhere
In most of the above examples, it is rather easy to point to the conditioning factors responsible for allophonic variation. However, note that these tendencies do not yield identical consequences in all languages! Furthermore, some instances of allophonic variation cannot be explained that easily.
Example from Korean:
- [r] word-initial or intervocalic
- [l] elsewhere
- 'similarity' of [r] and [l] not easy to justify
- note, however: [r] and [l] prone to confusion even in the English language, as in meteorological, corollary, irrelevantly, etc.
Another allophonic adjustment in English:
- [ɫ] post-vocalic (dark /l/ - velarized by raising of the back of the tongue towards the soft palate)
- [l] elsewhere (pre-vocalic)
Note that ...
- in extreme cases (dialects of London, South Australia) the raising of the back of the tongue virtually creates an [u] vowel
- type of assimilation not found in many of the world's languages (cf. German kalt, Italian caldo)
The range of allophonic variation encountered in natural languages means that it is not easy to predict which sounds can or cannot be allophones of a single phoneme.
- allophones = variations from a norm (the phoneme)
- frequently, one of all allophones suggests itself as the normal value/phoneme
- if English /w/ is voiceless after voiceless plosives (e.g twin, quit), and voiced elsewhere (i.e. under all other circumstances), then /w/ (rather than /w̥/) is the phoneme
- if the two allophones of a single phoneme are [ŋ] before a velar consonant, and [n] elsewhere, then /n/ (rather than /ŋ/) is the phoneme
- if the two allophones of a single phoneme are [ã] before a nasal consonant, and [a] elsewhere, then /a/ (rather than /ã/) is the phoneme
- complementary distribution = allophonic variation dependent on the phonetic environment the phoneme occurs in (e.g. [ɫ] vs. [l] in English)
- free variation = allophonic variation independent of the phonetic environment the phoneme occurs in; random interchangeability
You can see an example for free variation here:
Another example is the old song from the 1930s:
Here you can see an example for a case of complemtary distribution:
Example of free variation of a consonantal phoneme:
- realization of word-initial th (as in then, this, there) as either [ð] or [d] (possibly due to reasons of unawareness or indifference of choice)
- [ð] and [d] = free variants (freely fluctuating allophones) of the phoneme; unconditioned by their phonetic environment
Example of free variation of a vowel phoneme:
- realization of e in economics as either /ɛ/ or /i:/; realization of the ei in either as /i:/ or /aɪ/
- each of the vowel sounds above are separate phonemes (cf. head, heed, and hide, or men, mean, and mine) which are not interchangeable in most words
- variation often dependent on regional and stylistic preferences (e.g. (oversimplified!) categorization as 'American' vs. 'British' for pronunciation of either)
Careful: Allophonic variation that happens independently of the phonetic environment the sound occurs in is not always as free as it appears! The variation is often strongly dependent on regional or stylistic influences (shifting pronunciation: just as speakers shift between lexical style registers, they may also shift between phonetic registers for stylistic reasons).
(adapted from Ramers 1998, 47)
Consider these German words. In each of them, you will find an instance of either [x] or [ç].
Becher, Buch, Biochemie, Bucht, Chemie, Dach, doch, durch, euch, Flüche, Frauchen, hoch, ich , Küche, Löcher, Lache, manche, Milch, rächen, rauchen, reich, riechen
- Now, find out in which contexts German uses [x] and in which contexts it uses [ç]. To do that,
- group the instances of [x] and [ç] together
- state the segment each of these instances of ch is preceded by
- group segments together that have something in common
- try and find the rules which determine which allophone to use
- Can you think of a minimal pair that would distinguish [x] and [ç] as separate phonemes?
- What do you think of the proposed minimal pair Kuhchen (as in 'little cow') versus Kuchen? Would you accept this? Why (not)?
For background reading, you can have a look at Fromkin/Rodman (1998) and their section on phonology. Their general introduction is easy to read but mind that they don't always use IPA symbols for their transcriptions and that they mainly use examples in American English. In the exam at the end of term you will have to use IPA symbols and stick to either British English/Received Pronunciation or American English/GenAm.
In Clark/Yallop (1995) you will find information on
- phonemes in section 4.2, pp. 91-94
- allophones in section 4.3, pp. 94-99
- free variation in section 4.11, pp. 117-119
Vivian Gramley, last modified November 2009