Pike (1947) calls phonemics ‘a technique for reducing languages to writ-ing’. The writing Pike refers to is an alphabetical writing system in which each letter is a phoneme . Naturally, it is thought, the fewer phonemes the better. An analysis with fewer phonemes is said to have better phonemic economy. The desire to achieve better phonemic economy has often led scholars to ignore other considerations. We have seen two cases above, over-analysis and under-analysis, which reduce the phonemic inventory by blurring the notion of a sound. However, it is not obvious what the importance of phonemic economy is or how it should be measured. If the goal is to save fonts for the printer, it is a trivial matter, and counting the phonemic inventory is enough. If phonemic economy makes claims about the organization of sounds in the speaker’s mind, then it is a different matter, and the measurement of economy is more complicated. For example, compare the analyses in (7).
(7) Regular analysis Over-analysis
Phonetic [t, d, k, g, tw, dw, kw, gw] [t, d, k, g, tw, dw, kw, gw] Representation [t, d, k, g, tw, dw, kw, gw] [t, d, k, g, tw, dw, kw, gw]
Inventory 8 5
Surface 8 12
To utter the same set of sounds, the speaker would make eight sounds in the regular analysis but twelve in the over-analysis , since [tw, dw, kw, gw]
represent two sounds each. Thus, with regard to production, the regular analysis is more economical in that the speaker utters fewer sounds. In addition, the over-analysis must assume consonant clusters (e.g. [tw]) and a lack of transparency between the phonological representation and the phonetic output (e.g. [tw] are two sounds in the representation but one sound [tw] phonetically). In contrast, the regular analysis does not have
such problems. Thus, it is not clear whether the over-analysis gains better economy overall.
Sometimes phonemic economy is extended to the analysis of the syllable.
A maximal Chinese syllable is CGVC or CGVG. The status of the medial G is unclear. For example, Chan (1985: 67–8) suggests that although [kw]
is phonetically the same in both Cantonese and SC, it should be analysed as one sound in Cantonese but two in SC. The reason is that [kw] and [khw] are the only CG combinations in Cantonese, and counting them as single sounds can eliminate three syllable types, CGV, CGVG, and CGVC. In contrast, SC has many CG combinations, and counting CG as two sounds can reduce the consonant inventory by about thirty (no need to set up the series Cj, Cw, and Cɥ), with an increase of only three syllable types (that is, adding CGV, CGVC, and CGVG). A similar argument is made by R. Li (1983).
There are several problems with this reasoning. First, it assumes that two series of consonants, say C and Cw, are twice as complicated as one series. The assumption is correct if phonemes are the minimal units of speech. However, if distinctive features are the basic building blocks of sounds (see below), then adding a Cw series to the C series may only involve one more feature, say [+round] for [w]. Secondly, comparing phoneme count with syllable count is like comparing apples with oranges; they do not have the same unit of measurement. For example, what is the basis for deciding whether adding ten consonants is more (or less) costly than adding three syllable types? Third, there is the issue of the generality of linguistic structure. For example, in most languages the syllable is divided into onset and rhyme, where the rhyme starts with a nuclear vowel. If one is to postulate a rhyme that starts with a glide, one needs to provide independent evidence for it. An apparent reduction in the consonant inventory of a particular language is no compelling evidence.
2 . 3 . U S I N G S Y L L A B L E S T R U C T U R E I N P H O N E M I C A N A LY S I S
The problems with under-analysis and over-analysis can be resolved without compromizing phonemic economy if we take syllable structure into consideration. Every syllable has a nucleus , usually fi lled by a vowel.
The part before the nucleus is called the onset , the part after it, the coda . 2.3. S Y L L A B L E S T R U C T U R E I N P H O N E M I C A N A LY S I S 15
The nucleus and the coda together are called the rhyme . For example, in [man] the onset is [m], the nucleus is [a], the coda is [n], and the rhyme is [an]. In [mai] the onset is [m], the nucleus is [a], the coda is [i], and the rhyme is [ai]. A syllable with a nuclear vowel but without a coda conso-nant is called an open syllable , such as [ma] (with a short [a]) or [maa]
(with a long [a]). A syllable with a nuclear vowel and a coda consonant is called a closed syllable , such as [man]. A syllable like [mai], where [i]
is in the coda, can also be called a closed syllable. More discussion of the syllable is given in Chapter 4. A slightly simplifi ed representation of [maː] and [man] in SC is shown in (8).
(8) S S syllable
ONC ONC Onset-Nucleus-Coda m a n m a phonemes
[man] [ma:] transcription
In (8) there is one vowel phoneme [a], in agreement with under-analy-sis. However, given the syllable structure, the predictable vowel length is also represented. In [man] the vowel is associated with the nucleus, and so it is realized as a short vowel. In [maː] the vowel is associated with both the nucleus and the coda, and so it is realized as a long vowel.
Thus, the analysis captures both phonemic economy and phonetic accu-racy. Now consider the representation of [swan] ‘sour’ in SC, shown in (9).
(9) S syllable
O N C Onset-Nucleus-Coda
sw phonemes
[swan] transcription n
a
In (9) there are four phonemes, in agreement with over-analysis. How-ever, since [sw] share one onset slot, they are realized as a single sound [sw]. The analysis again captures both phonemic economy and phonetic accuracy. In (9) I have assumed that two sounds can merge into one.
Which sound pairs can merge and which cannot is specifi ed in feature theory, which concerns the representation of speech sounds, to which we now turn.
2 . 4 . F E AT U R E S A N D T H E R E P R E S E N TAT I O N O F S O U N D S
A fundamental discovery in phonology is that speech consists of a sequence of sounds (besides prosodic structures such as the syllable and stress), despite various phonetic interactions between neighbouring sounds. Each sound in turn is made of more basic elements called ‘features’ (or ‘dis-tinctive features’). In thi s section I discuss feature theory, including the representation of complex sounds and len gth. I also discuss the theory of underspecifi cation.