Andrew Radford - Linguistics An Introduction [Second Edition]

This is known as the Elsewhere Condition, and it states that where two rules

90

sounds

could apply to the same input and produce different outputs, then the rule which applies in the more specific set of contexts applies first, thereby preventing application of the second rule. In the present case, (71a) applies only when the plosive is preceded by /s/, whereas (71b) is written to apply anywhere. Thus, (71a) is obviously the more specific rule and will apply in preference to (71b) wherever its conditions are met. Subrule (71b) is called the ‘Elsewhere case’, or more generally the default case. It states that the default specification of [aspiration]

for voiceless plosives is [+aspiration] so that a voiceless plosive will be aspirated by default (i.e. other things being equal). The Elsewhere Condition with its associated notion of a default is an important component of UG, and its consequence in this case is that a child acquiring English does not have to learn that

(71a) must be applied before (71b) (exercises 7 and 8).

Constraints in phonology

We have characterised phonological alternations in terms of a basic

(sometimes rather abstract) underlying form which undergoes various opera-

tions or processes to emerge as a surface form. This way of thinking about

phonology has been very influential (and continues to be), but it’s not the only way to think of the organisation of a language’s sound system. Over the past decade, phonologists have developed an approach to phonology based on the

idea that phonological representations have to respect a certain set of constraints.

For instance, instead of a process which deaspirates an underlying voiceless plosive after /s/, we could propose two constraints. The first would say ‘voiceless plosives are always aspirated’ (we can call this ASPPLOS), while the second would say ‘no sound is ever aspirated immediately after /s/’ (we can call this NOASP(S)).

As they are stated, our two constraints clearly conflict with each other: when applied to a sequence such as /sp/, the constraint ASPPLOS would require the output /sph/, while the constraint NOASP(S) would require the output /sp/. In an approach to phonology known as Optimality Theory, this kind of conflict is

resolved by allowing one of the conflicting constraints to outrank or override the other. In English, the constraint NOASP(S) wins out over ASPPLOS, and we can impose the ranking NOASP(S) << ASPPLOS. On the other hand, in French, say, there are no aspirated sounds. We can therefore assume that there is another constraint, NOASP, which says ‘never aspirate a sound’. This is a more general case of NOASP(S). For French, therefore, we would assume that NOASP wins out over ASPPLOS.

In Optimality Theory, we assume that all phonological patterning is the result of ranked constraints. Moreover, we assume that all the constraints are available to all languages. For instance, in French we would assume that the constraint

ASPPLOS is valid, even though its effects will never be visible because of the

Phonemes, syllables and phonological processes

91

Input /pin/

Candidates

Other constraints

NOASP(S) ASPPLOS

NOASP

p=in

*

phin

*

sp=in

*

*

sphin

*

*

*

in

*

dog

***

Figure 28 An Optimality Theory tableau for the input /pin/ in English

precedence given to NOASP in that language. When we wish to account for the

pronunciation of a particular form, we assume that, in principle, any possible change can take place, but that, in fact, the ranking of constraints operative in the language ensures that only one output is permitted. This is the output that satisfies the set of constraints better than any other. This is the optimal candidate (hence the term ‘Optimality Theory’). An example for English, where the input is /pin/, appears in figure 28 (such figures are referred to as tableaux).

Consider first the ‘wild’ outputs /in/ and /dog/ (there are, of course, any number of these, but the considerations raised here apply to all of them). Amongst the highly ranked constraints are so-called ‘Faithfulness constraints’ (in figure 28,

these would fall under other constraints), which essentially say ‘don’t add or remove sounds gratuitously’. With /in/ we have gratuitously removed the first consonant, while with /dog/ we’ve gratuitously changed all the sounds to something else. Therefore, /in/ and /dog/ obviously violate such constraints to differing degrees, and such violations are marked by an asterisk in the appropriate column in a tableau. The candidates /sp=in/ and /sphin/, while perhaps slightly less bizarre, involve the gratuitous addition of /s/, so they also violate the Faithfulness constraints that we are assuming are highly ranked in all languages. So, we see that each of these candidates has one or more asterisks in the column corresponding to the highly ranked Faithfulness constraints.

The interesting cases are the candidates /p=in/ and /phin/, neither of which violates Faithfulness. However, both of them do violate one constraint, but we are supposing that in English ASPPLOS outranks NOASP. Thus, /p=in/ is less optimal than /phin/ as it violates a more highly ranked constraint. In a complete tableau, the best (optimal) candidate is the one that violates the least highly ranked constraint (if any), and so /phin/ emerges as the successful candidate (as shown by the pointing finger in the tableau) (exercise 9).

92

sounds

Exercises

1.

The sounds [ç h s] are in complementary distribution in native words

in the Olsk dialect of Even, a Tungusic language spoken in Yakutia,

Siberia. By examining the following Even words, decide what governs

this distribution ([ie] and [iæ] are diphthongs consisting of [i] + [e]/[æ]) bead

nɪsɑ

blows

huːn

bottom hɛr

cave

hor

foundation hat

his skill hɔːn

hot

hoːksi knife

çɪrqan

knows

hɑːn

pocket çiep

poplar hʊl

rotted

çiævʊs

sad

bʊlʊs sole

hɛssə

soup

çilj

Soviet

hɔvjɛːt

spectacles bʊsqʲɪ star

ɔsɪqam vein

hula

weapon us

2.

List all the theoretically possible combinations of two consonants in

English, then investigate how many of these could be onsets. Which of

the impossible combinations can be explained in terms of their sonor-

ity profile?

3.

Recall that the symbol = means an unaspirated consonant and the

symbol h means aspiration. Show how the pattern of data below can

be explained by the Maximal Onset Principle. Assume that separate

words are syllabified separately. (Note that it will be necessary for you

to generalise the text discussion of aspiration so as to take account of

the position of plosives in syllables.)

i.

stub

[st=ʌb]

(a) ii.

this tub

[ðɪs thʌb]

iii.

disturb

[dɪst=əːb]

i.

spare

[sp=ɛː]

(b) ii.

this pear

[ðɪs phɛː]

iii.

despair

[dɪsp=ɛː]

i.

scar

[sk=ɑː]

(c) ii.

this car

[ðɪs khɑː]

iii.

discard

[dɪsk=ɑːd]

4.

Break the following words into syllables, and, applying the Maximal

Onset Principle, identify the onsets, nuclei and codas by providing a

diagram such as that in (58). Some of these words may have more than one acceptable pronunciation, usually depending on rate of speech, so

there may be more than one correct answer for a given item.

(a) comfortable; (b) confessional; (c) secretary; (d) cooperative;