Sound Patterns of Language (2) Aspirated and Unaspirated Stops We now turn to consider another example of allophonic variation in complementary distribution in English. We begin by recalling the observation that the first sound in the word pill and the second sound in the word spill, although both voiceless bilabial stops, there is a significant acoustic and articulatory difference between them. The voiceless bilabial stop in pill is aspirated whereas that in spill is unaspirated. Following standard IPA convention, we have indicated aspiration by adding a modifier symbol, in this case the superscript h, to an IPA symbol (for example, p h, for an aspirated voiceless bilabial stop). The surface representations (SRs) of the above words would thus be [p h wl] and [spwl] respectively. That is using our informal (short-hand) IPA notation. What about the more precise mental representation in terms of phonological features? We assume that consonants, in addition to the phonological features previously discussed, can also have a feature "ASPIRATED. The feature-values for unaspirated and aspirated voiceless bilabial stops are as shown below: [p] [p h ] CONSONANTAL + + SONORANT!! SYLLABIC!! NASAL!! VOICED!! CONTINUANT!! LABIAL + + ALVEOLAR!! PALATAL!! ANTERIOR + + VELAR!! CORONAL!! SIBILANT!! ASPIRATED! + Let s now consider the distribution of aspirated sounds in English in greater detail. What sounds can be aspirated and in what phonological contexts do aspirated sounds appear? The answer to the first question is that voiceless stops in English have aspirated as well as unaspirated variants. This should not be surprising since this group of sounds constitute a natural class of sounds, namely, the set of sounds with the feature-values!continuant,!voiced,!sibilant (the feature!sibilant is required here to exclude affricates, which are never aspirated, from the set). Thus in the word pairs till ~ still and kill ~ skill, for example, we also find an aspirated voiceless 1
stop when it is the first sound in a word whereas we find an unaspirated voiceless stop appearing as the second sound in a word. The SRs of these words are [t h wl], [stwl], [k h wl] and [skwl] respectively. What about other phonological contexts? Do voiceless stops appear in any other contexts and, if so, are they aspirated or unaspirated? Consider the first and the last sounds of the word pop; the first sound is aspirated whereas the last sound is unaspirated, [p h ap]. Also, in the word popsicle, [p h apswkl], the word initial bilabial voiceless stop is aspirated but the two word-internal voiceless stops (one bilabial and one velar) are both unaspirated. It turns out that aspirated and unaspirated voiceless stops are in complementary distribution in English. There are no minimal pairs in English which differ only with respect to whether a voiceless stop is aspirated or unaspirated. So far, we have seen that a voiceless stop is aspirated when it occurs at the beginning of a word, but otherwise it is unaspirated. Actually, the precise description of the distribution aspirated and unaspirated stops turns out to be a bit more complicated: a voiceless stop is aspirated if and only if it is the first sound in a stressed syllable, which may not necessarily be the first syllable of the word; for example, in a word like repeat = [rip h it], the voiceless bilabial stop that occurs as the initial sound of the second syllable, which is stressed, is aspirated. Some examples illustrating the distribution of allophonic variation of voiceless bilabial stop [p] ~ [p] in English: Set A Set B Set C Set D Set E Set F [p h ] [p] [p h ] [p] [p] [p] pédigree petúnia empórium rápid spíll móp pérsonal patérnal compúter émpathy spórt líp pérsecute península rapídograph competítion splásh pép pílgrimage pecúliar compétitive computátional spráy gáp Voiceless Stops: Aspirated (p h, t h, k h ) at the beginning of a stressed syllable Unaspirated (p, t, k) elsewhere Since there are no minimal pairs and, furthermore, since it is entirely predictable whether a voiceless stop will be aspirated or unaspirated, according to the phonological context in which it appears, we will not assume that both the aspirated and unaspirated versions of each voiceless stop are phonemes. To do so would both add needless redundancy to our theory of mental grammar and would also fail to explain the fact that speakers of English do not recognize aspiration as being distinctive (it is not an accident that there are no minimal pairs; aspiration simply cannot be used to make to words minimally distinct, according to the phonology/grammar of the language). 2
Instead, we will assume that voiceless stops are invariably unaspirated in underlying phonemic representations and that the aspirated voiceless stops that appear in surface phonetic representations of words are derived by way of a context-sensitive phonological rule. Logically speaking, we could also try doing things the other way around; that is, take voiceless stops to be aspirated in the underlying representation, and derive unaspirated stops by a rule. But, if we did it that way, the specification of the context for the rule s application would be significantly more complex since the distribution of unaspirated stops is far more general than that of aspirated stops, which is very narrow and specific (beginning of a stressed syllable). Thus we will follow the general methodological strategy of taking the allophone that is the elsewhere case in terms of distribution to be the underlying phoneme. Informally speaking, keeping to the assumption that all voiceless stops are unaspirated in URs of words, we want a rule that makes a voiceless stop aspirated when it occurs at the beginning of a stressed syllable. Below is a formal statement of such a rule, where they symbol [ σ is used to indicate the beginning of a syllable and the feature [strong] to indicate that it is a stressed syllable: Aspiration Rule!CONTINUANT!VOICED +ASPIRATED [ σ!sibilant [strong] Consider now the derivations of the words pill and spill: UR: /pwl/ /spwl/ Nasalization NA NA Aspiration p h wl NA SR: [p h wl] [spwl] We have considered two phonological rules for English so far: Nasalization and Aspiration. In the derivations examined above, only one of these rules applies. Let s now consider a case where both rules apply, such as the derivation of the word pin: 3
UR: Nasalization Aspiration SR: /pwn/ pw#n p h w#n [p h w#n] An interesting question arises concerning the mapping of UR to SR. What happens when the input representation satisfies the context of more than one rule? The logical possibilities are that there is an established ordering or that the rules are unordered (that is, they can freely apply in any order). It is theoretically preferable not to have to say that the application of a particular set of rules is extrinsically ordered; a grammar that says nothing about ordering is obviously simpler than a grammar that must specify a particular ordering (and recall that for n rules there will n! possible orderings to choose from, which will be a fairly large number if there only a few rules to worry about). In the current case, it is easy to verify whether extrinsic ordering is necessary, since there are only two rules: just try both orderings and if both derivations give the same correct output, then there is no need to specify any rule-ordering. We already considered the derivation where the Nasalization Rule applied before the Aspiration Rule so let s now consider the derivation where these rules apply in the opposite order: UR: Aspiration Nasalization SR: /pwn/ p h wn p h w#n [p h w#n] As can be easily verified by comparing the outputs of the two different derivations, the correct SR is obtained regardless of which order the rules apply. Thus, at least as far as things now stand, we do not need to assume that the phonological rules are extrinsically ordered. It is very important to realize that a given sound may be a distinctive phoneme in one language but not in another. For example, we have seen that nasal vowels are not distinct phonemes in English. Yet in many other languages oral and nasal vowels must be treated as separate phonemes. Similarly, although aspiration on voiceless stops is predictable in English, this is not true in all languages. For example, consider voiceless stops in Thai. We find minimal pairs that differ only in terms of whether a voiceless stop is aspirated or not such as the following: 4
[paa] 'forest' ~ [p h aa] 'to split' [tam] 'to pound' ~ [t h am] 'to do' [kat] 'to bite' ~ [k h at] 'to interrupt' Here, we can easily see that aspirated and unaspirated voiceless stops must be distinct phonemes in Thai since it is not predictable whether a voiceless stop will be aspirated or unaspirated based solely on its phonological context. The SRs above must be derived from the following URs (although in these cases the mapping is trivial in the sense that no phonological rules apply): /paa/ 'forest' /p h aa/ 'to split' /tam/ 'to pound' /t h am/ 'to do' /kat/ 'to bite' /k h at/ 'to interrupt' In sum, the picture that arises is one where, in English, aspirated and unaspirated stops are allophones of a single unaspirated phoneme, whereas in Thai aspirated and unaspirated stops are related to separate phonemes at the phonemic level, as sketched out below: English Voiceless Stops: /p/ /t/ /k/ phonemic-level [p] [p h ] [t] [t h ] [k] [k h ] allophonic-level Thai Voiceless Stops: /p/ /p h / /t/ /t h / /k/ /k h / phonemic-level [p] [p h ] [t] [t h ] [k] [k h ] allophonic-level Korean is another language that behaves like Thai in the sense that there are minimal pairs differing only with respect to whether a voiceless stop is aspirated or not. Japanese, on the other hand, is often said to lack aspirated stops all together. 5
Next we will briefly examine the distribution of so-called long vowels in English. Consider the following data illustrating vowel length in English: Set A write neat rope lop lock tap pick roach treat sop wet Set B ride need robe lob log tab pig rose treed sob web In all of the Set B words, the vowel is longer (in terms of duration) than the vowel in the corresponding Set A word. In IPA, long vowels are indicated by a modifier symbol : placed immediately following the vowel symbol [V:]. In the case of a long high front vowel, for example, the symbol used would be [i:]. Note that there are no minimal pairs in English that differ only with respect to whether a vowel is long or not. Furthermore, as can be seen in the sets of data above, we can easily predict which vowels are long and which vowels are not in English; vowels are long if and only if they are immediately followed by a voiced consonant. It is therefore tempting to assume that long vowels never appear in the URs of words, but rather are derived in SRs by a phonological rule than lengthens a vowel when it occurs before a voiced consonant: Vowel Lengthening Rule!CONSONANTAL +LONG +CONSONANTAL +SYLLABIC!SYLLABIC +VOICED 6
There is a complication, however, since it turns out that tense vowels are always a bit longer than lax vowels in English, regardless of the context in which they appear. For example, the vowel in beat is a bit longer than the vowel in bit. But, at the same time, as we saw with the data above, the vowel is longer in bead than in beat and the vowel in bid is longer than that in bit. So it looks like we are dealing with two separate phenomena and that there is at least a three way contrast for vowel length. It could be that there is a vowel shortening rule that shortens the length of a lax vowel or that there are two independent vowel lengthening rules, one that lengthens a tense vowel (regardless of context) and the one above, which (further) lengthens any vowel that appears immediately before a voiced consonant. In this class we will not worry about the length differences between tense and lax vowels any further, and will only take into consideration the general (in that it applies to all vowels) context-sensitive rule stated above. Consider the derivations of the words tape and tame. UR: /tep/ /tem/ Nasalization NA te m Aspiration t h ep t h e m Vowel Lengthening NA t h e :m SR: [t h ep] [t h e :m] 7