Linguistics An Essential Introduction (Version 1.5)

The reason why our brain places them in the same category — why they sound “the same” to us, unless we focus our attention on them — is, that they cannot occur in the same phonetic environments, so that they cannot be used to distinguish different words. Look at the following set of representative words containing the two phones:

Looking at the first four words, we see a pattern: [pʰ] occurs at the beginning of words before a vowel, [p] occurs at the end of words after a vowel. The next two words suggest that [pʰ] also occurs at the beginning of words before a consonant, but [p] occurs before a consonant if there is another consonant preceding it. The next word confirm that [p] occurs before a consonant if it is preceded by a consonant, the next word shows that this is also the case if it is preceded by a vowel. Now we notice that all words so far are monosyllabic, so we might try to summarize our observations in terms of the following hypothesis: [pʰ] occurs as the first element of a syllable onset, [p] occurs anywhere else (i.e, as the second element of the onset or as any element in the coda. The next word seems to contradict this hypothesis: [p] occurs as the first element of a syllable onset. In contrast, the next word follows our prediction, with [pʰ] occurring as the first element of a syllable onset. We now notice that in the ninth word, which contradicted our hypothesis, the syllable is unstressed, while in the tenth word, it is stressed. In the first, second and fifth word, where [pʰ] also occurs, the syllable is also stressed, as these words are monosyllabic, so there is no other syllable that could be stressed. We can reformulate our hypothesis: [pʰ] occurs as the first element of the onset of a stressed syllable, [p] occurs anywhere else. If two speech sounds cannot occur in the same environment, they are said to be in complementary distribution.

Note that because the two speech sounds discussed here are in complementary distribution, we can predict, using a very simple generalization, where each of them will occur. It is this predictability that allows our brains to ignore the difference and place the two phones in the same category, treating them as different manifestations of the same entity. We call such manifestations allophones of the same phoneme.

Situations where seemingly different entities are actually different manifestations of the same entity are not restricted to language — an example from the physical world would be the substances we call water, steam and ice. Although they have different properties — water is liquid, steam gaseous, ice is solid — they are just different manifestations of the inorganic compound described by the chemical formula H₂O. As in the case of allophones, we can predict which form it will take in which environment: if the temperature is below zero degrees Celsius (32 degrees Fahrenheit, 273.15 degrees Kelvin), it will appear as ice, if the temperature is above 100 degrees Celsius (212 degrees Fahrenheit, 373.15 degrees Kelvin), it will appear as steam, and in all other cases it will appear as water!

One very widespread notation for capturing the relationship of allophones to a phoneme is in the form of a phonological rule. Such rules are typically represented in the following format, where P stands for “phoneme”, A stands for “allophone” and E stands for “(phonetic) environment”, with the underscore showing the position of the allophone:

Before we apply it to the distribution of [p] and [pʰ], let us apply it to our example of H₂O. Using the format of the phonological rule, the predictions about the form of this substance can be represented as follows:

Note that we are using the word elsewhere for one of the conditions: since the first two conditions only leave the range between 0°C and 100°C, we do not have to describe this range more precisely.

In order to write a phonological rule corresponding to our generalization about the distribution of [p] and [pʰ], we have a problem that we do not have when talking about H₂O — we need to determine a way of representing the abstract phoneme. In the case of chemical compounds, we can specify the molecular structure, but in the case of language, there is no equivalent for this.

Theoretically, we could represent the phoneme using an arbitrary label, for example, PHON230216. But it is customary in phonology to use one of the allophones between slashes, as we did in the previous section. In order to do so, we we need to decide which of the two allophones to choose to represent the phoneme. Again: neither of them is the phoneme, as the phoneme is an abstract category. There are two principles that can help us decide: a) we should choose the allophone with the least restricted distribution, b) we should choose the “simplest” allophone, i.e., the one characterizable with the least number of phonetic features. In the case of the aspirated and non-aspirated voiceless bilabial plosive in English, both criteria point to [p]: it occurs in almost all phonetic environments except one (at the beginning of an onset), and it has one feature less than [pʰ] (which is aspirated in addition to being voiceless, bilabial and plosive).

Thus, the phonological rule corresponding to our generalization would look something like this (the dollar sign $ is a common way of representing syllable boundaries, and we are using the word elsewhere as we did above, as a shorthand for “all other environments”):

The apostrophe before the dollar sign indicates that the syllable must be stressed, the underscore following the dollar sign represents the first segment of the onset, the C in parentheses indicates that there can be an optional second consonant, followed by the nucleus. The notation we use to capture the different aspects of the rule are not standardized in linguistics, they depend on the particular model of linguistics a researcher uses and/or on the level of detail they want to capture.

Things are not always as straightforward, however. Let us look at another example, the phenomenon of “Canadian raising” discussed in Section 5.6. Recall that in Canadian English, the vowel in pride is [aɪ], but in price, it is [ʌɪ], and likewise, the vowel in cloud is [aʊ] but in mouth, it is [ʌʊ]. This, too, is predictable: note that in the words pride and cloud, the consonant following the diphthong is voiced, while in price and mouth, it is voiceless. This is not an accident: the realization of the diphthongs in question depends precisely on this distinction: [aɪ] and [aʊ] occur before voiced consonants, [ʌɪ] and [ʌʊ] before voiceless consonants.

In principle, this generalization can easily be captured in a phonological rule. But which of the two allophones do we choose to represent the phoneme? Neither of them is more widely distributed than the other, and they both have the same number of features. Thus, both of the following representations are equally plausible (V theoretically stands for any vowel, but [ɪ] and [ʊ] are the only vowels occurring as the second part of a diphthong following [a]/[ʌ] in Canadian English, so we can use the character here):

or

Recall that it doesn’t really matter, as phonemes are abstract categories. But it would be nice to make a principled decision. The term Canadian raising suggests that [aV] is considered the more basic variant, that is “raised” to [ʌ] under certain conditions. But that is because [a] is the phone occurring in all environments in other North American varieties — [ʌ] is raised in comparison to other varieties. There may be one reason to prefer the first of the two rules: outside of the diphthongs, [a] is the more frequent vowel, occurring in many more words than [ʌ].

CC-BY-NC-SA 4.0. Written by Anatol Stefanowitsch