Context

Although we have talked about the characteristics of different sounds in a spectrographic display, there are no absolute criteria that could uniquely define what a sound /X/ is in all phonetic environments, even for a particular individual. Thus, the lay person’s assumption that phonemes exhibit ‘invariance’ can hardly be justified.

Several factors influence both vowels and consonants. Earlier (see Diphthongs Table 1), we saw that the duration of vowels is different in stressed and unstressed syllables. Besides this, other variables are important too. One of the frequently mentioned variables is the influence of the following consonant, because the same vowel may be significantly longer before a voiced consonant than before a voiceless consonant.

The effect of the following consonant on the duration of the preceding vowel is mentioned above. In fact, this effect is also observed in all preceding sonorant segments. If we look at the spectrograms of the pairs sent– send, and kilt– killed (figure 1), we can observe the differences in the durations of the vowels and the sonorant consonants before voiced and voiceless stops. As we can see, the durations of the vowels and the sonorant consonants are, expectedly, greater before /d/ than /t/.

Fricatives exert an even greater influence over vowels than stops. For ex ample, we get the following readings for the same vowel [Λ] depending on the nature of the following consonant. We found the following differences with the change in the following segment:

but [bΛt]    (vowel length before a voiceless stop: 104 ms)

bud [bΛd]  (vowel length before a voiced stop: 172 ms)

buzz [bΛz] (vowel length before a voiced fricative: 210 ms)

The length of the word, as well as its number of syllables, seems influential. For example, while we get a reading of 93 ms for the vowel [ɪ] in pick, it goes down to 56 ms in picky and to 37 ms in pickiness.

Position of the word in the phrase or in the sentence, the rate of speaking, the type of word, either topic or comment, all influence the duration significantly. For example, the vowels of the two words dog [ɔ] and man [æ] change their duration by between 20 ms and 65 ms in the following sentences:

Similarly, in a listing situation, the last item has greater duration. If we compare the following two sentences, we get different durational readings for peaches and oranges in their two locations:

(a) I like apples, oranges, and peaches.

(b) I like apples, peaches, and oranges.

For example, we get an average of 94 ms as the durational reading for the final [z] of oranges in sentence (a), which goes up to a 174-ms reading for the same sound in sentence (b). Pre-pausal (at the end of phrases, clauses, or sentences) stressed vowels seem to have the longest duration; this is diametrically opposed to vowels in function words. As a result, the duration of a vowel in the former position may be up to two or even three times that of the same vowel in the latter.

Consonants can also be influenced by the contexts in which they appear. We saw earlier that the aspiration of the voiceless stops varied according to the following segment. Specifically, there was a longer lag before a sonorant consonant (e.g. play [p^hl̥e]) than before a vowel (e.g. pay [p^he̥]). As for their duration, fricatives are the consonants that are more consistently affected by the context. For example, we get readings of 196–231 ms for the /s/ of sub [sΛb], which goes up to 309–325 ms in final position in bus [bΛs]. Yet, when we place it at the end of the phrase take the bus, /s/ gives us a reading of 328–365 ms.

Speakers adapt to various circumstances of communication and adjust their production patterns accordingly. Thus, it is no surprise that speech style and rate influence the production, and we see a clear difference between slow, careful speech and fast, colloquial speech. While the former is characterized by a slower tempo, avoidance of reductions and/or deletions, and an attempt to make the production as distinct as possible, the latter is replete with deletions and reductions, and the durations of the components necessarily get shorter. These contrasting phenomena, which are known as hyperspeech (or “overshooting”) versus hypospeech (or “undershooting”), should not be conceived of as a binary split, but rather as a continuum (hence the terms ‘very slow’, ‘slow’, ‘normal formal’, ‘conversational’, ‘fast’, ‘very fast’). Consequently, the different speech realizations and different acoustic readings are the results of these varying production patterns.