October 24th
We saw that English speakers (as well as speakers of other languages) have consistent intuitions about the syllable structure of the words in their language. Everyone in the class agreed about the number of syllables a word contains, as well as the boundary between the syllables. For example, everyone agreed that the word "extreme" contains two syllables, and that that the boundary between those syllables occurs between the [k] sound and the [s] sound.Since the syllable structure of a word in English is predictable, we argued trhat speakers do not need to memorize the syllable structure for the words of their language. Instead, there is some general process or rule that tells us what the syllable structure for any given string of phonemes should be. In order to capture this predictable pattern, we tried to make a syllable-building algorithm that gives us the right syllable structure for a string of phonemes.
The algorithm that we settled on contains three steps:
- Nucleus Formation
- For every sound segment in the word, check if that sound is a vowel.
- If the sound is a vowel, then connect that segment to a nucleus node (N). Connect the nucleus node to a Rhyme node (R). Connect the Rhyme node to a syllable node (σ) .
- Onset Formation
- For every syllable created in Step One:
- Check if there is an unattached consonant segment to the left of the syllable. If there is, then do the following:
- If the the unattached segment can be attached to the syllable, then connect it to an onset node (O). Now do the following:
- Check if there is another unattached consonant segment to the left of the consonant you've just added to the onset node. If there is, check if it can be attached to the syllable while still obeying the phonotactic constraints of the language. If it can, then attach it to the onset node you've just created. Repeat this step for every unattached consonant to the left, until you reach a consonant that cannot be attached according to the phonotactic constraints of the language.
- Connect the onset node you've just built to the syllable node (σ) of the current syllable.
- Coda Formation:
- For every syllable created in steps one and two:
- Check if there is an unattached consonant segment to the right of the syllable. If there is, then do the following:
- If the the unattached segment can be attached to the syllable, then connect it to an coda node (C). Now do the following:
- Check if there is another unattached consonant segment to the right of the consonant you've just added to the coda node. If there is, check if it can be attached to the syllable while still obeying the phonotactic constraints of the language. If it can, then attach it to the coda node you've just created. Repeat this step for every unattached consonant to the right of the syllable, until you reach a consonant that cannot be attached according to the phonotactic constraints of the language.
- Connect the coda node you've just built to the rhyme node (R) of the current syllable.
- Put every vowel segment into a nucleus, and put that nucleus into a rhyme, and that rhyme into a syllable.
- Put as many of the consonants to the left of each syllable as possible into the onset of that syllable. The onset node attaches to the syllable node.
- Put all remaining consonants into the coda. The coda node attaches to the rhyme node.
We saw in class that the order of the steps in our syllable-building algorithm is important. In particular, we saw that switching the order of steps two and three gave us different (and wrong) predictions about the boundaries between syllables.
I mentioned that this order (build onsets before building codas) seems to reflect a linguistic universal, rather than a fact peculiar to English phonology.
October 10th
We move now to the study of phonology. Your textbook defines phonology as the component of grammar that determines the selection of speech sounds and that governs both the sound patterns and the systematic phonetic variation found in language.This definition may seem a little abstract. To make it more concrete, we looked at some cases of systematic phonetic variation in English. For example, we discovered that the voiceless stops in English are sometimes aspirated, and sometimes unaspirated (recall that aspirated sounds are produced with a burst of air, while unaspirated sounds are not). This is seen in the following words (English spelling on the left, IPA on the right):
| spit | [spɪt] |
| pit | [pʰɪt] |
| stop | [stɑp] |
| top | [tʰɑp] |
| skew | [skju] |
| cue | [kʰju] |
If you hold your hand in front of your mouth while pronouncing the above words, you will notice a big puff of air after the word-initial stops in the words "pit", "top" and "cue", but not in the words "spit", "stop" and "skew". This is because word initial voiceless stops in English are always aspirated, while voiceless stops following the fricative [s] are always unaspirated (the actual situation is a bit more complicated than this, but this description will do for now). We use the superscript [ʰ] to indicate that the preceding sound is aspirated.
Now the question arises: Are the aspirated and unaspirated really distinct sounds in English? That is, does English make a distinction between [p] and [pʰ], [t] and [tʰ], or [k] and [kʰ]?
We answer this question by introducing the concept of phonemes and allophones. The phonemes of a language can be thought of as the inventory of sounds that "count" as being really different to speakers of that language. For example, we have every reason to believe that speakers of English treat the sound [p] as being distinct from [b], and thus have reason to think that these sounds represent separate phonemes.
More concretely, we can identify whether two sounds in a language correspond to different phonemes in a language by looking for minimal pairs in that language which differ only in those two sounds. For example, we know that [p] and [b] correspond to different phonemes in English because of the existence of minimal pairs like "bat" and "pat" which differ only in whether they contain a voiced or voiceless bilabial stop. Thus, these two sounds correspond to different phonemes.
On the other hand, there are no minimal pairs in English that differ only in the aspiration of a stop. That is, there are no minimal pairs like the following:
| [pit] | [pʰit] |
| [spit] | [spʰit] |
In fact, a native English speaker without training would likely be unable to pronounce a word like [pit], which begins with an aspirated voiceless stop, or a word like [spʰit], which contains an aspirated voiceless stop following an [s].
The thing to notice about sounds like [p] and [pʰ] is that everywhere one of these sounds can occur, the other cannot, and vice-versa. So we find [pʰ] at the beginning of words, but not [p]. On the other hand, we find [p] after [s], but not [pʰ]. We describe this situation by saying that [p] and [pʰ] are in complementary distribution.
When sounds are in complementary distribution like [p] and [pʰ], they are likely to be different allophones of a single phoneme. The idea is that, at some abstract level, there is really only one "p" sound in English. However, this "p" sound can show up as [p] or as [pʰ], depending on the surrounding environment.
It is important to realize that this is just a fact about English phonology. There are other languages in the world in which [p] and [pʰ] are different phonemes. For example, the language Bengali (aka Bangla), makes a phonemic contrast between [p] and [pʰ], as well as between other sounds differing only in their aspiration. This can be seen by looking at the following list of minimal pairs from Bengali (taken from the textbook Linguistics: An Introduction, by Radford et al):
| [kʰal] "canal" | [kal] "time" |
| [tʰaka] "to remain" | [taka] "to stare" |
| [pʰul] "flower" | [pul] "bridge" |