What is Phonology?
The study of phonology is the study of the patterned interaction of speech sounds. A fairly obvious observation about human language is that different languages have different sets of possible sounds that can be used to create words. For example, the sound is found in languages like Navajo, Coushatta, and Secwepemc, but not in English, Spanish, or French.
When one language borrows sounds from another language, the borrowing language must often adapt the words to fit the set of possible sounds in its inventory. For example, observe the following data, which illustrate borrowings into Hawaiian from English:
rice [ɹaɪs] [laiki]
wine [waɪn] [waina]
brush [bɹʌʃ] [palaki]
ticket [tɪket] [kikiki]
The data in the table above show that Hawaiian alters the English words in order to fit them into the possible inventory of sounds. For example, Hawaiian does not have the sounds [s], [ʃ], or [t]. Whenever the English word contains one of these sounds, it is replaced with the sound (eg. tɪket] > [kikiki] ). Also, Hawaiian does not have the sound . Whenever the English word contains this sound, it is replaced with the sound (eg. [ɹaɪs] > [laiki]). Similarly, the sounds [b] and [æ] are replaced with [p] and [a], respectively.
The following chart shows the sound inventory of Hawaiian:
However, besides the replacement of one sound for another, there are other differences between the English and Hawaiian words. In the Hawaiian forms, vowels are inserted that do not exist in the English forms. For example, you may note that in all the examples, a final vowel is added in the Hawaiian forms (eg. [waɪn] > [waina]). Also, a vowel is inserted whenever there are two consonants side-by-side in the English forms ([bɹʌʃ] > [palaki]). Finally, in the name Albert, there is a consonant added at the beginning of the word ([ælbɹt] > [ʔalapaki]).
This suggests that Hawaiian not only has restrictions on what sounds can occur in the language, but also conditions on how those sounds can be used in the formation of words. Based on the data above, we can propose three conditions on the interaction of sounds in Hawaiian:
Words in Hawaiian must not end in a consonant
Words in Hawaiian must not have two consonants in a row
Words in Hawaiian must begin with a consonant
A thorough study of Hawaiian words would show that these restrictions are not just restrictions on borrowed words, but also on all words in Hawaiian.
One of the goals of phonology is to describe the rules or conditions on sounds and sound structures that are possible in particular languages.
Another major goal of phonology is to account for the similarities among human languages. That is, even though the different languages have different sets of sounds and different ways of arranging and patterning those sounds, there are a number of similarities across human languages. The following are a few of these similarities, often called universals:
All consonant inventories have voiceless stops
All languages have syllables
All inventories can be split into vowels and consonants
There are also some near-universals, such as the following:
Only two languages in the University of California Segment Inventory Database (UPSID), Rotokas and Mura, have no sonorant consonants
All languages in UPSID have some kind of , except Hawaiian
91.5% of the languages in UPSID have
One of the goals of phonology is to define the space of possible sounds and sound structures that all human languages draw from.
What are Phonemes?
What are Syllables?
What are Distinctive Features?
What are Rules?
What is Optimality Theory?
What are Phonemes?
When you looked at the actual transcriptions, you should have seen that the transcriptions of the sounds represented by the letter p in the words were not the same. Take a look at the words with their transcriptions:
As you can see in the chart above, the letter p is transcribed in three ways:
You can demonstrate how these sounds are different on your own. Take your hand and place it in front of your mouth. Now say pat.
Did you feel the air against your hand? Now say spat.
When you say pat, you should feel more air against your hand than when you say spat. That extra puff of air is known as aspiration, and is represented in the transcription by the superscript "h".
Now, often, when one says tap, tip or top, there is no release of the "p" sound, and no air is present when the sound is articulated. This is known as an unreleased sound, represented by the "◌̚" next to the sound.
So now we know that the sound that English speakers hear as "p" can be pronounced in three different ways: regular, aspirated, and unreleased.
Look at the distribution of the different pronunciations across these data abovr. The distribution is not random. That is, we can predict where each pronunciation can occur. These facts can be shown as the following:
The "p" sound is pronounced as when it occurs at the beginning of the word.
The "p" sound is pronounced as when it occurs at the end of the word.
The "p" sound is pronounced as everywhere else.
How would you transcribe the "p" sound in each of the following words?
The facts thus far about the data that we've been looking at are the following:
The "p" sound in the data is pronounced in three different ways.
This difference in pronunciation is not random, but predictable, depending upon where the "p" sound occurs in the word:
Beginning of the word
End of the word
Anywhere except at the beginning or the end
It is important to note that the places in which the different pronunciations occur are unique, and do not overlap. That is, you never find [pʰ] at the end of the word, or [p̚] at the beginning of the word, or [p] in either place.
Since these different pronunciations never appear at the same place in words, they are said to be in complementary distribution.
This brings us to the concept of the phoneme. A phoneme is a mental representation of a sound that has predictable variants. Each of the variants of that sound is called an allophone.
For example, in the case of the English stops, the sounds [pʰ], [p̚], and [p] are all predictable variants of one sound. They are in complementary distribution, as discussed above. Therefore, they must all be allophones of a phoneme.
How do we represent this phoneme? For reasons which will become clearer in later lessons, the phoneme is represented by the sound which has the broadest distribution, or occurs in the most places. Since [pʰ] and [p̚] can only appear at the beginning or end, and [[pʰ], [p̚]] appears everywhere else, [[pʰ], [p̚]] has the widest distribution.
Therefore, the phoneme is . The /p/ indicates that the representation is a phoneme, not a single sound.
Truth, Justice and the Linguistic Way
Still confused? Ok, imagine Clark Kent and Superman. We know that Clark Kent and Superman are the same person, right? Each persona is a variant of a single person, even though they look and act differently.
Even if we didn't know they were the same person, we could figure it out (because we're smarter than Lois and Jimmy) by looking at the different contexts in which we see Clark Kent and Superman. Here are the contexts we find Clark Kent:
On the trail of a story
In the supermarket
Buying a new suit for a date with Lois
Clark Kent shows up during contexts of being a reporter, being the owner of a refrigerator, being a suitor for Lois, etc. Now what about Superman:
The world is going to explode in 5 seconds
Jimmy is being held captive by Braniac
So Superman appears only in the context of being a hero.
In other words, Clark Kent and Superman never appear at the same in the same place in the same context. This means that they are in complementary distribution.
Since they are in complementary distribution, if they were sounds, they would be allophones of the same phoneme.
Which is the phoneme, Clark Kent or Superman? Well, since Superman only appears in the context of being a hero, Clark Kent has the wider distribution. Therefore, the phoneme is /Clark Kent/.
Earlier, we discussed the term complementary distribution. Again, sounds are in complementary distribution if they never appear in the same contexts.
Sometimes, sounds do appear in at least some of the same contexts. When this happens, the sounds are in overlapping distribution.
For example, observe the following data:
Note in particular the sounds [pʰ], [b], and [kʰ]. In each form, one of those sounds appears at the beginning of the word. Therefore, each one of the sounds can appear in the context of the beginning of the word.
Based on that fact, the sounds [pʰ], [b], and [kʰ] cannot be in complementary distribution, because they can appear in the same context.
This leads to the conclusion that [pʰ], [b], and [kʰ] are in overlapping distribution, since in the context of the beginning of the word, each of those sounds can appear.
Furthermore, if these sounds are in overlapping distribution, they must be variants of separate phonemes. That's an important relationship. Say the following to yourself as a mantra:
complementary distribution = allophones of the same phoneme
overlapping distribution = allophones of separate phonemes
Repeat this to yourself as needed.
Holy Aspiration, Batman!
Let's go back to the superhero analogy for a moment. We'll now add Bruce Wayne and Batman into our data set. As with Superman, Batman appears in the context of being a hero, while Bruce Wayne appears everywhere else (i. e. they are in complementary distribution).
Hence [Bruce Wayne] and [Batman] are allophones.
Now, consider just Superman and Batman. Are they allophones of a single phoneme?
So, to answer the last question, Batman and Superman must belong to different phonemes, because they can both appear in the context of being a hero, and are therefore in overlapping distribution.
Another useful term to know is contrastiveness. When sounds are in overlapping distribution, they are contrastive. This is to indicate that the sounds can create lexical contrasts.
What does that mean?
Basically, it means that you can change the meaning of a word simply by changing one of the sounds to another. For example, if you have the word pat [pʰæt], you can change the meaning of the word to something else by changing the [pʰ] to a [b], giving you the word bat [bæt]. Therefore, the sounds [pʰ] and [b], are contrastive.
Take the word pat again. If you change the [pʰ] to [p], is the result a different word? Not in English, it isn't. It sounds a little strange, but it's still recognizable. Therefore, these sounds are not contrastive.
Summary of Phonology Lesson 1: What are Phonemes?
In this lesson, we looked at concepts of sound distribution. If sounds are in complementary distribution, they cannot appear in the same contexts. If sounds are in overlapping distribution.
Furthermore, if sounds are in complementary distribution, they are allophones of the same phoneme. If sounds are in overlapping distribution, they are allophones of different phonemes.
If sounds are allophones of different phonemes, they are contrastive.
What are Distinctive Features?
In this lesson, the goal is to learn about the motivations for distinctive features in phonological theory.
[± spread glottis]
[± constricted glottis]
In the first section, you will be introduced to the idea of a natural class.
Recall the previous lesson, which discussed the distribution of aspiration in English. In that lesson, it was shown that certain consonants are aspirated at the beginnings of words. However, observe the following data, remembering that a * refers to an ungrammatical form.
pat [pʰæt] sat [sæt] *[sʰæt]
tap [tʰæp] lap [læp] * [lʰæp]
cap [kʰæp] gap [gæp] * [gʰæp]
The data above illustrate that the consonants that are aspirated at the beginnings of words forms a subset of the English consonants.
The sounds [p], [t], [k] aspirate at the beginnings of words, but consonants such as [s], [l], [g] do not aspirate at the beginning of words.
Since there are rules of language that apply to only certain sets of words, it is useful to refer to such sets as being composed of a certain feature or features that are not shared by other consonants in the larger set. These subsets are known as natural classes.
Further, the features that define natural classes are known as distinctive features. The next sections will further elaborate the idea of distinctive features and introduce many of the specific features that have been proposed for human language.
So, we have the concept of distinctive features. These distinctive features allow us to define natural classes of sounds. But how do we define these features?
Well, first of all, the features we define should be adequate to define some natural class of sounds. But remember that a natural class is composed of sounds that share a certain feature or group of features. Clearly, different sounds should not share all of the same features. So, if the proposed set of distinctive features in human language is adequate, every sound should have a unique set of features.
In the phonetics lessons, it was discussed that different sounds are classified by the features associated with their specific articulations. For example, consonants are classified in terms of their place of articulation, manner of articulation, and voicing. Each sound was represented by a unique combination of these features.
So, a good place to approach distinctive features might be those used in the phonetic classifications discussed in the phonetics lessons. Are these features adequate? If they are, they should be able to define all natural classes of sounds, and every sound should be definable in terms of those features.
To test all the natural classes of human language is a long drawn out process, and such features are still being tested. However, we don't have to go too far afield to discover that places of articulation, manners of articulation, and voicing are not adequate. Observe the following data from Scottish English:
writhe [ræð] mile [mʌil]
nine [nʌin] beige [beːʒ]
tease [tiːz] road [rod]
love [lʌːv] horn [horn]
peace [piːs] food [fud]
boar [boːr] life [lʌif]
Question: Based on the above data, before which consonants do the vowels [i, e, a, o, u, ʌ, ʌi ] appear, and before which consonants do the vowels [iː, eː, aː, oː, uː, ʌː, æ] appear?
The previous section presented you with some data from Scots English. In that data, it was shown that the vowels and the vowels appeared before the following sets of consonants
[i, e, a, o, u, ʌ, ʌi ] [iː, eː, aː, oː, uː, ʌː, æ]
Question: In looking at these two sets of environments (environments are those sounds that follow or come before a particular sound), are the vowels [i, e, a, o, u, ʌ, ʌi ] and the vowels [iː, eː, aː, oː, uː, ʌː, æ] in complementary or overlapping distribution? If you're not sure, review the section on phonemes (hint: remember Superman and Clark Kent?).
Based on the data shown in the previous section, it is clear that the vowels [i, e, a, o, u, ʌ, ʌi ] and the vowels iː, eː, aː, oː, uː, ʌː, æ] are in complementary distribution.
Let's run through the argumentation.
The vowels [i, e, a, o, u, ʌ, ʌi ] can appear before the following set of consonants: [f, m, θ, s, n, d, l] .
The vowels [iː, eː, aː, oː, uː, ʌː, æ] can appear before the following set of consonants: [v, ð, z, ʒ, r].
The two sets of consonants are not the same.
Therefore, the vowels [i, e, a, o, u, ʌ, ʌi ] do not appear in the same contexts as the vowels [iː, eː, aː, oː, uː, ʌː, æ].
Therefore, the two sets of vowels are in complementary distribution.
Now that we know that the two sets of vowels are in complementary distribution, we can theorize that they are allophones of the same phoneme.
Question: if they are allophones of the same phoneme, which set represents the phoneme? (Which one is Clark Kent?)
Remember from the section on phonemes that the phoneme is represented by the allophone that has the broadest environment. In this case, the vowels from the set [i, e, a, o, u, ʌ, ʌi] would represent the phonemes /i/, /e/, /a/, /o/, /u/, /ʌ/, /ʌi/.
The other set of vowels represent allophones of those phonemes. When we discussed English stops, we defined the contexts in which the aspirated stops appear and when the unreleased stops appear. We want to do the same here for the vowels [iː, eː, aː, oː, uː, ʌː, æ].
This is where natural classes come in handy. In order to define where the vowels [iː, eː, aː, oː, uː, ʌː, æ] appear, we simply have to define the features of the natural class of consonants that can follow those vowels. These features should define this class uniquely; that is, the features that define that natural class should not also define the class of consonants that follow the short vowels, as that would defeat the purpose.
So, to begin with let's look at the set of consonants in question in terms of the phonetic classifications that we determined before.
Here's the set of consonants: [v, ð, z, ʒ, r].
Let's start with Place of Articulation. Do all these consonants share the same Place of Articulation?
Move on to the next section to see the answer.
So, the question is: do all of the consonants in the set [v, ð, z, ʒ, r] share the same place of articulation?
The answer is no. The consonant [v] is a labiodental consonant, the consonant [ð] is an interdental consonant, the consonants [z] and [r] are alveolar consonants, and the consonant [ʒ] is an alveopalatal consonant.
Therefore, we cannot define the natural class in terms of Place of Articulation. Now, let's see if we can define this set in terms of Manner of Articulation. Do the consonants in the set [v, ð, z, ʒ, r] share the same manner of articulation?
Try to determine this for yourself and then go to the next section for the answer.
So, the question is: do all of the consonants in the set [v, ð, z, ʒ, r] share the same manner of articulation?
The answer is almost, but no. The consonants [v], [ð], [z], and [ʒ] are all fricatives. However, the consonant [r] is a trill, not a fricative.
Therefore, we cannot define the natural class in terms of Manner of Articulation. Now, let's see if we can define this set in terms of voicing. Do the consonants in the set [v, ð, z, ʒ, r] share the same voicing?
Try to determine this for yourself and then go to the next section for the answer.
So, the question is: do all of the consonants in the set [v, ð, z, ʒ, r] share the same voicing?
The answer is yes. All of the consonants are voiced.
Therefore, we can define the natural class at least partially in terms of Voicing. However, the classification of voicing doesn't uniquely define this set of consonants as opposed to the consonants that define the other environment. That is, if we were to define the consonants [v, ð, z, ʒ, r] as voiced consonants, we would have to say that the vowels [iː, eː, aː, oː, uː, ʌː, æ] appear only before voiced consonants. However, the vowels [i, e, a, o, u, ʌ, ʌi ] appear before the consonants [f, m, θ, s, n, d, l] and four of those consonants are voiced.
Therefore we need some other feature besides place of articulation, manner of articulation and voicing to define the set [v, ð, z, ʒ, r] as a natural class of consonants. This is where distinctive feature theory comes in.
One can think of distinctive features as a set of binary "switches" that all sounds have as part of their mental representation. Every sound has a unique configuration of these "switches". For each sound, some of the switches may be turned on, while others are turned off. Which switches are turned on and which are turned off determine what kind of sound is produced.
To represent these switches, every switch is given a name, such as [voice], [distributed], etc. (See the last section of this lesson for a more complete list of the features).
To represent whether the switch is turned on or off, either a + or - is placed before the name. For example, if the switch [voice] is on, it is represented as [+voice]. If it is off, it is represented as [-voice].
We noted that the consonants in the set [v, ð, z, ʒ, r] are all voiced. Therefore, each of the sounds has the feature [+voice], which indicates that the switch [voice] is turned on.
The only real exceptional features are the place features [LABIAL], [CORONAL], [DORSAL], and [RADICAL]. For a number of linguists, these features do not have an on-off switch like other features. Instead, some sounds have the feature, and some don't. The reasons for this, and the implications of this will not be discussed here. You need only remember that these particular features act a little differently.
In the case of Scots English, what we need is to define a feature that is shared by all of the consonants in the set [v, ð, z, ʒ, r] that is separate from voicing. One such feature that has been proposed is the feature [continuant]. Sounds that are [+continuant], that is, that they are characterized has having this feature, are produced with continual airflow through the oral cavity. Sounds such as fricatives and some approximants have this feature, but stops and nasals do not. Thus this feature divides speech sounds into two sets that are not part of the classification system learned in the phonetics section.
Now if we look at the set [v, ð, z, ʒ, r] again, we see that they are all fricatives or approximants. So now we have two ways of defining this set as a natural class. These consonants are all both [+voice] and [+continuant]. Can any of the consonants in the set [f, m, θ, s, n, d, l] be defined by both of these features?
Well, the consonants [f, s, θ] are [+continuant], being fricatives, but they are not [+voice]. The consonants [m, n, d] are [+voice], but not [+continuant].
What about [l]? The consonant [l] is often discussed as being ambiguous. In some languages, [l] is [+continuant], and in some languages it is [-continuant]. It would appear that in Scots English, the consonant [l] is [-continuant].
Therefore, the features [+continuant] and [+voice] do, indeed, define the set [v, ð, z, ʒ, r] as a unique natural class.
In the lesson What are Rules?, you will learn about how to use phonemes and distinctive features to create phonological rules. In the next lesson, What are Syllables?, you will learn about the concept of the syllable, and the evidence to support this phonological entity.
The next section in this lesson gives a list of distinctive features that have been proposed for human language.
The following is a list of proposed distinctive features, compiled in Gussenhoven and Jacobs (1998):
Sounds which are [+ consonantal] are those which have some kind of constriction along the center of the vocal tract. This constriction must be at least as narrow as that required for a fricative.
Sounds which are [+sonorant] are those which are produced with a constriction in the vocal tract that allows the air pressure both behind and in front of the constriction to be relatively equal. This feature generally divides the sound system into sonorants ([+sonorant] sounds), which are nasals, approximants, glides, and vowels, and obstruents ([-sonorant] sounds), which are oral stops, fricatives, and affricates.
Sounds which are [+approximant] are those sounds whose constriction allows for a frictionless escape of air.
Sounds which are [+voice] are those which are produced with vibration of the vocal folds.
[± spread glottis]
Sounds which are [+spread glottis] are those produced with a glottal configuration that produces audible glottal friction. For example, the aspirated stops in English are [+spread glottis]
[± constricted glottis]
Sounds which are [+constricted glottis] are those which are produced with the vocal folds drawn together and tense.
Sounds which are [+continuant] are those which are produced without a central blockage in the vocal tract. For example, fricatives have a central constriction, but there is no complete blockage of the air, and they are therefore, [+continuant].
Sounds which are [+nasal] are produced with nasal airflow.
Sounds which are [+lateral] are produced with airflow passing through one or both sides of the tongue, which is in contact with the
central part of the oral cavity.
These features, [LABIAL], [CORONAL], [DORSAL], and [RADICAL] are features that are often characterized as not being + or -, but rather, either a consonant has the feature or not.
What are Syllables?
In this lesson, the goal is to learn what syllables are, and the evidence that motivates them.
Maximum Onset Principle
The Beat Goes On
To begin with, you should be made aware of the fact that you, as a speaker of language, have some intuitions about what a syllable is. In fact, you may already have learned about syllables in your previous education.
In rather basic terms, a syllable is a timing unit for language. Words in language take certain amounts of time to utter. This time can be measured in terms of syllables.
Each of the preceding words had differing lengths, each measureable in terms of syllables. The first word is composed of one syllable, the second word is composed of two syllables, and the final word is composed of five syllables.
In the next sections, there will be a discussion of the structure of the syllable, and the motivations for considering the syllable as a phonological entity.
How do we know that syllables exist?
One reason is that we can count them. On the previous page, we talked about syllables as timing units. Language users can perceive those units and even count them. If I give you the word antidisestablishmentarianism, you know that there is more than one syllable involved. You even know that it's probably more than two or three. So, you have some perception of some words being perceptually longer than others in terms of syllables.
Here's another reason.
In English, we use an alphabetic system to write sounds. More or less. An strict definition of an alphabetic system is one in which one symbol refers to one sound. For example, the word mat has the following phonological representation: [mæt]. In this word, the symbol m can be said to refer to [m], the symbol a can be said to refer to [æ], and the symbol t can be said to refer to [t].
Of course, when we get to words like rough, show, and the like, this breaks down a bit. But you get the idea.
Other languages use a system in which one symbol refers not to a single sound, but a group of sounds. Observe the following from Cherokee, and Iroquoian language spoken in Oklahoma and North Carolina:
ᎦᏁᎵ ganeli 'married person'
ᎦᏚ gadu 'bread'
ᏁᏩᏓ newada 'hominy'
ᏑᎵ suli 'buzzard'
The word gadu has one of the same symbols as ganeli, the symbol Ꭶ. Both ganeli and gadu have the 'ga' combination. The words newada and suli also have one of the same symbols as ganeli, the symbols Ꮑ and Ꮅ . Further, they each have one of the same sound combinations as ganeli, 'ne' and 'li'.
Therefore, one can see that there is a match between a Cherokee symbol and a combination of consonants and vowels.
If we separate consonant and vowel sequences into syllables, then we can say that Cherokee uses a writing system in which one symbol represents a syllable.
In the next section, another piece of evidence regarding the existence of the syllable is discussed.
Observe the following data from English:
In each of these examples, there is a sequence in the middle of a word.
However, in English, there is no way to begin a word with the sequence . So, words like *tlap, *tling, etc. are not possible words of English.
Using the idea of syllables, can you explain why?
The next section will discuss a possible solution.
Why Can't English Words Begin with "tl"?
Well, we could just say that this is a special rule of English, but that would only restate the problem. We still would want to find out what the underlying reason is.
But, if we have the concept of syllables, then there is a possible solution.
As we noted earlier, [tl] is perfectly acceptable in words like atlantic and hitlist. It is also possible to state that, in each such case, the [t] and [l] are conceivable part of two separate syllables:
Atlantic [æt] [læn] [tik]
hitlist [hɪ] [tɪst]
In a hypothetical word such as [tlap], there is only one syllable, and both [t] and [l] are part of that one syllable.
So, perhaps the rule is that a [tl] combination can't be in the same syllable.
Such an explanation is only possible if syllables exist as phonological entities.
On the next page, there is discussion about the parts of syllables.
What's in a Syllable?
What sorts of things are in a syllable? Let's take a simple example to start with: cat.
Let's take that one step further and use IPA to represent the word cat: [kæt].
It's not very hard to determine that [kæt] has one syllable. In some varieties of English, the word cat is pronounced with two syllables, but that would have a different IPA representation. The pronunciation [kæt], however, has one syllable.
So, what's in it? Well, there's a [k], a [æ], and a [t]. If we simplify that a little, there is are two consonants with a vowel in the middle, or we have a CVC sequence, where C is a consonant, and V is a vowel.
Based on this, we know that a syllable can be composed of a consonant followed by a vowel, followed by a consonant.
What other structures can a single syllable have ? Think of different single syllable words.
What's in a Syllable?
Observe the following monosyllabic words and decide what kind of structure each word has:
Since each of these is monosyllable, they should represent a possible syllable.
Thinking in terms of just consonants and vowels for the moment, what are the possible syllable structures represented above (for example, the word tea [ti] represented the structure CV, since it is composed of a consonant [t] and a vowel [i].
What's in a Syllable?
In terms of consonants and vowels, the following words represent the following structures:
tea [ti] CV
a [ʌ] V
at [æt] VC
tree [tri] CCV
ask [æsk] VCC