3.3 Vowels

The goals of linguists, if you remember, are to describe what people know about their language and to figure out how languages are similar and different. For word forms, specifically for phonemes, this means that we must describe both how speakers and hearers distinguish phonemes within a given language and how individual phonemes and systems of phonemes differ between languages. To satisfy both of these goals, we will be looking for ways to describe the variation between speech sounds. Just as we saw that categories of things could be described in terms of values along different dimensions, we will be looking at dimensions of sound, dimensions that allow us to make distinctions between phonemes within and between languages.

Phonetic symbols

Before we discuss the specific phonemes of languages, we need to decide how to represent the sounds. Let's briefly consider English. Like other spoken languages, English has a set of vowel and consonant phonemes that its speakers use to make the words of their language. Like other written languages, English also has a way of graphically representing spoken word forms. The English writing system is an example of an alphabetic writing system in which phonemes are represented by characters or combinations of characters. But, for various historical reasons, the English writing system does this very imperfectly. Consider the words way, weigh, wait, and wake. These words share the same vowel phoneme, but it is spelled in four different ways. That is, a single phoneme may be represented using different letters or combinations of letters. Now consider the words bother, brother, border, and voter. These words share the letter o, but it represents four different vowels, each a different phoneme. That is, a single letter may represent multiple phonemes. We can conclude two things from this.

• We must be careful not to confuse sounds with letters; the letter o is not a vowel, though it is used to represent vowels.
• We cannot rely on English spelling when we are concerned with the pronunciation of English words.

Because we will need a way to represent the phonemes of English and other languages unambiguously, we must rely on a set of symbols for this that are not used quite like the alphabets of any alphabetic writing systems. Symbols representing the basic sounds, or phones, of spoken languages, are called phonetic symbols. Linguists use a set of phonetic symbols called the International Phonetic Alphabet (IPA). The symbols in the IPA are based on the characters in the Roman alphabet, which is also the basis for the writing systems of many languages, including English, Spanish, Lingala, and Tzeltal. The IPA is overseen by the International Phonetics Association. Here is a full chart of the IPA symbols that you can click on to hear some of the sounds. Note that you probably won't understand this table until after you've studied this section, the next one on English consonants, and the one after than on consonants in other languages. I will use a subset of the IPA symbols in this book. (To view many of these symbols, you will need Unicode support in your browser. Go to this appendix to see if your browser has this support.)

I will use phonetic symbols in two ways. Much of the time they will be used to distinguish the phonemes within a language. The important thing will be to make sure each phoneme has a unique symbol; which symbol we use is not as important. When the symbols are used in this way, they will be enclosed by slashes. As an example, my pronunciation of the word phonemes is /'fonimz/. I will also be using symbols in another way: to represent the actual sound that is produced in a given situation, rather than a category of sounds (a phoneme, that is), and to represent sound differences between different dialects and different languages. For this purpose we will have to be more careful about which symbols are used for which phones. When the symbols are used in this way, they will be enclosed in square brackets. As an example, my (more detailed) pronunciation of the word phonemes is ['foʊnɪimz]. It is important to note that the symbols are at best only an approximation to the actual phone that is used in any given situation because the space of possible phones is enormous, possibly infinite, while we only have a relatively small, finite set of phonetic symbols.

Vowel features

A complete account of the sounds of a spoken language (or of spoken language in general) would have to make reference both to the way sounds are produced, articulation, and the way they are perceived, their auditory properties. In this book we will not have time to go into the auditory properties of speech in any detail, but we cannot neglect them completely. Differences in articulation are obviously pointless if they aren't reflected in the way the results sound to hearers. In fact, two quite different ways of articulating a sound can sometimes produce the same auditory effect. From the perspective of a hearer, those would have to belong to the same phoneme.

In order to understand how vowels and consonants work, we need to know a little bit about the physical apparatus that is used to produce them. The figure below shows a side view of the vocal tract, with labels for some of the parts that we'll be discussing in this section and the next sections.

The heart of the system is the larynx, located in the throat under the Adam's apple; the larynx contains the vocal cords, two pieces of flesh that can be loosened or tightened. When air from the lungs passes through the larynx, it may be allowed to pass unhindered, as when we are breathing out. Or the vocal cords may be tightened and brought together so the air causes them to vibrate. This vibration, or voicing, is what distinguishes normal speech from whispering and certain speech sounds from others. Here is a slow-motion movie, produced at the UCLA Phonetics Laboratory, of the vocal cords vibrating during speech.

Given an outward airstream and vibrating vocal cords, we have a device that is a little like a brass musical instrument. The vocal cords are like the vibrating lips of the musician, and the region of the vocal tract between the larynx and the opening of the mouth (the yellow region in the figure), the oral cavity, is like the body of the horn. What makes the vocal tract far more versatile than a trumpet or a trombone, however, is that speakers can change the shape of the instrument as they are playing it, producing a great variety of sounds.

What distinguishes vowels from each other auditorily is the precise shape and volume of the oral cavity, and the main organ involved in adjusting the shape and volume is the tongue. Given the constraints on the shape of the vocal tract and the way the tongue is manipulated, there are three extreme positions the tongue can take that lead to vowel sounds. If the body of the tongue is pushed forward and toward the roof of the mouth, we get this vowel, something like the vowel in the word beat and written with the symbol [i]. If the body of the tongue is lowered while the back of the tongue is pushed toward the roof of the mouth, we get this vowel, something like the vowel in the word boot and written with the symbol [u]. Note that in both cases the tongue cannot approach the roof of the mouth too closely, or we get the sound of friction near the contact, that is, a consonant rather than a vowel sound. Finally, if the body and back of the tongue are both moved away from the roof of the mouth, at the same time narrowing the region that is between the larynx and the mouth (the "pharynx") we get this vowel, something like the vowel in the word hot (as pronounced by most Americans) and written in this book with the symbol [ɑ]. (The IPA symbol for this sound is the variant of a in the Roman alphabet that resembles this Greek letter.) The configurations of the vocal tract for these three vowels are shown in the figures below.

How can we turn this informal description of what is going on in the vocal tract into a more compact description that makes it clear how vowels differ from one another? First, [i] and [u] share one property; they are both associated with a relatively narrow gap between the tongue and the roof of the mouth (and a relatively wide pharynx). They differ in this way from [ɑ], which has the widest possible gap between the body of the tongue and the roof of the mouth (and a relatively narrow pharynx). Just as we spoke of dimensions along which categories of meaning varied, we will speak here of dimensions along which these sound categories vary. The dimension that distinguishes [i] and [u] from [ɑ] we'll call height. So far we have seen two different values on this dimension, one for [i] and [u] and another for [ɑ].

Another, equivalent way to talk about the difference between these vowels is in terms of features; [i] and [u] have the feature high, while [ɑ] has the feature low. A feature always corresponds to some value on a dimension. For example, with the domain of personal pronouns we saw that the gender dimension had possible values of masculine, feminine, and neuter; each of these can also be seen as feature. All of the things having a certain feature can also be viewed as a category; thus there is the category of masculine pronouns and the category of high vowels.

Dimensions can be discrete, that is, with a finite set of possible values, or continuous, that is, with an open-ended and possibly infinite set of possible values. Gender is a discrete dimension — there are only three possible values — while vowel height is a continuous dimension because we can theoretically put the highest part of the tongue at any point between the lowest and highest possible points. The fact that height is a continuous dimension means that the values high and low identify approximate points along the dimension and that a given vowel is only relatively high or low. But note that the fact that a dimension is continuous does not imply that people can distinguish all of the possible values; they cannot. For example, speakers can probably reliably produce and hearers can probably reliably distinguish no more than about five different vowel heights. Note also that what is a continuous dimension in the world may be treated as discrete within language (or elsewhere in cognition). Thus within a given language, as we'll see soon, only a small number of height distinctions are made.

The dimension of height distinguishes [i] and [u] from [ɑ], but it does not distinguish [u] from [i]. Looking at the diagrams of the vocal tract again, we can see that these two vowels differ mainly in terms of where the narrowest gap between the tongue and the roof of the mouth is. For [i] it is as far forward as it is possible for the tongue body to move and still produce a vowel sound; for [u] it is as far backward as it is possible for the tongue body to move and still produce a vowel sound. This dimension is called backness; again we have seen (so far) two possible values on this dimension: front, as for [i], and back, as for [u]. But note that, like height, backness is a continuous dimension, so front and back are labels for approximate regions along the dimension. For [ɑ], the closest point of approach between the tongue body and the roof of the mouth is near the back of the mouth, so this is also considered a back vowel. Summarizing, the two articulatory dimensions of height and backness represent the location in the mouth of the closest point of contact between the tongue body and the roof of the mouth.

If every vowel has a height value and a backness value, then we can visualize all of the possible vowels in a two-dimensional space known as vowel space. Given the constraints of the mouth and tongue, it turns out that vowel space has roughly the shape of a trapezoid; this is shown in the figure below, along with some of the terms used to refer to different regions within vowel space. Recall that these terms are only for convenience; there are really very many, maybe even an infinite number of, possible positions within this space.

Vowels of Spanish

We'll start with a relatively simple vowel system, that of Spanish. Spanish has five vowel phonemes, represented in the Spanish alphabet by the letters a, e, i, o, and u. The same characters are conventionally used to represent the vowels themselves. Listen to the vowels in these five Spanish words, and try to pronounce them yourself (here and elsewhere in this book, we will focus on Western Hemisphere, rather than European (Castilian), Spanish): piso, puso, peso, poso, paso. You may not be aware of the position of your tongue with respect to the roof of your mouth, but you should be aware of how the position of your tongue changes as you change from one vowel to another and how similar particular pairs of vowels are to each other in terms of tongue position. The figure below shows the approximate positions of the Spanish vowel phonemes in the vowel trapezoid. The positions are approximate because each phoneme is a category with a fairly wide range of possible pronunciations, depending in particular on which consonants appear before or after the vowel. You can click on each vowel in the figure to hear its pronunciation.

We can describe each of the Spanish vowels along the two dimensions of height and backness. For Spanish we need three values of height and two values of backness to do this. There is a further dimension that by itself does not distinguish any of the vowels, but it will be useful for us in discussing how languages and dialects differ from one another. This is the degree of lip rounding that accompanies the vowel. In Spanish /u/ and /o/ are the only vowels accompanied by significant lip rounding. For /i/ and /e/ the lips take the opposite position: they are spread. I will treat lip rounding and spreading as a single dimension. The figure below shows roughly where each of the Spanish vowels falls along the three dimensions of height, backness, and rounding. Each vowel is represented by three small circles of the same color, one for each dimension.

Another way to depict the differences is in a table that gives a label (feature name) for the value of each vowel on each dimension. I will use the symbol "∅" to indicate that a phone has no value or a default value for a dimension.

From the table, we can see that height is a contrastive dimension for Spanish vowels. That is, if we change nothing but the height of a vowel phoneme, we get a different phoneme: lowering the height of /i/ gives us /e/; raising the height of /o/ gives us /u/. The situation is a little more complicated for the other two dimensions because they tend to change together. Changing only the backness of a Spanish vowel phoneme does not give us a Spanish vowel at all. If we make /u/ a front vowel, we also have to make it spread to get /i/. If we make /e/ a back vowel, we also have to make it rounded to get /o/. If we make /a/ a front vowel, we also have to make it mid to get /e/. The same holds if we attempt to change only the rounding of a Spanish vowel. To distinguish all of the Spanish vowels, we actually need only either backness or rounding, but not both; that is, all we'd need to know about a Spanish vowel would be its values on two dimensions to know which vowel it is (verify this for yourself from the table). So we can call backness contrastive or rounding contrastive, or we can call the combination of backness and rounding contrastive. The point is that the system is redundant; it provides more information than is actually needed to make the distinctions.

Redundancy is common in language. We will see further examples with English vowels and in the grammar of various languages. Redundancy is a Hearer-oriented feature of language. When two or more different dimensions agree with one another, the Hearer can still extract the meaning even if they fail to note what is going on one of the dimensions. So with Spanish vowels, if the Hearer detects that a vowel is high and back but misses the fact that it is rounded, they can still know which vowel it is (/u/). The same would be true if they only detected that the vowel was high and rounded, but not that it was back.

Vowels of Japanese

The vowels of Standard Japanese are quite similar to those of Spanish. There are five phonemes centered on roughly the same positions in the vowel trapezoid as for Spanish. There are two differences. First, the high back vowel, /u/ in Spanish, is not normally rounded in Japanese; in Japanese it sounds like this. When we need to make it clear that it's an unrounded high back vowel we are discussing, we use the symbol [ɯ], but /u/ is conventionally used to represent this Japanese vowel phoneme.

A second difference between Spanish and Japanese is more significant. In Spanish, vowels tend not to vary too much in length, and how much they vary depends on the dialect. In any case, if the length of a vowel is changed in a word, we get the same word. So if a Spanish speaker pronounces the word peso with an extra-long /e/, the difference will probably be noticeable to a hearer, who might find it a bit odd but would not interpret it as a different word. In other words, vowel length is not contrastive in Spanish. Japanese differs in this regard. If we take a Japanese word with a short vowel and lengthen that vowel, we get potentially a different word. For example, su means 'nest', suu, with the same vowel, roughly twice as long, means 'number'; koke means 'moss', kokee means 'solid', kookee means 'succession'. (Note that ee and oo here represent long versions of e and o, not the sounds these letters would represent in English.) That is, vowel length is contrastive in Japanese. Note that this does not mean that changing the length of any vowel in a Japanese word must yield a different Japanese word, only that it potentially does. Thus mise 'store' is a word, but miise is not; miira 'mummy' is word, but mira is not.

There are two ways to analyze Japanese long vowels. We could consider each to be a separate phoneme, making ten vowel phonemes all together. Alternatively we could consider each long vowel to be two short vowels in succession. For our purposes, the choice doesn't really matter, and I will go with the second alternative (which is the usual way to treat Japanese vowels).

Vowels of General American English

The first thing to note about the English vowels is that they are a great deal more complicated than the Spanish or Japanese vowels. This complexity has several consequences. First, there is considerable disagreement among linguists on the details of how to describe English vowels, and even some of the brief comments I have to make about them below may be disputed by one or another linguist. Second, there are considerable differences between dialects, and this makes an overall description of English vowels difficult, if not impossible. Third, the English vowel system (within a given dialect) is not as stable as a simpler system like that of Spanish, and it is more likely to change over time.

In this section, we will consider only the vowels of General American English (GA), specifically the variety that I speak. (If you are not a speaker of GA, some of your vowels may differ from those we'll be describing. Many of these differences are described in the section on English accents.) Let's begin with the front vowels. These are the vowels in the following words, each followed by the symbol I will use to represent the vowel: beat (/i/), bit (/ɪ/), bait (/e/), bet (/ɛ/), bat (/æ/). The approximate positions of these vowels in vowel space are shown in the figure below. Again you can click on a symbol to hear the vowel.

You can see that five vowels are squeezed into the front part of the vowel space, the same number of vowels that Spanish has all together. If the differences in height (and minor differences in backness) were the only dimensions distinguishing these vowels, this would put quite a burden on the hearer, who after all has to figure out that beat and bit are different words. It would also be a burden on the speaker, who would have to be very careful in placing the tongue for each vowel.

So not surprisingly, these vowels are also distinguished from each other along other dimensions. We can put the highest four vowels in two groups, one consisting of the those traditionally called "long", /i/ and /e/, the other consisting of those traditionally called "short", /ɪ/ and /ɛ/. (The fifth vowel, /æ/, tends to group with the "short" vowels.) There are at least three differences between these two groups. First, the "long" vowels really do tend to be longer than the short vowels. Second, the "short" vowels tend to be "pure"; that is, the position of the tongue does not change during the production of the vowels. The "long" vowels, on the other hand, involve some movement, in both cases toward the upper left corner of vowel space. This is more noticeable for /e/ than it is for /i/ (but in other dialects, it is quite striking for /i/ too). Third, though you probably can't feel this yourself, for the "long" vowels, the base ("root") of the tongue is pushed forward. Finally, the muscles in the tongue are tenser than for the short vowels. Because values on the last two dimensions tend to go together, they are often considered to be a single dimension called tenseness. The two extremes on this dimension are referred to as "tense" and "lax".

So we again have redundancy, as many as five different dimensions distinguishing /i/ from /ɪ/ and /e/ from /ɛ/. It appears that speakers do not always make use of all of these dimensions to achieve the differences that hearers hear in them.

Now let's look at the central and back vowels phonemes of GA. These are the vowels in the following words, with the symbols for each after them: boot (/u/), put (/ʊ/), boat (/o/), bought (/ɔ/), pot (/ɑ/), but (/ʌ/). (Note: many GA speakers do not make a distinction between /ɔ/ and /ɑ/; more on this in the section on English accents.)

There is an additional central vowel, symbolized by /ə/, that appears in two quite distinct environments. (The IPA symbol for this sound is an upside-down e.) First, it is the most common vowel in unstressed syllables in English, that is, syllables that are weakly articulated. An example is the first phone in the word about and the last phone in the word comma. Some people consider this to be the same phoneme as /ʌ/; the main reason for treating it separately is that it behaves differently from /ʌ/ in the history of the language as well as in a number of modern dialects. Second, it appears followed by the consonant /r/ (in stressed and unstressed syllables). This combination is worth mentioning because speakers normally combine the two phones into one vowel sound. Examples are the beginning of the word earth and the end of the word actor. Again there is disagreement about how to treat the phone; some people would prefer to treat /ər/ as a single phoneme in the language.

The approximate positions of the back and central vowels in vowel space are shown in the figure below. Again you can click on a symbol to hear the vowel.

Again we find a large number of vowels squeezed into a region of the height-backness space, and, not surprisingly, the vowels differ on other dimensions. Like /i/ and /e/, /u/ and /o/ belong to the set of "long" vowels. Both share the features of the other long vowels: length, tendency to "move" as they are pronounced, and tenseness. /ʊ/, /ɑ/, and /ʌ/ are clearly "short" vowels, sharing features with the front "short" vowels /ɪ/, /ɛ/, and /æ/. /ə/ is "short" in its unstressed form, but in the /ər/ combination, it resembles the "long" vowels. /ɔ/ can also be relatively long, although it is usually lax, so it doesn't fit neatly into the division between "long" and "short" vowels.

There are three other "long" English vowels, all of them involving significant movement during their production. Vowels of this sort, in which the position of the tongue changes a lot during their production, are called diphthongs. The symbols for diphthongs indicate both the beginning and the end positions. Diphthongs are different from pairs of vowels in that one of the two parts is pronounced more forcefully than the other; for these English diphthongs the more forceful part is the first. The English diphthongs are exemplified in the words bite, bout, and boy; they are represented by the symbols /ay/, /aw/, and /ɔy/. (Note that the symbol at the beginning of the first two has not been used elsewhere for English, though it was used for Spanish and Japanese. It represents a low vowel that is further front than /ɑ/; we'll meet it again later when we discuss English accents.) The symbols [y] and [w] are used for the second members of the diphthongs; these correspond closely to the vowels [ɪ] and [ʊ]. We will meet these symbols again in the section on English consonants. The use of these consonant symbols helps to show that the part of the diphthong that is "stronger" is the first element. The figure below shows the paths that each of the diphthongs traces in vowel space. You can click on the beginning symbol to hear the vowel it represents.

Note that we have already discussed six dimensions for describing vowels: height, backness, length, rounding, tenseness, and whether the vowel height and backness change during the production of the vowel. I'll call this last dimension diphthongization. For GA, all we need to specify for this dimension is whether there is movement in the direction of a high front position ([i]), in the direction of a high back position ([u]), or not at all. (There are other possibilities in other English dialects, as we will see later.) Now we can describe all of the English vowels in terms of their values on these dimensions. The table below does this (for all of the dimensions except length). The point here is not to memorize any of this — there is, after all, considerable disagreement about the details — the point is to see how articulatory dimensions (features) allow us to begin to capture what speakers and listeners seem to know about the vowel phonemes of a language.

Here are examples of each vowel, again as pronounced in my accent. You may notice that the vowels do not sound exactly the same in different words. Since each phoneme is a category, we should expect there to be different variants in the same way that not all apples look alike. I'll have more to say about this variation within phonemes in the section on phonetic contexts.

Problems