Phonetics is the study of speech sounds. Although language is obviously composed of sound, speech sounds came to be the main focus of linguistic investigation only in the 20th century. 19th century linguists were more interested in written rather than spoken language. Only with the work of Swiss linguist Ferdinand de Saussure in the early 20th century did linguists recognize the primacy of sound in human language and the secondary, superficial nature of writing.

Phonetics has three branches:

1) Acoustic phonetics is the study of the physical properties of sounds, the air wave frequencies of which sounds consist. The frequency of vibrations measured in hertz; volume of sound measured in decibels. Instruments used to measure and record speech sounds include the sound spectrograph, which produces readouts called sound spectrograms.

2) Auditory phonetics is the study of how sounds are perceived by the human ear or recognized by the brain. (Mention Oronyms, Mondegreens.)

3) Articulatory phonetics is the study of how sounds are produced by the vocal apparatus. Articulatory phonetics will be the main focus of our study. You need to memorize the anatomical features of the vocal tract shown on the back cover of your textbook.

The flow of sound during any given speech act can be divided into units of sound that recur in the flow of other speech acts. In phonetic transcription using the IPA, these separate sounds, called phones, are customarily enclosed in square brackets to distinguish them from letters: [o] = the sound of the vowel in the word rope, not the letter o.

The division of the speech continuum into separate sounds, or phones, is made on the basis of both articulatory and acoustic data. Auditorily, the sounds seem to the ear to change at a particular point. Acoustically, this change can be seen on a sound spectrogram. Auditorily, the change of one sound to another is the result of measurable changes in the movement of the speech organs.

Data from all three branches of phonetics have revealed another important point about speech sounds. The phone is not the smallest formal unit in language. Listen to the phones [v] and [u]. Auditorily, we perceive a single unit of sound. Articulatorily, however, more than one part of the vocal apparatus is at work making the sound (cf. f/v or i/u). This fact only became apparent during the work of the Prague School linguists in the earlier part of this century. Nikolai Trubetskoi noticed that each speech sound or phone is a complex entity. He called each component of a sound a phonetic feature (voicing, lip rounding are phonetic features). These features are produced more or less simultaneously rather than consecutively, which is why we tend to hear them all together as a single unit. After the Second World War, when sound spectrographs came into general use, the phonetic features posited by the Prague School linguists actually became visible. They appear as separate formants in the spectrogram of each sound ([v] has voicing and high static; [u] has a higher first formant and lower second formant, [i] has a higher second formant and lower first formant).

Thus, not all sounds are equally different, some share virtually no phonetic features, while others differ in only a single feature. (For instance [f] differs from [v] in only one feature, whereas [m] and [h] differ in several features). Sounds sharing a given phonetic feature can be grouped together into what are called natural classes (vowels, consonants, nasals, voiced sounds, etc.).

It has been found that a relatively small number of phonetic features were all that was necessary to describe all the sound in any given language (See textbook pages 235 and 236 for a sneak preview of the phonetic features which form the sounds of English. Notice that there are less than half the number of phonetic features as there are sounds in English. Phonetic features are usually based on articulatory and sometimes acoustic details (strident vs. non-strident).

One of the Prague linguists, Roman Jakobson, was intrigued by how few phonetic features there seemed to be in relation to speech sounds in the world's languages. He tried to come up with a set of universal phonetic features that could be used as tools to describe any sound in any language of the world. His theory of universal phonetic features failed, however, because as more languages were described, the number of features kept increasing.

In this class, we will concentrate on the natural classes of sounds rather than on the phonetic features. We will also learn the IPA alphabet, a special phonetic alphabet linguists use to write speech sounds.


Before launching into our discussion of articulatory phonetics, it would be useful to discuss the IPA and its use for transcribing English.

Writing uses graphic symbols to represent speech sounds. All systems of writing in use today represent the sounds of language. This connection can be viewed in two ways.

a.) From sound to writing. Every written language has a system of rules for how to spell the spoken word. These rules are called the orthography, the writing rules of the language.

b.) From writing to sound. Every written language has a system of rules for pronouncing correctly what is written. These rules are called orthoepy, or pronouncing rules.

Some writing systems come close to achieving a one-to-one correspondence between sound and written symbol. In Spanish, for instance, the rules of orthography and orthoepy are extremely simple. Once you have learned the Spanish alphabet, spelling and reading Spanish words is relatively unproblematical. Nearly every letter has one basic pronunciation, and most basic sounds can be written with one and only one letter (exceptions being the sound [s], which may be written s, c, z). The Georgian writing system comes even closer to perfect economy. There are 33 different sounds in Georgian and only 33 different letters in the Georgian alphabet. Whenever you hear a new Georgian word, you know exactly how to write it. And whenever you meet a new word in reading a Georgian text, you know exactly how it should be pronounced; and whenever you hear a new Georgian word pronounced, you know exactly how to write it.

Other languages, such as English, show far less economy of correspondence between sound and written symbol. The English alphabet has only 26 letters. However, English spells one and the same sound with several different letters depending upon the word. Consequently, one letter or group of letters can be read in any one of several possible ways, such as the gh in rough, ghost, night. The correspondence between sound and letter in English frequently involves meaning as well as sound. Often, one must know the meaning of an English word to pronounce or write it correctly: meet/meat; through/threw.

There are hundreds of alphabets and writing systems in use in the world. Many are based on the Latin alphabet, but the system of sound to symbol correspondences are different in each one. (Example, varieties of the Latin alphabet are used for most European languages, but the values of letters is often different: j in Spanish, English, and Hungarian.) Here are some more striking examples of the variety of writing systems:

English horse apple

Russian lowad; qbloko

Georgian cxeni vawli

Navaho 풠 bilasana

Cherokee 񠠠

To alleviate the confusion of having so many different writing systems, phoneticians have devised a single, unified phonetic alphabet capable of representing virtually any sound in any language of the world. This alphabet is called the International Phonetic Alphabet, or the IPA. The IPA tries to and comes close to achieving the ideal one-to-one correspondence between sound and written symbol-- not only for a single language, but for all languages. This allows linguists to write about different languages without studying their particular writing systems.

The need for a universal writing system finds an analogy in the development of standardized scientific names for plants and animals. Zoologists and botanists use a universal system of names called scientific names to denote genus and species. This system of scientific names, developed in the mid 18th century by the Swede Carl Linnus, is based mostly on Greek and Latin (the Latin Equus denotes the genus "horse" and Malus denotes the genus"apple".) Because it is universally accepted, this system of scientific names is understandable to scientists all over the world regardless of their native language.

Why do we need the IPA?

Because you will be expected to spend so much time learning and working with the IPA in the next two weeks, I feel the need to explain even more clearly why it is important that you do so.

You may ask: Why do we need a special alphabet like the IPA? Why can't we simply use the 26 letters of the English alphabet more systematically to sound out English words as well as words of other languages?

1) Well, there is a far greater number of different sounds in the world's languages than there are letters in the English alphabet. Let's transcribe the sounds given above.

English [hOrs] [pl]
Russian [lOSt'] [jAblk]
Georgian [tsxEni] [vASli]
Navaho [:i!≥] [bilAsA!nA]
Cherokee [sokwili] [s&tA]

Notice that to avoid confusion between the IPA transcription and other writing systems, IPA transcription is enclosed in square brackets. We write with the English alphabet; we transcribe with the IPA.

If linguists didn't have the IPA, they would have to resort to using the writing system of each individual language, as we did earlier. This would be quite difficult. There are 14 major writing systems in use, and some, such as the Latin alphabet, have hundreds of variants each using a somewhat different set of correspondences between sound and symbol. Also, many hundreds of languages are still not written at all.

Notice that an unmodified English alphabet could not be used to transcribe the many non-English sounds that occur in other languages. Using English letters to convey words of other languages is awkward and has in the past resulted in gross distortions of the pronunciation of foreign words: Peking, Ceylon, Cherokee.

To see how difficult it can be to write non-English words with the English alphabet, try to write the following words using English letter combinations: Georgian tsqali, water, ghori pig; Sahaptin huq'huq' pig; Navaho l', horse; Cherokee vtla, no.

2. The second question you may ask is this: While it is obvious why we need an IPA to transcribe other languages, can't we at least just use the English alphabet more systematically to write English? In fact, there are quite a few more sounds in English than there are in the English alphabet. There are 26 different letters, but something like 34 different sounds.

Obviously, we need some sort of phonetic alphabet to explain the pronunciation of the extra sounds in English. We have seen that:

a) One English letter is often equivalent to more than one different sound:

fat/father/fate book/ boot this/thistle

b) The same sound may be represented by different letters

ship/education/fissure oat/note/owe

And the problem with English orthography and orthoepy cannot be solved simply by basic education. Even most highly educated people would need to be warned how to pronounce such toponyms as Puyallup, Sequim, Spokane. Or last names such as Vajda. A phonetic alphabet is necessary for dictionaries and guides to help English speakers pronounce their own language.

Furthermore, a standard phonetic transcription is essential when describing different English dialects. Not every speaker of English pronounces the same word in the same way.

cat/cot/caught father/farther bird

It is obvious that some sort of phonetic transcription is necessary as a guide to pronouncing unfamiliar English words as well as a means of transcribing English dialects.

Go over English vowels and consonants on pp. 208-9.

(Although we will discuss un-English sounds found in languages, you will only be responsible for knowing the IPA symbols used in English. These are listed on pp 208-209 and again on the inside cover of you textbook. Notice the symbols for the vowels in cot/caught.)

Notice that, like the system of scientific names, the IPA also uses Latin as a base, adding extra symbols or diacritic marks for various modifications of the basic sounds. Because each IPA symbol stands for one basic sound regardless of the language being transcribed, it differs from the actual Latin alphabet, which often use the same symbol for different sounds in different languages (cf. "j" in Engl., Sp, Fr., Germ. Hungarian). Although it looks like English or Latin, the IPA is not truly biased in favor of English or of any other language. The phonetic transcription of English words may also differ radically from conventional English spelling: cf. the transcription of "thigh," "phrase."

NOTE: Before the IPA gained wide acceptance, American linguists employed a somewhat different set of symbols to transcribe English. The difference between IPA symbols and those used by many American linguists for several sounds can be found on p 185. The symbol used in your book for the sound [r] also differs from the IPA. You may use either symbol for these sounds, but the IPA variants are best.

Also, your book mixes dialects in its transcription. (Explain [O] and [A]; [A] and [a].) You should transcribe English words as you pronounce them in your dialect.

Although some of the IPA symbols are not yet accepted by all linguists (witness the choice of symbols for several American sounds), and although new sounds will no doubt be discovered that require new symbols, the IPA is still the most versatile writing system ever devised. Therefore, all linguists and linguistics students should learn and use it.

Limitations of the IPA:

Aside from the temporary, logistic problems of getting everyone to use the IPA, there is another more intrinsic limitation inherent to this or any other universal writing system.

At the present time it is true that all sounds which contrast to reflect differences in meaning in any known language are represented in the IPA. It is also true that new symbols can easily be added to accommodate the discoveries of new sounds.

The permanent, intrinsic problem with the IPA is that real sounds are infinitely variable. Sound spectrograms show that a single sound varies slightly each time it is pronounced. Thus, the IPA could only capture a part of each individuals unique accent, or way of speaking (called an idiolect)

The differences here, of course, are not meaningful. But when comparing sounds across languages, the minute differences become more important. What seems to be the same sound in different languages may actually vary in minute degrees. (cf. the degrees of aspiration in the sound "t" in English, German, Georgian, Mongolian; or the degree or type of glottalization in the sound "t" in Georgian, Navajo or even in English hatbox.. Even [m] differs in English and Russian in slight ways that the IPA does not distinguish. Thus, the IPA would not be able to transcribe all the phonetic detail of, say, a Russian accent in English.

The IPA ignores minute differences between sounds if those differences never contrast with one another in any single language. The IPA symbols, therefore, are generalizations. The sounds of speech, however, are more complicated.

Thus, when comparing nearly identical sounds from language to language one must be aware that the IPA is only accurate up to a point. To get a complete picture of native pronunciation, one must fill in the tiny phonetic details left out by the IPA. There is no real solution for this problem. One could invent a slightly different symbol for all the different examples of, say, aspirated "t" or glottal "t," but inventing different symbols for extremely similar sounds in each language would defeat the very purpose for which the IPA was invented. The ultimate uniqueness of sounds in each individual language renders the IPA an imperfect and incomplete alphabet, despite the fact that it is the closest thing to an ideal alphabet we will probably ever have. This inevitable problem derives from the very nature of speaking and writing. Sound can only be depicted with discrete graphic units to a certain degree of precision. The only thing more complete and accurate would be the use of sound spectrograms instead of symbols to denote actual sounds.

Because similar sounds are in reality minutely variable from language to language, the IPA is not a truly perfect alphabet. Because of the nature of language, no alphabet can truly be perfect--the more so if applied across languages. With a finite set of symbols an alphabet can only approximate the infinite number of actual sound variations in speech. Once again, a universal linguistic descriptive device cannot claim to be as accurate and universal as Mendeleyev's Periodic Table of Elements, where a finite number of units (the elements) actually can be used to describe an almost infinite number of combinations. This discrepancy, once again, illustrates the unique nature of human language.

Even if linguists devised a complete set of transcriptions symbols--perhaps based on sound spectograph readings rather than on the Latin alphabet--the phonetician's task would still be incomplete. This is because languages change over time, and new symbols would have to be devised as sounds changed. All human languages change over time--some quickly, others more slowly. Language is changing even now: your grandparents said [hw]; now most of you say [w]. The reasons for such changes are not always clear: language mixing (French borrowings caused the sounds v and z to become initial consonants; f into Slavic from Greek after 988). Languages isolated from contact with other languages tend to change more slowly, but change they do. Not all language change can by any means be traced to the interaction with different languages ([hw] to [w], for instance) [sk] became sh in Engl. from 5-6 cent AD. Words borrowed during that time, such as Latin disk, become dish. After 600 AD the rule stopped working, the new [sh] remained, but new words coming into the language retained sk. Large number of Danish borrowings occurred circa 700AD: sky, skull, ski. The Danish intrusion resulted in some interesting doublets: ship/skipper, shirt/skirt, (meanings change over time also).