The written "g" is part of a digraph "ng" for the velar nasal that is more properly transcribed [h] in the phonetic alphabet. (One reason to have special symbols like this is to avoid the confusion of things like the "g" in "ng".)

Your pronunciation will differ in some ways from that of your friends or the instructor. This is generally due to difference in regional dialect or sometimes a matter of age.

For example, for a dwindling number of English speakers, the two words in the name of this board game are distinct -- "wh" is voiceless, while plain "w" is voiced. That's a distinction that goes back to Old English and earlier. But for most speakers today (including me), they're homophonous and should both be transcribed with voiced [w].

In homeworks and exams, as long as you give an accurate transcription of how you pronounce something, you'll get full credit. If you think I've taken off credit unfairly for something like this, tell me! I'll ask you to pronounce a word for me, and decide on that basis whether your transcription is in fact a plausible transcription of the way you speak.

Phonology: the structure of sound

Recall the basic distinction mentioned earlier.

• phonetics: the physical manifestation of language in sound waves; how these sounds are articulated and perceived
• phonology: the mental representation of sounds as part of a symbolic cognitive system; how abstract sound categories are manipulated in the processing of language

Thus phonetics refers to the physiological and acoustic parts of the following diagram, while phonology resides in the brain.

From The Speech Chain

Now we'll focus on the more abstract side of things, and how sounds are organized in a particular language.

Phonemes

The phonological elements of a language are the basic, distinctive sounds, also called phonemes. In English, these are the following (for a dialect of Standard American English).

• consonants: p, t, k, b, d, g, č, , f, θ, s, š, h, v, ð, z, ž, m, n, ŋ, l, r, w, y
• vowels: i, u, I, U, e, o, , ə, :, æ, a, ay, aw, oy

These sounds are said to be "distinctive" because they can be used to make contrasts between different words. This can be illustrated for the stops, using minimal pairs (words that differ in exactly one sound).

pill
till
kill

bill
dill
gill

And for the vowels (We can't get an exact minimal set for the entire range of vowels in the context [h_d], so in some cases the initial consonant also differs. For each individual pair of vowels, however, we could come up with a minimal pair.):

heed
who'd
hid
hood
aid
ode
head
HUD
awed
had
odd
hide
how'd
Boyd

And for the nasals:

rum
run
rung

In English, the velar nasal [h] can't occur at the beginning of a word -- cf. map, nap, *ngap -- which will lead us to the next issue, the way these elements are organized into words.

But first, note that a basic way in which languages differ is their inventory of sounds, or phonemes. For example:

• German has the voiceless velar fricative [x], as in Bach "creek".
  • English has voiceless fricatives such as [s] and velars such as [k], but it doesn't have a single phoneme that has both of these properties.
• German also has the high front rounded vowel [ü], as in kühn "clever".
  • Again, English has high front [i] and rounded [u], but these properies are not combined in one vowel.
• English [θ] sets it apart from many languages, including German and French.
  • They have several voiceless fricatives, but not the interdental.

When you learn a new language, one of the things you have to do is learn the "list" or inventory of sounds. That's what children have to do also, when learning their native language.

Syllables

The phonological structure of a language -- the way these elements are organized -- includes the notion of syllable and its subparts. This structure is crucially involved in describing the possible words of a language.

• the onset or consonant(s) at the beginning of the syllable
  • English normally permits up to two consonants
  • but in addition, [s] can be added to the beginning of many syllables as well, making up to three consonants
  • all sounds can occur in this position except for [ŋ]
• the nucleus or vowel that is the core of the syllable
  • sometimes a consonant can serve as the nucleus, as in the second syllable of kitten.
• the coda or consonant(s) at the end of the syllable
  • English normally permits up to two consonants at the end (belt, jump, arc)
  • but in addition, certain sounds such as [s, t, θ] can be piled up (belts, sixths)

Here's a general schema of how syllables are constructed.

The category rhyme simply brings together the nucleus and the coda, so the rhyme part of the syllable blend is the nucleus [E] and the coda [nd]. The reason for this name should be obvious: in order for syllables to rhyme, what has to match is just this part of the syllable -- trend, end, spend, etc. (In longer words, rhyme is defined as matching this part of the stressed syllable and all the way to the end of the word: flower, power, shower, tower, hour, scour, etc.)

Sonority

Human speech, like many animal vocalizations, tends to involve repetitive cycles of opening and closing the vocal tract. In human speech, we call these cycles syllables.

A syllable typically begins with the vocal tract in a relatively closed position -- the syllable onset -- and procedes through a relatively open nucleus, then closing again while approaching the coda or the next syllable's onset. The degree of vocal tract openness correlates with the loudness of the sound that can be made.

Speech sounds differ on a scale of sonority, with vowels at one end (the most sonorous end) and obstruents (stops, affricates, fricatives) at the other end. In between are the liquids [l] and [r], and nasal consonants like [m] and [n].

Languages tend to arrange their syllables so that the least sonorous sounds are restricted to the margins of the syllable -- the onset in the simplest case -- and the most sonorous sounds occur in the center of the syllable -- most often a vowel.

Here are some typical English syllables that illustrate this pattern.

"soon"

"blend"

				*
		*		*
		*		*		*
*		*		*		*		*
[b]		[l]		[]		[n]		[d]

And in "pretending" each syllable corresponds to a peak in sonority.

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

[p]

[r]

[ə]

[t]

[]

[n]

[d]

[I]

[ŋ]

As a consequence of this sonority requirement, an English word such as film is one syllable:

But if we try to reverse the last two consonants, the hypothetical word fiml comes out as two syllables, since [l] is a new peak, higher in sonority than the preceding nasal. (This new word would end just like pummel.)

Similarly, if we change the [l] in film to an obstruent such as [z] in hypothetical fizm, once again we end up with a new syllable. (It would rhyme with prism.)

These syllabifications aren't something we need to learn for each word: they're a general property of the language. That's why we know how these hypothetical words would be pronounced.

In these last two words, the consonant serves as the sonority peak at the end of the word. The consonant is syllabic, serving as the nucleus in the absence of a vowel. English permits nasals and liquids to serve in this way, at least in unstressed syllables.

prism, bottom, sump'm (for "something"), cap'm (for "captain")
hidden, button, kitten, risen

bottle, little, towel
swimmer, higher, butter

For [r], the consonant can function as a vowel even in a stressed syllable.

bird, fur, word

In some dialects, such as Standard British, Boston, and Coastal Southern US, any [r] in the rhyme of a syllable (whether nucleus or coda) loses its r-ness and becomes a schwa-like vowel. These are called "r-less" dialects.

Another general property of English is that there are restrictions on what consonants can serve as an onset cluster -- i.e. the string of (two) consonants at the beginning of a syllable. It's not enough for the sonority to increase from the first consonant to the second: it has to increase by two steps.

• actual words with obstruent + liquid (two steps)
  • brick, true, free, crab; play, blue, flea, glib
• possible words with obstruent + liquid
  • blick, clee
• impossible words with obstruent + nasal (just one step)
  • *bnick, *fnee, *gmue, *dmay
• historical loss of initial consonant in obstruent + nasal (letter now silent)
  • knee, knight, gnat, gnaw

This too is part of our general knowledge of the language: we can distinguish blick and *bnick as "possible" and "impossible" even if we've never heard either word before.

But what about words like snow, with an obstruent + nasal onset cluster? You can take any ordinary English onset, and (subject to some restrictions) tack an [s] on the front of it, completely ignoring sonority. This includes clusters of two consonants that obey the general rule; if the first of these is a voiceless stop, [s] can be added to make three consonants.

snow (cf. no)

stop (cf. top)

spray (cf. pray)

This is a special property of [s] and no other obstruent in English. Essentially, it's because [s] is a perceptually salient sound with loud fricative noise: it doesn't depend in the normal way on syllable structure. Many other languages give similar special treatment to [s] and related sounds; in German (and Yiddish), for example, it's the (alveo)palatal fricative, as in Schmutz "dirt."

Once again, syllable structure is a way in which languages differ.

Hawaiian, for example doesn't allow any coda consonants at all, and a maximum of one consonant in the onset. This means that borrowed words get a lot of extra vowels, to create new syllables of the proper type.

ink > 'înika
Norman > Nolemana

Polish, on the other hand, allows more consonants at the beginning or end of a word than English does. This is why some Polish names are hard for English speakers to pronounce, such as Gdansk or Zbigniew Brzezinski.

bzdura "nonsense"
babsk "witch"
grzbiet [gzhbyet] "back"
marnotrawstw [-fstf] "of wastes"

A language learner, when exposed to lots of examples of words and syllables in a new language, comes to understand what structures are possible in that language by observing the attested patterns.

Allophones

There are often differences in the way a phoneme is pronounced in a specific context. The variant pronunciations are called allophones ("other sounds").

When it's important to make this difference:

• we'll use [square brackets] to indicate sounds from a phonetic point of view, i.e. focusing on their physical properties and the details of actual pronunciation;
• and we'll use /slashes/ to indicate sounds from a phonological point of view, i.e. as part of an abstract representation independent of potential differences in the way the sound in pronounced in specific contexts.
• In other words, in the ideal case, [ ] = allophone, / / = phoneme.

A classic example of sound alternation in English, which I mentioned in the first lecture, relates to the [s] found at the beginning of a syllable before a voiceless stop.

Although a word like spin is basically pin with [s] added, the /p/ in each case is pronounced differently.

• pin contains an aspirated version of /p/, with a puff of air after the stop is released; this is written [p^h]
• spin contains a plain /p/, without a puff of air after the stop; this is written just [p]

The same is true for pairs like pit~spit, pot~spot, pair~spare, etc.

A simple statement of this alternation is as follows:

But the same generalization holds not just for /p/ but for the other voiceless stops, /t/ and /k/. Compare these word pairs:

• top~stop, take~stake, tie~sty, etc.
• kin~skin, cope~scope, can~scan, etc.

So more accurately, there's a single general statement that covers all these cases, stated in terms of natural classes.

The aspirated and unaspirated versions of the voiceless stops are in complementary distribution: each occurs in its own context, which does not overlap with the contexts of the other.

The rule stated here assumes words of one syllable only. The full statement of where aspiration occurs in English is more complex: voiceless stops are aspirated when they occur syllable-initially and are followed by a stressed vowel (rápid, rap^hídity); as well as word-initially regardless of stress (p^hot^háto). At the beginning of a word, a preceding /s/ prevents the stop from being syllable- or word-initial.

If related words (containing the same morpheme, or meaningful element) have different stresses, then we often find alternations in whether the same underlying sound /t/ is pronounced phonetically as plain [t] etc. or aspirated [t^h] etc.

This process is completely unconscious for most speakers, and often quite hard to unlearn.

English speakers who learn a language like French or Spanish, in which all voiceless stops are unaspirated, typically impose aspiration according to their native rule; but that's wrong for these languages, and sounds foreign.

Similarly, a French or Spanish speaker learning English will typically fail to produce aspiration in the right places; this is part of what it means to have a foreign accent.

Aspiration in English is a small example of what phonological knowledge consists of:

• it's learned unconsciously by children imitating (quite accurately!) the details of the language around them
• it's systematic, applying to all words with voiceless stops, not just some random selection
• it's defined in terms of a natural class (here "voiceless stops") rather than some arbitrary set of three consonants

The study of phonology is largely the investigation of alternations like this -- what changes occur, what sounds undergo them, and in what contexts.

Flapping

A prominent feature of American English affects /t/ and /d/, and is called flapping. A flap is a quick motion with the tongue, in this case against the alveolar ridge. It's similar to the /r/ of Spanish in a word like para, although it's a separate phoneme in that language.

All these English words have flaps where "t" or "d" is written in the spelling (in the relevant dialects).

The proper phonetic symbol for a flap is - it's an "r" missing the top left serif.

For most speakers, in the right context a sound that is phonologically /t/ will end up sounding phonetically just like one that is phonologically /d/, since both become a flap []. Though /t/ and /d/ are distinguished by voicing, the flap [ ] is voiced.

Thus these words are all homophonous in flapping dialects, i.e. they're pronounced the same.

And the answer in this exchange is therefore ambiguous:

-- Do you want the ladder or the chair?

-- Give me the [lær].

It's possible for a more emphatic pronunciation such as [læt^hr] to avoid the ambiguity; but that's not the usual pronunciation.

Of course, /t/ and /d/ don't always end up as flaps. In many contexts they're distinct, as these minimal pairs illustrate.

The question, then, is what context causes flapping to occur.

There are two conditions:

1. The /t/ or /d/ has to be between vowels (this includes a syllabic [r] or [l])
   • so not in hit, melting
2. The following vowel has to be unstressed.
   • so not in tin, attain

If you compare the list of homophones (with flapping) vs. minimal pairs (without flapping), you'll see that only the homophones satisfy both these conditions -- and so flapping occurs, which is what makes them homophones.

The same basic word (or word root, or morpheme) will sometimes undergo flapping, sometimes not, as the context changes. This includes adding a vowel:

sit [t]	sitting, sitter [ ]
spot [t]	spotty [ ]
mad [d]	madder, maddest [ ]
bird [d]	birdy [ ]

As well as moving the stress (primary ´ or secondary `):

atómic [th]	átom [ ]
còmputátion [th]	compúter [ ]
prágmatìsm [th]	pragmátic [ ]
addíctive [d]	áddict [ ]
edítion [d]	édit [ ]

(If the /t/ is syllable- or word-initial, then it's also aspirated, as we should expect.)

And when English borrows a new word, the new word is subject to these patterns too, regardless of the exact situation in the original language.

tofu [th]	(Japanese [t])
tortilla [th], [th]	(Spanish [t], [t])
coyote [kh], [ ]	(Spanish [k], [t])
condor [kh]	(Spanish/Quechua [k])
panache [ph]	(French [p])

These facts show that aspiration and flapping are active parts of our knowledge of English; it's not just something we learn about individual words (such as flapped latter), but rather something that we know about the language.

One of the most interesting things about flapping is how it interacts with a process affecting the diphthong /ay/ in many dialects of English. When followed by a voiceless consonant, the diphthong is "raised" so the first part is more like the first vowel of mother than that of father.

regular diphthong at the end of a word, or before a voiced consonant

Ø, b, d, v, z, , ð, m, n, l, r
tie, jibe, hide, live, rise, oblige, tithe, time, line, tile, tire

raised diphthong before a voiceless consonant

p, t, k, f, s
hype, white, bike, life, ice

What is interesing is that this distinction between [ay] and raised [əy] is maintained even when the voicing distinction is eliminated by flapping. Thus if a speaker has raising in write, that pronunciation is maintained in writer, while rider will have [ai] just like ride.

Thus even though flapping eliminates the distinction between the consonants in these two words, they still do nor rhyme.

Most speakers are aware that the two words are pronounced differently, but they think that the difference lies in the consonants. In fact, as far as the actual sounds produced are concerned, the difference lies entirely in the vowels.

The reason why speakers "hear" the difference in the consonants is because, on an abstract level in their minds, the words are represented as /rayter/ and /rayder/, with the difference localized in the consonant. The raising of the /ay/ in the former, and the flapping of the consonants in both are subsequent unconscious processes.

What we must ask ourselves, then, is how raising keeps working properly to distinguish these two words when the conditioning factor on raising -- voicing on the following consonant -- has been obliterated.

We might imagine that speakers raise the /ay/ in writer on analogy to write, where the conditioning factor is still intact. However, this leaves it as mysterious why they treat the two words differently for the purposes of flapping. If phonological rules are simply steered by the properties of the basic root, then we would expect flapping in writer to fail because it fails in write.

Instead, it seems like what is happening is that speakers have an abstract representation of a word in their minds, and they apply phonological rules to these representations in some order, so that the output of one rule can be the input to another:

• writer

  /rayter/ --> raising --> /rəyter/ --> flapping --> /rəyer/

• rider

  /rayder/ --> raising /rayder/ --> flapping --> /rayer/

What does phonology do for us?

So we've gotten an idea now of how the sounds of language are produced, how they are classified, and how they fit together into systems. An interesting question to ask is why it should be as it is. In this last section we will consider this issue and see that the phonology of human language is an ingenious solution to a serious problem.

Apparent design features of human spoken language

We can start by listing a few characteristics of human spoken languages:

1. Large vocabulary: 10,000-100,000 items
2. Open vocabulary: new items are added easily
3. Variation in space and time: different languages and "local accents"
4. Messages are typically structured sequences of vocabulary items

1. Small vocabulary: ~10 items
2. Closed vocabulary: new "names" or similar items are not added
3. System is fixed across space and time: widely separated populations use the same signals
4. Messages are usually single items, perhaps with repetition

1. Vocalizations communicate individual identity
2. Vocalizations communicate attitude and emotional state

1. Easy naming of new people, groups, places, etc.
2. Signs for arbitrarily large inventory of abstract concepts
3. Language learning is a large investment in social identity

How can it work?

Experiments on vocabulary sizes at different ages suggest that children must learn an average of more than 10 items per day, day in and day out, over long periods of time.

A sample calculation:

• 40,000 items learned in 10 years
• 10 x 365 = 3,650
• 40,000 / 3,650 = 10.96

Most of this learning is without explicit instruction, just from hearing the words used in meaningful contexts. Usually, a word is learned after hearing only a handful of examples. Experiments have shown that young children can learn a word (and retain it for at least a year) from hearing just one casual use.

Let's put aside the question of how to figure out the meaning of a new word, and focus on how to learn its sound.

You only get to hear the word a few times -- maybe only once. You have to cope with many sources of variation in pronunciation: individual, social and geographical, attitudinal and emotional. Any particular performance of a word simultaneously expresses the word, the identity of the speaker, the speaker's attitude and emotional state, the influence of the performance of adjacent words, and the structure of the message containing the word.  Yet you have to tease these factors apart so as to register the sound of the word in a way that will let you produce it yourself, and understand it as spoken by anyone else, in any style or state of mind or context of use.

In subsequent use, you (and those who listen to you speak) need to distinguish this one word accurately from tens of thousands of others.

Note that the perceptual error rate for spoken word identification is less than one percent, where words are chosen at random and spoken by arbitrary and previously-unknown speakers. In more normal and natural contexts, performance is much better.

Let's call this the pronunciation learning problem. If every word were an arbitrary pattern of sound, this problem would probably be impossible to solve.

So what makes it work?
 

The Phonological Principle

In human spoken languages, the sound of a word is not defined directly (in terms of mouth gestures or acoustic wave patterns). Instead, it is mediated by encoding in terms of a phonological system:

1. A word's pronunciation is defined as a structured combination of a small set of elements
   1. The available phonological elements and structures are the same for all words (though each word uses only some of them)
2. The phonological system is defined in terms of patterns of mouth gestures and noises
   1. This "grounding" of the system is called phonetic interpretation
   2. Phonetic interpretation is the same for all words

How does the phonological principle help solve the pronunciation learning problem? Basically, by splitting it into two problems, each one easier to solve.

1. Phonological representations are digital, i.e. made up of discrete elements in discrete structural relations.
   1. Copying can be exact: members of a speech community can share identical phonological representations
   2. Within the performance of a given word on a particular occasion, the (small) amount of information relevant to the identity of the word is clearly defined.
2. Phonetic interpretation is general, i.e. independent of word identity
   1. Every performance of every word by every member of the speech community helps teach phonetic interpretation, because it applies to the phonological system as a whole, rather than to any particular word.
   2. Speakers of different dialects will have somewhat different phonetic interpretation of the phonological units, but because they are all dealing with essentially the same units, once you learn the basics of a different dialect's phonetic interpretation, you can learn new words from speakers of that dialect, and interpret them phonetically in your own.

Additional Online resources

These are not a required part of the course materials, but are presented for those students who are interested in further information.

• The Phonological Atlas of North America.
• Web site of the International Phonetic Association.
• On-line spectrogram reading tutorial (a spectrogram is a kind of visual display of the acoustic information in speech).
• An annotated demonstration of vowel synthesis.
• A self-study course at Sterling University on the phonetics and phonology of "received pronunciation" (standard British English).
• Examples of Canadian raising.