The vowel categories

English has an unusually rich and complex vowel system, and a great deal of
variation in vowel pronunciation across dialects. Standard English spelling
does not identify pronunciations clearly or reliably, as poems like this
one indicate. Therefore, the easiest way to start is with a list of vowel
categories or equivalence classes, each represented by a set of words whose
vowels are all pronounced alike. For each vowel category, we'll pick a single
word as a convenient way to name the class. This doesn't tell us how the
words in each vowel class are pronounced -- and of course the pronunciation
varies across dialects. Different dialects also differ in how many distinctions
they make. Therefore each phonological equivalence class tells us only that
the vowels in it behave the same way.
Across English dialects, we need something like 24 "lexical sets" to do the job.
This particular classification comes from J. C. Wells' Accents of English:
Number Name Other Examples Number Name Other Examples
1 KIT ship, rib, dim 13 THOUGHT Waugh, hawk, broad
2 DRESS step, ebb, hem 14 GOAT soap, robe, home
3 TRAP bad, cab, ham 15 GOOSE loop, mood, boom
4 LOT stop, odd, Tom 16 PRICE ripe, tribe, time
5 STRUT cup, rub, hum 17 CHOICE boy, void, coin
6 FOOT bush, look, good 18 MOUTH pouch, loud, noun
7 BATH staff, clasp, dance 19 NEAR beer, weird, fierce
8 CLOTH cough, long, gone 20 SQUARE care, air, wear
9 NURSE curb, turn, work 21 START far, sharp, farm
10 FLEECE reap, seed, seize 22 NORTH for, York, storm
11 FACE late, babe, name 23 FORCE ore, floor, coarse
12 PALM bra, Brahms, blah 24 CURE boor, tour, gourd
Because of splits and mergers, any particular choice of sets is likely to make
some distinctions that are unnecessary for a given dialect, and also to fail to
make some other useful distinctions.
For example, very few American dialects distinguish the TRAP and BATH sets
-- though many British dialects do. The NORTH and FORCE sets have
merged for most speakers on both sides of the Atlantic, though a few dialects
still distinguish them at least in part. You can see if they have merged for you
by asking whether "for" and "four" are pronounced the same, or "horse" and
"hoarse". There are some people for whom they are different!
In the Philadelphia area, the TRAP set (whose vowel is sometimes called
"short a") has split into two. According to Labov 1989:
Short a is tense in closed syllables before nasals (man, fan, stand, Dan, etc.) with the
exception of irregular verbs (ran, swam, began) and tense before voiceless fricatives
(pass, fast, path, laugh, etc.) but lax in all open syllables (hammer, passage) unless the
second syllable is an inflectional suffix where the vowel is tense (passing, laughing),
and tense before /d/ in three words mad, bad, glad but lax in all other words before /d/
and lax before learned words like alas and wrath.
Thus to cover Philadelphia pronunciation properly, we'd have to split Wells'
TRAP set into two. We could call one the MAD set (man, pass, mad, etc.) --
the cases that Labov describes as "tense" -- and the other the TRAP set -- the
residual cases that Labov describes as "lax". So far you have no basis for
knowing what these terms refer to, but you can conclude that for Philadelphia
speakers, the vowels of words in the MAD set are similar to one another, and
different from the vowels of words in the TRAP set.
[In fact the "tense" vowels are higher and fronter in the IPA chart given below].
The idea of naming an English vowel type by referring to a set of example
words is a helpful one. As we've indicated, though, this doesn't tell us anything
about how a vowel category is actually pronounced. One way to indicate
actual pronunciation is to give some audio examples. This is useful, but often
we want to say something more systematic about the nature of vowel
pronunciations, their relationships, how they are changing, etc. For this, use of
the IPA vowel chart is helpful. The full chart is given below.
It represents a sort of three-dimensional space:
degree of opening of the vocal tract runs from top to bottom;
fronter vs. backer position of the tongue runs from left to right;
spread vs. rounded lips is indicated by pairs of symbols at a given place in the
chart.

You can learn more about the acoustic value of points in this vowel space by
downloading the IPA Help program from SIL's web site.
In terms of IPA vowel symbols, we can give the nominal pronunciation of the
vowels in Wells' lexical sets as follows, for the varieties of British and
American speech that we might hear from a news reader in a
national
broadcast:
British American Keyword

British American Keyword
1

KIT 13


THOUGHT
2

DRESS 14

GOAT
3

TRAP 15


GOOSE
4

LOT 16

PRICE
5

STRUT 17

CHOICE
6

FOOT 18

MOUTH
7


BATH 19

NEAR
8

CLOTH 20

SQUARE
9


NURSE 21


START
10


FLEECE 22


NORTH
11

FACE 23


FORCE
12


PALM 24

CURE
Note that the colon-like character indicates extra length of the preceding
sound, while a sequence of two characters indicates a sound that starts one
way and ends another.
This presentation leaves out a great deal of variation in pronunciation. Some
of this variation just chances the sound quality of certain words without
changing any lexical class definitions. For example, some American have a
relatively front vowel in the LOT class (for instance, people from the Chicago
area), while others have a much backer vowel.The category covers the same
set of words, but is just pronounced differently. In other cases, particular
words or classes or words move from one class to another. For example,
some Americans pronounce the vowel in "king", "sing", "bring" like the vowel
in KIT or the vowel in FLEECE. Finally, there are many splits and mergers.
Nearly all Americans merge the CLOTH and THOUGHT classes, and also
merge the LOT and PALM classes; a large proportion of Americans also
merge all four of these classes into one. Some mergers depend on the
context -- additional mergers of vowel categories before /r/ and /l/ are
especially common. Some people pronounce "Mary", "marry" and "merry"
differently (with the vowels of the FACE, TRAP and DRESS classes,
respectively) while others merge them -- without of course merging examples
like "mate", "mat", "met" that don't involve a following /r/.
There are some annoying aspects of the IPA character set as it applies to
English, which we will modify for the rest of this discussion. Basically we'll use
the regular a in place of the upside-down a, and the regular r in place of the
upside-down r. You should do the same in your homework.
1. The IPA assigns the normal printed a character to a low front vowel which is
hardly ever found, and assigns the upside-down a (or "turned a ") character to
the very common open central vowel, such as the vowel in American English
"pot". It will be a lot easier to read and write IPA descriptions of English if we just
take the regular lower-case a to be the American "pot" vowel.
2. The IPA assigns the upside-down r character ("turned r") to the particular kind
of "bunched-tongue r" used by most speakers of American English, while
reserving the ordinary r for the "trilled r" as found in Spanish (and many other
languages). Since the trilled r is not found in American English, it's again easier
to read and write if we use the standard r symbol.
Given these changes, then the relevant lexical sets for "standard" American
vowels become something like this:
IPA Name

IPA Name
1

KIT 14

GOAT
2

DRESS 15

GOOSE
3 & 7

TRAP & BATH 16

PRICE
4 &
12

LOT & PALM 17

CHOICE
5

STRUT 18

MOUTH
6

FOOT 19

NEAR
8 &
13

CLOTH &
THOUGHT
20

SQUARE
9

NURSE 21

START
10

FLEECE
22 &
23

NORTH &
FORCE
11

FACE 24

CURE
The consonant categories
The full set of basic IPA consonant categories is shown in the table below:

The subset of 22 IPA symbols relevant to English dictionary-style
pronunciations is given below:

Of these, some have basically their normal value in English spelling: p, b, t, d,
k, g, m, n, f, v, s, z, l.
The "right-tailed n" is the velar nasal sound at the end of "hang" or "ring".
The alveolar tap "fishhook" is the sound that spelled "t" or "d" becomes (for
most American speakers) in "Peter" or "ladder" or "at all".
The interdental fricatives "theta" and "eth" are
the sounds
at the start of "thin" and "this" respectively.
The palatal fricatives "esh" and "yogh" are the sounds in the middle of
"ashen" and "azure" respectively.
The palatal approximant j is essentially the consonant spelled 'y' in English, as
in "yield" or "yes".
In addition to the consonants in this table, you will need a few other things:
1. The consonant w, as in "will" or "wallaby" -- due to a peculiarity of IPA
classification, it does not appear in the main consonant table.
2. To make "affricates" such as the initial sounds in "chunk" or "jest", you need to
combine a stop and a fricative. For English, there are two cases:
1. the voiceless palatal affricate (like the start of "chip") which in IPA is
written
2. the voiced palatal affricate (like the start of "jut") which in IPA is
written
3. Stress is marked before the affected syllable; primary stress is marked by a
raised vertical line, while secondary stress is marked by a lowered vertical line.
Thus "California" is written
Writing IPA with ASCII
For the vowels, given the obvious correspondences [a], [e], [i], [o], [u], the
following IPA subset should be sufficient:
IPA Character Description ASCII

script a A

ash (ae ligature) {

turned a 6

turned script a Q

epsilon E

schwa @

reversed epsilon 3

small capital i I

open o O

slashed o 2

oe ligature 9

capital oe ligature &

upsilon U

barred u }

wedge V

small capital y Y

For the consonants, given the obvious correspondences [p] [t] [k] [b] [d] [g] [m]
[n] [s] [f] [v] [l] [r] [j] [w] [h], the following should suffice:

eth D

left-tail n J

eng (right-tail n) N

esh S

theta T

ezh (yogh) Z

glottal stop ?
For main stress, use double quote marks ("), and for secondary stress, use
percentage sign (%).
Some IPA fonts can be downloaded from here. If you want more detail on the
use of ordinary "ascii" characters to represent IPA symbols, consult this
reference.
Several chapters of a definitive work-in-progress on regional variation in the
pronunciation of North American English can be found here.