~~~~e ]l~~~~...roses, but of the red rose, proud rose, sad rose of all my days. we can express our...
TRANSCRIPT
~~~
'; •=
;. ' ~~~
n
,,,
,,~,
~~
/'
r ems.
~
,~,1
1
'
'(
r 1~,~ r
te;
~ ~
~
'•
~`,
j~~.o
;~•
~l~ ~-' "
l
.-.
~4,,
.0
,,
ut~`
~y
, a~~
,,~......'
F,.,___~
'4vb
~
rte
,,~
`~"
`~
z,.-'
~~ ~~
~~ ~~~~E
]L~~~~
anguage comes so naturally to us that it is easy to forget what a strange
and miraculous gift it
is. All over the world members.of our species fash-
ion their breath into hisses and hums and squeaks and pops and listen to oth-
ers do the same. We do this, of course, not only because we Iike the sounds
but because details of the sounds contain information about the intentions of
the person making them. We humans are fitted with a means of sharing our
ideas, in
all their unfathomable vastness. When we listen to speech, we can be
led to think thoughts that have never been thought before and that never
would have occurred to us on our own. Behold, th
e bush burned with fire, and
the bush was not consumed. Man is born free, and everywhere he is in chains.
Emma Woodhouse, handsome, cl
ever, and rich, with a comfortable home and
happy disposition, seemed to unite some of the best k~lessings of existence.
Energy equals mass times the speed of light squared. I have found
it impossi-
ble to carry the heave burden of responsibility and to discharge my duties as
King without the help and support of the woman I love_
Language has fascinated people for thousands of years, and linguists have
studied every detail, from the number of languages spoken in New Guinea to
why we say rc~~.le-dazzle instead of da
zzle-rnzzle.Yet to me the first and deepest
challenge in understanding language is accounting for its boundless expressive
power. What is the trick behind our ability to
fill one another's heads with so
many different ideas?
Z
Words and Rules
The premise of this book is that there are two trick, words and rules. They
work by different principles, are learned and used in difFerent ways, and may even
reside in different parts of the brain. T
heir border disputes shape and reshape lan-
guages over centuries, and make language not only a tool for communication but
also a medium for wordplay and poetry and an heirloom of endless fascination.
The first trick, the word, is based on a memorized arbitrary pairing between a
sound and a meaning. "What's in a name?" asks Juliet. "That which we call a
rose by any other name would smell as sweet." What's in a name is that every-
o~e in a language community tacitly agrees to use a particular sound to convey
a particular idea. Although the word rose does not smell sweet or have thorns,
we can use it
to convey the idea of a rose because
all of us have learned, at our
mother's knee or in the playground, the same link between a noise and a
thought. Now any of us can convey the thought by making the noise.
The theory that words work by a conventional pairing of sound and meaning is
not banal or uncontroversial. In the earliest s
urviving debate on linguistics, P
lato
has Hermogenes say, "Nothing has
its name by nature, but only by usage and
custom." Cratylus disagrees. "There is a correctness of name existing by nature
for everything: a name is not simply that which a number of people jo
intly ab ee
to call a thing." Crarylus is a creationist, and suggests that "a power greater than
man assigned the first names to things." Today, those who see a correctness of
names might attribute it instead to onomatopoeia (words such as crash and oink
that sound like what they mean) or to sound symbolism (words such as sneer,
cc~ntankeroais, and mellifluous that naturally call to mind the things they mean).
Today this debate has been resolved in favor of Hermogenes' co
nventional
pairing. Early in this century Ferdinand de Saussure, a founder of modern lin-
guistics, called such pairing the arbitrary sign and made it a cornerstone of the
study of la
nguage. ~
Onomatopoeia and sound symbolism certainly e~dst, but they
are asterisks to the far more important principle of the arbitrary sign—or else we
would understand the words in every foreign language instinctively, and never
need a dictionary for our own! Even the most obviously onomatopoeic words—
those for animal sounds—are notoriously unpredictable, with pigs oinking boo-
boo in Japan and dogs barking gong gong in Indonesia. Sound symbolism, for its
part, was no friend of the American woman in the throes of labor who overheard
what struck her as the most beautiful word in the English language and named
her newborn daughter Meconium, th
e medical term for fe
tal eYcrement.-
The Infinite Library 13
Though simple, the principle of the arbitrary sign is a powerful tool for get-
ting thoughts from head to head. Children begin to learn words before their
first birthda}; and by their second they hoover them up at a rate of.one every
two hours. By the time they enter school children command 13,000 words, .
and then the pace picks up, because new words rain down on them from both
speech and print. A typical high-school graduate I:nows about 60,000 words; a
literate adult, perhaps twice that number.3 People recognize words swiftly. The
meaning of a spoken word is accessed by a listeners brain in about a fifth of a
second, before the speaker has finished pronouncing
it.`~ The meaning of a
printed word is registered even more quickly, in about an eighth of a second.
People produce words almost as rapidly: It takes the brain about a quarter of a
second to find a word to name an object, and about another quarter of a sec-
ond to program the mouth and tongue to pronounce it
.bThe arbitrary sign works because a speaker and a listener can call on identi-
cal entries in their mental dictionaries. The speaker has a thought, makes a
sound, and counts on the listener to hear the sound and recover that thought.
To depict an entry in the mental dicrionaty we need a way of showing the entry
itself, as well as its sound and meaning. The entry for a word is simply its ad-
dress in one's memory, li
ke the location of the boldfaced entry for a word in a
real dictionary. It
's convenient to use an English letter sequence such as r-o-
s-e
to stand for tI-ie entry, as long as we remember this is just a mnemonic tag that
allows us to remember which word the entry corresponds to; any symbol, such
as 42759, would do ju
st as well. To depict the word's sound, we can use a pho-
netic notation, such as [roz].~ The meaning of a word is a link to an eptry in
the person's mental encyclopedia, which captures the person's concept of a
rose. For convenience we can symbolize it with a picture, such as
t~33. So a
mental dictionary entry looks something like this:
rose so
und: r6<
meaning: ~
"This book uses a simplified phonetic notation similaz m that f
ound in dictionaries, i
n which
the long vowels d in bait, e in
beet, z in
bite, o in boat, a
nd u in Uoot are distinguished From the
shott v
on~els k in bat, e in
bei, i
in bit, o in pot, and u in I~ut. t
ln unadorned a stands for t
he fi
rst
vowel in fa
ther or yapa. T
he symbol ~ i
s used for [h
e neutral vowel in the suffix of melted and
Rose's (e
.g., nz
elti~, ro
z}z), a version of t
he vowel sometimes called s
chwa.
"Long
voti~el," "
short vowel," a
nd ocher technical terms in linguistics, ps
ycholinguistics, and
neuroscience are defined in the Glossaq:
4 ~
Words and Rules
A final component is the word's part of speech, or Grammatical category,
D which for rose is noun (N):
rose so
und: roz
meaning: ~
part of speech: N
And that brings us to the second trick behind the vast expressive power of
language.
People do not ju
st blurt out isolated words but rather cm~xbine them into phrases
and sentences, in which the meaning of the combination can be inferred from
the meanings of the words and the way they are arranged. We talk not merely of
roses, but of the red rose, proud rose, s
ad rose of al
l my days. We can express our
feelings about bread and roses, guns and roses, the War of the Roses, or days of
wine and roses. We can say that lovely is
the rose, ro
ses are red, or a rose is a rose
is a rose. When we combine words, their ananbement is crucial: Violets are red,
roses are blue, though containing all the ingredients of the familiar verse, means
something very different. We all know the difference between young women
looking f
or husbands and hxisbands loolzing for young women, and that looking
women husbands yoticng for doesn't mean anything at al
l.
Inside everyone's head there must be a code or protocol or set of rules that
specifies how words may be arranged into meaningful combinations. Modern
linguists call it a grammar, sometimes a generative brnmmar to distinguish it
from the grammars used to teach foreign languages or to teach the dos and
don'ts of formal prose.
A grammar assembles words into phrases according to the words' part-of-
speech categories, such as noun and verb. To highlight a word's category and
reduce visual clutter often it is convenient to omit the sound and meaning and
put the category label on top.
N Irose
Similarly, the word a, an article or cleteri~ainer, would look like this:
The Infinite Library
~ 5
det
a
They can then be joined into the phrase a rose by a rule that jo
ins a determiner
to a noun to yield a noun phrase (NP). The rule can be shown as a set of con-
nected branches; this one says "a noun phrase may be composed of a deter-
minerfollowed by a noun':
NP
/ \
det
N
The symbols at the bottom of the branches are like slots into which words may
be plugged, as long as the words have the same labels growing
o"ut of their
tops. Here is the result, t
he phrase a rose:
NP
/ \
det
N
a
rose
With just two more rules eve can build a complete toy grammar. One rule de-
fines apredicate or verb phrase (VP); the rule says that a verb phrase may con-
sist of a verb followed by it
s direct object, a noun phrase:
VP
/ \
V
NP
The other rule defines the sentence itself (S). This rule says that a sentence
may be composed from a noun phrase (the subject) followed by a verb phrase
(the predicate):
S
NP
VP
When words are plugged into phrases according to these rules, and the
phrases are plugged into bigger phrases, we get a complete sentence, such as
A rose is a rose:
6 ~
Words and Rules
S
/ \
NP
VP
det
N V
NP
1 ~
~ /
a rose is d
et N
1 ~
a rose
Other parts of the rules, not sho~m here, sp
ecify the meaning of the new com-
bination. For example, the complete NP rule says that the meaning of the }'
el-
low rose of Tomas is based on the meaning of rose, which is called the head of
the phrase, and that the other words modify the head in various ways: yellrnv
specifies a distinctive trait, T~s it
s location.
These rules, though crude, illustrate the fantastic expressive power made
available by grammar. First, the rules are prod2.cctiroe. By specifying a string of
kivuZs of words rather than a string of actual words, the rules allow us to assem-
ble
ne~~~ sentences on the fly and not regurgitate preassembled cliches—and
that allows us to convey unprecedented combinations of ideas. Though we
often speak of roses being red, we could talk about violets being red
if the
desire came over us (perhaps to announce a new hybrid), because the rule
allows us to insert violets into the N slot j
ust as easily as roses.
Second, the symbols contained by the rules are symbolic and hence abstract.
The rule doesn't say, "A sentence may begin with a bunch of words refer:inb to
a hind of flower"; rather, it
says, "A sentence may begin with an NP," where
NP is a symbol or variable that can be replaced by any noun, just as x or y in a
mathematical formula can be replaced by any number. Vve can use the rules to
talk about flowers and their colors and smells, but we can just as easily use
them to talk about karma or quarks or floob-boober-bab-boober-hubs (who, ac
-cording to Dr. Seuss, bounce in the water like blubbery tubs).
Third, the rules are combinatorial. They don't just have a single slot, like a
fill-
in-the-blank exam question; every position in the sentence offers a choice
of words from a lengthy menu. Say everyday English has four determiners (n,
any, on
e, and the) and ten thousand nouns. Then the rule for a noun phrase
al-
lows four choices for the determiner, f
ollowed by ten thousand choices for the
head noun, yi
elding 4 x 10,000 = 40,000 ways to utter a noun phrase. The rule
The Infinite Library ~
7
for a sentence allows these forty thousand subjects to be followed by any of
four thousand verbs, providing 40,000 x 4,000 = 160,000,000 ways to utter
the first three words of a sentence. Then there are four choices for the deter-
miner of the object (640 million four -word beginnings) followed by ten thou-
sand choices for the head noun of the object, or 640,000,000 x 10,000 =
6,400,000,000,000 (6.4 trillion) five -word sentences. Suppose it takes five
seconds to produce one of these sentences: To crank [hem a
ll out, f
rom The
abandovcment abased the abbey and T7ae abandonnze~~t abased the abbot, through
The abandoazment abased the zoologist, all the ~vay to The zoologist zoned the
zoo, would take a million years:
Nlany such combinations are ungrammatical of course, owing to various
complications I haven't mentioned—for example, you can't say T7ze Anron, a
abandonment, or The abbot abase the abbey. And most of the combinations are
nonsensical: Abandonments can't abbreviate, and abbeys can't abet. Yet even
with these restrictions the e,~pressive range of a grammar is astonishing. The
psychologist George Miller once conservatively estimated that if speakers keep
a sentence perfectly grammatical and sensible as they choose their words,
their menu at each point offers an average of about ten choices (at some
points there are many more than ten choices; at others, only one or two). %That
works out to one hundred thousand five-word sentences, one million sip-word
sentences, ten million seven-word sentences, and so on. A sentence of twenty
words is not at
all uncommon (the preceding sentence has twenty words be-
fore and so on), and there are about one hundred million trillion of them in En-
glish. For comparison, that is about a hundred times the number of seconds
since the birth of the universe.
Grammar is an example of a combinatorial system, in which a small inven-
tory of elements can be assembled by rules into an immense set of distinct ob-
jects. Combinatorial systems obey what Miller calls the Eacponential Principle:
The number of possib]e combinations meows exponentially (geometrically)
with the size of the combinations Combinatorial systems can generate incon-
ceivably vast numbers of products. Every kind of molecule in the universe is
assembled from ahundred-odd chemical elements; every protein building
block and
catalyst in the living world is assembled from just twenty amino
acids. Even when the number of products is smaller, a combinatorial system
can capture them all and provide enormous sa~~ings in storage space. Eight bits
define 23 = 256 distinct bytes, which is more than enough for all the numerals,
punctuation marks, and upper- and lowercase letters in our writing system.
This allows computers to be built out of id
enrical specks of silicon that can be
8 ~
Words an
d Ru
les
in jus
t tw
o st
ates
, ins
tead
of the dozens of pi
eces
of ty
pe that once filled ty
pe-
setters' ca
ses.
Billions of
years ago
life on Ear
th settled on a co
de in wh
ich a
string of three ba
ses in
a DNA molecule became the ins
truc
tion
for sel
ecti
ng
one amino ac
id when ass
embl
ing a pr
otei
n. There are fou
r kinds of bas
es, so a
thre
e-base string allows for 4 x 4 X 4 = 64 possibilities. T
hat is enough to giv
e
each of the twenty amino aci
ds it
s own str
ing,
with plenty le
ft ove
r for th
e start
and st
op ins
truc
tion
s th
at beg
in and
end the pro
tein
. Two bas
es would hav
e
been
too
few (4 x 4 = 16), f
our mo
re tha
n ne
eded
(4 x 4 x 4 x 4 = 256).
Perh
aps the most viv
id des
crip
tion
of the st
agge
ring
power of a combinator-
ial sy
stem
is in Jor
ge Lui
s Bo
rges
's sto
ry "The Lib
rary
of Babel."9
The lib
rary
is
a vast net
work
of galleries wi
th boo
ks composed of all the co
mbin
atio
ns of
twenty-two let
ters
, the comma, the period, and the spa
ce. Somewhere in th
e
libr
ary is a book tha
t contains the true hi
stor
y of
the fut
ure (including the
sto
ry
of you
r de
ath)
, a book of prophecy tha
t vi
ndic
ates
the act
s of every man in the
univ
erse
, an
d a book containing the cla
rifi
cati
on of the my
ster
ies of hum
anit
y.
People roamed the galleries in a fu
tile
search for th
ose te
xts from among the
unto
ld number of bo
oks wi
th fal
se versions of
eac
h re
vela
tion
, the millions of
facs
imil
es of a gi
ven book differing by a cha
ract
er, and, of course, the miles
and miles of gib
beri
sh. The narrator notes th
at eve
n when the
human species
hoes
ext
inct
, the li
brar
}; tha
t sp
ace of
combinatorial possibilities, wi
ll endure:
"ill
umin
ated
, so
lita
ry, infinite, perfectly mo
tion
less
, eq
uipp
ed with precious
volumes, use
less
, incorruptible, secret."
Tech
nica
lly,
Borges needn't have des
crib
ed the lib
rary
as "i
nfin
ite:
' At eighty
char
acte
rs a lin
e, forty lines a pag
e, and 410 pag
es a book, the number of
book
s is aro
und
10~~
soo,
000
or I
fol
lowe
d by
1.8
mil
lion
zeroes. That is
, to be
sure
, a very lar
ge number—there are
only 10%° particles in the vi
sibl
e uni-
vers
e—bu
t it is a fi
nite
number.
It is ea
sy to make a toy grammar tha
t is eve
n more pow
erfu
l than the
scheme tha
t ge
nera
tes The Lib
rary
of Ba
bel.
Suppose our
rule for the ve
rb
phrase is enriched to al
low a sentence (5) to
appear inside it,
as in I tol
d Mary
he was
a fo
ol, i
n wh
ich he was
a fo
ol comes after the object NP Mar
y:
VP
i~
V
NP
S
Now our
grammar is
rec
acrs
ive:
The rul
es create an
entity that can contain an
example of it
self
. In thi
s ca
se, a Se
nten
ce con
tain
s a Ve
rb Phr
ase which in
The In
fini
te Lib
rary
~ 9
turn can con
tain
a Sentence. An entity th
at contains an
exa
mple
of it
self
can
just
as ea
sily
contain an example of it
self
tha
t co
ntai
ns an example of
itself
that con
tain
s an
example of itself, and
so on:
/ \
NP
VP
i~
V
NP
S
NP
VP
i~
V
NP
S
/ \
NP
VP
i~~
V
NP
S
/ \
NP
VP
i~
V
NP
S
In thi
s case a sen
tenc
e can co
ntai
n a verb phrase, whi
ch can contain a sen
-tence, which can con
tain
a ver
b ph
rase
, wh
ich can co
ntai
n a se
nten
ce, ad in-
fini
tum.
For example, I think
I'll tel
l you th
at I jus
t read a news sto
ry tha
trecounts tha
t Stephen
Bril
l re
port
ed tha
t the pr
ess un
crit
ical
ly bel
ieve
d Ken-
neth
Sta
rr's
announcement tha
t Linda Tr
ipp
testified to him tha
t Monica
Lewinsky told Tripp th
at $ill Cl
into
n told Vernon Jo
rdan
to advise Lev~rinsky
not to
testify to Starr that she had
had
a sex
ual re
lati
onsh
ip wit
h Cl
into
n. That
stat
emen
t is a Rus
sian
doll with thirteen se
nten
ces in
side
sen
tenc
es-i
nsid
esentences. A rec
ursi
ve grammar can generate se
nten
ces of
any
length, and
thus
can
generate an
inf
init
e number of sentences. So a human being ossess-
bP
ing a recursive grammar can
exp
ress
or understand an infinite number of dis-
tinct th
ough
ts, limited in
pra
ctic
e on
ly by st
amin
a and mortality
10 ~
Wo
rds and
Rtii
les
~~
I
The idea that the cre
ativ
ity in
here
nt in la
ngua
ge can be e~cplained by a gra
m-
mar of combinatorial ru
les
is usu
ally
associated with the lin
guis
t Noam
Chomsky. Chomsky tra
ced the id
ea to Wilhelm von Humboldt, a nineteenth-
century pi
onee
r of
linguistics, who explained lan
guag
e as
"th
e infinite use of
fini
te media." Acc
ordi
ng to Chomsky, the idea is
even ol
der than that; Hum-
bold
t was the last in
a tradition of "C
arte
sian
lin
guis
ts" da
ting
back to the En-
lightenment.'
Enlightenment phi
loso
pher
s were cap
tiva
ted by the
dizzying ra
nge of
thou
ghts
made expressible by a com
bina
tori
al grammar. In his book The
Search for
the
Per
fect
Language the semiotician Umberto Eco recounts the
many Promethean schemes the
se phi
loso
pher
s came up with to
perfect and
i
harness their power.l~ Des
cart
es noticed tha
t the decimal system allows a
person to learn in
a day the names of all the quantities to infinity, and he sug
-
gested tha
t a universal art
ific
ial la
ngua
ge built on sim
ilar
principles could or-
ganize a
ll human thoughts. Leibniz, too, dreamed of a
univ
ersa
l logical
grammar that would gen
erat
e onl}~ valid sequences of ideas, banishing irra-
tionality and err
or forever.
Three hundred years lat
er we still are
fallible, and sti
ll take years to lea
rn a
Babel of local la
ngua
ges with [he
ir ten
s of
tho
usan
ds of ar
bitr
ary signs. Why
has no modern language used the
horsepower of
combinatorial grammar to the
fullest and abandoned the unp
rinc
iple
d, par
ochi
al, onerous-
to-memorize lau
n-
dry
list cal
led vo
cabu
lary
? The answer becomes clear when ~~~e look at th
e
most famous of the co
mbin
ator
ial schemes of the Enlightenment, the
philo-
soph
ical
lanwage of Bishop John Wilkins. The arbitrary name was an afF
ront
to VVilkins's se
nse of
good des
ign,
and he strove for a way to eliminate
it. He
wrote, "We sho
uld,
by lea
rnin
g ...the Nna
nes of thi
ngs,
be instructed likewise
in their Nat
ures
."
Wilkins's system, la
id out
in a le
ngth
y 166S opu
s, offered the user a noza-
arbi
trar
y name for
eve
ry thi
ng by div
idin
g th
e un
iver
se int
o ca
tego
ries
and
subc
ateg
orie
s and sub-subcategories, and assigning a vowel or consonant to
I:
every branch in the tre
e. The first syl
labl
e id
enti
fied
one of the forty Ca
te-
~I`
gories int
o which Wilkins had sorted
all th
inka
ble thoughts. For example, Z
stood fo
r "sensitive spe
cies
" (a
nima
ls).
and could be fol
lowe
d by i for "beasts"
f
(quadrupeds). The nex
t consonant picked ou
t a sub
divi
sion
; t, for
example,
stood fo
r ra
paci
ous terrestrial European canines. A fin
al vow
el pin
poin
ted th
e
species, yielding Zita as th
e name for
dog
s. By sim
ilar
computations one
The In
fini
te Library ~
I I
could deduce ano
ther
two thousand names for things. Zana is a scaly river
fish with reddish fl
esh,
in ot
her wo
rds,
salmon. Siba is a ty
pe of pu
blic
mil
i-ta
ry rel
atio
n, nam
ely,
defense. Deba is a portion of
the first of the te
rres
tria
lelements (fi
re),
to ~n
~t, fl
ame.
Coba is a con
sang
uino
us economic rel
atio
n of
direct asc
enda
nt, a.k.a. fat
her.
Wilk
ins'
s ph
ilos
ophi
cal language has
been ana
lyze
d in
sigh
tful
ly by Borges
and Eco, and we can see
why no one tod
ay spe
aks Wi
llis
h. ~'-
For one thing, it
forces use
rs to perform a chain of computations in their heads eve
ry time
they
want to refer to a dog. Ev
ery vowel and con
sona
nt is la
den with meaning
and act
s as
a premise in a le
ngth
enin
g de
duct
ion.
Spe
aker
s of the language
would have to
play a game of Twenty Que
stio
ns, in
ferr
ing an ent
ity from a de-
scription, for
every word in a sen
tenc
e. They cou
ld of course sim
ply memo-
rize the answers, such as th
at a portion of the first of
the terrestrial elements
is a fla
me, but that is
not much easier than memorizing that th
e word for
flam
e is
flnn
ae.
A second problem is th
at there are more things in heaven and earth than
were dreamt of in
Wil
kins
's philosophy, which ide
ntif
ied only two thousand
concepts. Wilkins understood the exponential principle and tried to cope with
the problem by len
gthe
ning
the wor
ds. He pro
vide
d su
ffix
es and connectors
that all
owed
cal
f, for
exa
mple
, to be e~tpressed as
cmv +young, and cutronomer
to be exp
ress
ed as artist +st
ar. But eventually he gave up and resorted to
using
synonyms for
con
cept
s hi
s la
ngua
ge cou
ld not
gen
erat
e, such as bo
x fo
r co
ffin
.Wilkins's dilemma was tha
t he cou
ld either expand his
sys
tem
to embrace all
conc
epts
, which would require even lo
nger
and more unwieldy st
ring
s, or he
coul
d fo
rce hi
s us
ers to remember the
nea
rest
synonym, rei
ntro
duci
ng the de-
spised memorization process.
A thud problem is that in a logical language words are
ass
embl
ed purely on
info
rmat
ion -
theoretic principles, with no regard to the problems that incar-
nate
creatuzes might have in pronouncing and und
erst
andi
ng the strings. A
perfect combinatorial la
ngua
ge is al
ways
in da
nger
of generating mou
thfu
lslike mxy
zptl
k or I~ftsplk, so Wilkins and other language -de
sign
ers of the En-
lightenment all had to make con
cess
ions
to pronounceability and euphony.
Sometrmes they defiled th
eir sy
stems with irregularities, for exa
mple
, revers-
ing avowel and consonant to make a word more pronounceable. At other
times th
ey hobbled the sys
tem with restrictions, such as that consonants and
vowels must alt
erna
te. Every even -numbered pos
itio
n in a word had to be
fill
ed by one of th
e nine vowels of En
glis
h_ and that rP
~rr;
~rA,
~ ,,,~,,.. ,.
..«,.
12
Words and Rules
gories, such as species in a genus, to nine apiece, regardless of how many
species exist in the world.
Another problem
is that Wilkins's words are .packed tight with information
and lack the safety factor provided by redundancy The slightest slip of the
tongue or pen guarantees misunderstanding. Eco catches Wilkins himself mis-
using Gade (barley) for Gape (tulip).
Finally, all that power is not being put to any sensible use. The beauty of a
combinatorial system is that it Generates combinations that have never before
been considered but that one might want to talk about some day. For exam-
ple, the combinatorial system known as the periodic table of the elements in-
spired chemists to look for hitherto unknown chemical elements that should
have occupied the empty slots in the table. Combinatorial grammar allows us
to talk about a combinatorial world, a world in which violets could be red or a
man could bite a dog. Yet familiar objects and actions around us often form a
noncombinatorial list of distinctive kinds. When we merely have to single out
one of them, a combinatorial system
is overkill. We never will have to refer to
fish with an enmity to sheep or to military actions with scales and reddish
flesh, and that's what a combinatorial, system for words like Wilkins's allows
us to do. To refer to everyday things it s easier to say dog or fi
sh than
to work
through a complicated taxonomy that is just a fancy way of singling out dogs
or fish anywa}c
The languages of Wilkins and other Enlightenment thinkers show that combi-
natorial grammar has disadvantages as well as advantages, and that illuminates
our understanding of the design of human language. No language works like
Wilkins's contraption, with every word compiled out of meaningful vowels and
consonants according to a master formula_ All languages force their speakers
to memorize thousands of arbitrary words, and now we can see why.1
3 Nlany
bodily organ systems are made from several kinds of tissue optimized for jobs
with contradictory specifications. Our eyes have rods for night vision and
cones for day vision; our muscles have slow-twitch fibers for sustained action
and fast-twitch fibers for bursts of speed. The human language system also ap-
pears to be built out of two kinds of mental tissue. It has a leacicon of words,
which refer to common things such as people, places, objects, and actions,
and which are handled by a mechanism for storing and retrieving items in
memory. And it has a grammar of rules, which refer to novel relationships
The Infinite Library
~ 13
amonb things, and which is handled by a mechanism for combining and ana-
1}~zing s
equences of symbols.
To a parsimonious scientific mind, however, two mental mechanisms can be
one too many. The poet Wiliiam Empson wrote of the Latin philosopher,
Lucretius could not credit centaurs;
Such bicycle he deemed asynchronous.
14
Today's skeptics also might wonder about atwo-part design for language. Per-
haps words and rules are two modes of operation of a single faculty. Simple, fa
-miliar thoughts need short noises, which we call words, and complicated,
unfamiliar thoughts need long noises, which we call phrases and sentences. A
single machine might make either short or long noises, depending on the kinds
of thoughts it is asked to express. Or perhaps there is a gradual continuum be-
tween memory and combination rather than two distinct mechanisms, with
words at the memory end of the continuum and sentences at the combination
end. To show that words and rules are handled by different machines we need to
hold the input anc~ output of the putative machines constant. We need side
-by -side specimens in which the same hind of thought is packed into the same
kind of verbiage, but one specimen shows the handiwork of a word regurgita-
tor and the other shows the handiwork of a rule amalgamator. I believe that
languages do provide us with such specimens. They are called regular and ir-
regular words.
English verbs come in two flavors. Regular verbs have past tense forms that
look like the verb with -ed on the end: Today I j
og, yesterday I j
ogged. They are
monotonously predictable: j
og jogged, waUz—wniked, play%played, kiss—kissed,
and so on. (
Regular nouns, whose plurals end in -s, such as cats and dogs, are
similar.) The list of regular verbs is also open-ended. There are thousands, per-
haps tens of thousands, of regular verbs in English (depending on. how big a
dictionary you consult), and new ones are being added to the language
all the
rime. When fc
~z came into common parlance a decade or so ago, no one had to
inquire about its past-tense form; everyone knew it vas f
c:xed. Similarly, when
other words enter the language such as Spam (flood with E-mail), snarf (down-
load afile), mung (damage something), nzosh (dance in roughhouse fashion),
and Bork (challenge a
political nominee for partisan reasons), the past-tense
forms do not need s
eparate introductions: We a
ll deduce that they are
spnmmed, sf
xarfed, nzzinged, noshed, and
Borleecl.
l4 ~
Words and Rules
Even young children do it. In 1958 the psychologist Jean Berko Gleason
tested four- to seven-year-old children with the following procedure, now
known as the wiig-test:
•
This is
a wug.
s
•
Now there is another one.
There are two of them.
These are t~vo
The children could have refused to answer on the grounds that they had never
heard of a wug and had never been told how to talk about more than one of
them. Instead, Serko Gleason wrote, "Answers were willingly, and often insis-
tently, b ven." Three-quarters of the preschoolers and 99 percent of the first-
graders filled in the blank with ti
inc~s. Similarly, when shown a picture of a man
who I:nows how to ricl or Bing or filing and did the same thing yesterday, most
children said that he ricked or binged or glinged.
The children could not have heard their parents say u~
tiigs or binged before
entering the lab, because these words had been coined especially for the exper-
iment. Children therefore are not parrots who just play back what they hear.
And the children could not have been previously rewarded by parents for utter-
ing those forms, because the children did not know the words before entering
the lab. Children therefore are not like pigeons in a Skinner boa, who increase
or decrease the frequency of responses in reaction to the contingencies of rein-
forcement. Noam Chomsky and Eric Lenneberg, pioneers of the modem study
The Infinite Library
~ 15
of language and contemporaries of Berko Gleason in the Harvard-NIIT com-
munity, pointed to children's ability to generalize construcrions such as the reg-
ularpast tense in support of their theory that language is actively acquired by a
special rule-forming mechanism in the mind of the child.~>
As it happens, al
l children are subjects in a version of Berko Gleason's ex-
periment. Children often make up words or mangle them and are happy to put
their ne~v verbs in the past tense. Here are some examples:
spidered
lightninged
smunched
poonked
speeched
broomed
byed (went by)
eat lunched
cut -upped e~~b
All children also make creative errors in their speech like these:
I buyed a fire dog for a grillion dollars.
Hev Horton heated a Who.
My teacher bolded the baby rabbits and we patted them.
Daddy I stealed some of the people out of the boat.
Once upon a time a alligator was eating a dinosaur and the dinosaur
vas eating the alligator and the dinosaur was eaten by the alliga-
tor and the alligator goed kerplunk. t-
Such errors bring us to the second flavor of a verb in English: irregular. The
past-tense form of an irregular verb
is not simply the verb decorated with an
-ecl ending. For example, the past tense of ]ncy is not bayed, but boaight. Simi-
larly, the past tense of hear, hold, st
eal, and go are heard, held, st
ole, and went.
Inewlar verbs contrast ~~ith reb lar verbs in almost every way. Whereas reb
ulars are orderly and predictable, irregulars are chaotic and idiosyncratic. The
past tense of sink is sanl ,and the past tense of ri
tag is rang. But the past tense
of cl
iszg is not clang, but clacng. The past tense of think is neither thai2l nor
thitnk, but thought. And the past tense of blink is neither blank nor blunt nor
bloicght, but a regular form, blinked. The language maven Richard Lederer
wrote a poem, "Tense Times with Uerbs," that beb ns
:
16 ~
Wo
rds and Ru
les
The ver
bs in En
glis
h ar
e a fright.
How can we lea
rn to read and write?
Today we spe
ak, bu
t first we spo
ke;
Some faucets lea
k, but
never lok
e.
Today we write, bu
t first we wro
te;
We bit
e ou
r to
ngue
s, but never bot
e.
Each day I tea
ch, fo
r years I taught,
And pre
ache
rs pre
ach,
but nev
er praught.
This tale I
tell; th
is tal
e I told;
I smell the fl
ower
s, but nev
er smold.
If knights sti
ll slay, as once they sl
ew,
Then do we play, as once we plew?
If I sti
ll do as once I did
,
Then do cows moo, as they once mid? ~ 6
Also
in co
ntra
st to the re
gula
rs, irregular ve
rbs form a closed
list. There are
only about 150 to 180 irregular verbs in modern Eng
lish
(depending on how
you cou
nt),
and the
re have been no rec
ent additions.
79 The youngest ir
regu
lar
is pro
babl
y snuck, which sneaked int
o the language ove
r a century ago and is
still not accepted by pur
ists
.'-0 And the freewheeling children in Berko Gle
a-
son's study were downright stodgy when it came to ir
regu
lar fo
rms:
Only one
out of
eig
hty-
six turned Bin
g into ban
g, and one oth
er turned fi
ling
int
o gl
ang.
'-1
These dif
fere
nces
sugbest a simple th
eory
. Regular past-tense forms are pre-
dict
able
in sound and gen
erat
ed fre
ely because th
ey are
products of
a rul
e that
live
s in the minds of ch
ildr
en and adu
lts:
"The pas
t te
nse of a ver
b may be
formed from the ver
b fo
llow
ed by the
suf
fix -e
d:' The rule would look ju
st like
the rules of
syn
tax in the toy
grammar we played with earlier,
past
/ \
V
suffix
and would generate a si
mila
r in
vert
ed-tree-like str
uctu
re:
Upast
V
suff
ix
walk
-ed
The In
fini
te Library
~ 17
Irregular verbs, in co
ntra
st, are unpredictable in form and res
tric
ted to a lis
tbecause they are memorized and ret
riev
ed as in
divi
dual
words. An irregular
form would loo
k ju
st like th
e lexical en
try we saw when considering the name
of the ros
e. It would be linked with the entry for
the pla
in form of the same
verb
and lab
eled
as
its pa
st tense:
hold
held
soun
d: hol
d so
und:
held
meaning: ~
meaning: r~
part of speech: V
part
of speech: V
tense: pas
t
Two mechanisms trying to do the same job
would get
in each oth
er's
way un-
less something adj
udic
ated
between them, and the
re is in
deed
a simple pr
inci
-pl
e: If a word can pro~~ide it
s own past tense from memory, the rule is
blocked;
elsewhere (by default), the
rule
appl
ies.
'-'-
The fir
st part e,
~pla
ins why we adults
don't say bo
lded
and stealed; ou
r knowledge of held and stole blo
cks the rule
that would have added -ed
. The second part e~tplains why bot
h children and
adul
ts say
Borl ed and nzoshed and ricl ed and bro
omed
; as long as a ver
b does
not ha
ve a form in memory, the rul
e maybe app
lied
. The abi
lity
of a ru
le to ap
-ply elsewkere or by default—that is
, to
any word that does not
alr
eady
have a
specified form in memory—is the
source of
its po
wer.
A speaker who needs to
express a past tense or pl
ural
is never left speechless, even when a search in
memorg comes up emptyhanded.
The theory th
at regular forms are
gen
erat
ed by rul
e and irr
egul
ar forms are
retr
ieve
d by rot
e is
ple
asin
g no
t on
ly bec
ause
it explains the
dif
fere
Ttce
s in
pro-
duct
ivit
y between the two patterns but als
o because it Fits nicely in
to the
larg
er pic
ture
of the design of language.
At first gla
nce ir
regu
lar verbs would seem to have no reason to liv
e. Whv
shou
ld lan
guag
e have forms that ar
e ju
st cus
sed ex
cept
ions
to a ru
le? What are
they
good for
, besides gi
ving
chi
ldre
n a way to make cut
e er
rors
, pr
ovid
ing
mate
rial
for
humorous verse, and making lif
e miserable fo
r foreign language
stud
ents
? In Woody Allen's story "The Kugelmass Epi
sode
" a hu
mani
ties
pro-
fessor in a mi
dlif
e cr
isis
finds a magic cab
inet
that projects him int
o any book
he tak
es in with him. Aft
er a tempestuous aff
air with Madame Bovary, Kugel-
mass tries again with another no
vel,
but thi
s time the
cab
inet
malfunctioned,
and the pro
fess
or "was pro
ject
ed into an old
te:
ctbo
ok, Remedial Spa~aisl2, and
was running for
his
lif
e over a barren, rocky terrain as th
e word ten
er (`t
ohave')—a large and hairy irregular verb—raced after him nn
its sn
;,,r
il~ i
PQ~ "'3
1 S
~ Wo
rds and
Rifl
es
But un
der the wo
rd-and
-rul
e theory we need no
t su
ppos
e th
at evolution fit-
ted us wit
h a sp
ecia
l gadget for irregularity. Irr
egul
ar for
ms are
jus
t words. If
our la
ngua
ge faculty has a knack for memorizing wo
rds,
it sho
uld have no in
hi-
biti
ons about me
mori
zing
pas
t-tense fo
rms at
the same tim
e. These are
the
verb
s we cal
l ir
regu
lar,
and
they ar
e a mere 1S0 add
itio
ns to a me
ntal
lex
icon
that
alr
eady
num
bers
in the te
ns or hundreds of thousands. Ir
regu
lar and regu-
lar fo
rms therefore wo
uld be the
inevitable outcome of two men
tal subsys-
tems, words and rules, tr
ying
to do the same thi
ng, name
ly, express an
eve
nt or
state that too
k place in the past.
Regular and ir
regu
lar fo
rms th
row a spotlight on the advantages an
d di
sadv
an-
tages of
wor
ds and
rules, because ev
eryt
hing
else about them is th
e sa
me: They
both
are
one word lo
ng, an
d bo
th convey the same mea
ning
, pa
st ten
se. The ad-
vantage of
a rule
is tha
t a va
st number of fo
rms are generated by
a com
pact
mech
anis
m. In English the savings ar
e si
gnif
ican
t. The rul
es for -ed
, -s,
and
-in
g
(the
thr
ee reg
ular
for
ms of th
e verb) cut.our mental storage heeds to a quarter of
what they would be if e
ach fo
rm had
to be
sto
red se
para
tely
. In other languages,
such as Tu
rkis
h, Ban
tu, a
nd many Native Am
eric
an languages, t
here
can
be hu
n-
dreds, tho
usan
ds, or eve
n mi
llio
ns of co
njug
ated
for
ms for every ver
b (f
or differ-
ent combinations of te
nse,
per
son,
. number, gender, mood, ca
se, an
d so on), an
d
the sa
~rin
gs are
ind
ispe
nsab
le. The rule also allows new wor
ds lik
e mosh, rare
words li
ke abase, and abstract words lik
e abet to be sup
plie
d with a past tense
(nao
slae
d, aba
sed,
abet
ted)
, even if there we
re no pr
evio
us opportunities for the
speaker an
d he
arer
to have committed the for
m to
memory. On the other han
d, a
rule
is mo
re powerful than needed for wo
rds we heaz so
often that retrieval fro
m
memory is ea
sy. As we sha
ll see, it
is th
e most common ver
bs, such as be, hc
eve,
do, go,
and say
, that tum out
to be irr
egul
ar in language aft
er lan
guag
e.
Rules have another shortcoming tha
t in
vite
s the word system to memorize
irre
gula
rs. Re
call
tha
t one of the nuisances pl
agui
ng John Wilkins as he de-
sign
ed his
per
fect language was tha
t fl
esh-and-
blood humans had to pro-
nounce and understand the pr
oduc
ts of the ru
les.
A sequence of sounds tha
t
encodes a concept pr
ecis
ely an
d efficiently may be unresolvable by the ear or
unpronounceable b}
' the ton
gue.
So it
is wit
h th
e rule for the past tense in En-
glish. The delicate tongue
-tap
tha
t gr
aces
the end
of a regular form may esc
ape
a listener and be om
itte
d when he re
prod
uces
it, res
ulti
ng in a so
leci
sm such
as sup
pose
to,
use to,
or cut and dry, or in signs and in
scri
ptio
ns lik
e th
ese:
Broil Cod
Use Books
The
hxfi
nite
Library
19
Whip Cream
Blac
ken redfish
Can Veg
etab
les
Bor sets
Handicap Fac
ilit
ies Available
In certain older exp
ress
ions
-ed was omi
tted
so of
ten th
at the
e~c
pres
sion
even-
tual
ly lost th
e -e
d al
toge
ther
, even among careful speakers and li
sten
ers.
That's
how ~t~e en
ded up with ice cr
eam (o
rigi
nall
y iced cream), so
tiu~ cream, mince
meat, and DanaTz Yankees.-'' Irregular verbs, in co
ntra
st, tend to use vowel
changes such as ri
tzy rang, st
rilze—struck, and b1o
u~—b
lew,
which are
as cl
ear as
a bell.
Simi
larl
y, the ver
y ob
livi
ousn
ess to
the det
ails
of the ve
rb that ma
kes a ru
leso po~~~erfi.tl (
it applies acr
oss the board to
all
verbs, wh
ethe
r they are familiar
soun
ding
or not) can let
it blindly jam a suf
fix onto the
end of an inh
ospi
tabl
esound. The result can be an uneuphonious ton
gue-
tv~r
iste
r such as ed
ited
or
sixths. Monstrosities like these are nev
er found among the
irr
egul
ars,
which all
have sta
ndar
d Anglo-Sakon word sounds such as grew and str
ode and clu
~zg,
which please th
e ea
r and roll off th
e to
ngue
.'-'
Language works by words and rul
es, each with strengths and ~~~
eakn
esse
s.Irregular and reg
ular
ver
bs are
contrasting specimens of words and rules in go
-ti
on. These are
the themes of this book, but with many twists to come. It
would be too good to be true if we reached a maj
or con
clus
ion about the most
comp
lica
ted object in th
e I:nown uni
vers
e, the
human brain, si
mply
by see
ing
how children name pic
ture
s of
litt
le bir
ds. The word -an
d-rule the
ory fo
r regu-
lar and irregular ver
bs is an ope
ning
statement in the la
test
round of a debate
on ho~v th
e mind ~vo
rls th
at has
raged for
cen
turi
es. It
has
ins
pire
d two alter-
native the
orie
s that are equ
ally
ingenious but
diametrically opposed, and in-
tens
ive
research showing what is
right
and
v~~rong
about each of
them—perhaps res
olvi
ng the deb
ate fo
r good. The the
ory has sol
ved many
puzz
les about the Eng
lish
lan
guag
e, and has
illuminated the ways tha
t chil-
I, dr
en lea
rn to ta
lk, th
e fo
rces
that make languages diverge and the for
ces th
atmake them alike, t
he way tha
t la
ngua
ge is processed in the bra
in, and even th
eI
natu
re of ou
r concepts about things and peo
ple.
But to reach th
ose condu-
sion
s we fir
st must put reg
ular
and irregular ver
bs under a more pow
erfu
l mag-
nify
ing glass, where we will find some unexpected fingerprints.
2
j ~
egular and irregular words have long served as metaphors for the law-
abiding and the quirky. Psychology textbook point to children's errors
I like breal ed and goecl as evidence that we are apattern-loving, exception-
hating species, e~cplaining everything from why children have trouble learning
simple laws of physics to why adults make errors when using computers or di-
agnosing diseases. In 1984 George Orwell has the state banning irregular
verbs as a sign of its determination to crush the human spirit; in 1989 the
writer of a personal ad in the Neiv York Review of Books asked, "Are you an ir-
regular verb?" as a sign of her determination to exalt it.
~; Science is not always I:ind to fo111ore from the natural world. Elephants do
forget, lemmings don't commit mass suicide, two snowflakes can be alike, we
use more than 5 percent of our brains, and Eskimos don't have a hundred
words for snow We had better give irregular and regular verbs a closer look be-
fore using them as evidence for a language faculty that works by words and
rules, or more generally, a mind that works by lookup and computation.
'~ Regular anduregular forms do not work in isolation; they are part of the in-
tegrated living system we call a language. This chapter will tease out regular
inflection from the linguistic organs and tissues in which it is
embedded. The
next chapter, on irregular verbs, will have a different feel. Living creatures can
be dissected, but creatures dead so long that only a trace of the living organs
21
23 ~
Wo
rds and Ru
les
remain must be excavated. Our tour of
the irregular ver
bs wil
l uncover them
from layers of
his
tori
cal sediment lai
d down ove
r th
ousa
nds of
yea
rs.
Does language even luive an anatomy? Many peo
ple th
ink about la
ngua
ge in
the fo
llow
ing way: We need to communicate, and lan
guag
e is
the fulfillment of
that
need. For every ide
a there is
a word and vic
e-ve
rsa,
and we utt
er the
words in an ord
er that re
flec
ts the
con
nect
ions
among ideas. If thi
s common-
sens
e view is tr
ue, th
ere would be little need to speak of la
ngua
ge bei
ng a com-
plex
sys
tem.
The com
plex
ity would res
ide in the
meanings, and language
would reflect that co
mple
xity
dir
ectl
y.
The point of th
is cha
pter
is to show that this vietiv is
mistaken. I ~n
~ill
put reb
ular ver
bs under a microscope to
reveal the delicate anatomy that makes them
~vork.'Language do
es express meaning as sound, of course, bu
t no
t in
a single
step. Sentences are put together on an assembly lin
e composed of mental
modules, shown on the fol
lo~~
ing pa
ge. One is a st
oreh
ouse
of memorized
word
s, the mental le.~icon. Another is a team of rules that combine words and
part
s of
words into bier ~n~ords, a component cal
led ~r
zarp
holo
gy. A thi
rd is a
team of ru
les th
at combine words int
o phrases and sentences, a component
called syntc~~. The thr
ee components pass messages about meaning back and
forth with the res
t of
the mind so th
at the words correspond to what the
speaker wants to say. This in
terf
ace between lan
guag
e and mind is called se-
naantics. Finally, the assembled wor
ds, ph
rase
s, and sentences are
massaged by
a se
t of
rules int
o a sound pattern that we can pronounce when speaking or ex-
trac
t from the stream of
noi
se when lis
teni
ng. Th
is interface between language
and the mouth and ear
is called phonology.
Many people are suspicious ofbox-and-arrow diagrams of
the mind. The
walls of
the boxes and the
pat
hs of the arrows oft
en seem arbitrary, and cou
ld
just
as easily have been drawn dif
fere
ntly
. In the case of
lan
guag
e, how
ever
,
thes
e components pop out as we tease apa
rt the phenomena, and at least
some of the di
visi
ons are now becoming visible in the li
ving
brain, as
eve wil
l
see in
chapter 9.1 Thi
s chapter wi
ll e,~
cplo
re the
kin
ds of discoveries that have
led li
ngui
sts to div
ide la
ngua
ge int
o parts, usi
ng onl
y th
e fa
cts of
reg
ular
and ir-
regular words. Fir
st, we will se
e why the
lex
icon
is di
ffer
ent from the two
boxe
s of
rules to the ri
ght,
then why morphology is in
a different box from syn
-
tat,
and fin
ally
, why phonology and sem
anti
cs each gets a bo
x.
The eas
iest
boxes to keep separate ought to be the
box
es containing words and
rules. From the discussion in the preceding chapter, it
should be clear tha
t a
Diss
ecti
on by Linguistics
~ 23
D4ou
th and Ear
s
2
Phon
olog
y(r
u]es
that de
fine
the sound pattern
of a language)
Lexicon
Morphology
Syntae
(sto
red en
trie
s ,_~
(rul
es for
for
ming
__,~,
(rul
es for
for ~+cords,
comp
lex wo
rds,
forming phrases
including irregulars)
including re
gula
rs)
and sentences)
Semantics
(mea
ning
s ex
pres
sed
through language)
1Beliefs an
d De
sire
s
simp
le ~~~ord like duck belongs in the lexicon to the left in
the dia
gram
. Just as
clea
rly,
a sen
tenc
e li
ke Daf
fy is a duc
l~ is
ass
embl
ed by the
rules of syntax in th
ebox on the
rig
ht. A
ccor
ding
to th
e ~n~ords-and-rules theory, in
e~ul
ar forms such
as swarrz are al
so words that come from the lex
icon
, because th
ey are
as ar
bi-
trary as dud
z. lA~
hat do we do the
n ~d
th reg
ular
forms lik
e ga
iack
ed? They look
like words and sound lik
e words, but
I have been insisting they don't have to
be sto
red in the
lexicon. They don't seem lik
e wo
rds,
but the}' don't seem like
sent
ence
s ei
ther
, which are
the cle
ares
t products of rules.
The problem is that the
ter
ms word and rule come from eve
ryda
y pa
rlan
ce and
are as
sci
enti
fica
lly fuzzy as oth
er vernacular te
rms,
like
~nc
g and rocH. On closer
exam
inat
ion,
the word word has
t~v
o ve
ry different sen
ses.
'- The first sense
matc
hes the ev
eryd
ay not
ion of
a wor
d: a stretch of sound that eg
ress
es a con-
cept, that is p
rint
ed as a string of le
tter
s between white spa
ces,
and that may be
combined with other wo
rds to form phr
ases
and sen
tenc
es. Some of these wo
rds
are st
ored
who
le in th
e lexicon, li
ke duck and su~ana; others ar
e assembled out of
24
~ Words acrd Rccles
,smaller bit
s by rules of morphology such as
q~iacl~ecl and dLi
clz-
bill
ed platy
ties
. A
technical term for a word in this sense is a nzor~hological object, to
be distin-
guished from phrases and sentences, which are syntactic objects.
The second sense of word is a stretch of sound that has to be memorized be-
cause
it cannot be generated by rules. Some- memorized chunks are smaller
than a word in the first sense, such as prefixes like un- and re- and suffices like
-able and -ed. Others are larger than a ~sord in the
first sense, such as idioms,
cliches, and collocations. Idioms are phrases whose meanings cannot be com-
puted out of their parts, such as eat yozar heart out and beat around the bush.
Collocations and cliches are strings of ~~~ords that are remembered as wholes
and often used together, such as gone with the wind or like two peas in a pod.
People know tens of thousands of these expressions; the linguist Ray Jackend-
off refers to them as "the Wheel of Fortune lexicon," after the game show in
which contestants guess a familiar expression from a few fragments. A chunk
o£ any size that has to be memorized—prefix, su
ffix, whole word, idiom, co
llo-
cation—is the second sense of word. It is the sense of word that contrasts with
rule, and the sense I had in mind when choosing the
title of this book. A mem-
orized chunk is sometimes called a
liste~~ze, that is, an item
that has to be
memorized as part of a list; one could argue that this book ought to have been
called Listenaes arul Rules.
So walked
is a word in the
first sense (a morphological object) and not a
word in the second sense (a listeme); its listemes are tival~ and -ecl. These
one-
part listemes—prefixes, suffixes, and the stems they attach to, such as
walk—are called morphemes, a term coined by the nineteenth-century lin-
guist Baudouin de Courtenay to refer to "that part of a word which is en-
dowed with psychological autonomy and is for the very same reasons not
further divisible."3
What about the rules? Why divide the rules of morphology, which build com-
plex words (including regular plurals and past-tense forms), from the rules of
syntax, which build phrases and sentences? Both are productive, recursive,
combinatorial systems, and some linguists see them as two parts of a larger
system.`'Yet all linguists recognize that they are not identical. This may seem
of no interest to anyone but a student cramminb for a Linguistics 101 final,
but in fact it has been a source of countless barroom arguments, late-night
dorm-
room debates, and irreconcilable differences.
Dissection by Linguistics
~ 25
What is the correct word for people who pass by: pcuserbys or passersby? Do
nervous fiancees dread the first meeting of the mother-in-laws or the n2others-
in-law> Who did Richard Nixon force to resign: a series of Attorney Generals,
or a series of At
torneys General? Here are a few real -
life examples:
Dear Ms. Grammar,
A member of the Fr
iday
Night Couples League ... had a hole in one on [he thi
rdhole and another on the fifrh. Did he have two holes in one or
two hole in ones?
One of us bel
ieve
s that the pattern should be the same as in attorne}~s general
and passersb~~. The other disagrees, believing that holes in one would indicate
that the golfer gained multiple holes in one shot. A Diet Coke has been wagered
on [his, and we have ab eed that Ms. Grammar sha11 be the final authorihcs
SPOONFULS
From a recipe: "Now throw in nuo tablespoons ful
l of chopped parsley and cook
ten minutes more. The quail ought to be tender by then." Never mind the quail;
how are we ever going to get those tablespoons tender? The word, of course, is
tablespoonfi~Is, no matter how illoa cal it seems. One dictionary contains the en-
try sp
oons
ful,
but phis is not generally accepted b
Gin and tonic season (no hyphens, please) is just about finished, but Joe Gale-
ota of West Roxbury would still like to know how to order when he's having more
than one. "
Friends advised me that the answer is 'd ns and tonic' because alcohol
is the main ingredient," he writes.%
Never has the U,S. faced a worse crisis than in 1887, afrer the invention of the
Jack
-in -the-Box. It had become a fad overnight, and everyone was having a
whale of a time when someone asked, "What is its plural?" "Jack
-in -the -Boxes!"
claimed some. Others hotly insisted, "J
acks-in-the-BoxP' Civil war seemed in-
evitable, when Zeke Kelp's Crusade won a compromise on "Jacks-
in-the-Boxes."
Unthanked for forty-three years, Kelp will be honored next week when N. 1.
City unveils a hydrant in his name s
All right, the last example isn't from real life; it
's from the Early Cartoons and
Writing of Dr. Seuss. The others are from well-known language columnists.
Hole
-in-one is from Ms. Grammar, th
e nom de plume of Barbara Walraff when
presiding over "Word Court" in
the Atlantic Monthly. Spoonful i
s from
Theodore Bernstein, the late New York Tinxes editor who wrote the syndicated
26 ~
Words and Riles
column "Bernstein on Words." Gin'and tonic is from Jan Freeman, who dis-
penses "The Ward" in
the Boston Globe.
People disagree on how to pluralize nouns, and they care about who is cor-
rect. Purists insist that the -s belongs on the noun in the middle of the expres-
sion (~zotaries publiq r
unners-zi.p), and those with the common touch are
content to leave it at the end (notar}~ pti+blics, runner-ups). "Ms. Grammar" ad-
vised her beseechers that holes in one is technically correct, but added,'"to say
`two holes in one' is
to ask to be misunderstood." Her Solomonic suggestion
was to say a Bole in one
ti.vice, and to buy t~vo Diet Cokes.
For my purpose—figuring out how the human mind deals with language—
there is no correct answer. Most disputes about "correct' usage are questions of
custom and a
uthority rather than g
rammatical logic (see "The Language
Mavens" in my book Tie Langiin~e Instinct), and in these disputes in parricular,
both parties have b ammatical logic on their side. Their agony highlights the dis-
tinctions among lexicon, morphology, and syntax, and illustrates the theme of
this book: that the mind analyzes every stretch of language as some mixture of
memorized chunks and rule-governed assemblies. How people pluralize an ex-
pression depends on how they tacitly analyze
it: a
s a word or as a phrase.
With a simple word the plural suffix goes at the end: one ~iri, t
~vo girls. Now
what happens in a compound word composed of t~vo simple words, such as cr
n~~-
girl? The plural
still goes on the end: tivo cowgirls, not tivo cmvsgirl ar tu~o cows-
girls. That is because the word,oirl inside crnvgirl is special. It
is called the head of
the word, and it stands for the word as a whole in determining its meaning (a
cowe 1 i
s a kind of girl) and in determining its plural: The -s
Goes on gi
rl. A
pl2rase
also has a head, and it too determines the meaning and gets the plural. b'ut now
~-ve discover the major difference between a word, the product of morpholow,
and a phrase, the product of syntax: In the phrase, the head is on the left, not the
right. If you meet more than one girl from Ipanema (head =girl), they are girls
front Ipnnenaa, not girl fr
o~~z I~anenxas. With a word the plural is
on the end (c
rnv-
girls); v
~~ith a phrase the plural can be in the middle (g
irls front I~asaenuz).9
The seeds of the mother-iss-Iativ dispute were sown by a special option of En-
glish: Occasionally a phrase gets repackaged into a long word. For example, a
hangover victim may complain of a bottorix-of-the-birdcage taste in her mouth;
the phrase bottom of th
e I~irdcnge has been packaged as a word that modifies
taste. When aword-made-from-a-phrase
is new and fresh, speakers still can
perceive the anatomy of the phrase inside the word. For example, we parse the
modifier bottom-
of-the-birdcage to understand that it means something as foul
as the bottom of a birdcage.
Dissectiotx I~~ L
inguistics
~ ?7
But when the phrase is used as a ti~ord repeatedly, the original meaning can
recede from
collective memory. The phrase boundaries melt into a glob, and
speakers no longer sense its parts. No one thinks of Thacrsclay as Tlzor's Da~~
anymore, or of breakfast as hrenlzing
cz fa
st. Modern English has thousands of
former phrases and complex words that have congealed into what people no~~
perceive as simple words, such as
bztsiness (busyness), Christnxas (Christ's
Mass), and spinster (one who spins). The meltdown, of course, does not hap-
pen o~~ernight or in
all speakers at once; there must have been a time when
some English speakers
still heard
Christti~acu as Christ's Mass and others heard
it as the arbitrary name of the holiday, ju
st as today's older speakers hear the
awe in awesome where younger speakers hear the whole word as a synonym for
good.
Most of our disputed plurals originated as phrases and then became words.
Long ago people might have thought, "she is not my mother
i~a reality; she is
only my mother in law" (that is, according to canon or Church law). But the
concept of a spouse's mother needs a word, and eventually the phrase got re-
analyzed as that word: "She is my mother-in-lmv." Similar meltdowns occurred
in these phrases:
Jack is in the boy —~ That is
a Jacl -in-the-box.
Phyllis completed that hole in one shot —+ She got ahole-in-ofze.
Barry passed by --
~ He is a passerby.
I set aside a spoon full of parsley —; I set aside a s~oonfi.cl.
If some speakers
still hear the phrase inside the word, they will be tempted to
put the plural marker on the head of the phrase: two mother +sin law, Jack + s
in a box, hole +sin one, pcuser + s Ir
}; spoon + s fu
ll. But if s
peakers glom the
words together in their minds, they will be tempted to put the plural marker at
the end: nz
otheri~xlcrw + s,
jnckinthel~ox + es, pa
sser]r~~ + s, holeinone + s,
spoon-
f241 -F S.
It's not that phrase hearers interpret these expressions literally (for example,
that amother-in-law is a mother as recognized by the law), or that the phrase-
deaf treat them as any old string of consonants and vowels; both surely recob
nine them as complex words built out of familiar words. It
's just that they grow
different kinds of connective tissue when piecing these e;cpressions together.
Those who would describe themselves as sons-in-lativ hear mother as the head
of a phrase inside the word (shown in the left tree in the diagram); those who
28 ~
Words and Rules
.would describe themselves as son-
in-laws hear a string of li
ttle words inside
the big word (right tree): N
N
NP
N
P N I
N
pP
mother i
n law
I / \
mother P
NP
I in
law
A proof that the in-iaw e~tpressions have congealed into words maybe found in
the umbrella word in-law, which can stand alone and be pluralized in the
usual way: The in-laws are conning over. It is a good bet that many of today's
commonly used phrases will also become opaque some day and turn into
words; the giveaway will be a plural at the end. Don't be surprised if one day
you hear about grant-in-aids, bill of ladings, or woriz, of
ans.
This ambiguity—one stretch of sound, two ways of building a tree in the
mind—also started the controversy raised by reports such as the following:
While Mo Vaughn should finish well over 300 with close to 40 home runs and
more than 100 RBIs, Mike Piazza has not been producing anywhere close to
what he did las[ season, when he
hit .362 with 40 homers and 124 RBIs.1
0
Baseball purists who deplore a
rtificial turf and the designated
hitter get
equally incensed by the plural form RBIs. RBI is
an acronym for run batted in,
a run scored by a teammate as a consequence of one's batting the ball. An RBI
and then another RBI are two runs batted in, and the acronym for runs batted
in is just RBI—so it should be I24 RBI, not 124 RBIs. (The purists are not
mollified by the sportscasters' common alternative, ribbies.) But the purists fail
to recognize that acronyms, like phrases, can [urn into bona fide words as a
language evolves, as in TV, VCR, UFO, SOB, and PC. Once an acronym has
become a word there is no reason not to treat it
as a word, including adding a
plural suffix to it. Would anyone really talk about three JP (justices of the
peace), five POW (p
risoners of war), or nine SOB (sons of bitches)?
An additional puzzle surrounds governors-generc~I, solicitors-general, and
attorneys-general. The speakers who bequeathed the plurals to us must have
Dissection by Linguistics
~ 29
analyzed the words as phrases, which have their heads on the left. Indeed, a
governor-general is a general governor, namely, one ~~ho has several governors
under him. The puzzle is, why didn't they simply call him a general governor?
After
all, the adjective comes before the head noun in English, not after it.
The ans~a~er is that these words, together with many other terms related to gov-
ernment, were borrowed from French when England was ruled by the Nor-
mans in the centuries after the invasion of William the Conqueror in 1066. In
French, the adjective can come rafter the head noun, as in Etats-Ufais (United
States) and chaise Iongue (long chair, garbled into the English chaise lounge).
The earliest citation in the Oxford English Dictionary is from 1292: "Tous at-
tomeyz general purrount lever fins et cinographer" (A
ll general attorneys may
levy fines and make legal documents). Anyone who insists that we eternally
analyze (hence pluralize) these words as they were analyzed in the minds of
the original speakers of Norman French
also should insist that we refer to
more than one major general as majors general, because amajor-general was
once a general major (from the French
nurjor-general). Long ago our linguistic
foreparents forgot the French connection and reanalyzed general from a modi-
fying adjective to a modified noun.
So if you are ever challenged for saying attort2ey-generals, nxother-in-laws,
pc+sserbys, RBIs, or Izole-in-owes, you can reply, "They are the very model of the
modern major general." They come from reanalyzing a phrase into a word, a
common development in the history of English, and a nice demonstration that
we treat stretches of language not as sounds linked directly to meanings but as
structured trees. People who put different trees on the same sound will use
the sound in different ways, even
if the meaning is the same.
Le['s now peer into the morphology box. Morphology maybe divided into de-
rivation—rules that form a new word out of old words, like duckfeathers and
acnlzissable—and inflection—rules that modify a word to
fit its role in a sen-
tence, what language teachers call conjugation and declension. The past tense
and plural forms are examples of in
flection.
English inflection is famous among linguists for being so boring. Other lan-
guages exploit the combinatorial power of grammar to generate impressive
numbers of forms for each noun and verb. The verb in Spanish or Italian
comes in about
fifty forms: fi
rst, second, and third persons, each singular and
plural, each in present, past, and future tenses, each in indicative, subjunctive
and conditional moods, plus some imperative, participle, and infinitive forms.
30
~ Words and Rules
Languages out
side
the Indo-European family, such as th
ose spoken in Af
rica
or the Americas, can be even more prolific. In the Bantu language Kiwnjo, fo
r
example, a ver
b is
encrusted with prefixes and suf
fi~:
es tha
t multiply out
to ha
lf
a mi
llio
n combinations per
ver
b. ~ ~
But Eng
lish
speakers subsist on onl
y four.
open
opens
opened
open
ing
Stra
ngel
y enough, Eng
lish
b ammar does no
t ha
ve onl
y four rol
es for
ver
bs to
play. It has
at least th
irte
en dif
fere
nt roles, but it sh
ares
the four forms among
them, as if
suf
fixe
s were e~c
pens
ive and the designers of the la
ngua
ge wanted
to economize.
The first suf
Fix is
a silent bit of
not
hing
, -~
d, which when added to the stem
opeiz turns
it int
o the inflected fo
rm. open. You may wonder: Why say tha
t
spea
kers
hallucinate an ima
gina
ry suffix at
the
end of a word? The reason
is
that it di
stin
guis
hes th
e root or stem—the irreducible nugget found in th
e
ment
al dictionar
y th
at cap
ture
s th
e essence of
a ver
b and upon which sufFi~ces
aze hung—from a particular incarnation of
tha
t ve
rb with
a_ particular pe
rson
,
number, and tense. In
Eng
lish
they can sound the same—to open and I open—
which disguises the fac
t th
at they are different versions of the verb. In oth
er
lang
uage
s th
e form of th
e ve
rb tha
t you loo
k up in a di
ctio
nary
can
not be pro-
nounced. For example, in Sp
anis
h you can say
canto, cn
nt~i
s, ca~aten, aid so
on, lea
ving
can
t- as the stem, but you can nev
er say
can
t- by itself. Stems are
ther
efor
e no
t the same thi
ngs as pronounceable ver
b forms, and that distinc-
tion is useful to preserve in English—to open ver
sus open0---even though the
two forms sometimes sound the same.
The suffix, -m is
used in
fou
r variations of th
e verb in En
glis
h:
Pres
ent tense, al
l but thi
rd-person singular: I, you, ev
e, the
y opefa
it.
Infi
niti
ve: They may open it,
They tried to op
en it.
Impe
rati
ve: Open!
Subj
unct
ive:
They insisted th
at it open.
The suffix -s is used for
onl
y one pur
pose
:
Pres
ent te
nse,
thi
rd-person singular: He, she
, it
opens the doo
r.
'. DtSS2Cf1011 ~ j,tnpulSliCS 3 ]
~ ~ -
iThe suf
fix -ing is used in at
least four wa
ys:
Pro~essive par
tici
ple.
He is op
enin
g it
.~
Pres
ent pa
rtic
iple
: He tried ope
ning
the
door.
Uerb
al noun (gerund): His incessant openi~zg of
the
box
es:
~ Uerbal adj
ecti
ve: Aq
uiet
ly-o
peni
ng doo
r.'i ~
Finally we come to ou
r friend -ed
, which has
four jobs:
Past
tense: It
ope
ned.
Perf
ect Pa
rtic
iple
: It has
ope
ned.
Pass
ive Pa
rtic
iple
: It was bei
ng ope
ned.
Verbal adj
ecti
ve: Ar
ecen
tly-
ape~
aed boa.''
Why make all the
se distinctions among ver
b forms that sound the same?
One rea
son
is tha
t the
list
of phrases ca
llin
g fo
r a form such as opened have
noth
ing in
common: To cap
ture
the
behavior of
-ed, we hav
e no choice bu
t to
list fou
r phrase typ
es sep
arat
ely.
Another rea
son
is that some dis
tinc
tion
s th
atare in
audi
ble for regular ve
rbs are audible fo
r irregular ones, and thi
s shows
that Eng
lish
spe
aker
s register these distinctions as the
y sp
eak.
About a thi
rd of
the irre u
lar ve
rbs have differ ent fo
g
rms for
the stem th
e asttense and the
Pperfect pa
rtic
iple
: I sing, I san
g, I have su
ng; I eat
, I ate
, I hav
e ea
teia
. A fe~
vI
make a fur
ther
distinction and have a sp
ecia
l form for the
ver
bal adjective--c
:j
saewly wedded co2tiple; c~
clrunlzen sailor; a sl2
runk
esa he
ad; ro
tten
eggs—which is
~, no
t used for
the par
tici
ple:
peo
ple say They hav
e wed, not
wedded; He has
drus
ak, no
t dr
a~nl
en; It
lza
s sh
runk
, no
t shrunlzen; The ems luive rot
ted,
not
rot-
t te
n. And one ver
b comes in ei
ght different forms:
~i In
fini
tive
; su
bjun
ctiv
e; imp
erat
ive:
To be or not
to be
; Let it be; Be
prep
ared
.
Pres
ent te
nse,
first -person singular: I ~m the
walrus.
I Pr
esen
t te
nse,
sec
ond -person sin
gula
r al
l pe
rson
s plural: You/we/they
h ar
e fa
mily
.
Pres
ent te
nse,
thi
rd-person singular: He/she/it is the
rock.
Past tense, first- and thi
rd-person singular: I/he/she/it was bor
n by
the river
~ Past tense, second -person singular, all per
sons
plural; sub
junc
tive
:
h' The way we/
you th
ey wer
e; If I were a rich man.
32 ~
Words and Rules
Progressive and present participle; gerund: Yo
u're being silly; It s not
easy being green; Being and Nothingness.
Perfect participle: I've been a puppet, a pauper, a pirate, a poet, a
pawn and a king.
With nouns, too, different grammatical forms have to dip into the same
small pool of suffixes. The naked stem dog must be distinguished from the sin-
gular dog + m because a dogcatcher doesn't catch just one dog and a dog Inver
doesn't l
ove just one. The dog inside these compounds refers to dogs in general
and thus differs in meaning from the singular form in a dog. The plural dogs
uses -s, which we have already met in the verb system in She opens the
-door.
The possessive forms dog's (singular) and dogs' (plural) use it too; the three
noun forms dogs, do
g's, and dogs' di
ffer only in punctuation.
All this redundancy suggests that regular inflection in English is remarkably
simple. All the inflections are suFfires; none of the grammatical roles call for a
prefix or some other way of decorating or tinkering with a word. And every
word has at most one inflectional suffix. We never get apensed or opensing, nor
do the plural -s and possessive
's stack up when several owners own something:
the dogs' b
1anl:et, not the dogs's (
dogzez) blanket. Finally, each niblet of sound
making up a suffix has a
life of its own and combines with several verb forms,
noun forms, or both, ra
ther than being a slave to only one role. This suggests
that instead of creditinb English speakers with seventeen verbose rules like
"To form the past tense, add -ed to the end of the verb;' we can credit them
with just one rule:13 "A word may be composed of a stem followed by a suffix,"
like the simple rule shown on page 16. Al
l the other details can be handled by
assuming that suffixes are stored in the mental lexicon with entries like those
for words, perhaps something like this:
-ed so
und: d
part of speech: suffix
use 1: past tense of a verb
use 2: perfect participle of a verb
use 3: passive parriciple of a verb
use 4: adjective formed from a verb
By factoring seventeen verbose rules into one austere rule and four lexical
entries, one per suf£i:c, we not only save ink but get some insight into the men-
Dissection Ins Linguistics 133
tal organization of language. English
~o~~ld have used seventeen different
forms for its seventeen slots in the noun declension and verb oonjUgaeion: Pre-
fi~es such as ib-, tra-, and Tza-, suffixes such as -og, -ig, and -cab, and so on. In-
stead
the
slots share a few sounds (
-~, -ed, -s, -zng) and one position
(immediately following the verb). This miserliness, called syncretism, is
found
in language after language. Syncretism su~;ests that the mind keeps separate
accounts for the templates that build words (for example, "word =stem + suf-
fi~'), fo
r the scraps of sound that maybe added to words (-
s, -ed, and -ing), and
for the roles these additions can play (for example, plural, participle, impera-
tive).~~ A particular construction like the English past tense is a. min -and-
match affair, assembled by hool:ina together parts
also used in other
constructions. No one knows why languages like to recycle their suffixes and
other ways of modifying words. It's certainly not to save memory space, be-
cause the savings are
trivial. Perhaps the reason
is to help listeners recognize
when a ~rord is composed of a stem and a suffix rather than being a simple
stem. Whatever its purpose, syncretism shows that in the language system,
combination is in the blood; even the tiniest suffixes are combinations of
smaller parts.
Syncretism—one form, several roles—is one I:ind of violation of the simplest
conceivable system in which every sound has one meaning and vice -versa. The
other hind of violation—one role, several forms—is rampant in languages as
well; linguists call it allomorphy.~~ Take the regular past -tense suffix---or is it
suffixes? Though always spelled -ed, it
is pronounced in three different ways.
In walked, it
is pronounced t. In jogged, it
is pronounced c~
. And in patted, it is
pronounced id, where i is
a neutral vowel called "schwa." We also find allomor-
phy in the regular plural: The suffvc -s has three different forms in cats, dogs,
and Izorses.
Are there in fact three past -tense suffixes and three plural suffixes? In some
languages, we are forced to this messy conclusion. Dutch speakers, for exam-
ple, select either -en or -s as the regular plural, depending; on the sound of the
end of the noun. But in English the three-~~vay variation has a simpler explana-
tion, worked out by the linguists Arnold Zwicky and Alan Prince.,_0~2e past
tense suffix is stored in the lexicon, not three, and a separate module fiddles
with its pronunciation: the rules of phonolog}; which define the sound pattern
or accent of a language.16
3-}
~ Wo
rds and Ru
les
Why do we pronounce the past tense suffix as t in
tiv
nike
d, d in joked, and~d
in pat
ted?
The choice is
com
plet
ely predictable, and can be stated as
a list of
rule
s:
1. Use ~d if
the verb ends in t or d (fo
r example, in pa
tted
and
pc~c~ied).
2.If it do
esn'
t, use t if the ve
rb ends in an unvoiced consonant—
that is,
a consonant in which the voc
al cords don't buzz, namely
p, k, f,
s, sh
, c1i, and th (for example, tap
ped,
wal
l ed, gassed,
snif
fed,
passed, bas
hed,
touched, and fr
othe
d).
3. Use d for all oth
er verbs: th
ose ending in vowels, such as played
and glowed, and those ending in the voi
ced consonants 1,
r, na, n,
b, g, v,
z, j,
zh, and th (for example, sme
lled
, 7~aarred, slan~naed,
plan
ned,
scrubbed, peed, sa
ved, bu.
<ed, ti
crged, cam
oufl
aged
, and
bath
ed).
This
sounds lik
e something out
of the ta
e co
de. Let's se
e if we can do bet
ter.
The first thing to no
tice
is th
at nothing in these rules is
specific to the pas
t
tense. Other constructions tha
t use -ed work the same way.
t d
-~
Past
tense:
kick
ed
flowed
patt
ed
Perf
ect participle:
has kicked
has flogged
has pa
tted
Pass
ive pa
rtic
iple
_ v~~as lucked
was flogged
was patted
Verb
al adjective:
a licked dog
a flogged horse
a pa
tted
cat
Outs
ide the ve
rb system entirely is ye
t another -ed construction tha
t comes in
the three variations; it
turns a noun tha
t means ':X" into an adj
ecti
ve tha
t
means "having Y":
c
Nominal adj
ecti
ve:
hooked
saber-toothed
pimple-f
aced
foulmouthed
thick-ne
cked
d
~d
lonb nose
d one-handed
horned
talented
winged
kind
hear
ted
moneyed
warm-blooded
bad-
temp
ered
bareheaded
Diss
ecti
on by
Livcgacistics
~ 35
The regular plu
ral -s als
o comes in three fo
rms,
~n~
hich
you can hear in
hC~1
.t~J
LS~ C~OaS, and Hor
ses.
The variation mir
rors
t}ie pact tense uncannily. Use -i=
when the noun ends in a sibilant sou
nd: s, ~, sh, ~h
, j, or ch. If it
doe
sn't
, use s
if the
noun ends in an unv
oice
d co
nson
ant.
Use z for
all
oth
er nouns. In fa
ct,
not on
ly does thi
s pattern appear with the plural, it
appears with th
e other -s
suff
ixes
as well:
'
~zPlural:
hawks
dogs
hors
es3rd pe
rson
sin
gula
r:
hits
sh
eds
chooses
Poss
essi
ve:
Pat's
Fred's
George's
The var
iati
on even appears in versions of -s
that ar
en't
genuine suf
fixe
s. En-
glis
h speakers commonly con
trac
t th
e ve
rbs has, is
, and does to the
ir final con-
sona
nt and glu
e it
ont
o the end of the su
bjec
t, as in
Mona's left or Da
d's home.
Sure
enough, the con
trac
tion
is pronounced in th
ree wa
ys, depending on how
the noun end
s:
s>-i
?12
as:
Pat's eaten.
Fred
's eat
en.
George's eat
en.
u~
Pat's ea
ting
. Fr
ed's
eat
ing.
Ge
orge
's eat
ing.
does:
What
's he want?
Where's he liv
e?
That
's riot all. Eng
lish
has
an affective -s th
at can be used to form nicknames in
some dialects and arg
ots,
as in
Pops, Nloms, Fats, Pats, and Wil
ls (th
e prince
seco
nd in line to th
e British throne). That -s can also show up in emotionally
colo
red sl
ang such as bo
~a~e
rs and nut
s, similar to the -y and -o th
at give us
Batt
}~ and wacko. (Sometimes the
two suffixes are even used tog
ethe
r, as in
Patr
y, Sac
gsy,
lVlxcgsy, foo
tsie
, fat
so, and Ratso.) St
ill another ve
rsio
n of -s
appears
in adv
erbi
al forms such as xa
nawa
res,
nowadays, bes
ides
, baclavards, thereaboicts,
and nmi
ds)z
ips.
A fin
al use for s is as
a meaningless link joining the ~n~ords in
compounds such as he
r-nt
snxa
n, st
ates
nxa~
a, kiv
esma
n, bon
dsnu
rn, Scotsman, and
gra~atsnucnship. And yes
, al
l of these -s'
s can be pronounced eit
her as s or as z,
depe
ndin
g on the pre
cedi
ng con
sona
nt (i
t's ha
rd to come up with examples for
the th
ird column):
s Z
~,Af
fect
ive:
Po
ps, Patsy
Wills, bon
kers
36 ~
Words aful Rul
es
Adverbial:
thereabouts
towards, now
aday
s
Link in compound:
hunt
sman
la
ndsm
an
So we have fifteen su
ffic
es tha
t show the same three-way or two-~vay varia-
tion
. Forty-one suffices tha
t ha
ppen
to fa
ll into fi
ftee
n parallel sets of
alterna-
tive
s is too
much of a coincidence to
sto
mach
. More lik
ely,
one
set
of ru
les
creates th
e three-way va
riat
ion,
and
the
set applies in at
leas
t fi
ftee
n situations.
There
is a second, equally striking set of coincidences th
at runs across the
suff
ixes
. If the
var
iati
on came From any old set of if ... then ru
les,
eve wo
uld
e~cpect to fin
d al
l ki
nds of
pai
ring
s between st
ems an
d su
f$ae
s: for example,
"Use s after the
vow
els a an
d e or
after the consonants th and g," "Use d after a
lz,"
and so on. But the rul
es are far mor
e la
wful
tha
n th
at. The t sou
nd comes
after un
voic
ed consonants, and the t its
elf i
s un
voic
ed. The ~i
sou
nd comes af-
ter voiced sou
nds,
and the
d itself is voiced. The -s suffixes show the
same
cham
eleo
nlik
e be
havi
or: We find un
voic
ed s aft
er unv
oice
d consonants, and
voic
ed z after voi
ced co
nson
ants
. It
looks as if som
ethi
ng is trying to ke
ep the
cons
onan
ts at the end of a word consistent: All
of them are
voi
ced,
or
all of
them are unv
oice
d.
Inde
ed, something is—the sou
nd pat
tern
of th
e En
glis
h la
ngua
ge. English
neve
r forces speakers to turn their vo
cal co
rds on for one
consonant the
n off
for the ne
xt, or ~~ice-versa. We see
the restriction in force in
one-piece words
that
end in a cluster of consonants. These words never rec
eive
d a suffix; they
just
hap
pen to be bu
ilt that way
, so any sou
nd pat
tern
they display ca
nnot
have
come from a su
ffix
rul
e, but
rat
her fr
om the way Eng
lish
speakers li
ke to pro-
nounce wor
ds in ge
nera
l. In
all but on
e of
these wor
ds, the vo
cal co
rd swi
tch
can be
left in
the ̀bff' po
siti
on:
Afte
r Iz (u
nvoi
ced)
: scan occ
ur
z can
not occur
Gll'
, fT,
X~ ~70.Y,
-
t ca
n oc
cur
cI can
not oc
cur
act,
fi~ct, pro
duct
—
Afte
r p (un
voic
ed):
scan occ
ur
z can
not occur
traipse, lapse, co
rpse
—
t can occur
c~ can
not oc
cur
apt, opt, abriLpt
—
Afte
r t (un
voic
ed):
scan occ
ur
z can
not oc
cur
blit
z, Iz
ibit
z, Pot
ts
—
Afte
r s (u
nvoi
ced)
: t can occur
d can
not oc
cur
post
, gho
st,
list
—
Diss
ecti
on ]~
y Lin
guis
tics
~ 37
In one
Eng
lish
wor
d, ad
>e, the vocal co
rd switch is left in the "on"position:
Afte
r d (v
oice
d):
s ca
nnot
occ
ur
z can occ
ur—
adze
In no English wo
rd is the vo
icin
g switch toiled on and off, in an en
ding
lik
ezt
, gs, 1:
z, or sd
.These
diff
icul
t -to-p
rono
unce clu
ster
s can, how
ever
, be cre
ated
by a dumb
rule
of morphology tha
t pi
ns a suf
fix onto the end of a ~r
ord without regard for
how the resulting train of consonants is to
be pr
onou
nced
. Th
at- is wha
t ha
p-pe
ns when a rule adds a d sou
nd to tivalk or an
s sou
nd to cl
og. En
glis
h cleans
up the
se awkward mismatches with a different kind of
rule. The rule says,
"When the
re is a cluster of
consonants at
the end of a syllable, ad
just
the voi
c-in
g se
ttin
g of the Iast co
nson
ant to make it
con
sist
ent with its
neighbor on the
left
." (I
n ot
her words, cha
nge hz
to lu
, pd
to pt
, an
d so on.
) The rule does not
care
whe
ther
the syllable was for
med by
a past -tense
suffi~c, a plural su
ffix
, a
contracted has
, a ni
ckna
me wit
h -s, o
r an
ythi
ng else. It ki
cks in aft
er the
sylla-
ble ha
s be
en ass
embl
ed, in the cle
anup
module we cal
l phonology.
Can we now tel
l wh
ethe
r the su
ffix
sto
red in
the
lex
icon
is -d
, an
d is con
-verted to a t when it finds it
self
at the end of wc~lh, or wh
ethe
r it is -t
and
is
conv
erte
d to
c~ when it f
inds its
elf a
t the end of
jog? A little detective work can
sett
le the que
stio
n. Not every sou
nd cares about the con
sona
nt tha
t follows
it.
Thos
e th
at do are consonants in wh
ich th
e ai
rstr
eam is obstructed, nam
ely ~,
b, t,
d, '
k, g, s
, sh, ch,
z, zh
, an
d tla. But the
vowels, and
the vow
el-l
ike co
nso-
nant
s r, 1,
sa, and na, are
indifferent to wh
at comes after the
m; they to
lera
te ei-
ther
s or z,
eith
er t or
d, as we see in these on
e-pi
ece wo
rds:
After
M1a:
scan occ
ur
z can also oc
cur
fence
leas
t can occur
d can also oc
cur
lent
lend
After
r:
scan occur
z can also oc
cur
force
fii re
t can oc
cur
d can also oc
cur
}ort
ford
After
1:
scan occur
z can also oc
cur
puls
e Stolz
t can occur
cl can also oc
cur
guil
t gacild
3S ~
Words and Rules
After a vowel:
scan occur
z can also occur
niece
snee<e
t can occur
d can also occur
goat
goad
Here we have laissez-faire environments in which the suffices can show
their true colors, untouched by rules of phonology. What do we find? That the
virgin suffices are pronounced -d and -z, not -t and -s:
After n:
After r:
After
I:
After a vo~~el:
we don't say s
w~e don't say t
we don't say s
~n~e don't say t
~.ve don't say s
we don't say t
we don't say s
eve don't say t
we say<
grins (grim), piers (Pinz)
we say c~
grinned
we say<
tivears (wen), co
res (kon)
we say c~
feared
eve say z
calls (
kolz), balls (bolt)
«~e say d
snaffled, w
ell-heeled
we say z
flees (flez), flecu (flez)
we say d
flrnved
The -t and -s ~~>e hear in words with choosy sounds such as walked and cats
must be the aftermath of the rule.
Finally, ~n~hat about the funny extra vowel in patted and korses? Here again
the change in sound is not some random act of vandalism. The vowel appears
when d follows t or d, and when z follows s or z. The ~~ord endings that trigger
the extra vowel are similar in pronunciation to the sufFixes themselves, and that
can't be a coincidence. Apparently a rule is trying to separate too-similar adja-
cent consonants by pushing a vowel between them: between t and d, d and d, s
and z, z and z, sh and >, and so on. In many languabes the rules of phonology do
something when a rule of morphology leaves two identical or near-identical
consonants in a row, presumably because there's no natural way to pronounce
them. Some languages drop the second consonant, others merge the two into
Dissection by Linguistics
~ 39
one long consonant, and s
till others, like English, wedge a vowel between
them. As with the rule that fiddles with voicing, the rule that inserts a vowel
must live in a phonology module separate from rules that stick on the various
suffixes, because the rule is oblivious to what hind of suffix it manipulates.
We even can deduce which of the two rules applies first, the one that
changes the voicing setring or the one that inserts the vowel_ The devoicing
rule is triggered by adjacent consonants; the vowel rule breaks up adjacent
consonants. If the voicing rule came first, it would convert pct + d to pat + t,
and only then would the vowel be inserted, yielding pant:
Morphology:
pat + d
Devoicing:
pat + t
Vo~~>el insertion:
pat + ~r + t
But that is not how we pronounce it; we say pdtid. This means that the vowel
rule must have come first, creating patted; now the voicinb rule. is
no longer
compelled to do anything, because the td sequence that would trigger it has
been broken up:
Morphology:
pat + d
Vowel insertion: pat + ~+ d
Devoicing:
not triggered
The ordering makes sense when you think about ho~~ the phonology module
should be organized. It has some rules that edit the string of vowels and conso-
nants composing a word (phonology proper), and other rules that comrert the
string into
actual sounds or muscle movements (phonetics). The vowel
-insertion rule makes a major change in the stuff that makes up a word, and be-
longs in the first subcomponent; the voicing rule does alast-minute adjustment
of pronunciation for the benefit of the muscles, and belongs in the second. ~
%This completes the analysis of the three versions of the past-tense
suffice.
When we started, eve needed forty-odd rules, each stipulating that some suffix
be placed next to some word ending. We have ended up with just two rules.
Best o£ all, what the rules do, why they do it, and in what order they do it all
40 ~
Wo
rds
arecl R
ifle
s
make sen
se in the li
ght of the sou
nd pattern of En
glis
h. Ind
eed,
this
kin
d of
layering may be found in languages
all ov
er the wor
ld.
Incidentally, t
here
is cor
robo
rati
ng evidence of a completely different ki
nd that
show
s that the thr
ee for
ms of -ed and
-s are
created on th
e fly by
a pho
nolo
gica
l
rule. Some psy
chol
ingu
ists
keep a pad and pen
cil in their pockets and
wri
te
down every sli
p of the
ton
gue they hear. Peo
ple make one
or t~vo such errors for
ever
y th
ousa
nd words the
y say, and many of the errors con
sist
in deleting, r
epeat-
ing,
or switching around vowels or consonants.is'Fhe last I
dnd of
err
or is called a
Spoo
neri
sm, in
honor of the Reverend Wil
liam
Spooner (1844-1930},warden of
New Col
lege
at Oxford, who came out w2th surprises such as O~ir que
er old
clean,
You hav
e hi
ssed
all
my r~z
yste
ry les
sons
and tas
ted th
e wh
ole wornz, and It is
naw
kistomary to
cus
s th
e br
ide.
They sound too goo
d to be tru
e, but I hav
e he
ard si
m-
ilar errors myself. Af
ter I sp
oke at
a scientific symposium th
e ch
air wr
appe
d up
the session by
saying I
would like to spank the
speakers, and
when I ask
ed a fri
end
how he li
ked hi
s new condominium, he said It se
ats my ni
icle
s.
Speech err
ors provide clues on
-how the speech sy
stem
is or
gani
zed.
For
ex-
ample, when a per
son intends to
say grapefruits but accidentally le
aves
out the
t, how does he pro
noun
ce the plural? If th
ere we
re a dis
tinc
t plural suf
fix pro-
nounced
-ss, he would say gra
pefr
ooss
, since th
is is wh
at the t in the gr
apef
rtii
it
entr
y would have demanded. In fact he says grapefrooz—pronouncing the
plur
al as z, which is appropriate to words end
ing in a vow
el.1
9 Si
mila
rly,
a per-
son may say The infant ti~.clzs—toaiches th
e ni
pple
, no
t tucl -zs, or may say Did
you bu
}' ~o~igh breakfasis?, no
t ~reczkfass. The errors show tha
t the form of the
suff
ix must be computed aft
er the vowels and consonants of
the
noun or ve
rb
were placed on the chute to th
e vo
cal tr
act.
Engl
ish di
d no
t al
ways
hav
e si
ngle-consonant suf
fixe
s and a rule that separates
them from ato
o-si
mila
r word ending. Our cur
rent
sys
tem
is the
res
ult of
a reor-
ganization tha
t began around the
time of the
ori
gin of
Modern Eng
lish
in th
e se
v-
enteenth century. Be
fore
that, -ed and -s su
ffic
es were pro
noun
ced (and spe
lled
)
with vowels
all the ti
me, no
t just with words ending in t or d or in
s or z. For cen
-
turies, En
glis
h speakers had been concentrating str
ess on the
fir
st syllables of
word
s, whi
ch shriveled the later syllables, an
d sp
eake
rs beg
an to le
ave ou
t the
vowe
ls in th
e su
ffix
es of many words. Wr
iter
s ca
lled
att
enti
on to the new, cl
ippe
d
pronunciations by sp
elli
ng them phonerically with an apostrophe in pl
ace of
the
dele
ted vowel, as in Sha
kesp
eare
's pla
y about "a
pair of st
ar-c
ross
'd lov
ers"
:
Deat
h, th
at has
suc
k'd the honey of thy
bre
ath,
Hath no power yet
upon thy
beauty:
Diss
ecti
on Icy
Linguistics
41
Thou art
not
conquer'd; be
auty
's ensign ye
tIs
crimson in th
y lips and in thy ch
eeks
.
The gua
rdia
ns of the En
glis
h la
ngua
ge deplored the change, as they do all
chan
ces.
In "A Pro
posa
l fo
r Correcting, Improving, and Ascertaininc th
e En-
glish Tongue," Jon
atha
n Sw
ift wrote:
What does yo
ur lor
dshi
p think of [he words "dr
udg'
d," "d
istu
rb'd
," "'
rebuk'd,"
"fledg'd," and a thousand others eve
rywh
ere to
be met ~a
~ith in prose as well as
I ve
rse?
Where, by leaving out a vowel ro
save a syllable, we form so
jarring a so
und,
and so difficult to ut
ter,
that I ha
ve oft
en won
dere
d how it
cou
ld eve
r obtain.
His con
temp
orar
y, Samuel Johnson, who was sta
ndar
dizi
ng the spellings of
Engl
ish words in a way tha
t re
flec
ted the morphemes tha
t composed them,
recognized that 'd and -ed were the same morpheme, and obliterated the dis-
tinction in th
eir spelling, making ed the spe
llin
g fo
r bo
th.'
-~ It is
unc
lear
why
he chose to le
ave th
e e in -ed across the board (mapped and mat
ted)
, bu
t op
ted
to spe
ll -s ei
ther
with or wit
hout
an e, depending on how it is
pronounced
(nui
ps and rruuses).
Today the old
syl
labi
c su
ffix
survives in a han
dful
of ad
ject
ives
:.ac
curs
ed,
aged
, beloti~ed, beaded (i
n th
e ex
pres
sion
ova
beaded knees), blessed, cro
olze
d,cu
ssed
, do
gged
, jc~
ooed
, le
arne
d, nalzed, rc~~~ed, wicl ed, and tivretched. (A few
more survive in rural dialects, such as f
orke
d, pea
ked,
streaked, and str
iped
.)''
Many of them are arc
haic
or po
etic
and are used mainly in sel
f-co
nsci
ous
speech. The psy
chol
ogis
t 1V
Ieli
ssa Bowerman, a researcher of
chi
ld lan
guag
e,had thi
s exchange with he
r four-year -ol
d da
ught
er about a class tri
p to
a nat-
ural his
tory
mus
eum:
'-'
n40T
HER
(play~ully): N(a
ybe you'll see
som
ethi
ng winged.
DAUGHTER: Maybe well see something snaked!
We've seen ~vhy the syntax box
, which builds ph
rase
s and sen
tenc
es, ha
s to be
separated from the morphology bo
x, whi
ch bui
lds words. We als
o have seen
why the
phonology box, which massages wor
ds into a pronounceable st
ream
of
42 ~
Wo
rds and
Rcales
sound, has
to be sep
arat
ed from syn
tax,
mor
phol
ogy,
and the
lex
icon
. But why
do eve
need separate boxes fo
r semantics (the tho
ught
s ex
pres
sed in
lan
guag
e)
and the
lexicon? Could we reduce the
dif
fere
nce between reg
ular
and irre~,ular
verb
s to a dif
fere
nce in meaning between the two kin
ds of ve
rbs,
rat
her than
putting one kind in the morphology box and the oth
er in the le
xico
n? Do we
even need to talk about an "en
try in
the
mental le
xico
n," th
e address in
mem-
ory that hol
ds a link to a sound and a link to a meaning? Or cou
ld we connect
thou
ghts
to sounds dir
ectl
y, eli
mina
ting
the
middleman? Here are some fac
ts
that
suggest tha
t we do need to credit the human mind with something like
dict
iona
ry entries.
Firs
t, the Eng
lish
irregular verbs could not
have ari
sen simply from a com-
munal effort to opt
imiz
e cl
arit
y. While irregular forms on ave
rage
are harder to
mistake fo
r th
eir base forms than regular forms are
(br
ing do
esn'
t sound lik
e
brought, nor take li
ke took.), many irregulars ar
e id
enti
cal to
the
ir base fo
rms:
Toda
y I hit
, ye
ster
day I hi
t; Tod
cry I
ptia
t, yesterday I pt
iit.
A sentence such as On
Wednesday I cut
the
gra
ss could mean las
t V~~ednesday, nex
t Wednesday, or
every Wednesday. If
citit were reg
ular
, the ambiguity would never ari
se: On
Wednesday I Gut
ted th
e grass would sin
gle ou
t the pr
eced
ing Wednesday. De-
spite the po
tent
ial am
bigu
ity,
how
ever
, twenty-eight Eng
lish
ver
bs insist on re-
mairiing unchanged in the pa
st tense.
Also
, ir
regu
lar forms do not
cor
rela
te with any kind of
meaning. Nlany ver
bs
are si
mila
r in meaning but have com
plet
ely di
ffer
ent pa
st-tense for
ms. For es
-
ample, hi
t, st
rilze, and skip
all refer to hitting. Hi
t is
an irr
egul
ar ~~e
rb tha
t do
es
not change in the pa
st tense: Today
tive
hit
golf balls; Yes
terd
ay use hit
golf bulls.
Strike is an irr
egul
ar verb that changes its vow
el, yi
eldi
ng stricck. And sla
p is
a
regu
lar ve
rb, with pas
t tense sl
appe
d.
Not only are th
ere verbs with sim
ilar
meanings and different past-tense
form
s, the
re are
ver
bs with different meanings and the same past-tense fo
rms.
Engl
ish has a class of ve
rbs li
ngui
sts ca
ll lig
ht ver
bs, such as come, ~o,
do,
talz
e,
have
, se
t, get, pt
ict,
and stnvul. Compared to or
dinary ver
bs they are le
ss filling; a
ligh
t ve
rb doe
sn't
have a meaning tha
t stays with it,
but takes on dozens of
meanings, especially in combination with pa
rtic
les such as in
, out, up, off
,
over
, and around:
come (move to he
re),
coi
~ze around (agree), come in to (inherit),
cone (re
ach or
gasm
), come off
cu (a
ppea
r), come ozi
t (d
ivul
ge ho-
mose:tuality), comae to (a~,vaken)
Diss
ecti
on 1~~
Linguistzcs
~ 43
go (move to th
ere)
, go out
with (date), go nut
s (dement), go i~a
for
(cho
ose)
, go of
f (emslode), g
o of
f (sp
oil)
do (act), do in (hill), do up (decorate), do a number on (ov
erwh
elm)
,do lainch (e
at together)
take (ca
use to go v«th), ta
lxe in (swindle), ta
ke off (launch), ta
pe in
(welcome), tak
e ov
er (us
urp)
, ta
ke icp (commence), tak
e a lea
k(u
rina
te),
tak
e a bath (lose money), tak
e c~
bath (b
athe
), tak
e cz
walk (wa
lk),
talz
e a loo
k (l
ook)
hai~e (possess), have (eat), luzve (se
duce
), hav
e a he
art (s
ympa
thiz
e),
have
over (entertain), hnve a coi
n (be ang
ry)
het (r
etri
eve)
, get
(become), ge
t over (su
rviv
e), ge
t ou
t (d
ivul
ge),
get
off on (en
joy)
, het a life (
self-i
mpro
ve)
set (p
lace
), se
t of
f (ig
nite
), se
t ac
p (arrange), se
t trip (tr
ick)
, se
t up (in
-troduce), se
t right (r
ectify
), se
t th
e st
age (p
repa
re)
put (cause to be'at), pzit off
(pro
cras
tina
te),
past
off
(off
end)
, ~ar.t one
over on (fool), pact do
nna (insult), put dawn (euthanize), put
i~x for
(req
uest
), put
oac
t extinguish), pact out (inconvenience), put
out
(con
sent
to sex)
stai
ad (rise), sta
nd out (impress), st
and up fo
r (defend), st
and in (re
-place), s
tand
off
(rep
el)
But in every in
stan
ce the
y retain their irregular past te
nse forms in the eY-
tend
ed meanings: Barney came aro
und,
Barney came out, Barney cnn
ae off as
(nev
er conzeci); Joan toolz hinz in, j
oczn took a bat
h, Joan to
ol over (never raked);
and so on
. Al
l the meanings march in lo
ckst
ep with the same irr
egul
ar past-
tens
e fo
rms,
no matter how tenuous the semantic th
read
tha
t li
nks them. The
mind links an irr
egul
ar sound such as took not
with the meaning of a word di-
rect
lybu
t with the word's root—a unique address in the mental lexicon, li
keth
e bo
ldfa
ced entry fo
r a word in a dictionary which can hav
e se
vera
l mean-
ings listed under it.'-3
An even more curious demonstration comes from families of
words with th
esame stem and dif
fere
nt pre
fixe
s. Words with prefixes keep the
past -tense
form of the stem: eat—ate becomes ove
reat
—ove
rate
; make—made becomes re-
nza~
Ze—r
emad
e. That is not surprising, because we a
ll hear th
e eat inside
overeat—overeating is,
aft
er all
, a kind of eating, na
mely
, eating too much.
What is surprising is that the same
- thing happens when the meaning of the
combination
is opaque. Few peo
ple se
nse the meaning of the stand inside
44 ~
Words and Rules
understand, the get inside fo
rget, or the come inside become. Nonetheless no
one is tempted to sa
y un
cier
stav
zded
, for
gett
ed, or becomed; the irregular forms
persist, giving us understood, forgot, and became. Here are some examples:
come~ance, become—became, overcome—overcame
go—went, undergo—undenuent
get bot, forget forgot
tnke—toolz, mi
stal
ze—m
isto
ok, overtczlze—overtook, partal:e~artook,
undertc~ ke
—%incle r
too k
set set, beset—beset, i
spset—ic~set
stand—stood, understand—understood, tivithstand ~,vithstood
draw-
-dre
w, withdrenv—withdrew
hold—held, hehoid—beheld, uphold—upheld, withholdwithheld
blue gave, forgii~e forgnve
Irregular forms stick like glue to their verb roots, even when reduced to
mean
ingl
ess
little tokens inside a biker verb. Speakers of English seem to an-
alyze become as be
- +come and understand as isn
cler
- + stazul, even though the
meaning of become is not computable from the meaning of be
- and the mean-
ing of come, and icn
ders
tasa
cl has nothing to do with st
andi
ng. Th
is is not some-
thing we have to Iearn in school. When we acquire language, our minds
analyze sets of wo
rds,
looking for their parts as if they were clues in a combi-
natorial puzzle. We mentally arrange them in a matri~t according to overlap:
be-
over-
under-
up-
with
cone
beco
v~2e
overcome
draw
withdrew
]col
d be
hold
tiipho2d
withhold
set
beset
upset
sta~ad
T-arulerstand
with
stan
d
talze
overtake
undertalze
and use the
common denominators in
the
rows and columns to make incisions
in the wor
ds, thinking of them the
reaf
ter as amalgams of parts: becrnrae =fie- +
covne>
withdraw = ur
ith- +draw, and so
on.'
-~
Of course, it
was English speakers of
cen
turies pas
t, our
linguistic ances-
tors, who first analyzed become as be- + con
2e and ext
ende
d the come—came
pattern to it, and i[ is
possible that to them the words were as transparently
Dissection IT
}~ Li
7zgu
isti
cs ~
45
built ou
t of parts as overeat or remake are to us today. Even so;
. it is
unlikely
t}]3[ VJfE ~13
Ve been StUplCl~y 111C'171oIiZing ~ie
ccz~
~ze,
ovesca„~e, w-<rl<drew, and so on
as structureless strings of vowels and consonants. If we were to come across a
new complex word, such as i~ndercov~ae, bestand, overbold, or withset, and were
unaware of it
s meaning, we would almost certainly use the irregular forms of
j th
e words inside them: ii
ndercanae (not undercomed), he
stoocl ove
rhel
d ,v
itlu
et.
~i Mo
reov
er, it is not pure sound that carries th
e irregular form: The past of su
.c-
cunab and encumber are
succicmbed and encumbered, not succnnae and encanze-
ber, because people don't perceive them as containing a prefix followed by the
word crn~ie, only th
e sound kung.
~i Clearly the perception of an embedded word comes from its spelling: be-
conxe contains c-o-nz-e; succicn2b doesn't. But spelling does not directly inform
speakers how to form the past tense; it merely assigns a distinct visual sig
na-
ture to ev
ery root, and speakers choose the past -tense form that goes with the
root. Samuel Johnson, who standardized the spellings of
thousands of modern
words, used pe
ople
's perception of
the anatomy of words as a rationale in
his
decisions, and that is
one of the reasons that the spellings of English words no-
tori
ousl
y do not always reflect th
eir sounds; often they reflect morphological
stru
ctur
e instead. We see thi
s in the many words that sound alike but are not
perceived as bei
ng the same word (that is, as having the
same root), and are
not given th
e same past -tense form:
nseet—tazet
versus
mete—nzetec~
nnp rang
versus
wring—wl
- asng
.bear—bore
versus
bare
—bar
edsteal—stole
versus
steel—steeled
frea
k—br
olze
versus
brake—bralzed
In the
last three ca
ses the spellings divulge the presence of words that are
rec-
ognizable in other guises—the adjective bare, the noun steel, th
e noun brake—
and we will see in chapter 6 that this makes an especially big difference in
how ~~e compute their past tense forms.'-'
The English system of inflection, we have seen, dissects cleanly into a few
simple components. Tie past -tense rule belongs to a component, morphology,
that bui
lds throbs out
of parts using rules. The rule itself is a masterpiece of
46
~ 4~
~ord
s arui Rul
es
minimalism—"a word can be composed of a stem and asuffix"—with all
oth
er
details distilled ou
t and col
lect
ed in the lexical en
try fo
r the suffix. The suf
fix
itself is shared among sev
eral
inf
lect
ions
(pa
st tense, participle, and so on),
and its var
iant
pro
nunc
iati
ons (t, d, -i
ci) do not
wastefully mu
ltip
ly listings but
are computed aut
omat
ical
ly by two ubiquitous rules of phonology. The distinc-
tion
ber
ivee
n the le
~tic
on (in
clud
ing ir
regu
lar inflection) and grammar (indud-
ing
regular in
flec
tion
) is
a d
isti
ncti
on between a
lis
t of entries and an
algo
rith
m fo
r combining them, rather th
an a side ef
fect
of a general ye
arni
ng
to distinguish mea
ning
s.
That lea
ves the irregular wo
rds.
Every irr
egul
ar tel
ls a sto
ry, and the
y ar
e the
topic of
the nex
t chapter.
3
~~~~1~` ~~~~~~~1`~~
n the
game known as Broken Telephone (or Chinese Whispers) a child
whis
pers
a phrase in
to the ear
o£ a second child, who whispers
it into the
ear of
a thi
rd child, and so on
. Distortions ac
cumu
late
, and when the
las
t child
announces the phr
ase,
it is co
mica
lly different from the original_ The game
~~orlcs because each child does no
t me
rely
degrade the phr
ase,
which would
culm
inat
e in
a mumble, but
recinalyzes it
, making a best guess about th
e words
the preceding ch
ild had in mind.
AlI la
ngua
ges change through the centuries. We do not
speak lik
e Shake-
spea
re (1564-1616), who did
not
speak lik
e Chaucer (1343-1400), who did
not speak like the au
thor
of Beowulf (around 750—s00). As the changes tak
epl
ace,
people fe
el the ground ero
ding
under the
ir feet and in every era have
pred
icte
d the imminent demise of the la
ngua
ge. Ye
t the tw
elve
hundred years
of changes since Beawaslf ha
ve not
lef
t us grunting like Tarzan, and tha
t is
be-
caus
e language change is a game of Broken Telephone.
A gen
erat
ion of speakers uses their lex
icon
and grammar to produce sen-
tences. The you
nger
generation li
sten
s to the sentences and tri
es to in
fer the
lexi
con and grammar, the remarkable feat we call Ia
ngua
ge acq
uisi
tion
. The
tran
smis
sion
of a lex
icon
and grammar in language acq
uisi
tion
is fairly hig
h in
fidelity—you pro
babl
y can communicate well
~n~i
th you
r pa
rent
s and you
r chil-
*For
a cha
rt th
at summarizes the history, da
tes,
and fam
ily affinities of t
he Eng
lish
language, se
epa
ge 212
.
47
~.z
;—
-~ ~~,"
~'~„~a
c.~.~
,~~ ,
,~.-
f-~, ~' .~ r,
~~
1 r~
r'L°°
~ ! ~. ~'I
J
n,,.,
L ~,
~~''' ~-
,9
;, s
r
''
`~
~~t~
~~
~~
~C,ID~]E~1E~~
anguage comes so naturally to us that it is easy to forbet what a strange
and miraculous b ft it i
s. All over the world members of our species fash-
ion their breath into hisses and hums and squeaks and pops and listen to oth-
ers do the same. We do this, of course, not only because we like the sounds
but because details of the sounds contain information about [he intentions of
the person making them. We humans are fitted with a means of sharing our
ideas, in
all their unfathomable vastness. When we listen to speech, we can be
led to think thoughts that have never been thought before and that never
would have occurred to us on our own. Behold, the bush burned with fire, and
the bush was not consumed. Man is born free, and everywhere he is in chains.
Emma Woodhouse, handsome, cl
ever, and rich, with a comfortable home and
happy disposition, seemed to unite some of the best hlessin~s of existence.
Energy equals mass times the speed of light squared. I have found it impossi-
ble to carry the heavy burden of responsibility and to discharge my duties as
King without the help and support of the woman I love.
Language has fascinated people for thousands of years, and linguists have
studied every detail, from the number of languages spoken in New Guinea to
why we say rc~zzle-daz`le instead of cI
a>zle-razzle. Yet to me the first and deepest
challenge in understanding language is accounting for its boundless expressive
power. WE~at is the trick behind our ability to
fill one another's heads with so
many different ideas?
2 ~
Words and Ru
les
The pre
mise
of th
is boo
k is that th
ere are two tricks, wor
ds and
rul
es. They
work by different pr
inci
ples
, are
Tea
med an
d us
ed in different ways, a
nd may eve
n
resi
de in different pa
rts of
the
brain. T
heir
bor
der di
sput
es sha
pe and
reshape lan-
guages over centuries, and
make lan
guag
e no
t on
ly a too
l fo
r communication bu
t
also a medium for wordplay an
d po
etry
and
an he
irlo
om of endless fascination.
The fi
rst trick, the wor
d, is
bas
ed on a memorized
arbitrary pa
irin
g between a
soun
d and a me
anin
g. "Wh
at's
in a name?" ask
s Juliet. "That whi
ch eve cal
l a
rose by any ot
her name wou
ld smell as sw
eet.
" What's in a name is th
at eve
ry-
one in
a lan
guag
e community tac
itly
ao~rees to us
e a pa
rric
ular
sou
nd to convey
a pa
rtic
ular
ide
a. Alt
houg
h th
e wo
rd rose does not
sme
ll sweet or have tho
rns,
we can
use it
to co
nvey
the idea of a rose because
all of
us have lea
rned
, at our
moth
er's
kne
e or
in th
e pl
aygr
ound
, th
e same lin
k between a noise and a
thou
ght.
Now any of us
can con
vey th
e thought by
mal
ting
the noi
se.
The the
ory that wor
ds work by
a convenrional pairing of
sound and
mea
ning
is
not ba
nal or unc
ontr
over
sial
. In the
ear
lies
t surviving deb
ate on linguistics, P
lato
has Hermogenes say
, "No
thin
g ha
s it
s name byn
atur
e, but onl
y by
usa
ge and
custom." Cra
tylu
s di
sagr
ees:
"There
is a correctness of name exi
stin
g by
nature
for everything: a name is not simply that which a number of people jo
intl
y ab ee
to cal
l a th
ing.° Cr
aryl
us is a creationist, and
su~ests that "
a power gr
eate
r th
an
man assib ed the fir
st names to things." Tod
ay, those ~vho see a cor
recr
ness
of
names mig
ht att
ribu
te it i
nste
ad to on
omat
opoe
ia (wo
rds such as cr
ash an
d oi
nk
that
sou
nd lik
e wh
at they mean) or to sound sym
boli
sm (words such as sneer,
cantnnkerous, and melliflurnis that naturally ca
ll to mind the
thi
ngs they mean).
Toda
y th
is debate ha
s been resolved in favor of Hermogenes' conventional
pair
ing.
Ear
ly in th
is cen
tury
Fer
dina
nd de Sa
ussu
re, a founder of
mod
ern
lin-
guistics, called such pa
irin
g the ar
bitr
ary si
gn and
made it a cornerstone of the
study of la
ngua
ge.'
Ono
mato
poei
a and sound symbolism certainly east, bu
t they
are asterisks to
the far
mor
e important pr
inci
ple of the
arb
itra
ry sign—or els
e we
woul
d understand the words in every fo
reig
n la
ngua
ge ins
tinc
tive
ly, a
nd never
need
a dic
tion
ary for ou
r ov
en! Even the
mos
t ob
viou
sly onomatopoeic words—
those for animal sou
nds—
are no
tori
ousl
y un
pred
icta
ble,
wit
h pigs oinking boo
-
boo in Jap
an and dogs barking go
ng gong in In
done
sia.
Sound sym
boli
sm, f
or it
s
part
, was
no fr
iend
of th
e Am
eric
an woman in the th
roes
of la
bor who ove
rhea
rd
what
struck he
r as the mos
t be
auti
ful word in the English la
ngua
ge and
named
her ne
wbor
n da
ught
er Meconium, th
e me
dica
l te
rm for fe
tal ex
crem
ent.
'
17xe In
finite Lib
rary
13
Though sim
ple,
the principle of the arbitrary sign is a po
werf
ul too
l for ge
t-ting thoughts from hea
d to head. Chi
ldre
n begin to lea
rn words bef
ore their
firs
t birthday, and by
their sec
ond they hoover them up at a rate of on
e every
two ho
urs.
By the time they ent
er school children command 13,
000 words,
and th
en the pace picks up, because new words rain down on them fro
m both
speech and pri
nt. A typ
ical
hig
h-sc
hool
gra
duat
e knows about 60,
000 wo
rds;
ali
tera
te adult, pe
rhap
s twice th
at num
ber.3 Pe
ople
rec
ogni
ze words sw
2ftly. The
mean
ing of
a spoken word is ac
cess
ed by a listener's bra
in in about a fifth of
ase
cond
, be
fore
the
speaker has fin
ishe
d pr
onou
ncin
g it.~ The meaning of a
printed wo
rd is registered even more quiclly, in
about an ei
ghth
of a second.>
Peop
le produce wor
ds alm
ost as rap
idly
: It
takes the
bra
in about a quarter of a
seco
nd to fi
nd a word to
name an object, and about another qu
arte
r of
a sec-
ond to prob am the mouth and ton
gue to
pronounce it
.bThe arbitrary sign wo
rks because a speaker and a li
sten
er can cal
l on identi-
cal entries in
their men
tal di
ctio
nari
es. The speaker has a thought, makes a
soun
d, and cou
nts on the
lis
tene
r to
hear the so
und an
d recover that thought.
To dep
ict an
entry in the me
ntal
dic
tion
ary we need a way of sh
owin
g the entry
itself, as wel
l as its
sou
nd and
meaning. The entry for a wor
d is sup
ply it
s ad
-dr
ess in
one's memory, li
ke the location of the
boldfaced entry for a wor
d in
are
al dic
rion
ary.
It's
con
veni
ent to use
an English letter sequence such as r-o -s-e
to sta
nd for the entry, a
s long as we remember thi
s is jus
t a mnemonic tag that
allo
ws us to
remember which wor
d the entry co
rres
pond
s to
; any
symbol, such
as 42759, would do
just
as well. To dep
ict the word's sound, we can use
a pho-
neti
c no
tati
on, su
ch as [roz].'~ The mea
ning
of a wo
rd is a li
nk to an
entry in
the pe
rson
's men
tal encyclopedia, which captures the
per
son'
s concept of
aro
se. For co
nven
ienc
e we can symbolize it with a picture, su
ch as 4~. So a
ment
al dic
tion
ary entry looks so
meth
ing like this:
rose so
und: roz
meaning: ~
*Thi
s bo
ok use
s a simplified pho
neti
c no
tati
on similar to that found in dictionaries, i
n which
the long vow
els a in bait, e in
beet, i in
bit
e, o in boat, and
u in bo
ot are
distinguished Ero
m the
short v
owel
s d in ]g
at, e in
bei, i
in bi
t, o in pot, and u in but. t1
n un
ador
ned a stands fo
r the
firs
tvowel in
fath
er or papa. The
sgm
bol s
is use
d for [he neu
tral
vow
el in
the
suffit of melted and
Rose
's (e
.g., nz
elti
ci, t
o_rz), a ve
rsio
n of th
e vowel s
omet
imes
cal
led s
chwa.
"Long vo
wel,
" "short vo
wel,
" and other te
chni
cal te
rms in
lin
guis
tics
, psy
chol
ingu
isti
cs, a
ndne
uros
cien
ce ar
e defined in
the Glossary:
4
~ Words and Rules
A final component is the word's part of speech, or b ammatical category,
which for rose is
noun (N):
rose so
und: roz
meaning: ̀~
part of speech: N
And that brings us to the second tri
ck behind the vast ex
pres
sive
power of
language.
People do not ju
st blurt out isolated words but rat
her combine them into phrases
and sentences, in which the meaning of the combination can be inferred from
the meanings of the words and the way they ar
e arranged. We talk not merely of
roses, but of the red rose, proud rose, sad rose of al
l my days. We can express our
feelings about bread and roses, guns and roses, the War of the Ro
ses, or days of
~adne and roses. We can say that lovely is
the rose, ro
ses ar
e red, or a rose is
a rose
is a rose. When we combine words, their arrangement is
crucial: Violets are red,
roses are blue, though containing
all the inb edients of the familiar verse, means
something very different. We all know the difference between young tivonzen
looking for husbands and husbands loolzing for young women, and that looking
women husbands yoticng for doesn't mean anything at all.
Inside everyone's head there must be a code or protocol or set of rules that
specifies how words may be arranged into meaningful combinations. Modern
linguists call it a grammar, sometimes a generative grnnzmnr to distinguish it
from the grammars used to teach foreign languages or to teach the dos and
don'ts of fo
rmal
prose.
A grammar assembles words into phrases according to the words' part-of-
speech categories, such as noun and verb. To highlight a word's ca
tego
ry and
reduce visual clutter often it is convenient to omit the sound and meaning and
put the ca
tegory
label on top:
N rose
Similarly, the word a, an article or cleternainer, would look like this:
THe Inf
init
e Library
I S
det
f aThey can then be joined into the phrase a rose by a rule that jo
ins a determiner
to a noun to yield a noun phrase (NP). The rule can be shown as a set of con-
nected branches; this one says "a noun phr
ase may be composed of a deter-
miner followed by a noun':
NP
det
N
The symbols at the bottom of the branches are
like slots into which words may
be plugged, as long as the words have the same labels growing out of their
tops. Here is the result, the phrase a rose:
NP
det
N
a
rose
With just two more rules we can build a complete toy grammar. One rule de-
fines apredicate or verb phrase (VP); the rule says that a verb phrase may con-
sist of a verb followed by it
s direct
object, a noun phrase:
VP
V
NP
The other rule defines the sentence itself (S). This rule says that a sentence
may be composed from a noun phrase (the subject) followed by a verb phrase
(the predicate):
/ \
NP
VP
When words are plugged int
o phrases according to these rules, and the
phrases are plugged into bigger phrases, we get a complete sentence, such as
A rose is a rose:
6 ~
Words and Rules
S
NP
VP
/ \ / \
det
N V
NP
/ \
a rose is det N
a rose
Other parts of the rules, not shown here, sp
ecify the meaning of the new com-
bination. For example, the complete NP rule says that the meaning of the yel-
lmv
rose of Tea:as is based on the meaning of rose, which is called the head of
the phrase, and that the other words modify the head in various ways: yellmv
specifies a distinctive wait, Texas its location.
These rules, though crude, illustrate the fantastic expressive power made
available by grammar. First, the rules are prod2~ctive. By specifying a string of
finds of words rather than a string of actual words, the rules allow us to assem-
ble new sentences on the fly and not regurgitate preassembled cliches—and
that allows us to convey unprecedented combinations of ideas. Though eve
often speak of roses being red, we could talk about violets being red
if the
desire came over us (perhaps to announce a new hybrid), because the rule
allows us to insert violets into the N slot j
ust as easily as roses.
Second, the symbols contained by the rules are symbolic and hence abstract.
The rule doesn't say, "A sentence may begin with a bunch of words referring to
a kind of flower"; rather, it
says, "A sentence may begin with an NP," where
NP is a symbol or variable that can be replaced by any noun, just as x or y in a
mathematical formula can be replaced by any number. We can use the rules to
talk about flowers and their colors and smells, but we can just as easily use
them to talk about karma or quarks or floob-boober-bab-boober-bubs (who, ac
-
cording to Dr. Seuss, bounce in the water like blubbery tubs).
Third, the rules are combi~aatorial. They don't just have a single slot, like a
fill-in-the-blank exam question; every position in the sentence offers a choice
of words from a lengthy menu. Say everyday English has four determiners (n,
any, one, and tlae) and ten thousand nouns. Then the rule for a noun phrase al-
lows four choices for the determiner, f
ollowed by ten thousand choices for the
head noun, yi
elding 4 x 10,000 = 40,000 ways to utter a noun phrase. The rule
Tlie In
finite Library
~ 7
for a sentence allows these forty thousand subjects to be followed by any of
four thousand verbs, providing x}0,000 x 4,000 = 160,000,000 ways to utter
the first three words of a sentence. Then there are four choices for the deter-
miner of the object (640 million four -word beginnings) followed by ten thou-
sand choices for the head noun of [he object, or 640,000,000 x 10,000 =
6,400,000,000,000 (6.4 trillion) five -word sentences. Suppose it takes five
seconds to produce one of these sentences. '~
o crank them a
ll out, from The
abantionnaent abased tl2e abbey and Tlxe nbandonme~zt abased the abbot, through
The abandonment abased the zoologist, all the ~vay to The zoologist coned the
zoo, would take a million years.
Many such combinations are ungrammatical of course, owing to various
complications I haven't mentioned—£or example, you can't say T7ae Anrofa, a
abandonment, or Tlae abbot abase tke abbey. And most of the combinations are
nonsensical: Abandonments can't abbreviate, and abbeys can't abet. Yet even
with these restrictions the e;cpressive range of a grammar is astonishing. The
psychologist George Miller once conservatively estimated that if speakers keep
a sentence perfectly grammatical and sensible as they choose their words,
their menu at each point offers an average of about ten choices (at some
points there are many more than ten choices; at others, only one or two). %That
works out to one hundred thousand five-word sentences, one million siY-word
sentences, ten million seven-word sentences, and so on. A sentence of twenty
v~rords is not at all uncommon (the preceding sentence has twenty words be-
foreand so on), and there are about one hundred million trillion of them in En-
glish. For comparison, that is about a hundred times the number of seconds
since the birth of the universe.
Grammar is an example of a combinatorial system, in which a small inven-
tory of elements can be assembled by rules into an immense set of distinct ob-
jects~ Combinatorial systems obey what Miller calls the Exponential Principle:
The number of possible combinations grows e~cponentially (geometrically)
with the size of the combination.g Combinatorial systems can generate incon-
ceivably vast numbers of products. Every kind of molecule in the universe is
assembled from ahundred-odd chemical elements; every protein building
block and catalyst in the living world
is assembled from just twenty amino
acids. Even when the number of products is smaller, a combinatorial system
can capture them all and provide enormous sa«ngs in storage space. Eight bits
define 23 = 256 distinct bytes, which is more than enough for all the numerals,
punctuation marks, and upper- and lowercase letters in our writing system.
This allows computers to be built out of idenrical specks of silicon that can be