1

Linguistic Theory
D. Terence Langendoen
University of Arizona
Prepublication version. Final version published in Bechtel, William, and Graham, George, eds.: A
Companion to Cognitive Science (Oxford: Blackwell), pp. 235-244, 1998..
1. Goals of Linguistic Theory
The goals of linguistic theory are to answer such questions as ‘What is language?’ and ‘What properties
must something (an organism or a machine) have in order for it to learn and use language?’ Different
theories provide different answers to these questions, and there is at present no general consensus as to
what theory gives the best answers. Moreover, most linguists, when pressed, would say that these
questions have not yet been satisfactorily answered by any theory.
In order to try to answer these questions, one strategy, originally employed by Joseph
Greenberg (1961), is to undertake a comprehensive study of the languages of the world, to determine
what properties they have in common and what distinguishes them from things that everyone agrees are
not languages. Another, advocated by Noam Chomsky (1980), is to examine a few particular languages
in depth to determine which of the intricate details that are found in one language turn up in all the others.
As each of these approaches is extended, they merge into one another, and can be expected, ultimately,
to converge on the same answer.
2. Expression and Meaning
Although we do not yet know enough to provide a definitive answer to the question ‘What is
language?’, what we do know enables us to say with certainty that every language is a system with
sufficient resources for communicating its speakers’ intentions, desires, and beliefs, no matter how
complex and unusual they may be. Let us call the spoken, signed, or written vehicle of communication
EXPRESSION, and what is communicated MEANING. For example, an American English speaker can
communicate the desire to find out what the people he or she is talking to talked about on a particular
occasion in the past by saying [w·d,olt'ok.Sbo¯t`], which is a transcription using the International
Phonetic Alphabet (IPA) of what would ordinarily be written What did you all talk about? In this
case, [w·d,olt'ok.Sbo¯t`] is the expression, and the desire to find out what the people one is talking
to talked about on a particular occasion in the past is the meaning.
An expression, when spoken, can be analyzed as a sequence of SYLLABLES, each said with a
particular degree of loudness (or stress) and pitch (or intonation). For example, the expression
[w·d,olt'ok.Sbo¯t`] consists of five syllables, the third of which has strongest stress and the fourth the
weakest. This stress pattern is indicated by the sequence of marks [ . ], in which [] indicates the
strongest stress, [.] the weakest stress, and [] an intermediate stress. The highest pitch also falls on the
third syllable, the lowest pitch on the second, fourth and fifth syllables, and the first syllable has an
intermediate pitch. This intonation pattern is indicated by the marks [ ], in which [ ] indicates the
highest pitch, [ ] the lowest pitch, and [ ] an intermediate pitch. The stress and intonation pattern of an
expression helps convey part of the meaning of that expression, as discussed further below.
2
A syllable, in turn, consists of a nucleus, usually a vowel, possibly flanked by consonants fore
and aft. For example, the syllable [w·] contains the vowel [·] preceded by the consonant [w], but with
no consonant following, whereas the syllable [.S] contains the vowel [S] alone, and the syllable [t'ok]
contains the vowel [o] which is both preceded and followed by a consonant. Consonants and vowels
can be further analyzed as bundles of phonetic features, specifying the movement, position or activity of
articulators, such as the lips, tongue, and vocal cords, or their acoustic effects. See Ladefoged and
Maddieson 1996 for discussion of phonetic analysis and notation.
A meaning can also be analyzed into component parts, though there is much less agreement
about semantic structure (the structure of meaning) than there is about phonological structure (the
structure of spoken expressions). Whatever representation is chosen for meaning, it must meet certain
criteria of adequacy, including the following. First, it must provide a way to determine what things it may
be used to talk about, for example, situations in a world of the speaker’s and hearers’ experience or
imagination. Second, it must provide a way to determine the logical properties of meaning, including
what it implies and what it is implied by. Third, it must indicate what act the speaker is performing when
expressing it (e.g. making a statement, asking a question, issuing a command). Fortunately, for our
purposes, it is not necessary to formulate meanings precisely, using a notation that meets these (and
other) criteria of adequacy. Except as needed, we represent the meaning of an expression by enclosing
its ordinary spelling in single quotes; for example, we represent the meaning of the expression
[w·d,olt'ok.Sbo¯t`] as ‘What did you all talk about?’. See Saeed (1997) and Larson & Segal
(1995) for discussion of semantic analysis and notation.
Given that a language provides a means of expressing meaning, we may add to the goals of
linguistic theory that it answer two additional questions, which we call the question of language
PERCEPTION and the question of language PRODUCTION. The question of language perception is: ‘How
does one determine a meaning for an expression one has heard?’. The question of language production
is: ‘How does one determine an expression for a meaning one intends to convey?’.
3. Morphs and Morphemes
The simplest expressions of a language are those that cannot be divided into meaningful parts, other than
the entire sequence of syllables that makes up each such expression and its stress and intonation
contour. For example, the expressions [no:] ‘No?’ and [no:] ‘No!’ are among the simplest expressions
of English, since their only meaningful parts are the syllable [no:] itself and the stress and intonation
contours [ ] (rising) and [ ] (falling). The instances of the consonant [n] and vowel [o:] that combine to
form the syllable [no:] are not meaningful, even though in other expressions, that consonant by itself, and
that vowel by itself, are meaningful (e.g. in [o:sis.t'az.nbr·ò.az] ‘Oh sisters and brothers’). The fact
that [no:] is meaningful only in its entirety is the basis of the challenge “What part of ‘no’ don’t you
understand?” emblazoned on T-shirts and bumper stickers throughout the English-speaking world.
The syllable [no:] with the associated meaning ‘no’ in English is a theoretical construct known as
a MORPH. A morph is a specific pronunciation associated with a specific meaning such that the
pronunciation cannot be broken down into meaningful parts whose meanings combine to form the
meaning of the whole. Behind the morph is an even more abstract theoretical construct known as a
MORPHEME. A morpheme is an association of pronunciation and meaning such that the pronunciations
and meanings of an entire class of morphs can be determined from it. In the case of [no:] ‘no’, there is
3
no distinction between the morph and the morpheme, since there is no other determinable pronunciation
with the meaning ‘no’, and no other determinable meaning associated with the pronunciation [no:].
However in many other cases the distinction is quite sharp.
The syllable [no:] with the associated meaning ‘no’ in English is a theoretical construct known as
a MORPH. A morph is a specific pronunciation associated with a specific meaning such that the
pronunciation cannot be broken down into meaningful parts whose meanings combine to form the
meaning of the whole. Behind the morph is an even more abstract theoretical construct known as a
MORPHEME. A morpheme is an association of pronunciation and meaning such that the pronunciations
and meanings of an entire class of morphs can be determined from it. In the case of [no:] ‘no’, there is
no distinction between the morph and the morpheme, since there is no other determinable pronunciation
with the meaning ‘no’, and no other determinable meaning associated with the pronunciation [no:].
However other morphemes differ phonologically or semantically from one or more of the morphs which
are derivable from them. For example, the morpheme /ju:/ ‘you’ underlies a dozen or more morphs
including [ju:] ‘you’, [ju:] ‘you (sg)’, [ju:] ‘you (pl)’, [jS] ‘you’, [jS] ‘you (sg)’, [jS] ‘you (pl)’, [] ‘you’,
[] ‘you (sg)’, [] ‘you (pl)’, [] ‘you’, [] ‘you (sg)’, and [] ‘you (pl)’. In the expression [ju:k'old] ‘Did
you call?’, the morph [ju:] ‘you’ represents /ju:/ ‘you’, just as [no:] represents /no:/. However, in
[jS0inkso:] ‘Do you think so?’, the morph [jS] ‘you’ represents /ju:/ ‘you’; in [w·d,olt'ok.Sbo¯t`]
‘What did you all talk about?’, the morph [] ‘you (pl)’ represents the morpheme /ju:/ ‘you’. Finally, in
[hclp.jasclf] ‘You (sg) should take what you (sg) want’, the morph [] ‘you (sg)’ (occurring at the very
beginning of the expression) represents /ju:/ ‘you’.
In describing the relation between morphs and morphemes, we have used the terms ‘underlie’
and ‘derivable from’ to emphasize the fact that that relation is systematic both phonologically and
semantically. Given the phonological form of the morpheme /ju:/ ‘you’, the various phonological forms of
the morphs derivable from it can be determined by well-understood processes of vocalic weakening
(from [u:] to [S]) and deletion, and consequent consonantal weakening (from [j] to []) and deletion.
Given the semantic form of the morpheme /ju:/ ‘you’, the various semantic forms of the morphs
derivable from it can be determined by similarly well-understood processes of contextual delimitation (in
the expression [w·d,olt'ok.Sbo¯t`], the interpretation ‘you (pl)’ is required for consistency with the
cooccurring morpheme /ol/ ‘all’, whereas in the expression [hclp.jasclf], the interpretation ‘you (sg)’ is
required for consistency with the cooccurring word yourself).
On the other hand, even though the morph [ju:] ‘ewe’ has the same phonological form as the
morpheme /ju:/ ‘you’, it is not derivable from that morpheme, because the relation between ‘ewe’ and
‘you’ is not systematic semantically. Similarly, even though the morph [òo¯] ‘you (sg)’ has the same
semantic form (in those dialects in which that morph occurs) as the morph [ju:] ‘you (sg)’, it is not
derivable from the morpheme underlying the latter morph, because the relation between [thou] and /ju:/
is not systematic phonologically. We conclude that /ju:/ ‘you’, /ju:/ ‘ewe’, and /òo¯/ ‘you (sg)’ are
distinct morphemes in English. Henceforth, whenever neither the exact phonological nor the exact
semantic form of a morpheme is at issue, we represent it by its ordinary spelling in italics, for example
no for [no:] ‘no’, you for /ju:/ ‘you’, ewe for /ju:/ ‘ewe’, and thou for ‘you (sg)’.
4
4. The Lexicon
The particular pairings of expression and meaning represented by individual morphemes have to be
learned individually. The totality of all the pairings in a language that must be individually learned is
known as its LEXICON, and its individual members are called LEXEMES (or LEXICAL ITEMS). The
lexemes of a language consists of all of its morphemes together with those combinations of lexemes
whose meaning or expression cannot be systematically determined from the lexemes of which it is
comprised. For example, the meanings of the word unusual and of the phrase touch base cannot be
systematically determined from the meanings of their constituent lexemes; hence those combinations are
also English lexemes. Similarly, the word went is an English lexeme, even if it is analyzed as containing
the two lexemes /go:/ ‘go’ and /d/ ‘in the past’; although its semantic form can be predicted from that
combination, its phonological form cannot be. (Compare the word planned, consisting of the lexemes
/plæn/ ‘plan’ and /d/ ‘in the past’, which is not an English lexeme, since both its phonological and
semantic forms are derivable from those of its component lexemes.)
5. Morphology and Syntax
The vast majority of the pairings of expression and meaning represented by combinations of lexemes,
whether these pairings are words, phrases, or sentences, do not have to be learned, because they can
be systematically determined from the lexemes of which they are composed and the way they are
combined and arranged. For example, the expression and meaning of the sentence What did you all
talk about? can be can be determined from the expression and meaning of the individual lexemes
about, all, did, talk, what, and you, and the way they are combined and arranged. The importance of
how the component lexemes are combined and arranged can be seen from the fact that when they enter
into other combinations or are rearranged, the result is always a different expression, which may have an
entirely different meaning, or no meaning at all. For example, the sentences What all did you talk
about? and What did you talk all about? consist of the same lexemes, but differently combined and
arranged, and have different meanings from the original. The sentence About what did you all talk?
also has a different arrangement, but the same meaning as the original. Finally, the arrangement in You
what talk did about all is meaningless. Linguists generally call the expressions which are meaningless in
a language UNGRAMMATICAL, and indicate that status by prefixing an asterisk to the expression, as in
*You what talk did about all.
The study of how lexemes combine to form words is called MORPHOLOGY (see Spencer 1991),
and of how lexemes (typically, but not necessarily words) combine to form phrases and sentences is
called SYNTAX (see Culicover 1997). The boundary between morphology and syntax varies widely
from language to language; what can be expressed in a word in one language requires a phrase in
another. For example, the meaning ‘from our hands’ can only expressed by a phrase made up of at least
three words and four lexemes in English (from, hand-s, and our; we use a hyphen to separate lexemes
within a word), but can be expressed by a single word made up of five lexemes in Turkish (el-ler-im-iz-
den).
6. Recursion
The combination of lexicon, morphology, and syntax gives every language its expressive power, its
ability to express any desired meaning. That power is, as far as linguists have been able to determine,
5
the same for all languages. If a fluent speaker of a language lacks a word or simple phrase to express a
particular meaning, he or she can do so by means of a more elaborate phrase. Moreover, if a particular
meaning which can only be expressed in a complicated way by the members of a community becomes
important to them, they will come up with simpler expressions for that meaning.
Since lexemes must be learned individually, no language can have more than a finite number of
them, and a relatively small number at that. (No language has been observed to have more than 10
6
lexemes, though the problem of deciding exactly how to individuate lexemes makes estimating a more
exact upper bound extremely difficult.) Hence in every language there must be meanings which can only
be expressed by words, phrases, and sentences which are not single lexemes. Moreover, it also
appears to be the case for every language, that there are meanings which can only be expressed by
phrases or sentences which are not single words. That is, the full expressive power of every language
appears ultimately to depend on its syntax.
How this task is accomplished is most easily explained by means of examples. First, the syntax
of English permits us to form two-word phrases such as these cats, these dogs, and these hamsters,
as well as five-word phrases in which these two-word phrases are joined by the lexeme and, such as
these cats and these dogs, these dogs and these hamsters, and (allowing for repetition of the two-
word phrases) these cats and these cats. Let us call the two-word phrases ‘simple’ and the five-word
ones ‘coordinate’. In addition, the syntax of English permits us to form longer coordinate phrases by
joining a simple phrase to a coordinate one with a short intonation break in between, such as these
hamsters, these cats and these dogs. Then since the latter phrase is itself coordinate, it can be joined
with another simple phrase, as in these cats, these hamsters, these cats and these dogs, and so on
without limit. By this means, we can form an infinite number of coordinate phrases in English, each of
which is distinct both phonologically and semantically from the others. This syntactic device of
coordinate-phrase formation does not by itself provide a means for expressing all possible meanings in
English, but it’s a start.
Second, the syntax of English permits us to form simple sentences such as I could fly, as well as
‘complex’ sentences in which a sentence is subordinated to a larger one, by the addition of the structure
I thought that or You thought that at the beginning, as in I thought that I could fly and You thought
that I could fly. Then since any sentence, not just simple ones, can be subordinated to a larger one, the
syntax of English also permits us to form longer complex sentences in which complex sentences are
subordinated, such as You thought that I thought that I could fly and I thought that you thought
that I thought that I could fly, and so on without limit. Again this syntactic device of complex sentence
formation does not by itself provide a means for expressing all possible meanings in English, but now
we’re on the road.
Our two examples illustrate RECURSION, the formation of phrases of a certain type (we assume
from this point on that a sentence is simply a certain kind of phrase) out of phrases of exactly the same
type. Each illustration starts with BASE CASES: phrases of a given type which do not contain any phrase
of that type within it. In the coordination illustration, these cats and these dogs is a coordinate phrase
which does not contain any coordinate phrase within it. In the subordination illustration, I could fly is a
sentence which does not contain any sentence within it (i.e., a simple sentence). It then proceeds to a
RECURSIVE STEP, according to which a larger phrase of a given type can be constructed out of parts
which include a smaller phrase of the same type. In the coordination illustration, the recursive step
6
consists of adjoining a simple phrase to a coordinate phrase to form a larger coordinate phrase. In the
subordination illustration, the recursive step consists of subordinating a sentence to a larger sentence by
adjoining the structure I thought that or You thought that to the original sentence.
The expressive power of language depends on the existence of many different recursive devices
in its syntax, and their ability to freely combine. For example, the coordination and subordination
processes just illustrated can be combined to permit the formation of coordinate complex sentences
such as I thought that I could fly, you thought that I thought that I could fly and I thought that
you thought that I thought that I could fly (coordination of complex sentences) and I thought that
you could navigate, she could steer and he could make dinner (subordination of coordinate
sentences). Each of these sentences expresses a meaning not expressible by coordination or
subordination alone.
7. Movement and Deletion
Languages employ various syntactic devices to prevent expressions for complex meanings from
becoming inordinately long. Two devices which appear in the grammars of all languages are known as
MOVEMENT and DELETION. We have already given an example which illustrates movement, namely the
question What did you all talk about?, in which the lexeme what is displaced from its ‘normal’
position following about, as in You all talked about what?. In this case, the movement of what serves
the purpose of focusing or highlighting what kind of answer the speaker is expecting (the name of a thing
rather than of a person, for example). In more complex sentences such as What did she tell him that
you all talked about? and Did she tell him what you all talked about?, it makes a difference
semantically where the lexeme is moved to. By the same token, it also makes a difference semantically
where the lexeme originates; compare What did you all put the baskets in? (which is equivalent to
You all put the baskets in what?) and What did you all put in the baskets? (which is equivalent to
You all put what in the baskets?).
Deletion may be employed in many situations in which the repetition of phrases can be
‘reconstructed’, as in the coordinate sentence You thought that I could fly, but I didn’t, which has
two meanings depending on what is reconstructed, either think that I could fly, resulting in the meaning
‘You thought that I could fly, but I didn’t think that I could fly’, or simply fly, resulting in the meaning
‘You thought that I could fly, but I didn’t fly’. If we exchange the second occurrence of I with you in the
original sentence, yielding I thought that I could fly, but you didn’t, the result has three meanings: ‘I
thought that I could fly, but you didn’t think that I could fly’; ‘I thought that I could fly, but you didn’t
think that you could fly’; and ‘I thought that I could fly, but you didn’t fly’. The additional meaning
results from the fact that the second occurrence of the lexeme I in I thought that I could fly can
DEPEND on the first. If this dependency is reconstructed in determining the meaning of but you didn’t,
the result is ‘but you didn’t think that you could fly’. See Fiengo & May (1995) for analysis of
dependency and its interaction with deletion.
8. Universal Grammar
A complete analysis of the lexicon, morphology and syntax of a language is called its GRAMMAR. The
theory of grammar, which is a significant part of linguistic theory as a whole, has as one of its goals the
answer to the question ‘What is the class of all possible grammars?’ The properties of that class is
7
called UNIVERSAL GRAMMAR. Related to the goal of understanding universal grammar is the solution to
the problem of language acquisition: ‘What enables a child upon limited exposure to a language to learn
the grammar of that language?’ One widely held assumption is that people are innately endowed with
knowledge of universal grammar, so that the task of language acquisition is really one of selecting the
best grammars for the languages of one’s experience.
9. Generative Grammar
The first complete statement of the theory of grammar in the form presented here was the
‘standard theory’ of GENERATIVE GRAMMAR presented in Chomsky (1965), henceforth ST.
(Chomsky’s 1957 presentation of the theory of grammar, which is widely recognized as having launched
the modern enterprise of linguistic theory, was less well developed inasmuch as it neither attempted to
account for the meaning of expressions, nor dealt with language acquisition.) According to ST, a
grammar consists of a number of interacting components including a lexicon, two syntactic components,
a semantics, and a phonology. For each meaningful expression, the grammar provides a DERIVATION,
which starts with a member of a set of AXIOMS such as #S# (for ‘Sentence’) and terminates with a pair
<Mng, Exp>, where ‘Mng’ is its semantic structure (representation of meaning) and ‘Exp’ is its
phonological structure (representation of expression).
The derivation of a meaningful expression in ST begins in one of its syntactic components called
PHRASE STRUCTURE, which constructs a hierarchical representation of the underlying syntactic structure
of the expression. The phrase-structure component is responsible for, among other things, recursion.
The output of phrase structure is the input to the lexicon, which inserts lexical items into the underlying
syntactic structure. Chomsky called the result of that operation DEEP STRUCTURE. The deep structure of
an expression is the input both to the semantics and to the second of the syntactic components, known
as the TRANSFORMATIONAL STRUCTURE. The semantics is responsible for constructing the meaning of
the expression, whereas transformational structure is responsible for, among other things, movement and
deletion. Finally, the output of the transformational structure, called SURFACE STRUCTURE, is the input to
the phonology, which determines the form of the expression proper, including its stress and intonational
contours. The overall structure of ST is outlined in Figure (1).
(1) Outline of the ‘standard theory’ of generative grammar (Chomsky 1965)
à Sem à Mng
Ax à Phr à Lex à DS
à Trn à SS à Phn à Exp
Legend: Ax = axiom, Phr = phrase structure, Lex = lexicon, DS = deep structure, Sem
= semantics, Trn = transformational structure, Mng = semantic structure (meaning), SS
= surface structure, Phn = phonology, Exp = phonetic structure (expression). Inputs
and outputs are italicized.
ST does not attempt to answer the questions of language perception and production posed at
the end of section 2. Rather than directly relating meaning and expression, it relates them indirectly,
through deep structure. Thus, given an expression, the only way to determine its meaning according to
ST is first to recover its deep structure, and then to compute its meaning, using semantics. However, the
recovery of deep structure from expression is not an operation which is well defined within ST. The
operations of the transformational and phonological components of ST map deep structure onto
8
expression, but are not capable of mapping expression onto deep structure. Similarly, given a meaning,
the only way to determine its expression according to ST is again to recover its deep structure.
However, just as the operations which map deep structure onto expression are not reversible, neither
are the operations that map deep structure onto meaning. Chomsky justifies his decision to limit linguistic
theory to merely providing a means of relating meaning and expression, rather than to providing a means
of determining one from the other, by maintaining that the first a matter of linguistic COMPETENCE (what
people know about language), and hence is properly within the scope of linguistic theory, whereas the
second is a matter of linguistic PERFORMANCE (how people use language), and hence lies outside its
scope. Not everyone agrees with Chomsky on this point, and in any event a total theory of language
(whatever it is called) would have to encompass both linguistic competence and linguistic performance.
The theory of generative grammar has evolved considerably over the past thirty years, but
continues to retain its ‘forked’ structure, in which meaning and expression are related indirectly through
an intermediate construct. For example, the version of generative grammar known as ‘government and
binding’ theory (Chomsky 1981) treats deep structure (renamed ‘D-structure’) as the starting point of a
derivation, and surface structure (renamed ‘S-structure’) as the intermediate construct from which both
meaning and expression are determined, as in Figure 2.
(2) Outline of the ‘government-binding’ theory of generative grammar (Chomsky 1981)
à Sem à Mng
DS à Trn à SS
à Phn à Exp
Finally, the ‘minimalist program’ of generative grammar (Chomsky 1995) treats the lexical items which
contribute to both meaning and expression as the starting point of the derivation, eliminating both D-
structure and S-structure entirely. Nevertheless, following application of transformational rules of
‘merge’, ‘move’ and ‘check’, derivations are still branched, with strictly phonological rules (now called
‘spellout’) applying on the branch leading to phonological representation (now called ‘phonetic form’ or
‘PF’), and strictly semantic rules applying on the granch leading to semantic representation (now called
‘logical form’ or ‘LF’), as in Figure (3).
(3) Outline of the ‘minimalist’ theory of generative grammar (Chomsky 1995)
à Sem à LF (= Mng)
Lex à Trn
à Phn à PF (= Exp)
It is of course not a necessary feature of the theory of generative grammar that it fail to provide
answers to the questions of linguistic perception and production. One theory that maps meaning to
expression in the grammar. and hence attempts to answer the question of linguistic production, is
‘generative semantics’, which originated in the 1960s as an alternative to ST; see Huck & Goldsmith
(1995). As diagrammed in Figure (4), the mapping from meaning to surface structure is carried out by
transformational rules, including rules which replace semantic substructures by lexical items represented
in terms of their underlying phonological properties, and the mapping from surface structure to
expression is carried out by phonological rules.
(4) Outline of the ‘generative semantics’ theory of generative grammar
Mng à Trn à SS à Phn à Exp
9
10. Optimality Theory
However, only very recently has a theory of generative grammar been proposed that attempts
to answer both the question of linguistic perception and of linguistic production, namely OPTIMALITY
THEORY (Archangeli & Langendoen 1997; Prince & Smolensky 1997, forthcoming). Optimality theory
assumes that the universal classes of expressions and meanings (that is, the class of all expressions of all
possible languages and the class of all meanings) can each be defined by a GENERATOR, which works
rather like the phrase-structure component of ST, or the rules for forming well-formed formulas in a
system of logic. Then, given a particular meaning, one determines its expression in a particular language
by evaluating the members of the universal class of expressions against a universal set of CONSTRAINTS
(henceforth Con). Similarly, given a particular expression, one determines its meaning (if any) in a
particular language by evaluating the members of the universal class of meanings against Con. That is,
the problems of language production and language perception are considered problems of selection, just
like the problem of language acquisition as described above in section 8. Figure (5) provides an outline
of optimality theory.
(5) Outline of the ‘optimality’ theory of generative grammar
a. Mng à Con à Exp
b. Exp à Con à Mng
If, as optimality theory proposes, the classes of meanings, expressions and constraints are
universal, how do languages differ? The answer is in the RANKING of the constraints; with each language
corresponding to a possible ranking of the members of Con; for numerous illustrations of how this
works, see Archangeli & Langendoen (1997). Presumably included among the members of Con is the
universal class of lexical items, so that the determination of which lexical items belong to which languages
is also a matter of constraint ranking.
Optimality theory has had its greatest success so far in accounting for the relation between
lexical items and their expression, and is only now beginning to deal with the major phenomena of
syntax, including recursion, movement, deletion, and dependency, so whether it will succeed in
extending the scope of linguistic theory to the narrow concerns of linguistic competence to the broad
concerns of linguistic performance remains to be seen.
Bibliography
Archangeli, Diana, and Langendoen, D. Terence, eds.: Optimality Theory: An Overview (Oxford:
Blackwell, 1997). [collection of articles which introduce optimality theory]
Chomsky, Noam: Syntactic Structures (The Hague: Mouton, 1957). [the book that launched the
enterprise of generative grammar]
Chomsky, Noam: Aspects of the Theory of Syntax (Cambridge, MA: MIT Press, 1965). [presents
the standard theory of generative grammar]
Chomsky, Noam: Rules and Representations (New York: Columbia University Press, 1980).
Chomsky, Noam: Lectures on Government and Binding (Dordrecht: Foris, 1981).
Chomsky, Noam: The Minimalist Program (Cambridge, MA: MIT Press, 1995). [represents
Chomsky’s current thinking about linguistic theory]
Culicover, Peter: Principles and Parameters: An Introduction to Syntactic Theory (Oxford: Oxford
University Press, 1997).
10
Greenberg, Joseph, ed.: Universals of Language (Cambridge,m MA: MIT Press, 1961). [the article
by Greenberg in this collection continues to be of interest]
Fiengo, Robert, and May, Robert: Indices and Identity (Cambridge, MA: MIT Press, 1995).
Huck, Geoffrey, and Goldsmith, John: Ideology and Linguistic Theory: The Deep Structure Debates
(London: Routledge, 1995).
Ladefoged, Peter, and Maddieson, Ian: The Sounds of the World’s Languages (Oxford: Blackwell,
1996).
Larson, Richard, and Segal, Gabriel: Knowledge of Meaning (Cambridge, MA: MIT Press, 1995).
Prince, Alan, and Smolensky, Paul: ‘Optimality: From neural networks to universal grammar’, Science
275 (1997), 1604-1610.
Prince, Alan, and Smolensky, Paul: Optimality Theory: Constraint Interaction in Generative
Grammar (Cambridge, MA: MIT Press, forthcoming). [technical introduction to optimality
theory]
Saeed, John I.: Semantics (Oxford: Blackwell, 1997).
Spencer, Andrew: Morphological Theory (Oxford: Blackwell, 1991).

tongue. including the following. Whatever representation is chosen for meaning. and that vowel by itself. using a notation that meets these (and other) criteria of adequacy. issuing a command). The syllable [no:] with the associated meaning ‘no’ in English is a theoretical construct known as a MORPH. A morpheme is an association of pronunciation and meaning such that the pronunciations and meanings of an entire class of morphs can be determined from it. Second. The question of language perception is: ‘How does one determine a meaning for an expression one has heard?’. In the case of [no:] ‘no’. A meaning can also be analyzed into component parts.g. it must meet certain criteria of adequacy. even though in other expressions. we represent the meaning of the expression [®w¿Û®dZÆO~l"tîO!k. 3. other than the entire sequence of syllables that makes up each such expression and its stress and intonation contour. Fortunately. that consonant by itself. position or activity of articulators. it must indicate what act the speaker is performing when expressing it (e.tîã~z. For example. in [®oÛ:"sIÛs. situations in a world of the speaker’s and hearers’ experience or imagination. we represent the meaning of an expression by enclosing its ordinary spelling in single quotes. possibly flanked by consonants fore and aft. consists of a nucleus. Morphs and Morphemes The simplest expressions of a language are those that cannot be divided into meaningful parts. and the syllable ["tîO!k] contains the vowel [O] which is both preceded and followed by a consonant. For example. A morph is a specific pronunciation associated with a specific meaning such that the pronunciation cannot be broken down into meaningful parts whose meanings combine to form the meaning of the whole.ã~z] ‘Oh sisters and brothers’). but with no consonant following. though there is much less agreement about semantic structure (the structure of meaning) than there is about phonological structure (the structure of spoken expressions). The instances of the consonant [n] and vowel [o:] that combine to form the syllable [no:] are not meaningful. First. it is not necessary to formulate meanings precisely. The question of language production is: ‘How does one determine an expression for a meaning one intends to convey?’. The fact that [no:] is meaningful only in its entirety is the basis of the challenge “What part of ‘no’ don’t you understand?” emblazoned on T-shirts and bumper stickers throughout the English-speaking world. since their only meaningful parts are the syllable [no:] itself and the stress and intonation contours [" %] (rising) and [" ß] (falling). such as the lips. or their acoustic effects. we may add to the goals of linguistic theory that it answer two additional questions. including what it implies and what it is implied by. whereas the syllable [. Third.g. it must provide a way to determine what things it may be used to talk about. Except as needed. for example. there is 2 . Consonants and vowels can be further analyzed as bundles of phonetic features. in turn. for our purposes. and vocal cords.A syllable. specifying the movement.n~"b|¿!D. asking a question. See Saeed (1997) and Larson & Segal (1995) for discussion of semantic analysis and notation. it must provide a way to determine the logical properties of meaning. the syllable [®w¿Û] contains the vowel [¿] preceded by the consonant [w]. Given that a language provides a means of expressing meaning.«Ù] contains the vowel [«] alone. See Ladefoged and Maddieson 1996 for discussion of phonetic analysis and notation. usually a vowel. which we call the question of language PERCEPTION and the question of language PRODUCTION.«~®bA~Wt}] as ‘What did you all talk about?’. making a statement. the expressions ["no%:] ‘No?’ and ["noß:] ‘No!’ are among the simplest expressions of English. are meaningful (e. Behind the morph is an even more abstract theoretical construct known as a MORPHEME. for example.

you for /ju:/ ‘you’. it is not derivable from the morpheme underlying the latter morph. in [®w¿Û®dZÆO~l"tîO!k. the interpretation ‘you (pl)’ is required for consistency with the cooccurring morpheme /Ol/ ‘all’. whenever neither the exact phonological nor the exact semantic form of a morpheme is at issue. However.no distinction between the morph and the morpheme. the morph [Æ] ‘you (pl)’ represents the morpheme /ju:/ ‘you’. Given the phonological form of the morpheme /ju:/ ‘you’.«~®bA~Wt}]. and no other determinable meaning associated with the pronunciation [no:]. [Æ] ‘you’. The syllable [no:] with the associated meaning ‘no’ in English is a theoretical construct known as a MORPH. we represent it by its ordinary spelling in italics. and /DAW/ ‘you (sg)’ are distinct morphemes in English. [] ‘you (sg)’. In describing the relation between morphs and morphemes. /ju:/ ‘ewe’. [j«] ‘you (sg)’.jã~"sEÞlf]. the various semantic forms of the morphs derivable from it can be determined by similarly well-understood processes of contextual delimitation (in the expression [®w¿Û®dZÆO~l"tîO!k. However in many other cases the distinction is quite sharp. and [] ‘you (pl)’. In the case of [no:] ‘no’. For example. in [j«~"TI)ÚNk"so!:] ‘Do you think so?’. Given the semantic form of the morpheme /ju:/ ‘you’. since there is no other determinable pronunciation with the meaning ‘no’. the various phonological forms of the morphs derivable from it can be determined by well-understood processes of vocalic weakening (from [u:] to [«]) and deletion. ewe for /ju:/ ‘ewe’. On the other hand. [ju:] ‘you (pl)’. Similarly. the morph [j«] ‘you’ represents /ju:/ ‘you’. [j«] ‘you’. [Æ] ‘you (sg)’. in [®hEÛlp. Henceforth.«~®bA~Wt}] ‘What did you all talk about?’. the morph [ju:] ‘you’ represents /ju:/ ‘you’. we have used the terms ‘underlie’ and ‘derivable from’ to emphasize the fact that that relation is systematic both phonologically and semantically. A morph is a specific pronunciation associated with a specific meaning such that the pronunciation cannot be broken down into meaningful parts whose meanings combine to form the meaning of the whole. even though the morph [DAW] ‘you (sg)’ has the same semantic form (in those dialects in which that morph occurs) as the morph [ju:] ‘you (sg)’. because the relation between ‘ewe’ and ‘you’ is not systematic semantically. We conclude that /ju:/ ‘you’. it is not derivable from that morpheme. and consequent consonantal weakening (from [j] to [Æ]) and deletion. Finally. there is no distinction between the morph and the morpheme. since there is no other determinable pronunciation with the meaning ‘no’. [ju:] ‘you (sg)’.jã~"sEÞlf] ‘You (sg) should take what you (sg) want’. [j«] ‘you (pl)’. [Æ] ‘you (pl)’. Behind the morph is an even more abstract theoretical construct known as a MORPHEME. just as [no:] represents /no:/. and thou for ‘you (sg)’. the morph [] ‘you (sg)’ (occurring at the very beginning of the expression) represents /ju:/ ‘you’. the interpretation ‘you (sg)’ is required for consistency with the cooccurring word yourself). However other morphemes differ phonologically or semantically from one or more of the morphs which are derivable from them. In the expression [®ju~:"kîO%ld] ‘Did you call?’. [] ‘you’. and no other determinable meaning associated with the pronunciation [no:]. 3 . even though the morph [ju:] ‘ewe’ has the same phonological form as the morpheme /ju:/ ‘you’. whereas in the expression [®hEÛlp. A morpheme is an association of pronunciation and meaning such that the pronunciations and meanings of an entire class of morphs can be determined from it. the morpheme /ju:/ ‘you’ underlies a dozen or more morphs including [ju:] ‘you’. for example no for [no:] ‘no’. because the relation between [thou] and /ju:/ is not systematic phonologically.

phrases. whether these pairings are words. and syntax gives every language its expressive power. the meaning ‘from our hands’ can only expressed by a phrase made up of at least three words and four lexemes in English (from. even if it is analyzed as containing the two lexemes /go:/ ‘go’ and /d/ ‘in the past’. hence those combinations are also English lexemes. the arrangement in You what talk did about all is meaningless.4. what can be expressed in a word in one language requires a phrase in another. what. The boundary between morphology and syntax varies widely from language to language. but differently combined and arranged. which is not an English lexeme. did. Recursion The combination of lexicon. the sentences What all did you talk about? and What did you talk all about? consist of the same lexemes. the meanings of the word unusual and of the phrase touch base cannot be systematically determined from the meanings of their constituent lexemes. For example. and our. The Lexicon The particular pairings of expression and meaning represented by individual morphemes have to be learned individually. 4 . For example. Morphology and Syntax The vast majority of the pairings of expression and meaning represented by combinations of lexemes. hand-s. although its semantic form can be predicted from that combination. its ability to express any desired meaning. morphology. and you. and of how lexemes (typically. The sentence About what did you all talk? also has a different arrangement. the expression and meaning of the sentence What did you all talk about? can be can be determined from the expression and meaning of the individual lexemes about. 6. consisting of the lexemes /plæn/ ‘plan’ and /d/ ‘in the past’. The importance of how the component lexemes are combined and arranged can be seen from the fact that when they enter into other combinations or are rearranged. talk. and indicate that status by prefixing an asterisk to the expression. as far as linguists have been able to determine. That power is. all. (Compare the word planned. The lexemes of a language consists of all of its morphemes together with those combinations of lexemes whose meaning or expression cannot be systematically determined from the lexemes of which it is comprised. the word went is an English lexeme. but not necessarily words) combine to form phrases and sentences is called SYNTAX (see Culicover 1997).) 5. since both its phonological and semantic forms are derivable from those of its component lexemes. Similarly. The study of how lexemes combine to form words is called MORPHOLOGY (see Spencer 1991). and its individual members are called LEXEMES (or LEXICAL ITEMS). or sentences. Linguists generally call the expressions which are meaningless in a language UNGRAMMATICAL. the result is always a different expression. as in *You what talk did about all. its phonological form cannot be. The totality of all the pairings in a language that must be individually learned is known as its LEXICON. because they can be systematically determined from the lexemes of which they are composed and the way they are combined and arranged. For example. and have different meanings from the original. and the way they are combined and arranged. Finally. but can be expressed by a single word made up of five lexemes in Turkish (el-ler-im-izden). do not have to be learned. which may have an entirely different meaning. we use a hyphen to separate lexemes within a word). For example. or no meaning at all. but the same meaning as the original.

no language can have more than a finite number of them. First. In the coordination illustration. I could fly is a sentence which does not contain any sentence within it (i. though the problem of deciding exactly how to individuate lexemes makes estimating a more exact upper bound extremely difficult. such as You thought that I thought that I could fly and I thought that you thought that I thought that I could fly. By this means. and so on without limit. the recursive step 5 . each of which is distinct both phonologically and semantically from the others..) Hence in every language there must be meanings which can only be expressed by words. This syntactic device of coordinate-phrase formation does not by itself provide a means for expressing all possible meanings in English. Then since the latter phrase is itself coordinate. Each illustration starts with BASE CASES : phrases of a given type which do not contain any phrase of that type within it. can be subordinated to a larger one. (No language has been observed to have more than 106 lexemes.the same for all languages. that there are meanings which can only be expressed by phrases or sentences which are not single words. these hamsters. and (allowing for repetition of the twoword phrases) these cats and these cats. these cats and these dogs is a coordinate phrase which does not contain any coordinate phrase within it. as well as ‘complex’ sentences in which a sentence is subordinated to a larger one. the syntax of English also permits us to form longer complex sentences in which complex sentences are subordinated. the syntax of English permits us to form longer coordinate phrases by joining a simple phrase to a coordinate one with a short intonation break in between. not just simple ones. That is. these cats and these dogs. In addition. If a fluent speaker of a language lacks a word or simple phrase to express a particular meaning. such as these hamsters. it also appears to be the case for every language. a simple sentence). Moreover. by the addition of the structure I thought that or You thought that at the beginning. the syntax of English permits us to form two-word phrases such as these cats. and so on without limit. Second. phrases. these dogs and these hamsters. according to which a larger phrase of a given type can be constructed out of parts which include a smaller phrase of the same type. as in I thought that I could fly and You thought that I could fly. It then proceeds to a RECURSIVE STEP. the syntax of English permits us to form simple sentences such as I could fly. as in these cats. he or she can do so by means of a more elaborate phrase. Let us call the two-word phrases ‘simple’ and the five-word ones ‘coordinate’. such as these cats and these dogs. Then since any sentence. Since lexemes must be learned individually. they will come up with simpler expressions for that meaning. Again this syntactic device of complex sentence formation does not by itself provide a means for expressing all possible meanings in English. Moreover. In the subordination illustration.e. but it’s a start. How this task is accomplished is most easily explained by means of examples. the full expressive power of every language appears ultimately to depend on its syntax. In the coordination illustration. Our two examples illustrate RECURSION. we can form an infinite number of coordinate phrases in English. if a particular meaning which can only be expressed in a complicated way by the members of a community becomes important to them. it can be joined with another simple phrase. these cats and these dogs. and these hamsters. but now we’re on the road. and a relatively small number at that. and sentences which are not single lexemes. the formation of phrases of a certain type (we assume from this point on that a sentence is simply a certain kind of phrase) out of phrases of exactly the same type. these dogs. as well as five-word phrases in which these two-word phrases are joined by the lexeme and.

morphology and syntax of a language is called its GRAMMAR. it makes a difference semantically where the lexeme is moved to. For example. 8. We have already given an example which illustrates movement. the movement of what serves the purpose of focusing or highlighting what kind of answer the speaker is expecting (the name of a thing rather than of a person. in which the lexeme what is displaced from its ‘normal’ position following about. Deletion may be employed in many situations in which the repetition of phrases can be ‘reconstructed’. as in You all talked about what?. resulting in the meaning ‘You thought that I could fly. Each of these sentences expresses a meaning not expressible by coordination or subordination alone. but I didn’t fly’. but I didn’t think that I could fly’.consists of adjoining a simple phrase to a coordinate phrase to form a larger coordinate phrase. the coordination and subordination processes just illustrated can be combined to permit the formation of coordinate complex sentences such as I thought that I could fly. and their ability to freely combine. Two devices which appear in the grammars of all languages are known as MOVEMENT and DELETION. the result has three meanings: ‘I thought that I could fly. namely the question What did you all talk about?. or simply fly. but you didn’t fly’. If we exchange the second occurrence of I with you in the original sentence. the result is ‘but you didn’t think that you could fly’. In the subordination illustration. it also makes a difference semantically where the lexeme originates. but you didn’t think that I could fly’. yielding I thought that I could fly. which has two meanings depending on what is reconstructed. In this case. The theory of grammar. the recursive step consists of subordinating a sentence to a larger sentence by adjoining the structure I thought that or You thought that to the original sentence. but you didn’t think that you could fly’. Movement and Deletion Languages employ various syntactic devices to prevent expressions for complex meanings from becoming inordinately long. but I didn’t. compare What did you all put the baskets in? (which is equivalent to You all put the baskets in what?) and What did you all put in the baskets? (which is equivalent to You all put what in the baskets?). By the same token. as in the coordinate sentence You thought that I could fly. which is a significant part of linguistic theory as a whole. resulting in the meaning ‘You thought that I could fly. she could steer and he could make dinner (subordination of coordinate sentences). 7. ‘I thought that I could fly. In more complex sentences such as What did she tell him that you all talked about? and Did she tell him what you all talked about?. either think that I could fly. See Fiengo & May (1995) for analysis of dependency and its interaction with deletion. but you didn’t. Universal Grammar A complete analysis of the lexicon. and ‘I thought that I could fly. for example). The expressive power of language depends on the existence of many different recursive devices in its syntax. If this dependency is reconstructed in determining the meaning of but you didn’t. The additional meaning results from the fact that the second occurrence of the lexeme I in I thought that I could fly can DEPEND on the first. has as one of its goals the answer to the question ‘What is the class of all possible grammars?’ The properties of that class is 6 . you thought that I thought that I could fly and I thought that you thought that I thought that I could fly (coordination of complex sentences) and I thought that you could navigate.

Phr = phrase structure. Exp>. two syntactic components. among other things. is the input to the phonology. which constructs a hierarchical representation of the underlying syntactic structure of the expression. so that the task of language acquisition is really one of selecting the best grammars for the languages of one’s experience. including its stress and intonational contours. where ‘Mng’ is its semantic structure (representation of meaning) and ‘Exp’ is its phonological structure (representation of expression). The derivation of a meaningful expression in ST begins in one of its syntactic components called PHRASE STRUCTURE. was less well developed inasmuch as it neither attempted to account for the meaning of expressions. called SURFACE STRUCTURE. Chomsky called the result of that operation DEEP STRUCTURE. For each meaningful expression. The phrase-structure component is responsible for. through deep structure. and then to compute its meaning. (1) Outline of the ‘standard theory’ of generative grammar (Chomsky 1965) à Sem à Mng Ax à Phr à Lex à DS à Trn à SS à Phn à Exp Legend: Ax = axiom. the recovery of deep structure from expression is not an operation which is well defined within ST. However. nor dealt with language acquisition. Lex = lexicon. the grammar provides a DERIVATION.called UNIVERSAL GRAMMAR. among other things. Thus. The semantics is responsible for constructing the meaning of the expression. Exp = phonetic structure (expression). Finally. The overall structure of ST is outlined in Figure (1). the output of the transformational structure. which inserts lexical items into the underlying syntactic structure. a semantics. Mng = semantic structure (meaning). which starts with a member of a set of AXIOMS such as #S# (for ‘Sentence’) and terminates with a pair <Mng. which is widely recognized as having launched the modern enterprise of linguistic theory. The deep structure of an expression is the input both to the semantics and to the second of the syntactic components. Sem = semantics. recursion. DS = deep structure. Trn = transformational structure. whereas transformational structure is responsible for. the only way to determine its meaning according to ST is first to recover its deep structure. Rather than directly relating meaning and expression. it relates them indirectly. a grammar consists of a number of interacting components including a lexicon. Phn = phonology. Related to the goal of understanding universal grammar is the solution to the problem of language acquisition: ‘What enables a child upon limited exposure to a language to learn the grammar of that language?’ One widely held assumption is that people are innately endowed with knowledge of universal grammar. The output of phrase structure is the input to the lexicon. Generative Grammar The first complete statement of the theory of grammar in the form presented here was the ‘standard theory’ of GENERATIVE GRAMMAR presented in Chomsky (1965). The operations of the transformational and phonological components of ST map deep structure onto 7 . SS = surface structure. which determines the form of the expression proper.) According to ST. 9. given an expression. ST does not attempt to answer the questions of language perception and production posed at the end of section 2. using semantics. known as the TRANSFORMATIONAL STRUCTURE. (Chomsky’s 1957 presentation of the theory of grammar. and a phonology. henceforth ST. Inputs and outputs are italicized. movement and deletion.

and hence attempts to answer the question of linguistic production. Similarly. Nevertheless. as in Figure (3). For example. is ‘generative semantics’. given a meaning. following application of transformational rules of ‘merge’. and strictly semantic rules applying on the granch leading to semantic representation (now called ‘logical form’ or ‘LF’). the version of generative grammar known as ‘government and binding’ theory (Chomsky 1981) treats deep structure (renamed ‘D-structure’) as the starting point of a derivation. (3) Outline of the ‘minimalist’ theory of generative grammar (Chomsky 1995) à Sem à LF (= Mng) Lex à Trn à Phn à PF (= Exp) It is of course not a necessary feature of the theory of generative grammar that it fail to provide answers to the questions of linguistic perception and production. just as the operations which map deep structure onto expression are not reversible. One theory that maps meaning to expression in the grammar. (4) Outline of the ‘generative semantics’ theory of generative grammar Mng à Trn à SS à Phn à Exp 8 . ‘move’ and ‘check’. and hence lies outside its scope. but are not capable of mapping expression onto deep structure. as in Figure 2. As diagrammed in Figure (4). the mapping from meaning to surface structure is carried out by transformational rules. Chomsky justifies his decision to limit linguistic theory to merely providing a means of relating meaning and expression. the ‘minimalist program’ of generative grammar (Chomsky 1995) treats the lexical items which contribute to both meaning and expression as the starting point of the derivation. and surface structure (renamed ‘S-structure’) as the intermediate construct from which both meaning and expression are determined. including rules which replace semantic substructures by lexical items represented in terms of their underlying phonological properties. and in any event a total theory of language (whatever it is called) would have to encompass both linguistic competence and linguistic performance. Not everyone agrees with Chomsky on this point. which originated in the 1960s as an alternative to ST. whereas the second is a matter of linguistic PERFORMANCE (how people use language). the only way to determine its expression according to ST is again to recover its deep structure. with strictly phonological rules (now called ‘spellout’) applying on the branch leading to phonological representation (now called ‘phonetic form’ or ‘PF’). However. rather than to providing a means of determining one from the other. but continues to retain its ‘forked’ structure. eliminating both Dstructure and S-structure entirely. in which meaning and expression are related indirectly through an intermediate construct.expression. The theory of generative grammar has evolved considerably over the past thirty years. see Huck & Goldsmith (1995). and hence is properly within the scope of linguistic theory. derivations are still branched. by maintaining that the first a matter of linguistic COMPETENCE (what people know about language). neither are the operations that map deep structure onto meaning. (2) Outline of the ‘government-binding’ theory of generative grammar (Chomsky 1981) à Sem à Mng DS à Trn à SS à Phn à Exp Finally. and the mapping from surface structure to expression is carried out by phonological rules.

Similarly. 9 . for numerous illustrations of how this works. or the rules for forming well-formed formulas in a system of logic. Noam: Rules and Representations (New York: Columbia University Press. Mng à Con à Exp b. 1981).: Optimality Theory: An Overview (Oxford: Blackwell. Optimality theory has had its greatest success so far in accounting for the relation between lexical items and their expression. Then. just like the problem of language acquisition as described above in section 8. Terence. the class of all expressions of all possible languages and the class of all meanings) can each be defined by a GENERATOR. D. Optimality Theory However. That is. the classes of meanings. 1995). how do languages differ? The answer is in the RANKING of the constraints. (5) Outline of the ‘optimality’ theory of generative grammar a. see Archangeli & Langendoen (1997). MA: MIT Press. Diana. one determines its expression in a particular language by evaluating the members of the universal class of expressions against a universal set of CONSTRAINTS (henceforth Con). namely OPTIMALITY THEORY (Archangeli & Langendoen 1997. which works rather like the phrase-structure component of ST. Optimality theory assumes that the universal classes of expressions and meanings (that is. and dependency. and Langendoen. Exp à Con à Mng If. [collection of articles which introduce optimality theory] Chomsky. the problems of language production and language perception are considered problems of selection. Chomsky. Peter: Principles and Parameters: An Introduction to Syntactic Theory (Oxford: Oxford University Press. deletion. one determines its meaning (if any) in a particular language by evaluating the members of the universal class of meanings against Con. only very recently has a theory of generative grammar been proposed that attempts to answer both the question of linguistic perception and of linguistic production. forthcoming). eds. Presumably included among the members of Con is the universal class of lexical items. including recursion. MA: MIT Press. 1980). [presents the standard theory of generative grammar] Chomsky. given a particular expression. so whether it will succeed in extending the scope of linguistic theory to the narrow concerns of linguistic competence to the broad concerns of linguistic performance remains to be seen. Prince & Smolensky 1997. Noam: Syntactic Structures (The Hague: Mouton. Noam: The Minimalist Program (Cambridge. 1965). given a particular meaning. movement. 1997). Chomsky. 1957). Figure (5) provides an outline of optimality theory. Bibliography Archangeli. [the book that launched the enterprise of generative grammar] Chomsky.10. 1997). Noam: Aspects of the Theory of Syntax (Cambridge. so that the determination of which lexical items belong to which languages is also a matter of constraint ranking. and is only now beginning to deal with the major phenomena of syntax. as optimality theory proposes. [represents Chomsky’s current thinking about linguistic theory] Culicover. with each language corresponding to a possible ranking of the members of Con. expressions and constraints are universal. Noam: Lectures on Government and Binding (Dordrecht: Foris.

Paul: ‘Optimality: From neural networks to universal grammar’. 10 . Science 275 (1997).Greenberg. and May. 1995).: Semantics (Oxford: Blackwell. Andrew: Morphological Theory (Oxford: Blackwell. Joseph. forthcoming). 1604-1610. MA: MIT Press. Prince. Robert. [the article by Greenberg in this collection continues to be of interest] Fiengo. 1996).m MA: MIT Press. 1961). Prince. Paul: Optimality Theory: Constraint Interaction in Generative Grammar (Cambridge. Alan. Robert: Indices and Identity (Cambridge.: Universals of Language (Cambridge. Ladefoged. John: Ideology and Linguistic Theory: The Deep Structure Debates (London: Routledge. 1991). Geoffrey. and Segal. and Maddieson. MA: MIT Press. Richard. Gabriel: Knowledge of Meaning (Cambridge. and Goldsmith. and Smolensky. 1995). MA: MIT Press. Larson. 1997). Huck. 1995). Ian: The Sounds of the World’s Languages (Oxford: Blackwell. John I. Spencer. Peter. and Smolensky. ed. Alan. [technical introduction to optimality theory] Saeed.

Sign up to vote on this title
UsefulNot useful