as a notion in linguistic part i the history of the phrase
TRANSCRIPT
As a notion in linguisticPart I
The History of the Phrase
The PhraseToday phrases are an essential and
undeniable truthHornstein, Nunes and Grohmann:
F4: Words are composed into units with hierarchical structure i.e. phrases.
Given this, they should have been discovered a long time ago
So when did linguists start referring to phrases?
Brief Overview of the History of LinguisticsThe history of linguistic investigation goes
back more than 2500 years:Before 600 BCE Indian linguistics
only ‘discovered’ in Europe many centuries laterSocrates (469-399 BCE),
one of the first to write on Greek linguisticsonly through secondary sources (e.g. Plato 428-348)
Study of Latin, Arabic and Hebrew after Greek
Other European languages only in late Middle Ages
Indian LinguisticsStarted as a way of preserving an
understanding and correct pronunciation of Vedic SanskritChanged over the centuries (as languages
do)Was the language of the scriptures and of
religious ritualsWanted to maintaine an understanding and
correct pronunciation of Vedic for religious purposes
So they started to write grammars
PaniniOne of the most important of the Indian
LinguistsNot because he was one of the earliest
His grammar was written sometime between 500 and 300 BCE
But because it is the most extensiveContains nearly 4000 rules
And because of it’s remarkable advanced natureIt contained notions not found in western linguistics until
the 20th centuryThe phonemeNull phonemesFormal rulesRule ordering
The Nature of SanskritSanskrit was a highly inflecting language
A lot of Case and agreement morphologyWord order was much freer than in
languages like EnglishRaamah pustakam pathati
Ram books readsRaamah pathati pustakamPustakam Raamah pathatiPustakam pathati RaamahPathati Raamah pustakamPathati pustakam Raamah
What did Panini say? Not surprising that most of the grammar is
about phonology and morphologyHad very little to say about syntax
Phrase Structure Rules?According to Kiparsky Panini’s grammar contained
rules like:A AA /C _ D
‘A’ becomes ‘AA’ in the context of a preceding ‘C’ and a following ‘D’
This looks very similar to a phrase structure rule:S NP VP
‘S’ is made up of an ‘NP’ followed by a ‘VP’But
This is not Panini’s representation of this rule – it’s Kiparsky’s
It isn’t a phrase structure rule, but a morphological oneIt is a reduplication rule
A is reduplicated (AA) in a certain context
Why Panini didn’t discover the phraseAlthough very advanced, it understandably
concentrated on the most obvious aspectsMorphology – because there are many forms of
many wordsPhonlogy – because the written system had a
phonetic base (unlike Chinese)Because of the free word order, syntax was not
so obviousBecause Indian grammarians studied no other
languagesdidn’t compare Sanskrit to other systems and so
couldn’t notice patterns in syntactic phenomena
GreekAncient Greek was a written language over 1400
BCEMycenaeans - Linear B
Based on the Minoan syllabic systemWasn’t ideal for Greek and we still can’t work out how this
early version of Greek soundedThe first written system was forgotten until about
900 BCEThen a new written system was developed
Based on Phoenician scriptPhonetic representation – more suited to Greek
pronunciationDeveloped over the yearsNow is the basis of most European written systems
We can therefore assume that the Greeks were interested in their language (at least phonetic aspects of it) from about this time
Greek Language PhilosophersSocrates (469-399 BCE)
No written records of anything that Socrates saidWe have the reports of other Greek philosophers
(e.g. Plato)Unclear whether these accurately portray Socrates' ideas
or whether they portray Plato’s ideas
Plato (428-348) also had things of his own to say about language
Aristotle (384 – 322 BCE) furthered these ideasThe Stoics (310 – 50 BCE) probably gave the
most thorough treatment
What the Greeks had to sayGreek linguistics was bound up in Greek
philosophyMathematics; Physics; Astronomy; Biology; Ethics;Logic (study of ‘valid arguments’); Rhetoric (the
study of persuasive speaking); Epistemology (the study of understanding)
E.g. whether words derive in nature or are man made?the nature of objects themselves are the origin of
wordsman gave names to objects randomly
Obviously these sorts of questions are not likely to reveal much about syntactic structure
Specific Greek Ideas on LanguagePhonetics, but only to the extent that it involved
the written systemNo written system is purely phonetic and so what they
had to say on the subject was not particularly interesting
Semantics, but more as a part of logicDistinguished subject and predicate from a semantic
point of view, but didn’t realise that this had any realisation in the sentence
Plato was first to split sentence into two major elements: onoma (noun) subjectrhema (verb) predicate
Words, not phrases identified morphologically
They were interested in words and their ontologyled to questions of categories
Specific Greek Ideas on LanguageThey were interested in sentences and
their functionsdeclaratives, interrogatives, imperatives
Not so much in their formsThey were interested in morphology
created paradigms of all the different forms of types of word
And terms for these forms
The Nature of GreekIt is not surprising that the Greeks were not
so interested in syntaxGreek, like Sanskrit, was highly inflectional
and had free word orderThe obvious aspects of the language were
morphological and not syntacticStudied just one language (Greek)Their philosophical interests lead them in a
different direction
On SubordinationPlato identified two basic elements (words)
of the sentenceGreeks also knew other words were
subordinate to theseIs this the beginning of the phrase?Today subordination = something is part of
somethingGreek notion more like ranks in the Army
Captains are subordinate to GeneralsDoesn’t mean that captains are part of generalsHierarchy – but not constituent structure
DependencyX is subordinate to Y = X depends on Y
Dependency marked morphologically (i.e. By agreement)Not semantically – meaning was the domain of logic,
not linguisticsFree word order meant that there was no
compulsion to see this relationship in terms of phrase structure
LatinThe Romans impressed by Greek intellectualism
encouraged the Greeks to continue their studiesStudy of Latin very much influence by Greek
linguisticsMost attempted to make what the Greeks said fit
LatinLatin similar to GreekBut not the same as Greek
categories and terms did not fit perfectly
This dependency on Greek linguistics meant that the Romans did not produce much innovation
PriscianPriscian (500 CE): one of the most influential
grammars of Latin basis of the study of Latin even after the fall of the
empireAlmost entirely based on Thrax’s (170 – 90 BCE)
Greek grammarThrax’s grammar did not even mention syntaxPriscian had 2 volumes on it
Nothing of interest in these chapterscontained rather dubious philosophical arguments
o Latin is basically S V because existence precedes actionsHe did mention the notion of a subordinate clause
But again, the use of subordination was that of the Greeks rather than the modern linguistic one
The Middle AgesThe middle ages starts with the fall of the
Roman Empire and ends with the start of the Renaissance
From a linguistic point of view we can split it into two partsScholasticism
Not particularly interesting – mostly based on Priscian grammar
Speculative GrammarMore innovative – the start of the study of other
languages
ScolasticismThe main idea to bring all knowledge under
one (Christian) frameworkIt strongly rejected ‘pagan’ Greek learningLatin became a lingua franca
Most linguistic work = producing Lain grammars for teaching
Ironically this maintained Greek influencebased on Priscian grammar, which itself was
based on Thrax
Speculative GrammarInterests in other languages than Latin
started to gather strengthThrough influence of Arabic linguistics from
Moors in Spain(influenced by Aristotelian ideas)
Through influence of Hebrew linguistics from Biblical studies
Perhaps because of this, more emphasis was put on syntax
Speculative grammarians first to unify syntactic and semantic notions of subject and predicate
Thomas of Erfurt (1310)Socrates albus currit benne
Socrates white runs well“white Socrates runs well”
‘albus’ is dependent on ‘Socrates’ and ‘benne’ is dependent on ‘currit’
Does this mean subject made up of ‘Socrates albus’ and a predicate made up of ‘currit benne’?i.e. phrase structure analysis?
But he took the meaning of dependency to be the traditional one“one part of a construction stands to another either
as depending on it or satisfying its dependency”
Thomas of Erfurt (1310)Also took Greek idea that dependency is based on
morphologyverb agrees with subject, so verb is dependent on
subjectTherefore the subject is the main element of the
sentenceSocrates
albus currit benne
Very different analysis to modern phrase structure based one
There is no evidence here of the notion of a phrase
The Renaissance (starting 14th C)The Renaissance was both forward and
backward lookingRevived interest in classical studies (Greek)Rejected current wisdom
Rise of the idea of nation more emphasis placed on national languages and less on the lingua franca
Surprisingly not much innovative thinking from a linguistic point of viewMost attempted to force the analysis of national
languages into the framework of Priscian who had already tried to force the analysis of Latin into the
framework of Thrax!
Port Royal GrammariansOne group of Renaissance linguists did have some
interesting ideas about syntaxConnect syntax to meaning unfortunately
rather than the traditional (Greek) view that morphology was its basis
the start of the ‘notional definition’ of categories (a noun is a name, etc.)
More interestingly, developed a new view of subordinationSubordination is a superficial representation of
independent constructs:The invisible God created the visible world God is invisible. God created the world. The world is visible.
This view necessarily precludes the idea of phrases the complex sentence is not made up of connected
elements, but of independent simple ideas.
Comparative LinguisticsDuring the Renaissance more languages than
ever before came to be studiedBut not much thought about possible
connections between languagesAssumed system devised by the Greeks was
applicable to other languages, it was Not assumed that other languages had
developed from GreekThere had been several theories, from Biblical
sources, that held that European languages all descended from a single parent language
Scythian HypothesisAccording to the Bible, God made everyone
speak different languages after the tower of Babel incident
But, then the flood killed everyone, except for Noah and his family
Legend has it that Noah’s son, Japheth, was the father of Europe he spoke a language from which all European
languages have descendedThis language, apparently, was called
Scythian (Goropoius 1569)Scythian died out a long time ago
Confirming the Scythian HypothesisOnce enough attention had been turned to
European languages, it started to be noted that there were similarities between them
Furthermore, regularities in the differences between languages could be noted
These mainly concerned the forms of words:Latin English
piscesfishpede footpater fatherpinna featherpugnam fight
So it seemed that the Scythian Hypothesis could be correct. Throughout the 17th century it gained popularity
Indian linguistics comes to EuropeThe Indian linguistic tradition had been known about
in Europe since the 1500sWalter (1733) added Sanskrit to Scythian family
real interest in the Indian tradition startedComparative Linguistics virtually swept all other
linguistic interests awayThe notion of the Indo European language family
really took offOn a scientific basis more than a religious one
It had been thought that Hungarian was not an Indo European language since the early 18th Century, 1770 Sajnovics - Hungarian and Lapp are related and 1799 Gyármathi - Hungarian and Finnish are from the
same family
NeogrammariansTowards the end of the 1800s a group of
particularly aggressive young linguists in Germany dominated linguistics
They criticised previous comparative linguistics for being too wishy-washy
Believed that sound laws should be inviolablelaws that govern how languages change over
time and hence can be related to common ancestors
if there were exceptions to these laws, they cannot have been stated right
Why Comparative Linguistics did not discover the phraseThe main focus of the comparativists was on the
lexiconeasy to collect and compare vocabularies of languages
to discover similarities and patterns in differencesnot easy to do the same with syntactic phenomena
Some languages have free word order, others have more fixed orders
Some languages have basic SVO organisation, others VSO or SOV
These things don’t seem to have much of a connectionTherefore, not much attention was paid to syntactic
phenomena, no surprise that no new discoveries were made in the
area for the nearly 200 years that comparative linguistics dominated the study of language
European StructuralistsThe dominance of comparative linguistics
was finally broken Ferdinand de Saussure’s ‘A course in
general linguistics’ published 1916 three year after his death
This ushered in a period known as StructuralismA hopeful title, from our perspective
What Structuralists BelieveThe elements that make up a language
form a system which can only be understood in terms of each otherThe elements are signs
An arbitrary link between form and meaningThe meaning of the signs is dependent on
the meaning of other signs in the systemThe 8.25 to Paris
Did the Structuralists discover phrase structure?Unfortunately not
Like the comparativists, the structuralists were concerned more with wordsTheir pronunciation (the Prague school)Their meaning
Saussure had virtually nothing to say about syntax as we conceive of it today
ConclusionFrom the beginnings of the study of language
Prior to 600 BCETo the Structuralist period in Europe
Until after the 2nd World WarNo one came up with the notion of a phraseReasons
The languages which were studied initially did not make syntactic discoveries easy
The later study of other languages were concerned with applying old ideas to new situations, rather than being innovative
The interests of the times (philosophical issues, language families, languages as systems) were more easily satisfied through looking at words
As a notion in linguisticsPart II
The History of the Phrase
Linguistics in the USA before 19th CSpecifically American linguistics is a relatively new
thingPreviously American linguists were doing European
linguistics (especially Indo-European based Comparative linguistics)
America has a wealth of its own languages (Amerindian)But until the late 1800s there was little intellectual
interest in these culturesAmericans were interested in expanding their own European
based culture and the native populations were a hindranceThe European attitude towards native peoples of the Americas
had been one of cultural imperialism from the startSo linguistic efforts were always geared to teaching the
natives European languages rather than learning theirs
Boas: the father of American linguisticsFranz Boas (1858-1942) was a German
anthropologistHe started as a physicist and geographer
became deeply interested in Amerindian culture on an expedition to Canada
Moved to America in 1887He saw the study of Amerindian cultures and
languages as urgent as they were fast disappearingmany had already died out
Devised ‘field methods’ to train linguistic students so they could rapidly form grammatical descriptions without having to learn the language
Discovery ProceduresBoas’ discovery procedures were a set of tools
linguists could apply to discover the grammar of a language
The best known of these are based on the notion of distributionThe minimal pair test
If two sounds appearing in the same context produce different words then they are distinct sounds (= phonemes) [khæt] – [phæt] [kh] and [ph] are distinctive [phæt] – *[pæt][ph] and [p] are not
In cases where sounds are in complementary distribution, they are non-distinctive [spæt] – *[sphæt]
This is different to the case where a sound is not part of a language *[Xæt], *[sXæt],
Boas’ Philosophical AssumptionsBoas was a believer in cultural relativity
Against the idea that cultures pass through stages of development with Western culture as the highest
Cultures cannot be evaluated against one another as higher or lower
Cultures develop their own ideas from which they view the world and thus cannot be compared on any external measure
To understand a culture you need to study it from its own position
Linguistic RelativityHumboldt (1767-1835)
Inner form of a languageLanguages don’t necessarily differ in the concepts they can
expressThey differ in how they combine concepts in order to view
the worldBoas combined Humboldt’s inner form with
cultural relativity in ‘linguistic relativity’The belief that we cannot make generalisations
about a language based on what we learn of another languageGoes against European linguistic tradition since the Greeks
Linguists must study a language in its own termsHence discovery procedures
Implies that languages can vary without limitSupported by the fact that Amerindian languages seemed
very different from European langauges
Boas ReprieveLinguistic Relativity
Languages can only be studied in their own terms
Discovery ProceduresThere are methods we can apply to discover
the units that languages make use ofHowever, Boas did not discover the phrase
BloomfieldLeonard Bloomfield (1988-1949) was the founder
of American StructuralismBut he wasn’t always a structuralistHe trained in Germany under the NeogrammariansHe became very impressed by the Wilhelm Wundt
(1832-1920)Wundt is said to be the father of experimental
psychologyHis aim was to make psychology scientificHe used introspection
Bloomfield wanted to make Linguistics more scientific
Bloomfield and BoasBloomfield was also influenced by Boas
Particularly in terms of linguistic relativity and the use of discovery procedures
Bloomfield 1914His influences from Boas and Wundt were
brought together in his short book An Introduction to the Study of Language (1914)This became a popular course book for
linguistics in AmericaIt contained chapters on all aspects of
linguistics, including syntaxIt contained the word ‘phrase’ twice
Both times referring to what we would call an idiom
Clearly at this time he did not know about phrases
Bloomfield and StructuralismIn the1910s, Wundt became heavily
criticised for his methodsIntrospection was not really scientific
Bloomfield needed something else to base his linguistic science on
In 1923 he published a seemingly positive review of Saussure’s workThis book still had no mention of the phrase
But even then, it was apparent that he had replaced Wundt with Behaviourism
BehaviourismBehaviourist Psychology took an extreme
empiricist stanceTo account for human behaviour all one needed was
to directly connectThe environment (stimuli)The subsequent behaviour (response)
There is no need to refer to unobservable things like ‘mind’
Behaviourism in Linguistics is a bit trickyThere is plenty of stuff that we cannot directly
observeBloomfield thought it was possible
If every abstract level was ultimately based on what is observable – sound
Phonetics phonology morphology syntax
Bloomfield’s LinguisticsIn 1933 Bloomfield rewrote his earlier textbook in
a much enlarged version LanguageIn this book the chapter on Syntax concerns itself
with constituent structureThe term phrase is used to mean a constituent
Therefore we know that the notion of the phrase was introduced at some point between 1914 and 1933, probably in the 1920s
The origins are a combination of:Bloomfield’s empiricismHis application of discovery procedures
Phone phoneme morpheme word phrase sentence
Bloomfield on PhrasesThe chapter on syntax is not very long It is rather superficial
Containing a few not very detailed examples Poor John
And no attempt at representationSome discussion of the notion ‘head’ of phrase
Head is defined distributionally A head is a word which has the same distribution as the phrase
Endocentric phrases have heads Poor John John
Exocentric phrases don’t have heads In the park; if John ran away
But the man doesn’t have a head by this definition the man has the same distribution as poor John, so they are of
the same category So the head does not determine the category of the phrase –
unlike current view
Open, closed and partially closed phrasesAn open phrase is something that can be added
to and still be the same type of phrase:students – interesting students – these interesting
studentsA closed phrase cannot be added to without
changing into another phrasethese interesting students – * polite these
interesting students – saw these interesting studentsA partially closed phrase can be added to, but not
by everything that could be added to an open phraseblack dogs – big black dogs – the big black dogsbig dogs – * black big dogs
This is not very illuminatingJust gives names to phenomena without explaining it
Neo-BloomfieldiansAfter Bloomfield’s death (1949) his
students carried on his workZellig Harris (1909-1992), Charles Hockett
(1916-2000)Hockett gave more detail to the Immediate
Constituent AnalysisHarris formalised the system to a greater
extent
HockettHockett’s A Course in Modern Linguistics
(1958) contains 3 detailed chapters on the Immediate Constituent (IC) analysis
It is interesting to go into detail about these ideas as they show that Hockett’s view of the phrase is not the same as today’s
Like Bloomfield, he used distributional devices to define syntactic notions such as ‘head’With the same problems
He did try to represent structureThough he said this was not important in itself –
just useful for linguists
Chinese Box representationHockett represented structure like this
This representation allows things that is not allowed in current phrase structure diagrams (trees)
Disappearance of elementsMarkers – add no meaning
Discontinuous ConstituentsConstituents split by material which is
not part of them
The Start of Modern PhrasesChomsky (1928- ) was a student of Harris
during the 1950s1957 Syntactic Structures
A small book for computer science studentsContained his own version of the Structuralist
IC analysisPhrase Structure Grammar
This was a ‘straw man’Designed to show what Chomsky thought was wrong
with the IC analysis
Phrase Structure GrammarMade up of rewrite rules, e.g.
S NP VPVP V NP
These produce phrase structures, represenatable as a tree
This is now how we conceive of a phrasePhrases have positions inside other structuresPhrases have structures of their own
How do phrases differ from ICsIt is impossible to have discontinuous
constituents in a phrase structure grammarRewrite rules concern mothers, daughters
and sistersS NP VP
S is the mother of NP and VPNP and VP are daughters of SNP and VP are sisters
Mothers and daughters stand in a dominance relation
Sisters stand in a precedence relation
Discontinuous constituents and crossing branchesA structure with discontinuous constituents
must contain crossing branches
This involves a precedence relation between Aux and NPThese are not sistersNP is the great aunt of AuxPhrase structure rules do not refer to
grandmothers, aunts, great aunts, etc. and so cannot produce these structures
Why is there a difference?The structuralists did not adopt a restrictive
theory of phrasesThe Chinese box representation placed very few
limitations on what could be representedEven if there were things that were difficult to
represent, this would not have matteredRepresentations were not important – only tools for the
convenience of the linguistChomsky’s PSGs are restrictive theories of
phrase structureThere are clearly things that they cannot do
(restricted)The representations come directly from the theory
(rules) and so are important for making the restrictions obvious
Was Chomsky right to criticise ICs?PSGs were a straw man – meant to be knocked
downPSG and IC analysis were clearly not exactly the
sameBut the structuralists had no theory
So they were immune from criticismBut at the same time, they made no real claims
In order to see the problems with the IC analysis, Chomsky was forced to invent the theory
One of his points was that the lack of formal theories of language is one of the weaknesses of linguistic investigation since the classical periodWith the exception of Panini
ConclusionThe notion of a phrase first came into being in the 1920s
If the history of the study of Language were put into 1 hour, the phrase would have been discovered less than 15 seconds ago
It probably developed out of two ideasRadical empiricismDiscovery procedures
Everything must be based on what is observable and can be discovered by observing the distribution of elements at various linguistic levels
Bloomfield didn’t really develop the notion much beyond the basics
Hockett went into more detail and tried to represent itChomsky introduced the current view in the 1950s
Differs from the original view Based on phrase structure rules More restrictive
As a notion in linguisticsPart III
The History of the Phrase
Are phrases a linguistic fact?Today, most syntacticians assume the existence
of the phraseA lot of them think of this as unquestionable
It has been demonstrated beyond doubtIt is so obvious
But in empirical science, nothing should be beyond doubtAll facts are theory dependentProof is impossible
In order to see how viable the assumption of phrases is we must review evidence for and against them
Evidence for the existence of phrasesThere are three types of argument that
have been put forward to support the assumption of the phraseArguments that descriptions based on non-
(phrase)structural grounds are inadequateTheoretical arguments that phrases are
necessaryEmpirical evidence for the existence of
phrases
Arguments against non-phrase based descriptionsChomsky (1957) demonstrated that a grammar
that did not assume phrases could not account for syntactic phenomena
Finite State GrammarRepresents a network of ‘states’ connected to each
otherThe grammar travels from one state to another as it
parses a sentenceThe sentence is grammatical if the grammar enters
the final state at the same time as parsing the last word of the sentence
Two things allow the grammar to move from one state to anotherConditions on the stateThe word of the sentence currently being parsed
An example of a FSG
SI is the initial state, SF is the final stateFrom SI the grammar can move to S1 or S2
If the first word is a determiner, it moves to S1If the first word is a proper noun, it moves to S2
From S1, it can move to S1 (if the next word is an adjective = recursion) or S2 (if it is a noun)
From S2 it can move to SF if the next word is a verb
Sentences this FSG can parseJohn leftThe boy leftThe old man leftThe old confused man leftEtc.
What a FSG can’t doEmbedding
Sentences can be part of sentencesThe man [who John met] left
Two ways to account for thisAfter the noun we add a further set of states
which allows another sentence to be parsedWe allow the network to recurs back to the
initial state
Including sentence networks inside sentence networks
But this reduplicates exactly the same network
There can be an infinite number of embedded sentence, so the grammar would have to be infinitely big
Recursing to SI
This uses the same network again, so it is better than the other idea
But it won’t work as the final state will be arrived at too soonThe man who John met (end of sentence)How can we get the final verb?
Conclusion on Finite State GrammarsThis is just one problem faced by a Finite
State GrammarThere are many othersThey are not adequate models for parsing
human languages
Arguments for the necessity of phrasesChomsky has argued that phrases are necessary
because all syntactic processes are dependent on phrase structure
E.g. Auxiliary inversionThe man is being watched is the man being watchedTake the first auxiliary and move it in front of the subjectBut
The man who is tall is being watched * is the man who tall is being watched
The reason this doesn’t work is that the auxiliary that moves must be the one of the main clause, not any embedded clause
Thus the process is sensitive to the structure of the sentence
Empirical Arguments for PhrasesBoas’ discovery procedures are still in common
use as a way to determine phrase structure (distribution tests, pronominalisation test, coordination test, etc)
The tests may be turned round and used as evidence for the existence of phrasesIf there were no phrases, why do the tests work?
John ran up the hill John ran up a bill John ran there * John ran thereUp the hill, John ran * Up a bill, John ran John ran up the hill and * John ran up a
bill anddown the road up a debt
How good are these arguments?FSGs are in adequate for modelling human
languageBut FSGs are not the only non-phrase based grammar
Other models do not suffer the same problems
Structure based processesIt is clear that syntactic phenomena are limited by
something that is not simply linearBut that doesn’t mean they must be limited by phrase
structureEmpirical evidence
Certainly shows somethingBut if it can be accounted for without phrases, it isn’t
an argument for phrases
Against the phraseThe phrase developed out of 2 structuralist ideas
Radical empiricismDiscovery procedures
Chomsky has been severely critical the American Structuralist movementThe empiricist stance is not scientific
Behaviourism involves after the fact explanations How do we know what any instance of human behaviour
is a response to – given a situation, a person in principle might say anything
It cannot account for certain facts How can children learn language
There is no reason to believe that there should be discovery proceduresThere are no discovery procedures employed in any other
scienceWe get our data from wherever we can get it
The big questionIf Chomsky is dismissive of empiricism and
discovery procedures, why did he accept phrase structure which followed from these?
Non-structure based theoriesDependency Grammar
Similar to traditional view of subordinationUsually semantically based, not morphological
Always one word which is not dependentAll other words are dependent
Not very good at accounting forword orderIn principle any word order
would be possibleNo reason why branches shouldn’t cross
Absolute vs. Relative linearityA FSG organises sentences in an absolute linear
wayThere is a first position followed by a second
position, etc.But linear order can be relative
A is in front of BC is in front of B C D A B or C A D B or A C D B, etcD follows C
As there are a choice of possible orders, we need some way to choose which one of them is the best
Optimality Theory is a way to determine the best of a set of alternatives
Optimality TheoryWe start with an inputWe generate a number of possible candidate
expressionsWe evaluated the candidates against set
conditionsConstraints
Constraints areConflicting (nothing can satisfy them all)Ranked (in the case of conflict the higher ranked
one is adhered to)The candidate which best satisfies the set of
constraints is optimal (grammatical)
How it works
Does this answer criticisms?Embedding
Assuming that the input is a dependency arrangementOrdering of words done with respect to dependent
wordsThe man who John met leftLeft – manMan – theMan – metMet – John, who
The same conditions will be relevant for both the main clause and the embedded one, so there is no redundancy
All words will find there position in the sentence with respect to the words they are dependent on
Structure dependency of linguistic processesAgain, the dependency relations in the
input are enough to determine the relevant information without phrasesThe man who is tall is being watched
Is – tall...Tall – man...Man – watched...Is – watched this is the auxiliary that
inverts
Empirical evidencePhrases
Have distributionCan be pronominalisedCan be coordinated
Distribution Words are ordered with respect to the
words the are dependent on the dependent word will appear wherever the
superordinate word appears:The man likes Mary
Man is subject – in front of verbDeterminer is in front of noun
Mary likes the manMan is object – behind verbDeterminer is in front of noun
Thus they will behave like a unit even if the grammar does not define them as suchThe phrase is an epiphenomenon
PronominalisationPronouns stand instead of nouns
They are positioned by the same constraints (as subjects, etc.)
Pronouns cannot be modified, nouns canTherefore nouns can have more
dependentsThe pronoun seems to replace more than
the nounPhrases are epiphenomena
CoordinationWords of the same type can be coordinatedSome of these words will have dependentsThese dependents will be positioned with
respect to themThe coordinated words will be positioned
with respect to the coordinationJohn and the tall woman
And – John, womanWoman – the, tall
This gives the appearance that what is being coordinated is bigger than wordsPhrases are epiphenomena
Which is better:a theory with phrases or without?Too soon to decideBut the issue is in danger of not being
addressedToo many linguists have dismissed
phraseless theoriesThey are not being investigated
So we are not discovering what they are capable of
ConclusionThe phrase is a relatively modern notion
First introduced in the 1920s and developed in the 1950s
Since its introduction syntacticians have enthusiastically embraced itTo the detriment of the opposite assumption
Real linguistic theory started after the 1950sSo phraseless theories have not really been
exploredThough the assumption that sentences are organised
without phrases is a much older ideaUntil such investigation takes place, we will
never really know whether phrases are a necessary part of human languages