291 Chapter 11 MOVE α, MINIMALISM 0. INTRODUCTION Let's look at what we've done so far. At the beginning of this book, we looked at rules that generate the basic phrase structure of human syntax. These rules generated trees which represent hierarchical structure and constituency. These trees have particular mathematical properties (investigated in chapters 3 and 4). In chapter 5, we saw that stipulated phrase structure rules missed some very basic generalizations, and developed the X-bar Schema. The X-bar schema, being very general, allows us (informed by parameter settings) to generate a wide variety of trees, and capture structural differences between heads, complements, adjuncts and specifiers. In chapter 6, we extended the schema to various clause types, complementizers, and DPs. In chapter 7, we saw that in fact the X-bar schema actually generated too many structures, and that we had to constrain its power. The device we use to limit it is a semantic one: the thematic properties of predicates (stored in the lexicon) and the theta criterion. What results from the output of the X-bar schema and the lexicon is called D- structure. The theta criterion holds of D-structures (as do the binding conditions). In chapters 8, 9, 10, we saw a variety of cases where we saw lexical items that either could not be generated where they surfaced by X-bar theory
Transcript
291
Chapter 11MOVE α, MINIMALISM
0. INTRODUCTION
Let's look at what we've done so far. At the beginning of this book, we looked atrules that generate the basic phrase structure of human syntax. These rules
generated trees which represent hierarchical structure and constituency. Thesetrees have particular mathematical properties (investigated in chapters 3 and 4).In chapter 5, we saw that stipulated phrase structure rules missed some verybasic generalizations, and developed the X-bar Schema. The X-bar schema,
being very general, allows us (informed by parameter settings) to generate awide variety of trees, and capture structural differences between heads,complements, adjuncts and specifiers. In chapter 6, we extended the schema to
various clause types, complementizers, and DPs. In chapter 7, we saw that infact the X-bar schema actually generated too many structures, and that we hadto constrain its power. The device we use to limit it is a semantic one: thethematic properties of predicates (stored in the lexicon) and the theta criterion.
What results from the output of the X-bar schema and the lexicon is called D-structure. The theta criterion holds of D-structures (as do the bindingconditions). In chapters 8, 9, 10, we saw a variety of cases where we saw lexicalitems that either could not be generated where they surfaced by X-bar theory
(eg. head adjunct complement ordering in French) or where lexical itemsappeared in positions other than the ones predicted by theta theory. Wedeveloped a new kind of rule: the transformation, which moved items around
from their base position in the D-structure to the actual position they appear inon the surface. There are three movement transformations: Head-to-headmovement (Ià C and Và I), NP movement and Wh-movement. In each of
these cases movement happened because it had to. Each movement had atrigger or motivation. Heads move to fill empty [Q] features or to take aninflectional suffix. NPs move to get case. Wh-phrases move to be near the[+Wh] feature. The output of the transformations is called S-structure, which is
itself subject to several constraints: the case filter, the EPP and the subjacencyconstraint. The model (flowchart) of the grammar looks like (1)
S-Structure (constrained by EPP,case filter, and Subjacencyconstraint)
Grammatical Sentences
This is a remarkably simple system that at the same time can generate a largenumber of sentence types. Not only that, it is a system with explanatoryadequacy, which makes specific predictions about how a child goes about
learning their language (via parameters).
Now, this said, a number of questions arise. Can the system be madeeven simpler? Are we missing any generalizations here? Recent work in
syntactic theory answers this questions with a resounding yes.
In this book we’ve proposed the following motivations for thetransformations.
2) a) Head movement -to get a suffix or fill Null [+Q]
b) NP movement -to appear in a position where case is assignedc) Wh-movement -to appear near a [+WH] feature.
Notice that while there are significant differences between the motivations for
the various types of movement, there is one overwhelming similarity. Themovements all occur so that one item can appear near another. In the case ofHead-movement the V or INFL head needs to appear as part of the same word
as the place it moves to. With Wh-movement, the Wh-phrase needs to be nearthe [WH] feature. NP movement is a slightly bigger stretch, but you can see thatthe NP needs to be near its case assigners (INFL, V). All the motivations formovement then seem to be locality constraints. That is, two items must be near
or local to one another.
If this is the case, then there isn't really a significant difference betweenthe rule types. Perhaps we can unify them into a single rule. This unified rule is
The minimalist programThe system of grammar described in this chapter is a verycursory look at some of the principles underlying the mostrecent version of generative grammar: The MinimalistProgram (Chomsky 1993, 1995). The minimalist programis motivated not only by the search for explanatoryadequacy but also for a certain level of formal simplicityand elegance. What is outlined here is by no means acomplete and accurate picture, but is meant to give you ataste of what current work is striving for
called Move α.. Move α says simply "move something" (but only if you have
to):
3) Move α (very informal version)
Move something somewhere
Now of course, this is a bit vague and we'll have to sharpen it up in some way.
In particular, we will want to constrain this so there isn't just random movementall over the sentence! So the next step is to formulate a constraint that motivatesand forces this transformation to apply (in all the different circumstances.)
Let's take wh movement as our paradigm case. In wh-movement thewh-phrase moves to the specifier of CP so as to be local with a [WH] feature.Another way to think of this is to say that both the wh-phrase and the
complementizer have a [WH] feature, and they need to compare them, or checkthem. Checking of features can only occur in a local configuration. In this casebetween a specifier and its head.
4) CP
NP C'
N C . . . [+WH] = [+WH]
checking configuration
The constraint that requires this movement is called the Principle of FullInterpretation (Chomsky 1993, 1995)
5) Full Interpretation
Features must be checked in a local configuration.
6) Local Configuration: WH features: Specifier/Head configuration.
We can extend this to the other cases of movement too. Imagine that Case is notsimply an ending, but is also a feature. An subject NP bears a [+NOM] casefeature. Imagine also that the heads of the phrases that assign case (INFL and V)
also bear this feature (although they don't show it morphologically).
We can thus reduce the case filter to the Full Interpretation: Nominative case isa feature checking like that in () and accusative case is like that in (8):
7) IP
NP I'
N INFL . . . [+NOM] = [+NOM]
checking configuration
Case and Agreement
The notion that INFL bears some kind of case feature often troubles
people, since Case is an inherently nominal kind of inflection and
INFL seems to be associated with verbal material. One clever
solution to this problem is to claim that verbal items like INFL do in
fact bear case, we just call case on verbs "agreement". In fact,
crosslinguistically there does seem to be some kind of correlation
between the kinds of agreement markers that are found on verbs and
the case marking on the subjects. So we could claim that [NOM]
when on a noun is case, but when on INFL or a V is agreement, thus
(Notice that this also allows a very elegant account of the passive. The passivemorpheme absorbs the V's external theta role and its [+ACC] feature). Because
of the configuration in (8) we are going to have slightly modify our definitionof local:
Finally, this can be extended to the head movement cases. Instead ofclaiming that verbs move to pick up inflectional suffixes in Và INFLmovement, let's claim that both the V and the INFL head bear some kind of
abstract inflectional features (e.g. [±past]), when the verb and INFL checkagainst one another then the suffix representing that tense feature (or agreementfeature) is allowed to surface on the verb. The local configuration in this settingis within the head itself (a relationship which is called a head-head
configuration):
10) INFL (morphology è V+ed)
V + [+past] [+past]
checking configuration
Similarly, both INFL and C bear a [+Q] feature, and they must be in a head-head checking relationship:
The system outlined above in section 1, is extremely simple and elegant. It does
however, make the unfortunate prediction that all languages will have exactlythe same set of transformational rules (although differ in phrase structure due tothe parameters). This is clearly not the case. English does not have Và I
movement. Many other languages lack passive and raising. Still others lack Wh-movement. Take the case of Chinese.
13)
a) ni xiang chi shemeyou want eat what"what do you want to eat?"
b) *sheme ni xiang chiwhat you want eat"what do you want to eat?"
c) ni kanjian-le sheiyou see-asp who"who did you see?"
d) *shei ni kanjian-lewho you see-asp"who did you see?"
Chinese appears to have no wh-movement. Languages like this are called wh-in-situ languages. Why then is it the case that the [WH] features on the WH-
phrases don't violate Full Interpretation? They are not in a local configurationwith their C. Full Interpretation predicts that (13a) should be ungrammatical and(13b) should be grammatical -- the exact opposite of the facts.
One possible solution to this problem is to slightly develop our modelof grammar. Consider what actually happens when you say a sentence.Presumably after you construct the sentence in your head you pronounce it. So
the next step after creation of S-structure is the phonological component of thegrammar. Similarly, the sentence is not fully interpretable (or semantically
comprehensible) until after S-structure. The semantic/interpretive part of thegrammar must also follow S-structure. The name given to the phonologicalcomponent in minimalist approaches to grammar is Phonetic Form or PF. Th e
name given to the semantic/interpretive component is Logical Form or LF.With these two additions the model of the grammar now looks like:
14) The Computational Component
The Lexicon X-bar rules “the base”
D-structure (constrained by Thetacriterion)
Transformational RulesMove α
S-Structure (constrained by EPP,
and Subjacency constraint)
LF PF Full Interpretation.
Chomsky (1993) makes two important claims. First he claims that FullInterpretation is a constraint that holds of sentences at LF, not S-Structure.
Second, he claims that the operations that happen between D-structure and S-
structure can also happen between S-Structure and LF. That means thattransformations can apply on the left arrow between S-structure and LF in (14).
Notice what this does, it means that some kinds of movement couldhappen after you say the sentence. At first glance this may seem very counter-intuitive, but it actually allows us to make the following remarkable claim about
cross-linguistic variation: All feature-checking movement happens in everysingle language. The differences between languages are in when that movementoccurs: before you start to say the sentence or after! Essentially there are twokinds of movement, movement that happens between D-structure and S-
structure (called overt movement) and movement that happens between S-structure and LF (called covert movement). Since covert movement happensafter the branching off to the PF (phonology) component you simply can't hear ithappen. The differences between languages then, are in when they time specific
instances of the general operation move-α. English times those instances
involving [WH] features before S-structure, Chinese times those samemovements covertly (after S-structure). This can be simply encoded in aparameter:
15) WH-parameter: Overt/Covert(English sets Overt, Chinese sets covert)
This kind of story also allows us to get rid of the odd-man-out oftransformations: Affix (INFL) lowering. This was the only movement that welooked at that went downwards. It also appeared to be in complementary
distribution with V-movement. If a language has one, it doesn't have the other.With the system described above we can get rid of the affix lowering account ofEnglish. English doesn’t have affix lowering. Instead, it has V à INFLmovement like any other language, only in English it is timed covertly, so you
never hear it.
3. CONCLUSION
The story I've spun you here covers a remarkable amount of material. It alsounfortunately has many holes. This is partially because this is only meant to be
an introductory textbook in syntactic structure. We couldn't possibly get into all
the details of all the material, nor could we possibly consider all the possiblearguments for (or against) a proposal like the overt/covert movement approach.In order to properly give such ideas their due, you'll need to take more advanced
courses in Syntax. Some good books that will take you on from where I've leftyou here are listed below, in the further reading section. I hope that while othersections of this book have given you a firm foundation upon which you can
build your syntactic knowledge, this chapter has given you a taste of currentresearch in syntax and the future of the discipline.
IDEAS, RULES AND CONSTRAINTS INTRODUCED IN THIS CHAPTER
i) Move α (very informal version)
Move something somewhere
ii) Full InterpretationFeatures must be checked in a local configuration.
iii) Local Configuration: [WH], [NOM] features: Specifier/Head configuration.
