2How do we explain things? In lay terms an explanation answers the question why:A: “Why are you carrying that large heavy suitcase?”B: “I just bought a Japanese wrist watch”A: “What’s that got to do with the suitcase?”B: “The watch is Japanese but the batteries are Russian”.
3In science, we also want to answer the question ‘why’ In science, we also want to answer the question ‘why’. But things are more difficult.The reductionist problem:If X explains Y, what explains X?The only natural end to this would be to map everything back to the initial event (the big bang).But we can’t do this – not enough evidenceEven if we could, we wouldn’t understand it and what’s the point of an explanation that you can’t understand?
4Explanation in Linguistics There are two ways that explanation comes into linguistics:The normal one – what explains X?A more technical one: given two theories how do we decide which one is best?
5Levels of adequacyFor any set of data, there are an infinite possible grammars that capture the dataSo how do we decide which one is the right one?
6A demonstration of infinite possible grammars Suppose a simple language with one word, “a”The sentences of this language contain any number of instances of the word:aaaaaaaaaaetc.
7S a S SaThis grammar will generate all and only all the sentences of this languageBut so will:S a or S a etc. S aS S Sa S aSMoreover so will:S’ Sb (obligatory deletion rule) S a Delete b at the end of a S Sa sentence
8All these grammars are distinct, but they all generate languages which have sentences made up of any number of ‘a’sWhich one is correct?Chomsky (1965) proposed that different grammars attain different levels of adequacy:observational adequacydescriptive adequacyexplanatory adequacy = the highest
9Observational adequacy: a grammar which predicts all and only the grammatical sentences of a language (e.g. all of the grammars we previewed)Descriptive adequacy:Native speakers have intuitions about how sentences are structured:S S S a a S a aA grammar which conforms to these intuitions is descriptively adequate
10Explanatory adequacyA theory which sheds light on the logical problem of language acquisition is explanatorily adequatetwo descriptively adequate theories of two different languages do not amount to an explanatorily adequate theory of language if they are very disparateif both are possible human grammars, how would a child be able to learn any one?a theory which contributes to a coherent notion of Universal Grammar therefore is one which has explanatory adequacy
11Normal ExplanationExplanatory adequacy does not in itself guarantee explanationit is just a method to use to distinguish between different grammars and to guide researchExplanation in linguistics comes through restrictionif grammatical principles are as complex as the data, then we have descriptionthe simpler the grammar, the more explanationsimple does not necessarily mean easy to understandit means structurally simplerthe fewer and more general the principles the simpler
12Thus, suppose we have a grammar whose principles are in a one-to-one relation with linguistic phenomena:Here the grammar is just as complex as the data and doesn’t help us understand it any betterBut if two or more of these principles can be collapsed into a more general one, the grammar is simpler than the dataP1P2P3etc.Phen1Phen2Phen3P1P3etc.Phen1Phen2Phen3
13But this is a reductionist argument We may be able to achieve more and more explanationBut we can never achieve the ultimate explanationPerhaps this is enoughPerhaps not
14The Minimalist Programme Chomsky has argued (since 1990s) that we can achieve a greater degree of explanationIf the theory we produce is built on only absolutely necessary assumptions, then it cannot be reduced any furtherWhat is absolutely necessary?the set of assumptions that if they were not made, we wouldn’t have a theory of language
15Suppose language is the mental system that links the part of the mind concerned with thinking and the part of the mind concerned with articulation (bodily movements)thoughtactionlanguage
16Suppose that is all there is The two interfaces have requirements for language in order for it to do its job:the products of the linguistic system must be interpretable in the relevant waysSuppose that is all there isthe linguistic component consists of only the things that are required to enable interpretation by the conceptual and phonetic componentsif anything else is needed to account for linguistic phenomena, this will require extra explanation (and we are back to where we started)
17A Minimalist Demonstration Why do things move?In GB there were different reasons why things move:to satisfy the Case Filterto bind bound morphemessemantic reasonsBut if movement is part of the linguistic system it must have a reason motivated by the output conditions (conceptual and phonetic interpretation)
18Language shows a number of phenomena which involve semantically interpretable features coupled with similar features which are not interpretede.g. verb agreementfeatures are interpretable on the subject (person, number, gender)features a purely grammatical (uninterpretable) on the verbPresumably, uninterpretable features are a problem at the output: what would the conceptual and phonetic components do with them?
19The minimalist claim is that movement serves the purpose of ‘checking off’ uninterpretable features uninterpretable features are generated in some position (e.g. on the verb)the verb moves to be in a certain structural relation with the subject (specifier-head) where the features of one are checked against those of the other
20IP DP3.s I’ I VPIf the features check, they are deleted and therefore not present at interpretationIf the features do not check they remain and cause the structure to be uninterpretableCheckingV3.s
21Phrase Structure in the Minimalist Programme The output systems require a single structure to be formed from individual words for interpretation:how can a sentence be interpreted either semantically or phonetically (order?) if unconnected words are presentedSo it is necessary to have a structure building part of the grammar
22Structure building proceeds as follows: take two wordsput them together to form a new object (= ‘merge’)choose one to label the new object (= head)lovesMaryloves
23Do we need any other principles? The structure building process is a series of mergers which builds a structure step by stepDo we need any other principles?how do we know which words to merge?how do we know which one to select as head?No other principle is needed other than that the structure that is built must be interpretableif we merge two incompatible words, it will not be interpretableif we choose the wrong head, it will not be interpretableThe system then distinguishes grammatical (interpretable) from ungrammatical (uninterpretable)