1.9k likes | 1.91k Views
From GB to minimalism. Marcel den Dikken ELTE & RIL-HAS Budapest. Government & Binding Theory. Chomsky (1981), Lectures on Government and Binding , revolutionised generative- syntactic theorising, by directing the field away from a system of rules (typically
E N D
From GB to minimalism Marcel den Dikken ELTE & RIL-HAS Budapest
Government & Binding Theory Chomsky (1981), Lectures on Government and Binding, revolutionised generative- syntactic theorising, by directing the field away from a system of rules (typically language- and/or construction-specific) and (typically ad hoc) filterson rules and towards a system of universal principles andparameters capturing variation Marcel den Dikken @ EGG 2016
Government & Binding Theory though widely referred to at the time as ‘Government & Binding Theory’ (also by its inventor himself: see e.g. Chomsky 1982), Chomsky has come to prefer the designation ‘Principles & Parameters Theory’ (P&P): ‘government’ and ‘binding’ are important ingredients of the theory but they are not in any way their essence Marcel den Dikken @ EGG 2016
Government & Binding Theory Chomsky (1995:29–30): ‘The P&P approach is sometimes termed Government-Binding (GB) Theory. The terminology is misleading. True, early efforts to synthesize current thinking in these terms happened to concentrate on the theories of government and of binding ..., but these modules of language stand alongside many others: Case theory, θ-theory, and so on. It may turn out that the concept of government has a kind of unifying role, but there is nothing inherent to the approach that requires this.’ Marcel den Dikken @ EGG 2016
Government & Binding Theory the essence of GB/P&P theory: • principles • parameters • modules Marcel den Dikken @ EGG 2016
A Modular Theory P&P theory is modular in that (a)it features different levels of representation (D-structure, S-structure, LF, PF) (b) it features multiple subtheories (Case theory, θ-theory, trace theory, bounding theory, binding theory, etc.) Marcel den Dikken @ EGG 2016
A Modular Theory to determine whether a particular syntactic construct is grammatical or ungrammatical, the theory consults all the relevant modules that bear on the construct, and determines for each of them whether or not any of its principles and parameters is violated to any construct that violates something in one or more modules the grammar gives a * Marcel den Dikken @ EGG 2016
A Modular Theory a construct that violates multiple principles is expected to be more severely penalised than one that violates just one there has never been a principled theory about the possibility that a violation of a principle in module A could be more costly than a violation of a principle in module B Marcel den Dikken @ EGG 2016
Subjacency & ECP it has been standard practice, however, to treat a violation of the Subjacency Condition (the core principle of bounding theory) as less severe than a violation of, say, the Empty Category Principle (at the heart of trace theory) it is often said that a Subjacency violation ‘feels less bad’ than an ECP violation Marcel den Dikken @ EGG 2016
Subjacency & ECP Subjacency Condition movement must not traverse more than one bounding node (English: IP, NP) at a time Empty Category Principle a trace must be properly governed (a) head government, or (b) antecedent government Marcel den Dikken @ EGG 2016
Subjacency & ECP Subjacency Condition movement must not traverse more than one bounding node (English: IP, NP) at a time ugly because it counts (‘more than one’), and stipulates bounding nodes — and does so in different ways for different languages Marcel den Dikken @ EGG 2016
Subjacency & ECP Empty Category Principle a trace must be properly governed (a) head government, or (b) antecedent government ugly because it is disjunctive and because by their nature, ‘head government’ and ‘antecedent government’ are two very diverse notions (i.e., the ECP is a fake unification) Marcel den Dikken @ EGG 2016
Barriers — The Mission recognising that LGB’s approach to the licensing and bounding conditions on movement (‘Move α’) is far from optimal, Chomsky (1986, Barriers) seeks to both unify and simplify the theories of trace licensing (ECP) and bounding (Subjacency) and to eliminate the stipulativeness of the latter by replacing it with an algorithmic approach to the computation of barriers Marcel den Dikken @ EGG 2016
Barriers — The Mission Subjacency Condition movement must not traverse a barrier Empty Category Principle a trace must be antecedent governed NB1: Barriers uses ‘0-subjacency’ esp. in the analysis of parasitic gap constructions; but elsewhere it works perfectly fine, too NB2: Barriers keeps the LGB-style disjunctive ECP; but it can actually do without head gov’t Marcel den Dikken @ EGG 2016
Barriers — The Mission Subjacency Condition movement must not traverse a barrier Empty Category Principle a trace must be antecedent governed compare this to what we had in LGB… Marcel den Dikken @ EGG 2016
Flashback… Subjacency Condition movement must not traverse more than one bounding node (English: IP, NP) at a time Empty Category Principle a trace must be properly governed (a) head government, or (b) antecedent government Marcel den Dikken @ EGG 2016
Barriers — The Mission Subjacency Condition movement must not traverse a barrier Empty Category Principle a trace must be antecedent governed Q1: how do we identify a ‘barrier’? Q2: how do we define ‘government’? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics government αgovernsβ iff (i) and (ii) (i) αc-commandsβ (ii) no barrier intervenes between α and β Q2: how do we define ‘government’? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and βiffΔ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β (iii)Δ is the immediate projection of γ, a governor of β Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and βiffΔ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β (iii)Δ is the immediate projection of γ, a governor of β Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β Q3: how do we identify a ‘blocking category’? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics blocking category a projection Π is a blocking category for β iff Π dominates β and is notθ-governed by a lexical category (‘L-marked’) by clause (i), a blocking category for β is almost always a barrier for β as well — but an exception is made for IP Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) if IP were automatically a barrier for β by being a blocking category for β we would threaten to exclude all syntactic relations across an IP in a non-θ-governed position Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) since sentences of the following type are grammatical, a proper antecedent government relation must be establishable across IP [CPhow [C=did [IPshe [VPdo it t]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) in this example, we don’t need to worry about VP as a potential barrier: t is adjoined to VP, hence not dominated by VP, so VP ≠ BC for t [CPhow [C=did [IPshe [VPdo it t]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) and because VP ≠ BC for t, IP cannot inherit barrierhood for t from VP, via clause (ii), either [CPhow [C=did [IPshe [VPdo it t]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β → so we’re good for local adjunct extraction [CPhow [C=did [IPshe [VPdo it t]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β but now what about sentences such as: [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β how can t be properly governed??? [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β Q1: VP = BC and inherent barrier for t? YES! [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β Q2: does IP inherit barrierhood for t? YES! [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics as analysed, this sentence should be woeful … … but it’s perfectly acceptable … … so something isn’t quite right Chomsky’s solution: intermediate adjunction to VP [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics as analysed, this sentence should be woeful … … but it’s perfectly acceptable … … so something isn’t quite right Chomsky’s solution: intermediate adjunction to VP [CPwho [C=did [IPshe [VPt′ [VPfind[APt nice ]]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β now VP ≠ BC for either R(t,t′) or R(t′,who) [CPwho [C=did [IPshe [VPt′ [VPfind[APt nice ]]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β … hence VP ≠ barrier for R(t,t′) or R(t′,who) [CPwho [C=did [IPshe [VPt′ [VPfind[APt nice ]]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and β iff Δ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β and IP cannot inherit barrierhood for R(t′,who) [CPwho [C=did [IPshe [VPt′ [VPfind[APt nice ]]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics Chomsky doesn’t provide empirical arguments for intermediate adjunction to VP in Barriers there is a theoretical advantage: all legitimate traces are now antecedent governed Chomsky’s solution: intermediate adjunction to VP even for cases of mov’t of an internal argument (θ-governed by V), there is a local antecedent that can license the original trace for the ECP [CPwho [C=did [IPshe [VPt′ [VPkisst]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics Chomsky doesn’t provide empirical arguments for intermediate adjunction to VP in Barriers he also uses it in his analysis of parasitic gap constructions (truly the linchpin of Barriers) Chomsky’s solution: intermediate adjunction to VP it is interesting that in current minimalist syntax intermediate adjunction to vP again is the norm — and again is defended with parasitic gaps [CPwho [C=did [IPshe [VPt′ [VPkisst]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics Chomsky’s solution: intermediate adjunction to VP BUT there is what appears to be a simpler solution to the ‘VP as BC and inherent barrier’ problem [CPwho [C=did [IPshe [VPt′ [VPkisst]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics blocking category a projection Π is a blocking category for β iff Π dominates β and is notθ-governed by a lexical category (‘L-marked’) if we replace ‘θ-governed by a lexical category’ with ‘selected’ (Chomsky & Lasnik 1993), and recognise that I selects VP, a VP in the com-plement of I no longer qualifies as a BC Marcel den Dikken @ EGG 2016
Barriers — The Mechanics blocking category a projection Π is a blocking category for β iff Π dominates β and is notselected if we replace ‘θ-governed by a lexical category’ with ‘selected’ (Chomsky & Lasnik 1993), and recognise that I selects VP, a VP in the com-plement of I no longer qualifies as a BC Marcel den Dikken @ EGG 2016
Barriers — The Mechanics as it happens, Chomsky (1986:20) himself presents an argument for considering the relation between I and its VP-complement to be one of θ-government he points to the relative acceptability of the VP-fronting example below as an indication that I θ-marks its complement ?(?)fix the car, I wonder whether he will t Marcel den Dikken @ EGG 2016
Barriers — The Mechanics θ-government per se does not exempt the VP in the complement of I from being a BC given the definition of BC in Barriers — but it does if we adopt the definition in terms of selection θ-government of VP by I compels us to give up intermediate adjunction to VP if we accept Chomsky’s (1986:6) condition on adjunction: adjunction is possible only to a maximal projection that is a non-argument ?(?)fix the car, I wonder whether he will t Marcel den Dikken @ EGG 2016
Barriers — The Mechanics on the standard assumption that something that is θ-governed is an argument (more gen-erally, anything that receives a θ-role is an argument), the VP in the complement of I, which Chomsky says is θ-governed by I, is an argument, hence by the constraint below not a legitimate intermediate adjunction site adjunction is possible only to a maximal projection that is a non-argument ?(?)fix the car, I wonder whether he will t Marcel den Dikken @ EGG 2016
Barriers — The Mechanics Chomsky provides a plausible rationale for his condition on adjunction if this condition stands, and VP-fronting shows that the complement of I is an argument, then this also rules out the intermediate adjunction to vP popular in current minimalism adjunction is possible only to a maximal projection that is a non-argument ?(?)fix the car, I wonder whether he will t Marcel den Dikken @ EGG 2016
Barriers — The Mechanics back to Who did she find nice? … if “Chomsky’s solution” targets a non-problem as far as VP’s inherent barrierhood and IP’s barrierhood by inheritance are concerned … Chomsky’s solution: intermediate adjunction to VP … don’t we still have to reckon with clause (iii) of the definition of a barrier? [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics barrier a domain Δ is a barrier for a relation R between α and βiffΔ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β (iii)Δ is the immediate projection of γ, a governor of β Marcel den Dikken @ EGG 2016
Barriers — The Mechanics Q: does find govern the trace of who? → depends on one’s definition of government Chomsky’s (1986) definition of government is based on m-command — so nice governs its own specifier, and thereby blocks find from governing the trace of who so there shouldn’t be a minimality problem with the derivation below: VP ≠ a minimality barrier [CPwho [C=did [IPshe [VPfind[APt nice ]]]]]? Marcel den Dikken @ EGG 2016
Barriers — The Mechanics what does Chomsky use clause (iii) for? barrier a domain Δ is a barrier for a relation R between α and βiffΔ dominates β and excludes αand(i) or (ii) or (iii) (i) Δis a blocking category for β (Δ≠ IP) (ii)Δ is a maximal projection immediately dominating a blocking category for β (iii)Δ is the immediate projection of γ, a governor of β Marcel den Dikken @ EGG 2016
Barriers — The Mechanics what does Chomsky use clause (iii) for? → the that-trace effect the lexical complementiser that is supposed to set up a minimality barrier, by (iii), for proper antecedent government of the initial trace of whoby the intermediate trace in SpecCP but if finite I governs its own specifier (as Chomsky assumes it does), this can’t be right! [CPwho do you think [CPt′ [C=*that [IPtdid it ]]]]? Marcel den Dikken @ EGG 2016