Features 26 October 2015 We left with a system that included the following elements. (1) Head Movement onstraint IfanX 0 movestoay 0,therecanbenoZ 0 thatisc-commandedbyy 0 and c-commandsx 0. (2) Like(s) Attract(s) Like(s) IfX 0 moves,thenitmustadjointoanx 0.IfXPmoves,thenitmustadjoin toanxp,ormovetoaspecifierposition. (3) Upwards Ifαmovestopositionβ,thentheprojectionofβ ssistermustdominateα. (4) Argument Movement Move an argument XP to Specifier position. (5) Wh Movement Move a wh-phrase to Specifier position. (6) Head Movement MoveanX 0. (7) eement a. FiniteI 0 assignsnominativeasetoitsspecifierposition. b. GenitiveaseisassignedtoSpecifierof. c. Accusative ase is assigned to the position adjacent to, c-commanded by,andklosetoaverborpreposition. d. FiniteI 0 assignstensetoaverbitc-commandsandiskloseto. (8) α c-commands β iff: a. Everyphrasethatdominatesαalsodominates β,and b. α doesnotdominate β. NB: dominates is not reflexive. (9) hex 0 riterion AnX 0 cannotcontainmorethanonewordinitats-structure. (10) he ase Filter An argument must have been assigned ase by S-structure. (11) he EPP A Specifier of IP must have been occupied by something by S-structure. he structure of simple clauses is something like: (12) P XP AdvP not P Whatwe relookingforisawaytocontrolheadmovement.wewanttoforce ittoapplysothatauxiliaryverbsinenglishmovethroughto,andwewanta similar outcome for all verbs in French(and perhaps German?). For main verbs in English, we want to prevent Head Movement from occurring and use eement instead. And in French non-finite clauses we want auxiliary verbs to do the same thingdoinfiniteclauses,butwewanttopreventmainverbsfrommovingto 0. We need a system that uses the main verb/auxiliary verb contrast and the kind of inflection to determine what happens. P XP
Let s consider English. If Pollock is correct, we want the ent tense morphologyin 0 toinflectthefollowingverbbyeementwhenthatverbisn tan auxiliary.butwewantittoforceheadmovementoftheverbwhenitisanauxiliary verb. At ent, our system is designed for the situation with auxiliary verbs. We ve made tense morphemes an affix on verbs, and allowed that requirement to bemetonlywith 0 movesto 0.oincludenon-auxiliaryverbs,weneedtolet theaffixalnatureoftensebesatisfiedbyeement.hatwillbethefirststepin our revision. But first, let s set up the mechanics of eement. Let s let lexical items be made up of two kinds of things: roots and features. herootswillbethatpartofthewordthatmakestheworddistinct.featureswill encode that roots inflectional class. Finite verbs will be made up of the root and the features that cause it to exs tense(among other things). (13) =has =eats (16) P P P P P P have eat Wecanthinkofthefeaturesintheseverbsasawhollymorphologicalthing. It gives an instruction to the morpho-phonology about how the item it is part of should be exsed. We ll put a similarly named feature in 0 position. But this feature will have a denotation. It will contribute the information that makes a clause semantically tensed. You can think of these two features as reenting the two parts of what make a word: its morphology and its denotation. In these scenarios, those two parts are separated. When words have been broken up in this way, eement must hold between them. We ll call the feature that is just morphological unvalued, and the feature that has a denotation, valued. (14) An unvalued feature must be valued by eeing with a valued feature by S-structure. A valued feature must ee with a matching unvalued feature if it is not morphologically exsed. (15) α s with β ifαc-commands β andtheirfeaturesmatch. Wewillthinkof asarulethatmakesaderivation.sowenowhavederivations like(16). P like u u u P like P natto natto like u u natto I ve marked the unvalued features with u, and indicated when ve been valued with. I ve also now spelled out the features relevant for the 0 part of the morphology. hese are the person and number features whose ultimate source 2
isthesubject.we llcomebacktohowtheinformationgetsfromthesubjectonto 0 ;atent,we llfocusonhowthisinformationgetsfrom 0 totheverb. he system I ve designed here uses what are known as privative features. You ll see other feature systems in the literature. Another popular system is one that uses feature classes, along with particular values. So, for instance, instead of the reentation in (13), we have (17) (now adding the person and number information). (18) P PRO agr P PRO P agr P (17) =has =eats P P have tense: person: number: eat tense: person: number: inf AP inf AP On this system, unvalued features are reented by not giving the feature class itsvalue(e.g.,tense:)andisconceivedofaspassingthosevaluesfromone feature class to another. Yet another popular system allows for features to have valencies, so that we have things like - and +. A related system employs a feature geometry, that allows some features to be composed of other features. hese last two systems are designed to exs relationships among features. I don t anticipate running into phenomena that require recourse to any of these more sophisticated feature systems. A feature+ system now allows us to capture the situation with nonauxiliary verbs in English. It can also handle the situation with non-auxiliary verbs in French, although in this situation what the features are that belong to 0 is obscure. French infinitival verbs do not inflect for person or number. I ll simply use agr to reent this mystery feature. sembler uagr inf heureux sembler agr inf heureux Whatwedon tcaptureinthefrenchcaseisthattheverbmovesto 0.It snotperfectlyclearfromwhatwehavethatthismovementisrequired,butlet sassumeit is.previously,weforcedheadmovementbyputtingamorphemeinanx 0 positionwhichwetreatedasaboundaffix,andletthataffix sneedtobepartofaword only be satisfied if it shared its X 0 position with its stem. We ve broadened how morphology sneedtobespokencanbesatisfiednowwith.sothatwon t be strictly possible any longer. But perhaps we can allow for the same requirement nonetheless. Let s mark features that have that requirement with. Let s assume that all the morphemes that reside in 0 in French have. We ll adopt (19) and this will force the derivation in(18) to include(20). (19) ByS-structure,anX 0 thatcontains mustbewithinay 0 thatcontainsthe features it s with. 3
(20) P PRO agr sembler agr inf inf P P AP heureux he will be our way of forcing movement. How will we prevent it? he X 0 onstraint is of use only in cases where two words try to occupy one head position, and are not able to form a single word thereby. But that won t be of help when the heads involved can form a single word. For instance, why can t English finite clauses involve movement of non-auxiliary verbs. (21) P here should be nothing wrong with this. After all, if the verb were an auxiliary, itwouldbeabletoresultinthisstructure.sowhatblocksthis? homsky (1995) suggested that we might block this by invoking a very general condition on derivations that requires every operation that is invoked to be required.heideahereisveryclosetotheonethatliesbehindfaithfulnessconstraints in Optimality heoretic frameworks. Every departure from a D-structure must be warranted by some requirement that prevents that D-structure from beingagoods-structure.i mgoingtoformulatehomsky sideainawaythatdavid Pesetsky1 suggested. His name for the constraint is Earliness. (22) Earliness Let U be a fixed D-structure reentation and D ={D,D 1,...,D n } be all the grammatical derivations from U to some reentation that satisfiestheconditionsonwell-formeds-structures.every D i D isungrammaticalif D i isgreaterthansomeother D in D,where D isthe numberofelementsin D. his will prevent the derivation that leads to(21), since there is a shorter derivation that just involves ment operations that inflect the verb. HowcanweensurethatauxiliaryverbsinEnglishmovetofinite?Obviouslythereisnothingaboutthefeaturesthatlivein 0 inenglishthatrequires this, as main verbs don t move. It must be something about auxiliaries. I suggest, then,thatwelettherootsofauxiliaryverbscomewith.becausethefeaturesthat arebundledwiththisrootareinanrelationwiththefeaturesin 0, will have to be brought together with Head Movement. P like P natto 1 See Pesetsky(1989). 4
(23) P (25) P P P vp AdvP not P P eaten natto Now let s consider what happens when negation is ent. In these contexts, wewanttheauxiliaryverbtobeabletomoveinto,butblockaltogether between and every other verb. hat s not easy to do with the tools available to us, and indeed, there is not a method that works without stipulation. I ll propose something here that is a combination of several ideas in the literature. I don t believe there is a standard solution to the problem, so this is a parochial analysis. First, let s try to prevent between and the verb when is ent. One way this can be achieved is by building the Head Movement onstraint into. (24) αswithβifαc-commandsβ,theirfeaturesmatchandthereisno γ 0 thatc-commands β andisc-commandedbyα. hiswillpreventfrombeingabletovaluethepersonandnumberfeatures across negation. like u u natto If failsinthisway,thenwecanblockmovementoftheverbbymakingit contingent on. Let s adopt(26). (26) IfX 0 movestoy 0,then mustholdbetweenthem. It will also, wrongly, prevent from holding between 0 and 0 even when 0 isn tent.hatis,itwouldblockthethirdstepinthederivationin (16).hat sbecause 0 willbeintheway.onewayofsolvingthatproblemistolet 0 alsohavepersonandnumberfeaturesthatgetvaluedby 0 andwhichthen with 0. Okay,butnowhowarewetoallowauxiliaryverbstomoveto 0?hiswon t be pretty. Suppose that English auxiliaries have two versions. One with and another without negation built into them. 5
(27) =haven t =have =wasn t =was (29) P uneg u ual uent u ual uent be uneg u upast be u upast Only auxiliary verbs (and modals) have these negative forms. Note that the neg feature contain is unvalued. hat is, it is a morphological instruction not a semanticone.heplacethatnegationshouldbesemanticallyinterpretedisin 0. So here s what a D-structure will look like with a negated auxiliary verb. Jill sing AdvP not neg P (28) P P Jill sing AdvP not P uneg u vp eaten natto neg P can now hold between 0 and the higher of the two 0. hat allows the neg feature in 0 to value the neg feature in 0, and that licenses Head Movement to combine them. (Again, the requires that Head Movement combinethem.) uneg u u vp eaten natto can hold between 0 and 0, and this allows Head Movement to bring them together.(moreover, because of the, must combine.) 6
(30) P (31) P Jill Jill AdvP sing not P neg sing AdvP not P neg u neg P vp eaten natto neg sing P vp eaten natto Andfinally,canholdbetween 0 andthehigher 0,andthislicenses movementof 0 to 0. Finally, let s consider auxiliary verbs in French. In finite clauses, these behave justlikeanyotherverbinbeingforcedtomovethrough 0 to 0.Wecancapture this by letting the features in 0 (and perhaps also 0 ) have in French. Putting on the features in these positions will drive any verb in an relation up to them, and this is what we find. And recall that this is possible even whenisentbecausetheheadofmovesoutoftheway. 7
(32) P (33) P PRO Jean AdvP AdvP ne agr pas P ne agr pas P P P 0 aime u u Marie Fromthereentationin(32),itispossible for 0 to with 0,andthis willallowtheverbtoheadmoveto 0.Oncein 0,itcan with 0,and thisallowsmovementof 0 to 0. But consider now the situation with infinitival clauses. Here we find that movementoftheauxiliaryverbcanhappento 0 butthatitneednot.atent we would expect S-structure parses for infinitival clauses with auxiliary verbs to look like(33). aver inf uagr inf I don t know what the features in 0 are, so I ve simply put agr in there. If we assumethatarenecessarytodeterminewhattheformofthe 0 is,thenwe canletanunvaluedfeatureonaver with 0.hatshouldbepossiblein (33). And we know that 0 doesn t have, because otherwise all verbs would obligatorilymoveupthere.wealsoknowthatauxiliaryverbsdon thave,asthat wouldrequireeveryauxiliaryverbtomoveto 0,and(33)isagrammaticalSstructure. So the problem is, then, how does Earliest allow 0 to move to 0 when it isn t required? his seems to be a genuine counterexample to a system that has something likeearliest.idon tknowwhattodoaboutit.earliesthassomeutilitybeyondthe cases we re looking at now, so we shouldn t jettison it. Pollock mentions that the two word orders belong to different registers. Perhaps we could speculate that registers constitute different grammars, and that speakers of French command both. Ifweimaginethatoneoftheregistershas onauxiliaryverbs,thenthiswillforce movementfrom 0 to 0 inthesecontexts. You can probably see that this system has some similarities to the complementizer system. We saw that English has two complementizers that force movement ofmaterialtothem.woofthosecomplementizersforcemovementof 0 to 0 P 8
when occur. hese complementizers show up in root 0 positions only. o bring these cases into our system we will have to assume that the complementizers inthesepositionshaveafeaturethatvalueon 0,andthatcarry.I don tknowwhatthisfeatureis,soi lljustcallitq. (34) P Q P Shirley should uq P eat natto We can recast the generalization we discovered about English complementizers earlier as: (35) A complementizer that heads a root clause must have. he other question complementizer comes with a feature that forces whphrases to move to its Specifier. Let s call this feature wh. And let s assume that anunvaluedwhisfoundonthosesthatarewh-phrases.soascenarioinvolving this complementizer might look like(36). (36) P Q wh Smith P P likes P uwh who As it s ently formulated, won t allow the wh feature on the object to be valued. hat s because ently builds in the Head Movement onstraint itdoesn tallow toskipahead. (24) αswithβifαc-commandsβ,theirfeaturesmatchandthereisno γ 0 thatc-commands β andisc-commandedbyα. Wedidthisaspartofourexplanationforwhy 0 blocks from 0 to 0.Inthisscenario,though,weneedtolet crossseveralheads.ifwecan release from the Head Movement onstraint, then from(36), we ll have a derivation like that in(37). 9
(37) P P (39) P Q wh P Smith P wh who Q wh Smith P P Q wh uwh who P P P P P likes wh who likes likes uwh who (ImaginethatthePwearelookingatisembedded,sothatthereisno onq,and consequentlynomovementof 0 to 0.)heresultofmovementheredoesn t satisfyourconditionon,whichisdesignedjustforthecaseofheadmovement. (19) ByS-structure,anX 0 thatcontains mustbewithinay 0 thatcontainsthe features it s with. Solet srewritethissothatitcanapplytothesecasesaswell. (38) LetF be a bulleted feature and Y 0 be the head position containingf. eementmustpairf withsomethingthatisadaughterofanxprojectionofy 0. But can we rewrite so that the Head Movement onstraint is bled from it? If we simply strike the clause that exses the Head Movement onstraint, we ll have to find another account for why blocks. Interestingly, there is evidence that is subject to a locality constraint in these contexts of the questions. his locality condition shows up when there are two swiththewhfeature,asin(39). he S-structure that emerges from(39) is(40a), and not(40b). (40) Salasked... a. who likes who. b. *whowholikes. hecontrastin(40)illustrateswhatisknownas Superiority. (40b)issaidtobe a Superiority violation. Roughly speaking, the generalization about Superiority isthatiftherearetwowhphrases,αandβ,andαc-commandsβ,thenqcanonly with α. If we look at Superiority and the Head Movement onstraint together, then onewayofstatingonerestrictionon thatcapturesthembothis(41). (41) α s with β only if α c-commands β, α and β have matching features,andthereisnoγthatc-commandsβwhichαcouldhaveed with. 10
. References homsky, Noam. 1995. he minimalist program. ambridge, Massachusetts: MI Press. Pesetsky, David. 1989. he earliness principle. Paper ented at GLOW. 11