Discourse configurational languages 26, 79-123
By admin
It has been argued that functional categories other than C, I, and D exist. The Infl skeleton has been “split” into categories that include several Agr nodes, Tense, and Aspect; likewise, the Det skeleton has been argued to present various Agr nodes, Number, or a Possession node. Splitting the Comp skeleton has a long tradition. Reinhart’s (1979) analysis was based on the premise of two such nodes in certain languages, the origin of what is now commonly referred to as “CP recursion”. And again several nodes have been argued to correspond to the Comp skeleton, a matter I want to discuss here.
But first I want to lay out my ground-rules. One, this is a research program. Surely there are alternatives in terms of features in single categories, instead of positing multiple categories each corresponding to a feature. I will not explore this alternative view here.
* Early versions of this paper were presented at the New Jersey Circle in Princeton and GLOW in Lisbon. I appreciate comments from Josep Maria Fontana, Charlotte Galves, Elena Herburger, Norbert Hornstein, Katalin Kiss, David Lightfoot, Ana Maria Martins, Jairo Nunes, Eduardo Raposo, Dieter Wanner, and my students and colleagues at the University of Maryland.
Two, both a formal and a substantive issue arise the minute one posits a new category. The formal one is proving the existence of the category in terms of some empirical evidence. The substantive one is determining what that category is, how it differs from others, etc. The second issue is too ambitious at this stage, and I will only have some conjectures to offer about the nature of the category I am positing. The first issue itself is non-trivial. It has to be determined whether the hypothesized category is a single category or an array of categories, is a head or a maximal projection, is higher or lower than other categories within the Phrase Marker. The main purpose of this paper is to deal with these sorts of questions for a given category I hypothesize.
Three, evidence adduced for a given proposal is relative to a framework, and it may count only within that framework. I assume the essentials of Chomsky’s (1993) Minimalist program. This means, to start with, that I cannot assume levels of representation other than PF and LF, operations that are not economical, parameters that are not trivial, etc.
The third of my ground-rules has consequences that should be clarified, particularly given the context where this article is appearing. At issue is whether there should be a level of Information Structure (IS), as posited by Valduvi (this volume). I will not assume such a level, and my reasons are of two sorts.
A separate level of IS is supposed to encode representations that have no consequence for the truth conditions of a sentence. The assumption behind this is that LF is a level that is truth-theoretical in nature. I disagree with both the proposal and the assumption.
In the minimalist framework as well as on the works leading to its definition of LF, it is not the case that LF has anything to do with truth, or for that matter notions that enter into the calculations of truth, such as arguments, predicates, operators, variables, and the like. The primitives of LF are syntactic, and include chains of different sorts, the conditions of their formation, operations of ellipsis, and little else. LF interfaces with semantic levels (call those Intentional/Conceptual structure, ICS) where perhaps a truth-functional apparatus has its place, as well as other non-syntactic devices. In other words, it is perfectly legitimate to map LFs to structures encoding matters of focus, background, and so on, without a separate level of IS.
In fact, we could not map such discourse matters otherwise in the minimalist framework. The system is designed around levels of competence that, in Chomsky’s words, follow from “virtual conceptual necessity”. It is far from clear that conceptual necessity forces the postulation of a level that is not even necessarily a competence level, but seems more like an interface with the performance systems that Chomsky places outside of the syntax proper (i.e., the optimal mapping of PF and LF).
Within the minimalist framework, the notion of discourse configurationality is coherent only as a metaphor, since discourse is in essence a matter of performance. Of course, performance-dependent does not mean vague, idiosyncratic, or senseless. For instance, for the minimalist program semantics is performative, and few domains have been shown to be so precise, rule-governed, and full of sense as semantics. However, this systematicity does not have to be expressed configurationally, and very possibly is not expressed this way. This is all to say that if any effects of discourse in fact play a role in the configurationality of a sentence, these must be expressed by way of a category that is recognizable at LF, and is born in the syntax proper. Apart from not being able to accept a separate level of IS, I also have difficulties with the assumption that it expresses matters which are not truth-functional. As is well-known from the literature on focus, a sub-class of sentences exhibiting this phenomenon have truth-theoretic consequences (see the Introduction to this volume). In turn, it might seem as if standard topicalizations have no truth-theoretic consequence, but this may not be true–at least for non-contrastive topics that create their own categorical judgement (see Raposo and Uriagereka (1993)). This is all to say that core phenomena within what we may call Information Theory are as truth-functional as others.
The point just made does not on itself mean that information theoretic phenomena must be represented at LF, anymore than any other phenomenon that has effects on the truth conditions of the sentence must. But even if the minimalist view of LF were given up in favor of a level of logical form in the traditional philosophical sense, matters of Information Theory would still have to be expressed in that would-be level. That is, information theoretic matters do not carry enough conceptual weight for us to posit a separate level of representation, and either LF or logical form seem perfectly fit to do the job of representing them.
However, I do accept the empirical fact that discourse affects configurations. Hence, as a consequence of the ground-rules I set up for myself, I am virtually forced to posit a syntactic category that encodes information-theoretic issues. I promised a conjecture in this respect, and with it comes a disclaimer. I have not found any conclusive evidence that there are separate functional categories to express matters of topic, focus, emphasis, contrast, etc. All of these have an aspect in common: they encode the point of view of a speaker or some other subject, in a manner to be clarified immediately. I therefore assume that one category alone serves as an all-purpose device to encode a point of view.
I will call the category in question “F”. Uriagereka (1988) borrows the insight behind this category from Chomsky (1977). I explicitly took F to merely stand for “functional”, and in other places I used F as a mnemonic for a “further” projection, or “focus” with the vague import of emphasis, contrast, information-encoding device, etc. All I mean is this: F encodes point of view. The claim is that all information theoretic operations need to be mediated through a point of view. That is, when emphasis appears in a sentence, someone is responsible for that emphasis. Old or new information is old or new for someone. Even the usage of a referring expression presupposes a speaker who assumes responsibility (mistakenly or otherwise) for calling someone Smith’s murderer or Jones, and similar issues arise for deixis, anaphora, etc.
The general view is this: a syntactic theory formally expressing dependencies goes as far as establishing valid chains and nothing else. These chains have indexation mechanisms, but the indices carry no semantic value. Value assignment is a matter that is dependent on a discourse, which is to say the point of view of a speaker or some other subject. This is what pragmatics is about. My proposal is that the syntax has one designated node that enters into the determination of what are valid value assignment for formal indices (i.e. relations) that the syntax creates. There are syntactic rules determining possible points of view in a sentence, and only a sub-class of the logically possible points of view that a pragmatic theory allows satisfy the syntactic derivation.
Take the sentence John’s mother believes that Mary likes NOTHING WHATSOEVER. It has an interpretation where the emphasis on the object of likes has a characteristic illocutionary force. Interestingly, the force in question may be related to either the speaker of the sentence (say, I), or the subject, John’s mother–but not to John. That is, I may be reporting a belief that John’s mother has: “Mary likes NOTHING WHATSOEVER!”. But John’s mother might also have an array of beliefs: “Mary doesn’t like peanuts; she doesn’t like candy, either; in fact, Mary doesn’t even like cookies or ice-cream.” Then I am entitled to conclude that John’s mother believes Mary likes NOTHING WHATSOEVER! John’s mother never had that belief, but had enough beliefs about Mary for me to conclude what I did. Even if we recreate the scenario for John instead of his mother, the sentence in point cannot express John’s illocutionary force–only the speaker’s and the mother’s.
This difference in emphasis can be captured in different ways. My proposal is to do it through a syntactic device: the emphasized element moves to the spec of an F category, which is to be interpreted in the performative levels as expressing the point of view of whichever subject is relevant (the speaker or an embedded subject). Covert or overt rules of movement allow certain chains involving F, and not other chains. Ideally, this alone will predict the class of derivations that feed the ICS level(s) for an appropriate point of view to be expressed.
Perhaps it is worth emphasizing that movement of an element to the spec of F does not translate into emphasizing that element at LF. The movement itself is a (covert or overt) operation to satisfy some LF requirement of the sort discussed by Chomsky (1992). However, only elements that are appropriately placed in F will be visible for the ICS level(s) to be interpreted as emphasized by whichever relevant subject is responsible for the emphasis. To illustrate this, consider John thinks that NOTHING does Mary like and NOTHING does John think that Mary likes. Crucially, in the first of these two sentences it is John who is responsible for the emphasis, while in the second instance it is I who is responsible for it. A way to express this is to freeze the displaced NOTHING in the place where it has been moved–say the specifier of F. While in the sentences seen before, at LF the object of likes could move to the lower [spec, FP] or the higher one, in this instance the movement has taken place overtly, and minimalist considerations would make it stay where we see it. This is presumably mapped into ICS in a trivial manner, with the element in the spec of FP expressing the point of view of its closest c-commanding subject (John if NOTHING is in the lower spec, myself if NOTHING is in the higher spec, assuming some version of John Ross’s performative analysis.) The syntax has decided the possible ICS interpretations, even if the interpretation proper happens at that level.
Examples of this sort can be created with focalization, topicalization, the mode of presentation of referential expressions (see Uriagereka (forthcoming), Raposo and Uriagereka (1993)). In the present context this means two things. First, when I argue that an element has raised to F this as such has no interpretation; it is a formal mechanism that will have various consequences, depending on the nature of the moved element (whether it is an operator, an argument, a predicate, etc.; see Uriagereka (forthcoming b)). Second, the last few paragraphs are intended as a conjectural motivation on the substantive issue behind F.
1. Or alternatively, an empty operator moves in the overt syntax, as proposed by some for Wh-movement, topicalization, tough and parasitic constructions, among others.
Finally, the minimalist program imposes a view on parametric variation. Within the Romance languages explored here it is reasonable to assume that while overt matters associated to F were quite productive in the Low Middle Ages and still are in contemporary dialects that I will call “archaic”, these matters are residual or non-existent in other dialects. More generally, examples presented in this volume show that variation exists with respect to discourse configurationality.
The minimalist framework captures intralinguistic differences through trivial morphological specifications. Assuming a hypothesized category F, we have three logical possibilities. First, F has “strong” features that must be checked by PF, or F does not have them. If it does, some head X must raise to F to check these features, which given the strategy of Greed (an element moves only to satisfy its own requirements) entails that the head that raises to F must itself contain F features that are matched in association to the F projection. It is then logically possible that the head X that raises to F may or may not bring strong features from the lexicon that are visible at PF. The second possibility is incompatible with a strong F: an X head with weak features will not be able to rise to F without violating Greed, and then the derivation will crash at PF, for F remains visible at that level in violation of the principle of Full Interpretation. However, an X head with weak features is compatible with a weak F. Finally, an X head with strong features is compatible with a weak F. In this instance, association of X to F is not in order to meet the requirements of both F and X, but only those of X.
The proposal yields three logically possible languages, which I will suggest are paradigmatically represented by Galician-Portuguese, French, and Spanish–within Romance. But quite aside of the empirical adequacy of this state of affairs, the theory forces us in this direction. There is little else we can do within the system, which extends in principle to parametric differences pointed out throughout this volume.
One of the fundamental differences among Romance languages lies in where pronominal clitics are placed in each language. For instance, within Western Romance, the patterns in (1) occur in tensed clauses:
(1) Patterns of Western Romance
A: <cl,V>
French, (spoken) Brazilian Portuguese(?)
B: <cl,... V >
[+tense]
Castillian Spanish, Aragonese, Catalan, ...
C: <0, V ,cl>; <Governor, cl,... V >
[+tns] [+tns]
Portuguese, Galician, Leonese, most "archaic" dialects
I will not explain these patterns (see Uriagereka (forthcoming) for an analysis and various references). For now, let me point out a few other, apparently unrelated properties associated to the third pattern:
(2) Properties apparently associated to Pattern C:
(i) Overt F(ocus) elements.
(ii) Overt Focus movement.
(iii) "Recomplementation".
(iv) "Sandwiched" Dislocations.
(v) Overt expletives.
(vi) Personal infinitives without Aux-to-Comp.
(vii) Interpolation of elements between clitics and V.
Examples of (2i) can be seen in the er element appearing in medieval Galician texts, as described by Mattos e Silva (1989):
(3)
a. Quando chegou... e vio que non podia passar,...
when arrived.III and saw.III that not could.III pass
na segunda vegada er meteu o manto en aquela agua
in-the second time ER inserted.III the robe in that water
'When he arrived and saw that he could not pass, the second
time he did insert the robe in that water'
b. E esto podes provar que Deus ouvio a ta oracon
and this can.II prove that God heard.III the your prayer
se me achares aqui depois que aqui er veeres
if me found.II here after that here ER came.II
'And you can prove that God heard your prayer if you found me
here after here you did come'
Traditional grammarians noted that this form “expresses contraposition”, “provides a pleonastic reinforcement”, or “reinforces the sense of the verb”. These linguists equate er to even, only, and similar adverbials, which obviously makes this element look like a focus marker. Similar “extra” elements can be seen in several contemporary dialects:2
2. For instance, Latin American Spanish dialects reported by Kany (1963):
(i)
a. como uno es sufrido, nadita que se queja [Chile]
since one is thick-skinned nothing that self complains.III.sg
'Since one is very thick skinned, one doesn't complain AT ALL'
b. entonces siempre te casas [Bolivia]
then always self marry.II.sg
'So you ARE marrying?'
c. eso jay me tocaba decir a mi [Bolivia]
that JAY me corresponded.III.sg say to me
'THAT was for me to say'
d. que no mas has traido [Peru, Ecuador]
what not more have.II.sg brought
'What THEN have you brought?'
(4) Xa eu sabia que el ias caer na poza!
Already I knew.I that it were.II fall in-the puddle
'I knew that you were going to fall in the puddle'
This Galician xa (lit: ‘already’) has nothing to do with completion of an event or anything associated to a time past; all it does is add illocutionary force to a more neutral statement.
Whether this sort of element is adverbial or integrated into the skeleton of the sentence is a harder question. Kiparsky (1986) argues that such elements as Tense are grammaticalizations of adverbial elements, and the same is arguably true of other functional categories. But regardless of whether this is generally the case, it does not answer the real question: of the two possibilities that UG grants for focusing elements–adverbials or functional heads–which one is at issue here? Property (iib) suggests that indeed a head is at stake. As is well-known, many Romance languages show the Focus movement in (5):
(5) moitas cousas lle eu dixera! (Galician)
many things dat I said.had.I
'I had said MANY THINGS to him/her'
The question is where the focused moitas cousas ‘many things’ is moving. Of course, this could be an adjunction–but it could also be an instance of movement to a spec. If the latter is the case, the obvious question is what this spec is, and having an F head would then clearly be handy. Notice, incidentally, that if mere adjunction is at issue, we probably should expect the possibility of several focus movements (which does not obtain), and that the sort of movement in (5) be fairly general across languages–which is not. Therefore, adjunction here is highly suspect.
Note also the position of the clitic in (5). This is entirely impossible in most modern Romance dialects, but it is quite systematic in archaic forms–this is property (2vii). Being deliberately naive about the clitic in (5), we can explain its position if there is an extra host for the element in the relevant dialects. A question I return to in section 4 is whether this could be F.
Another archaic trait of languages is productive in some Romance languages is recomplementation, which as Higgins (1988) notes is associated to dislocations we may refer to as “sandwiched”:
(6)
a. <dixeron que> a este home <que> non o maltratemos
said.III.pl that to this man that not him badly.treat.I.pl
'they said that this man that we should not treat him badly'
b. <Din que> o que e ver <que> non veu res
say.III.pl that the what is see that not saw.III nothing
'they say that what you may call seeing that he did not see
anything'
c. <Din que> se atopan a saida <que> han marchar
say.III.pl that if find.III the exit that will.III go
'they say that if they find the exit that they shall go'
In embedded clauses, Left Dislocated and peripheral phrases can be followed by an overt que; of course, they are also preceded by a complementizer. These are properties (2iii) and (2iv). Facts of this sort follow if the highest que is “old-style” C, and the lowest, F. Here I depart from analyses of “CP recursion”, as in Iatridou and Kroch (1993)–but only nominally. That is, I accept their insights, but I would like to suggest that the second element in the recursion is not a recursion of C at all, but a separate F head with properties different from those of C.
We may ask, though, how we know that it is the highest element that is C, and not vice-versa. Consider (7):
(7) (Din que) muitas cousas (*que) veu
say.III.pl that many things that saw.III
'They say that many things did he see'
Focused phrases are incompatible with overt que, unlike Left Dislocated phrases. Although que can occupy the head of an FP, in these languages it does not have the appropriate features to agree with a focused phrase in its spec–much like that is not compatible with Wh-phrases.3 This account holds only if it is the second que that is relevant for Focus, since the upper one is fine in these instances.
3. A Left dislocated phrases is possible, though. This indicates either that no agreement requirements hold for Left-dislocation, or else (less likely) that a further site is at issue in those instances. I will not pursue this matter here, but see Duarte (1987) for a presentation of relevant presentential sites--which include also Hanging Topics ("As For" constructions), which I suspect do involve a further site.
Property (2v) is rather powerful in archaic Romance, and one would also hope that it is related to the patterns above.4 The distribution of true overt pleonastics appears to be limited to sentences of a certain sort: mostly, those without a logical subject, or where this element does not need overt morphological Case (Raposo and Uriagereka (1990)):
4. In (i) we have several Spanish examples with overt expletives. According to Henriquez Urena (1939), these were common in several dialects up to the nineteenth century:
(i) a. Ello es facil llegar [Santo Domingo, Mexico,
it is easy arrive Colombia, and Caribbean
'it is easy to arrive' dialects in general]
b. Ello hay dulce de ajonjoli? it has sweet of sesame-seed
'Is there any sesame cake?'
c. Ello dice que no es muy buena
It says that not is very good
'it is reported that it is not very good'
(8)
a. El chove (Galician)
it rains
b. Isto sao dez horas ja (Portuguese)
this are ten hours already
c. el e sabido que... (Galician)
it is known that
d. El era unha lanterna de papel (Galician)
it(masc.) was a.fem lantern of paper
e. elle ha marotos muito grandes na tropa (Portuguese)
it has rascals very great in-the troops
f. el sabedes cando chegaran (Galician)
it know.II.pl when arrive.will.III.pl
'Do you know when they'll arrive?'
These examples contrast with others where the overt expletive is associated to a definite specific NP, a matter I return to in section 4. Overt expletives are also incompatible with focused phrases (9a), and in some dialects are used to express clausal emphasis (9b):
(9) a. *moitos convidados el chegaron (Galician)
many guests it arrived.III.pl
b. ello que yo lo vi (Old Spanish)
it that I it saw
'Of course I saw it!'
(9b) suggests that the overt expletive may occupy the spec of F, perhaps rising from the spec of IP.5 In section 4 I return to what may prevent phrases from being focalized to F over the expletive (9a).
Another known property of Western Iberian dialects is that they present inflected infinitivals. This would seem peculiar to these dialects6, but following Maurer (1978) we may need to extend our horizons. There are two separate issues here: whether infinitivals are inflected and whether they are personal. Maurer shows that personal infinitivals extended across medieval Romania, and got progressively lost in modern dialects except for those I am calling archaic–just like overt pleonastics were. Thus:
5. This is reminiscent of a sort of expletive often seen in Germanic, which has occasionally been argued to be in Comp. The natural step to take would be to try and unify both in terms of a common position, which at least for Romance would have to be after Comp:
(i) Dizque el habia moita xente na rua
it.is.said.that it has much people in-the street
'It is said that there was a lot of people in the street'
For Yiddish, Danish, and related Germanic languages it would be fairly natural to propose a post-comp site, particularly given that V-2 is possible in embedded causes even when Comp is filled, and given other "CP recursion" facts discussed, for instance, in Iatridou and Kroch (1993).
6. In particular, Galician, Portuguese, and the Mirandese variant of Leonese, within Iberian, plus Sardinian, and 17th century Neapolitan.
(10)
a. Senza vederla egli, passo ... (Medieval Italian)
without see-her he went.III
'Without he seeing her, she went...'
b. Por le vilain crever d'envie, chanterai... (Old French)
For the villan die of-envy sing.will.I
'For the villain to die with envy, I will sing...'
c. Mas por vos mandarlo, lo hare... (Rhetorical Spanish)
But for you order-it it do.will.I
'But because of you ordering it, I will do it'
Maurer argues that this construction is the basis of the inflected infinitivals. There are two important properties that inflected infinitivals exhibit. The infinitival form agrees with the subject; and subjects are possible here without “Aux (or more generally V) to Comp”. The traditional term “personal infinitive” makes reference to the latter possibility, namely preverbal subjects in infinitivals. Now, the examples in (10) do not seem to involve verb raising, and in that sense are standard personal infinitives. As for the matter of overt inflectional forms, Maurer notes correctly that overt agreement surfaces even in non-infinitival clauses. Take for instance (11a) or (11b):
(11)
a. Nu sei cumo tanto devamos, ganhandomos tanto dinheiro
not know.I how so.much owe.I.pl earning.I.pl so-much money
'I don't know how we owe so much, earning so much'
(Portuguese, Ervedosa do Douro)
b. la mujer esta muriendase
the woman is dying.fem.self
(Puertorican Spanish, Utuado, Lares)
c. No por querersen mucho...
not for love.eachother.pl much
(Sub-standard Spanish, turn of century)
(11a) is common in areas of Northern Portugal and Galicia. Here, agreement shows with a gerundial form. Similar facts have been reported (e.g. by Zamora Vicente (1966)) within Puerto Rico, as in (11b). Likewise, in 17th century Napolitan inflected non-finite forms extend to gerundials. This is all to say that there is nothing peculiar to the fact that the overt agreement surfaces with infinitivals. What is important is that it surfaces where it can, that is, where there is an “argument” and a “predicate” to agree and a syntactic category encoding the agreement overtly or covertly. This is why it is crucial that personal forms show up in a given dialect: if they do, overt, spelled-out agreement may follow. Note also that although the standard claim is that inflected infinitivals are not seen in mainland Iberian dialects, Borao (1908) records forms like (11c). In sum, personal infinitives, with or without a strong spell out of agreement features, are another trait of medieval Romance that disappeared in most dialects, though not in all.
So far, I have noted that dialects with personal infinitivals and dialects with strong F coincide. Synchronic arguments can also be given for the correlation. As Hernanz (1982) notes for Spanish, personal infinitives typically involve emphasis or contrastive focus of the overt subject (12a). Also, note the Galician (12b):
(12) a. al ella tener que pagar, y no los otros, Maria grito
on she have to pay and not the others Maria shouted
'Since she had to pay, and not the others, Mary shouted'
b. Nos queremos (*moitas cousas) comer(*mos)
We want many things eat.I.pl
In those contexts where personal infinitives are impossible, so are Focused phrases: for instance, volitional embedded clauses prevent both. Furthermore, as Raposo (1987) notes, in epistemic contexts the inflected infinitival must undergo verb movement, which in his theory follows from the need that the infinitival form has to receive Case:
(13)
a. *Xa me extranaba teus irmans non viren a festa
emph. me surprised your brothers not come.III.pl to-the party
b. Xa me extranaba non viren teus irmans a festa
emph. me surprised not come.III.pl your brothers to-the party
'Your brothers not coming to the party had surprised me indeed'
But Ambar (1989) notes that if the subject between the epistemic verb and the embedded verb is focalized, verb movement is unnecessary:
14)
extranabame so os teus irmans non viren a festa
surprised-me only your brothers not come.III.pl to-the party
'Only YOUR BROTHERS not coming to the party had surprised me'
Given facts of this sort, we have to come back to an explicit theoretical vehicle for the observed correlations. What should be clear, though, is that it is desirable to find a unified account to all these phenomena, since they appear (and disappear) simultaneously in a class of dialects.
Consider again the clitic patterns in (1), in particular for tensed clauses. For reasons that do not concern me now, I shall assume that in the unmarked instance clitics are hosted in a position outside VP. From this view point, pattern C is a nuisance. Traditional linguists have a line for it; in the words of Dieter Wanner (1988): “In [the Mussafia-Sorrento-Ulleland] approach, proclisis is seen as basic, . . ., so that any enclitic occurrence needs a specific reason for its existence.” In fact, the derivation has always been clear. Again, in Wanner’s words:
"[Within Vernacular Latin,] one of the reasons for the
increase in frequency [of the attachement of clitics to
verbs] is the leftward appearance of the verb in certain
emphatic clause types characteristic of spontaneous speech:
imperatives, quotatives, presentatives, verb focus, and
emotively marked utterances all have clause initial verbs.
The pronoun does not change its place actively in such
situations, but it remains stable in the lefthand portion of
the clause, preferably in second position."
In the examples that concern us here, if the verb is able to move in the relevant sites past the clitics, then we get the right syntax. With it, though, a few questions arise: why has the verb moved, where has it moved to, why has it not moved in the other instances? And so on.
Uriagereka (forthcoming) reports on various “Verb Second” approaches to these questions. But aside from the fact that the present paradigm does not align with V-2,7 a minimalist account cannot allow for “free” V2, prevented in some instances if a more economic derivation exists. V-2 accounts typically state that for unclear reasons the verb must raise past the clitics in matrix clauses, and then argue that in embedded clauses the raising is not necessary and therefore is impossible. The problem, though, is the first claim. In the minimalist program if the verb moves it must be because it itself has a reason to move, otherwise it would violate the Greed strategy. Uriagereka (1988) proposes that the verb moves to F for the same reason that it moves to I in languages with a strong I. In current terms, we must instantiate this idea by assuming that the verb has strong I features that are checked against the features in I, thereby eliminating I at PF in satisfaction of Full Interpretation. Following the same reasoning, I suggest that the archaic Romance verb has a strong F feature in its morphology, which is checked against the strong F position I am hypothesizing. I take this feature to be what shows up as the inflection in infinitivals. Upon raising V to F, F is made invisible for the PF level, and the derivation does not crash.
7. Germanic and Romance do not pattern alike in standard V-2 (ia/b):
(i) a. recht gut/wie fand ich sie (German)
very good/how find I her
cf. *moi boa/como encontroa eu (Galician)
very good/how find.I-her I
b. *rech gut/wie ich sie fand (German)
very good/how I her find
cf. moi boa/como a eu encontro (Galician)
very good/how her I find-I
c. (Eles) dironlle un libro a Maria
they gave.III.pl-dat a book to Maria
d. eu coido que a min levaronme todos algun garrido
I think.I that to me took.III.pl-me all some present
'I think that they all carried a present for me'
As in Celtic languages, in archaic Romance we do not need any topic in initial position for the verb to be first (ic). As in Greek, the verb movement in question is also possible in embedded clauses when the complementizer is present (id). Also, consider (ii):
(ii) a. (?*moitas/algunhas veces) a quen foi ver Xan
many some times to whom went.III see Xan
'(Many/some times) who did John go see'
b. Pedro estaba pensando (*moitas/algunhas veces)
Pedro was.III thinking many some times
que Xan non tuvera razon
that Xan not had.III reason
'Pedro was thinking that (many/some times) Xan was wrong'
a'. (moitas/algunhas veces) visiteu-no eu
many some times visited.I-him I
'Many times I visited him'
b'. Pedro estaba pensando que
Pedro was.III thinking that
My suggestion takes Wanner’s philological observation seriously, thus assuming the position the verb is moving to, his site for emphatic, focused, or emotively marked utterances, is a strong F head. In as much as movement to F exists only in archaic dialects, we must assume that F is strong only in the latter. In turn, in as much as movement to F in archaic languages is only in certain contexts, we must have a reason for V not raising in all other contexts. The contexts themselves are the following, again according to Wanner:
(15) Context: % of enclisis
a. Relative marker 0
b. Subordinate complementizer 0
c. question/exclamation word 0
d. main verb focus 0
e. predicate focus 0
f. argument focus 0
g. scrambling 0
h. negative element 0
i. S[ X[adverb] cl V... --> S[ X[adverb] V cl... 10
j. S[XP[adverb] cl V... --> S[ XP[adverb] V cl... 20
k. S[ S[subject] cl V... --> S[ S[subject] V cl... 50
l. TOPIC[XP] S[cl V... --> TOPIC[XP] S[V cl... 50
m. [conjunction] cl V --> [conjunction] V-cl 80
n. S[cl V... --> S[V-cl... 100
This is the general archaic pattern, with V movement in the instances of 100% enclisis, none in the instances with 0% enclisis, and mixed results elsewhere. The “elsewhere” case is easy, if not to isolate, at least to idealize. The conjunctions where proclisis is possible are instances of subordination elements–the majority yield enclisis. The rest are elements that have either a topic or a focus reading, such as subjects and adverbs; presumably the focused reading yields proclisis, and the Left-dislocated (or more generally, topic) reading yields enclisis.
The situation just described is similar (among others) to Zwart’s (1993) characterization of V movement to C in some Dutch variants or Cardinaletti and Robert’s (forthcoming) V movement to a node after C, which they call Agr-1. More generally, we may expect Rizzi’s (1990) treatment of complementizers agreeing with subjects as instances of I to F. The general question is this: why is this phenomenon restricted to instances where F is ungoverned?
In the spirit of Zwart’s minimalist approach, we can propose that a strong F is eliminated in various ways by PF: (i) an element can appear it its spec if it matches in the relevant features (in my terms it is a point of view element); (ii) F can cliticize to a governing element, thereby disappearing as an independent head by PF; (iii) an element incorporates to F, burying F inside the raised element’s morphology (much like arguments raised to expletives make them invisible for LF). Economy considerations make (i) the most desirable route. However, it is not always the case that the route in (i) can be taken, given that not all sentences involve material associated to a point of view–a matter I deal with immediately. Then, (ii) is straightforward, even if it involves one further derivational step: the cliticization of F. But here too we have a limitation: this strategy would work only if there is a site governing F, not otherwise. The elsewhere case is then (iii), which is the least economical route: the V must move to support the strong F, a rather cumbersome process and the least preferred one.
Recall that in order for the last move not to invoke a violation of Greed, the verb itself must have strong F morphology. That it does is of course not obvious in all Romance languages (though it is in Western Iberian, where this morphology shows up as inflection in infinitivals). It should be noted though that the notions of “strong” and “weak” morphological PF features are quite technical in the minimalist project, and do not correlate directly with pronounceable morphology. Be that as it may, a technical wrinkle is left by the fact that the “strong” (though abstract) feature in V is directly checked only in those instances where V raises to F, whereby F replaces the appropriate features if we assume Chomsky’s (1993) specific mechanism–an issue for Zwart’s analysis as well. I suggest that this is as a result of the verbal morphology in this instance being “zero strong”; that is, present enough to allow the verb to move to a site where the feature is checked against a corresponding functional head, but not strong enough to force the verb to always move by PF. In other words, zero morphemes allow a category to move exactly to the domain where they are checked, but do not force this move since they themselves are invisible at PF. The reason why this movement does not violate Procrastinate is that if the verb did not move in the relevant instances the derivation would crash at PF, in those languages where the F category itself is indeed strong. In turn, in instances where the F morphology is not “zero-strong”, but fully strong, as in inflected infinitivals, V must raise to F in the general instance (cf. (13)).8
8. This gives a new rationale for Raposo's V movement of the sort in (13), although it still creates an issue for Ambar's (14).
As for the strength of F, I suggest that it be relativized to invoking point of view features. Recall that it is clitics that allow us to see the verb cruising past them. In sentences without these elements or without any special focalization, is there any reason to believe that the verb has moved upwards? I strongly suspect not, for normal declarative sentences involving indefinites can have the format SVO, as in the Galician un home veu unha muller ‘a man saw a woman’. Implementing semantic ideas of Heim (1982), Diesing (1991) has interpreted the idea from the literature on Scrambling that indefinites mark the VP boundaries. If this is correct, it is doubtful that the verb has raised in this instance, and then the question arises as to why the derivation does not crash at PF because of F’s presence. The answer must be that F is not active, and hence is invisible in this domain. Again following a suggestion from Zwart (1993), we may take functional categories to be activated when their checking domain is invoked. That is to say, the F projection is activated only if an element moves to its spec or if an element moves to its head, and not otherwise.
Substantially, I have interpreted movement to the F projection as invoking a point of view. In a neutral thetic statement like the sentence above involving indefinite arguments and an unspecified event, it is not obvious that a point of view is crucial, for neither the illocutionary involvement of the speaker, nor its categorial judgement, or even its commitment to a given reference are involved.9 If the analysis of cliticization in Uriagereka (forthcoming) is correct, sentences with clitics are different, since they involve specific information of a referential sort (the clitics themselves) and hence a point of view. My argument there is that, in a class of languages, clitics end up in F by PF (after the verb moves to F for morphological reasons of the sort outlined above) and hence that F is indeed active in those instances.
9. As noted above, topicalizations/Left Dislocations trigger enclisis, which must mean that the presence of a topic in the spec of FP, while activating F, is not enough to check it--unlike focalization. This is plausible if only the latter process involves an LF operator and thus triggers a matching agreement between the spec of F and its head. If the agreement is not invoked for topics, V must move to F or the derivation will crash at PF. See Raposo and Uriagereka (1993) for further discussion of these matters and relevant references.
One last issue that remains is how the verb can move to F over the clitics (prior to their placement to F) without violating the Head Movement Constraint. Baker and Hale (1990) and Roberts (1991), among others, are devoted to answering this in various ways. It is not obvious, however, that within the minimalist program this is a real problem. The Head Movement Constraint was taken to follow from the Empty Category Principle, which is now seen as a descriptive generalization following from the way in which the Phrase Marker is constructed in a cyclic fashion. If so, movement over a head per se is not a problem, and in fact Chomsky (1992) discusses some such movements. The question is whether in such a move some principle of economy is violated, or the resulting structure violates the architecture where the checking of features is feasible. I will not go into this now, but see Uriagereka (forthcoming) for a suggestion that no violations takes place in these instances.
I want to argue that F heads assign abstract Case and allow for the realization of different sorts of morphological and abstract Cases in their spec, under head-spec agreement.10 The view that a nominative, citation Case is available in peripheral positions comes from Indo-Europeanists, who noted that several languages show either a form with case concord (the displaced phrase has the case that it would get if it were in an argument position) or a form with nominativus pendens.
10. It would be interesting to try and reconcile this idea with proposals arguing for nominative being assigned from Comp in Indo-European languages (e.g. Platzack (1986) and many others afterwards). Also, note that what I am saying is that Case may be realized under spec-head agreement, but not that it is assigned under these conditions. See below for a concrete proposal on assignment. For extensive discussion of these and related issues, and traditional sources of the general idea, see Zwart (1989). He argues for a default assignment of nominative Case, which is a different position from the one taken in Raposo and Uriagereka (forthcoming), in terms of a default realization. I am a bit sceptical about default assignments in as much as, all other things equal, they predict no Case Filter violations.
Under this view, Ambar’s examples (14) do not constitute a counterexample to Raposo’s (1987) claim that personal infinitives require Case: F may be the Case assigner. A similar source of Case can be at issue in personal infinitives more generally.11 It might be said that what assigns Case to the subject in instances like (10) above are the prepositions. This however is unlikely. A property of Exceptional Case assignment in English and other languages is that the case-marked NP shows the morphological case the preposition assigns. In contrast, in the Romance instances at stake the case surfacing in the subject of the infinitival is not the one that the preposition would assign, but nominative–even in the absence of the sort of I that, at least in modern dialects, assigns nominative Case. Instead, assignment from F would provide a valid source of nominative.
11. This adds a twist to Raposo's treatment. For him, it is crucial that infinitivals appear where Case is available. If F can assign Case, a further issue arises as to where F can be licensed--i.e., which heads license FP as a complement. This also has consequences for the matter of CP recursion, since not all verbs allow it, a matter I come back to. Notice finally that (i) confirms the ideas introduced now:
(i) Extraname so a min/el non convidares pra esa festa
surprises-me only to me.acc/him not invite-you to that party
'It surprises me that you did not invite only me/him
to that party
?*Extraname so eu/el non convidares pra esa festa
surprises-me only I.nom/he.nom not invite-you to that party
I show below that, in principle, phrases in the left periphery can receive nominative, or alternatively can carry along the case that corresponds to them inside the clause (accusative, dative, and even some other prepositions). However, the focalized element displaced in front of the verb in (i) must in this instance receive a Case other than nominative, which is by hypothesis assigned to the infinitive. In other words, a Case conflict arises here if the displaced phrase wants nominative, given that the infinitive needs Case and only this extra nominative can be assigned to it. See below for more details.
Note also that if F is generally present in personal infinitives, we will have an explanation for clitic interpolation in these contexts, which is possible only in archaic dialects:
(16) pra lle (ti) (enton) falar(es),...
to him you then talk-II
'in order for you to talk to him,...'
Not only is it possible for clitics to come pre-verbally in Galician infinitivals, but they can even appear before a subject and/or adverbial.12 The example is entirely parallel to (5) above.
12. My suggestion, though, creates the same issue raised in fn. 8 for Ambar's (14), if overt agreement in infinitivals signals a strong F morphology. Note also that the process here is very different from what happens in French. I believe it is a mistake to relate French and archaic Romance dialects with respect to this matter. The position of clitics in French infinitivals is always preverbal and always proclitic, with the verb as the host; in archaic dialects it is optionally preverbal and interpolation is possible, with different kinds of hosts being at issue--see Uriagereka (forthcoming).
It is reasonable to conjecture, also, that this assignment of Case is what licenses overt expletives. Alvarez (1981) reports examples of the following sort, from Minho Portuguese and Galician, respectively:
(17)
a. O tia Getrudes, ele esta i a minha aboa?
Hey aunt Getrudes it is there the my.fem grandma
'Hey, aunt Getrudes, is my grandma there?'
Here the logical subject of the sentences is specific, yet is compatible with the expletive.13 However, the word-order in (17a’) and (17b’) is not possible with the overt expletive. That is, the overt expletive is apparently incompatible with subjects in the spec of IP. This follows if the expletive starts out in this position, where it can get Case from F. That we need two sources of Case here is shown by the fact that both the expletive and the logical subject are showing overt nominative Case. The extra source comes from F, by hypothesis.
13. This is a type of example that Raposo and Uriagereka (1990) do not deal with. We were unaware of these facts, which seem far from perfect in the standard dialects.
There is an apparently unrelated paradigm that behaves like (17). Though stigmatized, the forms in (18a), with nominativus pendens showing up in the dislocated phrase are certainly common in Romance dialects:14
(18) a. meu fillo n'o mata a fame
my son not-him kill the hunger
'My son, hunger doesn't kill him'
(cf. 'a meu fillo n'o mata a fame'
to my son not-him kill the hunger
b. * meu fillo a fame n'o mata
(cf. a meu fillo a fame n'o mata)
14. Kany (1963) reports (as careless) the following Spanish examples:
(i)
a. el muy burro se le dio por jugar [Argentina]
the very ass self him.dat gave.III.sg for play
'The ass, it occurred to him to play'
b. el animal le di permiso [Chile]
the animal him.dat gave.I.sg permit
'The beast, I allowed him to'
c. el negro no le hace nada la culebra [Venezuela]
the black not him.dat make nothing the serpent
'the guy, the serpent doesn't bite him'
Interestingly, the nominative Left-dislocated phrase is incompatible with an overt subject in the spec of IP (18b)–for that word order, the Left dislocated phrase must bear whatever case it bears in the A-position it is associated with.15 Again, this sort of facts follow if there are two sources of nominative Case in an example like (18a)–for meu fillo ‘my son’ and a fame ‘the hunger’–quite independently of the accusative Case assigned to the pronominal clitic o ‘him’. Also, (18b) suggests that the dislocated element, like the expletive in (17), is at some point in the spec of IP.16 This would follow given Raposo and Uriagereka’s (1990) definition of government under c-command, and assuming Case assignment only under government: the only way a phrase can receive Case from F is if this phrase is strictly c-commanded by F (in the course of the derivation). This of course is the case if the phrase is in the spec of IP. Then case can be realized in the spec of a category (e.g. FP, but also IP, etc.), as discussed in Raposo and Uriagereka (forthcoming).
15. This dismisses a general approach to the phenomenon in terms of the "As for" construction (see Kany (1963))--restrictions of the sort noted are unexpected in those terms. The process is somewhat reminiscent of Maling's (1980) Icelandic stylistic fronting. An important difference is in terms of the elements which can front, which in Icelandic seem to be much more restricted.
16. Perhaps as in Diesing's (1989) analysis of Yiddish embedded clauses. Notice also the possible correlation between these instances and examples with locative inversion, as in Bresnan and Kanerva (1989).
Recall also that nothing I have said yet predicts why the expletive is incompatible with focused phrases (9a). An account is possible within the proposal about barriers in Raposo and Uriagereka (199): the expletive remains in the spec of IP in those instances, and hence this category is a barrier preventing focus movement across it. This yields (19):
(19)
a. *moitos convidados el parece que viron na festa
many guests it seems that saw.III.pl in-the party
b. estes convidados el parece que os viron na festa
these guests it seems that them saw.III.pl in-the party
'these guests it seems that they saw them in the party'
c. huu logar...en que er podesse chorar
a place in which ER could.I cry
'a place in which I could cry'
Not just any phrase in peripheral site is incompatible with the overt expletive (19b): it is exactly those that involve movement that are impossible (19a). Furthermore, Wh-movement is not incompatible with focus heads (19c). I therefore conclude that the position of the expletive in these instances is still the spec of IP, that only full phrases move to the spec of F, and that this movement cannot be across IP fully specified, in support of the Raposo/Uriagereka approach.
We have seen a host of properties separating different Romance dialects within what we may call the “CP-system”. The difference between B and C dialects (of the sort in (1)) is arguably trivial: whether F features in V are present (C) or not (B). This difference has important consequences, since movement of V to F would violate Greed and is thus generally impossible, yielding the clitic pattern in (1B). If we assume that (i) (overt F heads) and (v) overt expletives correspond to a strong F morphology, the lack of these in (1B) dialects also follows. But a merely syntactic F may not be strong enough to host clitics in isolation, thus predicting the absence of (vii) (the interpolation of material between clitics and V) in (1B) dialects. In contrast, (ii) overt focus movement, (iii) recomplementation, and (iv) sandwiched dislocations are still possible in (B) dialects while F has a syntactic realization, assuming these processes all invoke the strength of F once a point of view is at stake. Finally, consider (vi): although personal infinitives are not lost in (B) dialects (even if they have a rhetorical flavor), inflected ones remain only in a few archaic variants. Non-morphological F can still assign nominative Case, but the overt realization of F features in the verb is not present.17
17. While I take it that overt agreement morphology in an infinitival means strong F morphology in V, recall that I do not imply by this that only overt morphology is interpreted as strong--for it can be also "zero-strong". Nevertheless, it should be expected that the paradigm with overt morphology is the more stable one, if only because of learnability considerations.
As for the difference between A dialects and the rest, it may be more dramatic: neither is F strong there, nor is it the case (consequently) that strong F morphology shows up in the verb. Certainly, it is difficult to attest the properties in (2) in standard modern French, with the clear exception of (2v). Then again, I am not convinced that the expletive we see in French is of the same sort as the expletive in archaic dialects, the French one being a consequence of the impossibility of pro-drop.18
18. To be fair, I gave just the idealized picture. See Martins (forthcoming), Cardinaletti and Roberts (forthcoming), and Fontana (forthcoming) for a more detailed description and analyses of French, Spanish, and Portuguese diachronic facts.
Accepting “discourse configurationality” as an empirical fact does not have to entail rethinking the grammar in non-minimalist ways. The picture I have presented is minimalist in its operations, its assumptions about levels of representation, and its treatment of variation. The key to it is the postulation of an F category that encodes point of view, and which I believe we need across languages, even Indoeuropean languages where it is not obvious prima facie, or it has been given other names. In languages with overt focalization strategies of the sort discussed in this volume, it is hard to see how to avoid the postulation of such a category, which many of the chapters in this book do.
Apart from the various patterns that I relate here in terms of F, in work in progress I develop some ideas discussed in Uriagereka (1988) to relate F to the matter of Subjacency parameters, which are clearly non-minimalist. That is, parameters of that sort involve global domains, instead of lexicon items (see Rizzi (1989)). The line I pursue, instead, is Reinhart’s (1979) original insight that languages with F have an extra escape-hatch for Wh-movement, with complex associated properties depending on whether F is morphological, as discussed above. I believe that comparative grammar is of help in deciding these matters. In as much as it is not a single phenomenon, but a class of phenomena that appear and disappear together across closely related languages, we should in principle try and relate the array of facts to a single variable.
More generally, all matters of “CP recursion” should be amenable to an F treatment. This means that the circumstances under which Iatridou and Kroch (1992) propose a recursion of CP should follow from deeper matters. For instance, it is known that the phenomenon is restricted to certain epistemics, declaratives, and the like, in an affirmative version. This is exactly what we expect if in the relevant contexts we must select a category expressing a point of view, which is not expressible through desiderative or volitional verbs, the complements of nominals and adjectivals, or even the negative versions of declaratives and epistemics introducing a de dicto complement clause. Thus, we can keep the essential insights of Iatridou and Kroch, but instead of introducing a peculiar recursion (which is not seen elsewhere for I, D, or any other category) we can reinterpret their ideas in terms of a standard selection.19
19. For a more detailed presentation of these issues, see Torrego and Uriagereka (in progress).
I realize that positing a new category is costly, and the burden of proof is on the hypothesis for its existence. This was true for hypotheses about Comp, Infl, Det, Agr, etc., which at the time of their initial discussion met considerable skepticism from the field–as should be the case. As usual, the deciding factor is empirical. In the case that concerns us here, if we accept that information matters affect configurationality, the minimalist system forces us roughly in the direction outlined above. In as much as that direction seems to find empirical confirmation in a number of patterns, yielding a variety of inter and intra-linguistic facts, the system itself that forces the hypothesis is considerably strengthened.
* Early versions of this paper were presented at the New Jersey Circle in Princeton and GLOW in Lisbon. I appreciate comments from Josep Maria Fontana, Charlotte Galves, Elena Herburger, Norbert Hornstein, Katalin Kiss, David Lightfoot, Ana Maria Martins, Jairo Nunes, Eduardo Raposo, Dieter Wanner, and my students and colleagues at the University of Maryland.
References
Alvarez, R. (1981), O Pronome Persoal en Galego, University of Santiago de Compostela doctoral thesis.
Ambar, M. (1989), Para Uma Sintaxe da Inversao Sujeito Verbo em Portugues, University of Lisbon doctoral thesis.
Baker, M. and K. Hale (1990), ‘Relativized Minimality and Pronoun Incorporation,’ in LI 21.
Borao, J. (1908), Diccionario de Voces Aragonesas, Diputacion Provincial de Zaragoza.
Bresnan, J. and J. Kanerva (1989), ‘Locative Inversion in Chicewa,’ in LI 20.
Cardinaletti, A. and M.T. Guasti (eds.) (forthcoming) Small Clauses, Academic Press.
Cardinaletti, A. and I. Roberts (forthcoming), ‘Clause Structure and X-second,’ in W. Chao and G. Horrocks (eds.).
Chao, W. and G. Horrocks (eds.) (forthcoming) Levels of Representation, Foris.
Chomsky, N. (1977), Essays on Form and Interpretation, North Holland.
Chomsky, N. (1989), ‘On The Economy of Derivations,’ in MITWPL X.
Chomsky, M. (1992), ‘A Minimalist Program for Linguistic Theory,’ in Occasional Papers in Linguistics 1, MIT.
Diesing, M. (1989), ‘Verb Second in Yiddish and the Nature of Subject Position,’ in NLLT 8.
Diesing, (1991), The Syntactic Roots of Semantic Partition, UMass doctoral thesis.
Duarte, I. (1987) A Construcao de Topicalizacao Na Gramatica do Portugues, University of Lisbon doctoral thesis.
Fontana, J.M. (forthcoming) Phrase Structure and The Syntax of Clitics in the History of Spanish, UPenn doctoral thesis.
Heim, I. (1982), The Semantics of Definite and Indefinite Noun Phrases, UMass doctoral thesis.
Hellan, L. and K. Christensen (eds.) (1986), Topics in Scandinavian Syntax, Kluwer.
Henriquez Urena, P. (1939), ‘Ello,’ Revista de Filologia Hispanica I.
Hernanz, M. L. (1982), El Infinitivo en Espanol, Universidad Autonoma de Barcelona.
Higgins, R. (1988), ‘Where The Old English Sentence Begins,’ ms. UMass at Amherst.
Iatridou, S. and A. Kroch, ‘The Distribution and Licensing of CP
Recursion,’ ms. University of Pennsylvania.
Kany, Ch. (1963), American Spanish Syntax, University of Chicago Press.
Kiparsky, P. (1986), ‘Tense and Mood in Indoeuropean Syntax,’ Foundations of Language, 4.
Lapesa, R. (1942), Historia de la Lengua Espanola, Gredos.
Maling, J. (1980), ‘Inversion in Embedded Clauses in Modern Icelandic,’ Islenskt Mal 2.
Martins, A. M. (forthcoming), Untitled University of Lisbon doctoral thesis.
Maurer, T. H. (1978), O Infinito Flexionado Portugues, Compania
Editora Nacional, Sao Paulo.
Mattos e Silva, R. V. (1989), Estruturas Trecentistas, Estudos Gerais.
Platzack, C. (1986), ‘Comp, Infl, and Germanic Word Order,’ in L. Hellan and K. Christensen (eds.).
Pollock, J. I. (1989), ‘Verb Movement, Universal Grammar, and The Structure of IP,’ in LI 20.
Raposo, E. (1987), ‘Case Theory and Infl-to-Comp,’ in LI 18.
Raposo, E. and J. Uriagereka (1990), ‘Long-Distance Case Assignment,’ in LI 21.
Raposo, E. and J. Uriagereka (1993), ‘Two Types of Small Clauses,’ presented at GLOW, Lisbon, to appear in Cardinaletti and Guasti (eds.) (forthcoming).
Raposo, E. and J. Uriagereka (forthcoming), ‘Indefinite Se‘, to appear in NLLT.
Reinhart, T. (1979), ‘Two Comp Positions,’ paper presented at GLOW, Pisa.
Rizzzi, L. (1989), ‘On The Format of Parameters,’ in BBS 12.
Rizzi, L. (1990), Relativized Minimality, MIT Press.
Roberts, I. (1991), ‘Excorporation and Minimality,’ in LI 22.
Santorini, B. (1991), ‘The History of the Verb Second Constraint in Yiddish,’ ms. UPenn.
Torrego, E., and J. Uriagereka (in progress), ‘Indicative Dependents,’ ms. UMass and UMCP.
Uriagereka, J. (1988) On Government, UConn doctoral dissertation.
Uriagereka, J. (forthcoming), ‘Aspects of Clitic Placement in Western Romance,’ to appear in LI.
Uriagereka, J. (forthcoming b), ‘Government Restrictions and Basque Movements,’ to appear in NLLT.
Wanner, D. (1988), The Development of Romance Clitic Pronouns, Mouton de Gruyter.
Zamora Vicente, A. (1966), Dialectologia Espanola, Gredos.
Zwart, J.W. (1989) The First Case, U. Groningen MA dissertation.
Zwart, J.W. (1993) ‘Verb Movement and Complementizer Agreement,’ in MITWPL XV.