Uninterpretability is incompatible (in morphology) with other minimalist postulates. Consequences for Agree and Move. 1. The nature of morphology. Much work in generative morphology supports the conclusion that syntax and morphology share some significant principles of structural organization, for instance a rule forming headed constituents (Williams 1979). In the model of the Latin verb by Embick and Halle (forthcoming), schematized in (1), the morphological categories of aspect and tense as well as the hierarchies they enter into have an exact parallel in the syntax, as schematized in (2); thus common (computational and interpretive) principles must underlie both. Given the extent of the overlapping, we take it that optimally there should be a single morpho-syntactic component, submitted to all and only the same principles. (1) [T [Asp [v laud-a] v-e] r-a] mus (2) Subj [TP T [AspP Asp [vP v VP]]] 2. The agreement morphology of the verb. If we were to reason on the basis of results such as those in (1)-(2), it would be natural to suppose that the agreement inflection of the verb (-mus in (1)) furthers the parallelism with the sentence, lexicalizing the same category in the morphology as the subject in the syntax. However in the minimalist model of Chomsky (1995, 2000, 2001), it is axiomatic that the agreement morphology of the verb corresponds to an uninterpretable set of features. Thus the agreement features of the verb cannot correspond to a syntactically merged category, but must be valued in the course of the derivation (setting it in motion, as a matter of fact) and then eliminated from the syntax. Accordingly, Embick and Halle (forthcoming) assume that the agreement inflection of the Latin verb is merged not in the (morpho-)syntax as the other categories in (1), but rather in a specialized morphological component. The first price to pay for this analysis is the existence of the morphological readjustment component itself, hence the impossibility of a completely unified morpho-syntax, as well as the potential availability in the morphology of a series of mechanisms forbidden by the restrictive principles of syntax. The second price, itself a consequence of morphological readjustment, is Late Insertion, which runs counter the minimalist core idea that all syntactic structure is projected by lexical material (essentially the principle of Inclusiveness). 3. Empirical evidence regarding the agreement morphology of the verb. The theoretical argument against the treatment of verbal agreement via morphological readjustment is strengthened by the available empirical evidence. In particular we argue that there is evidence in favor of a treatment of the agreement inflection as a verb-internal pronominal subject. Thus consider Romance languages with subject clitics, i. e. Northern Italian dialects (and arguably Romantsch dialects, French); identical lexicalizations of the subject clitic and of the corresponding verb morphology include at least (3). (3) 1 sg: i e.g. i dorm-i ‘I sleep-1sg’ (La Pli de Mareo, Alto Adige) 2 sg: t e.g. te mangia-et ‘you ate-2sg’ (Strozza Valle Imagna, Lombardy) 1 pl: n(e) e.g. mangiau-ne ‘ate-1pl’, ne llavamu ‘ourselves wash’ (Giurdignano,Apulia) 2 pl: v e.g. vi lavava-vu ‘yourself washed-2pl’ (Camporeale, Sicily) 3 pl: n e.g. in l a-n ciamau ‘they him have-3pl called’ (Airole, Liguria) Furthermore, so-called agreement inflections of the verb can be interleaved in the pronominal clitic series, notably in the plural imperative, both in Romance and in Albanian dialects. (4) a. da-me-te-lle (Albidona, Calabria) give-me-2pl-it (‘Give it do me!’) b. o-m-ni-e (arberesh, Carfizzi, Calabria) give-me-2pl-it (‘Give it to me!’) Evidence similar to (4) is known to Halle and Marantz (1994) from Caribbean Spanish. Their treatment, consistent with the model of minimalist syntax/Distributed Morphology, involves a syntactic structure where the agreeing verbal form is followed by an enclitic group, and a morphological readjustment rule infixing the clitic group betwen the verb stem and its inflection. But consider the fact that the phenomenon in (4) is restricted to imperatives. It is a widely held theory that the imperative occupies a very high position in the sentence (Rivero 1994), say a high C position in a Rizzi (1997)-type hierarchy. This high position of the imperative can be argued to leave the syntactic space necessary for the stranding of the agreement inflection, effectively a subject clitic, and hence the space necessary for the appearence of enclitics both to the verb stem and to its inflection, roughly within the dotted lines in (5). Note that the verb stem corresponds to the 2 person singular imperative and can therefore independently lexicalize the Cn position. (5) [Cn imperative stem … [imperative inflection … In our presentation we shall work out the syntactic account in (5) in full detail. What is immediately relevant is that the alternative morphological account would need to stipulate a number of properties that already follow from the syntax as part of the morphology. Thus there is no morphology-internal reason why the agreement inflection of the verb should be splittable from the stem, to the exclusion for instance of the modal inflection of the infinitive -- which in Southern Italian dialects of the type of (4a) equally involves enclisis. Similarly, there is no reason why a morphological rule that has the power of infixing (part of) an enclitic group shouldn’t have the power of infixing (part of) a proclitic group. A further line of argument can be developed starting with the observation that the split between clitics always has 1 and 2 person clitics infixed and 3 person ones as enclitics. The dissociation between 1/2 and 3 is independently known in the syntax (where it underlies split ergativity and more) and understandable in interpretive terms as a dissociation between discourse- and event- anchored arguments. But in the morphology such a split can only be stipulated. In short, a treatment of the agreement inflection of the verb as as a pronominal subject inside the verbal structure is best suited to explain the evidence in (3)-(4); in this respect as well, we uphold the unification of morpho-syntax. 4. Potential problems for our argument. Before coming back to the consequences of our conclusions for the minimalist model, we review (as time allows) some more specific, morphologyrelated potential objections. Embick and Halle (forthcoming) treat so-called tematic vowels in terms of morphological insertion as well -- these are the vowels (hyphenated in (1)) that close off the consonantal verbal root as well as the tense and aspect morphemes. The model envisaged here commits us to an interpretable treatment of thematic vowels as well. Thus we know that the verbal root augmented by the thematic vowel enters into nominal compounds in Romance languages (French tire-bouchon, Italian cava-tappi ‘cork screw’). On such evidence (and other) we take it that thematic vowels are nominal morphology, representing essentially the internal argument. Another objection to which we lie open is that something like morphological readjustment and Late Insertion is independently needed. One reason for Late Insertion is ‘syncretism’, i.e. the fact that one and the same piece of morphology can lexicalize several different bundles of features. In this respect we take the line of Manzini and Savoia (2002, 2004), who consider such phenomena in the clitic system of several Romance dialects. Thus ‘syncretism’ between the (masculine) plural and the ‘dative' is the byproduct of a common quantificational features, that allows for both the plural interpretation and the distributor (dative) one. Analogously the ‘syncretism’ between dative and locative issues from the fact that the ‘dative’ can be equally lexicalized as a directional. In a nutshell, there is no abstract underlying ‘dative’ feature, therefore there is no ‘syncretism’, but simply a different range of interpretation of quantificational, locative categories (and more), which parametrically includes the distributor, directional (etc.) argument of certain events. 5. Agree and Move. Let us then return to the conclusion that the finite agreement of the verb corresponds to an interpretable morpho-syntactic category, as a pronominal subject does. The obvious consequence is that the agreement of such a category with a lexical subject cannot be treated in terms of minimalist Agree. Now, generative grammars typically allow for a different Agree, whereby the agreement between two nominal categories both introduced by Merge, allows for their coreference. This interpretive construal of Agree, independently needed for pronominalization, can be extended to subject-verb agreement. To the extent that Agree is a subcomponent of Move (up to Chomsky 2001:23), the computational construal of Move itself is then called into question, and an interpretive construal (Brody 1999) potentially favored.