Size matters

It is a characteristic of every theory using non-terminal spellout that depending on how many terminals they realize, morphemes are of different syntactic complexity and size. In an indirect way, this size issue interacts with syntactic phenomena such as movement, pied-piping and co-occurrence restrictions. These interactions are discussed from a Nanosyntactic point of view in F´abregas (2009), Caha (2009) and Starke (2011), among others.

The interaction of size and the functional sequence

Suppose that the functional sequence defines the order of four features as A > X > B > C, and a language has a lexical item likeLI1.

(34) f-seq

A X

B C

(35) lexical entry in the lexicon

A B C

LI1

LI1can be used to spell out A, B and C together only in case X is not projected in the language.

If X is projected, as in (34), then it acts as an intervener between A and B and prevents A from being lexicalized together with B and C; A on the one hand and B and C on the other have to be spelled out by different morphemes. I call this as‘disruption effect’. What happens in these cases depends on what other lexical entries are available in the lexicon. For instance, if the language possesses anLI2 spelling out A, an LI3 spelling out X, and an LI4 that can spell out B and C, then the tree is lexicalized in the following way.

(36)

A LI2 X

LI3 B C LI4

Consider now the situation in which there is noLI4that could spell out only B and C, and there are no lexical items that could spell out B and C separately either. Due to the Superset Principle, in this case B and C can be spelled out byLI1: LI1‘shrinks down’ to B and C and Underassociates its A feature.

(37)

A LI2 X

LI3 B C LI1

In Chapter 4, I will argue that non-inflecting demonstratives and certain Hungarian quantifiers display precisely the disruption effect in (37).

The interaction of size and movement

As movement changes constituency and the contiguity of terminals, extraction has an effect on how the structure can be lexicalized. Suppose that the functional sequence is X > A > B > C, and a language has LI1 (for A, B and C) and LI3 (for X). LI1 is specified such that it cannot Underassociate its C feature. In addition, the language has no other lexical item that could spell out any further combination of A, B, C and X. If in this language CP moves above X, then the structure becomes unlexicalizable because there will be no felicitous spellout for A and B. (C could be spelled out byLI1via the Superset Principle.)

(38) structure non-lexicalizable with the available entries of the language¹⁶

CP . . . C . . .

X A

B C

This language has two possibilities to create a lexicalizable tree. It can leave C(P) in its base position, or it can move the whole constituent containing A, B and C. That is, in order to create a structure that can be lexicalized, this language can have an in situ construction or it can apply pied-piping. In Chapter 5, I will show that certain pied-piping phenomena in the domain of Hungarian PPs receive a natural explanation along these lines.

In Phrasal Spellout lexicalization interacts with movement in a perhaps even more spectacular way. Movement creates constituents that can be the target of spellout, thus it can feed the use of certain lexical items. Movement also destroys constituents, which can bleed the use of certain lexical items. For more discussion and examples, I refer the reader to F´abregas (2009); Pantcheva (2011) and Starke (2011) (the latter also contains an insightful Nanosyntactic rendering of covert movement as overt movement).

By necessity, movement interacts with lexicalization and potentially changes the lexicalization possibilites in any theory that uses non-terminal spellout. See Newson (2010) for demonstration that (the OT-equivalent of) movement changes the lexicalization possibilities in Alignment Syntax.

As F´abregas (2009) points out, in a theory that uses terminal spellout, movement is not supposed to change the lexicalization possibilities of the structure: terminals are still terminals after movement.

16Traces are generally assumed to be ignored by the lexicalization algorithm.

Therefore if empirical evidence is found that movement changes spellout, it must be seen as strong evidence in favour of non-terminal spellout. For specific examples that support non-terminal spellout in this way, see F´abregas (2009), Caha (2009) and my Chapter 5 on Hungarian PPs.

The interaction of size and co-occurrence restrictions

The size of a morpheme also delimits the possible modifiers of that morpheme. When the same class of lexical items systematically admit a certain type of modifier in one language but not in the other, then a straightforward rendering of this fact in non-terminal spellout is that the modifier is spelled out together with the modifyee in the second type of language.

The reader will recall that Weerman and Evers-Vermeul (2002) explain the inability of certain Dutch pronouns to co-occur with DP-internal modifiers in the same manner. They argue that these pronouns spell out the whole KP, hence they‘eat up’ the positions that could be occupied by modifiers in the DP.¹⁷ In Chapters 4 and 9 I will touch upon the modification possibilities of Hungarian pronouns, and argue that the severely restricted modification of overt and covert pronouns can be explained along these lines.

The interaction of size and the apparent non-rigidity of the functional sequence

While some researchers contend that every projection of the functional sequence is present in every language, claims that certain languages do not appear to have a projection X or Y abound. As pointed out in Starke (2011), the idea that every projection is present in every language can be maintained without predicting that every projection has an overt realization in every language. If the vocabulary of a language is such that some projection is always spelled out together with some other terminal, it will appear to be missing from the functional sequence of the language.

In a variety of the missing functional projection claim, two phrases are said to be co-projected in some languages. This is applied to cases where there is evidence for both projections, but these never have co-occurring exponents and appear to be in complementary distribution. Analyses that make use of co-projection include, among others, Thr´ainsson (1996); Bobaljik and Thr´ainsson (1998); Giorgi and Pianesi (1997) and De Belder (2011). Co-projection introduces a form of cross-linguistic variation into the functional sequence (even if this is not a variation in order). This can be avoided in non-terminal spellout in a straightforward way. If two projections never have co-occurring exponents in a language but there is syntactic and semantic evidence for the presence of features associated to those projections, then this language does not possess lexical items that could spell out the two projections separately. Instead, this language only has lexical items that span both projections. In other words, the projections in this language undergo co-spellout, as opposed to co-projection. This puts the source of variation in co-occurrence into the lexicon rather than the syntax. In Chapter 9 I will argue that the plural marker does not co-occur with classifiers in Hungarian because of such a co-spellout.

Where have Fusion and Fission gone?

As already discussed, Nanosyntax does not have a morphological component and it does not have space in its architecture for post-syntactic operations like Fusion and Fission. The effects of Fusion and Fission, however, fall out as a side-effect of spellout. Fusion merges two adjacent heads and creates one head from them with one terminal. The effect of this operation is that two terminals which are separate in syntax are spelled out by a single morpheme. Non-terminal spellout captures this effect without further ado.

Fission works in the opposite direction as Fusion: it takes a single terminal and splits it up into multiple pieces, with a different morpheme lexicalizing each piece. The effect of Fission is that one terminal in the syntax is spelled out by several morphemes. Nanosyntax has no way of splitting up terminals. Instead, the analysis of apparent cases of Fission is that the relevant features have never been bundled in the first place. Instead, they were merged in separate terminals from the beginning. The reason why Fission seems to have taken place is that those terminals are spelled

out together in most other languages. In a theory with terminal spellout, that gives the mistaken illusion that they came from just one terminal.

2.5 Summary

This section introduced the lexicalization problem: the omnipresence of polysemous lexical items that appear in various positions in f-seq. In order to deal with this problem, I adopted the Nanosyntactic lexicalization algorithm. Nanosyntax operates with non-terminal spellout, and heav-ily constrains systematic polysemy and appearance in multiple positions. Nanosyntax assumes no morphological module. Its basic lexicalization principles are the Exhaustive Lexicalization Princi-ple and the Superset PrinciPrinci-ple. Competition between lexical items is regulated by two economy principles, Minimize Junk and Maximize Span. The lexicalization of non-terminals interacts with the functional sequence, pied-piping and co-occurrence restrictions in intricate ways. Chapters 3 through 5 will provide ample illustration of all these principles and interactions at work.