Revisiting agent pseudo-incorporation in Turkish: a dependent case theoretic perspective

An anonymous reviewer rightly points out that bare adverbs are incompatible with non-specific indefinites as shown in (30a) while specific indefinites as well as bare NPs are compatible with them as shown in (30b) and (30c), respectively. Given that bare adverbs have been used to diagnose object pseudo-incorporation, in particular to show that pseudo-incorporation is a VP-level process, the idea that non-specific indefinites remain within the VP is challenged by their incompatibility with bare adverbs.

To declutter the syntactic representations in the main text, we did not fully articulate our assumptions about where objects/themes enter the derivation. Although the issue of where they are merged in syntax is somewhat orthogonal to the case theoretic puzzle concerning subject pseudo-incorporation in Turkish, we will sketch an analysis here that readily accommodates the contrast in (30). An analysis of this type calls for a distinction between the merge positions of pseudo-incorporated NPs and argumental DPs, reflecting the original insight in Öztürk (2005a, 2005b, 2009, where she argues that while pseudo-incorporated objects are introduced under VP, DP arguments are introduced via separate functional heads on top of the VP. Translating this proposal to our representations where we maintain a fully decompositional approach in the verbal domain, we propose that a theme head is responsible for introducing argumental DPs which are themes. Notably, occurring immediately below the verbalizer, the theme head is not a verbalizer, and directly combines with the root phrase in syntax as shown in (31).

(31)

As we shall see, this addresses one half of the puzzle concerning the contrast in (30). The other half concerns the attachment site of bare adverbs. We argue for a lower attachment site for bare adverbs than regular VP modifiers. In particular, we propose that bare adverbs attach at the root level, i.e., before any verbalization applies. Our evidence comes from the so-called deverbal nouns that are able to license theme NPs as their complements. Arguably, these deverbal nouns are not verbal at any point in the derivation as shown by their incompatibility with manner adverbials in (32a). Crucially, however, they, along with their incorporated theme complements, are able to co-occur with bare adverbs as shown in (32b).

The contrast between (32a) and (32b) supports the idea that bare adverbs attach to a position before verbalization, which we hypothesize to be the root phrase. Combining the two proposals, namely that argumental DPs are introduced in [spec, theme], and that bare adverbs attach to the root phrase, we account for the inability of non-specific indefinites to occur under bare adverbs. As shown in (33), the string where the bare adverb precedes the argumental non-specific indefinite is ruled out on the grounds that there is simply no position for non-specific indefinites to occur under the root phrase.

Under this analysis, there is still a pattern that is expected to be ruled in, but is nevertheless unacceptable. As shown in (34), the non-specific indefinite DP bir kitap, preceding the bare adverb, should be able to remain without an overt accusative case given that it is still within the first phase. We argue that what rules out (34) is not a syntactic, but a prosodic restriction that disallows any material between the verb and non-specific indefinites.

While we do not know why non-specific indefinites, albeit being argumental DPs, exhibit this behavior, unlike other argumental DPs, there is evidence that the nature of the restriction is prosodic. Turkish allows right adjunction of bare adverbs as well as left adjunction as shown in (35). We hypothesize that the left versus right adjunction of bare adverbs uniformly targets the same level in the structure, i.e., the root phrase.

When we also employ right adjunction in (34), the sting becomes acceptable.

To be able to accommodate non-specific indefinites into the picture, we had to make two assumptions, namely that bare adverbs modify root phrases, and that there is a prosodic constraint active in the grammar of Turkish which can be obviated by the right adjunction of the bare adverb. Although it may seem that these are ad hoc assumptions, they are in fact independently supported. In what follows, we discuss the interaction of bare adverbs with agent pseudo-incorporation, and argue that the two assumptions needed for non-specific indefinites are independently justified. The relevant piece of data is the inability of bare adverbs to occur in agent incorporation structures. For example, although the semantic content of the adverbial is compatible with the event described by the verb as illustrated in (37b), the bare adverb with the same semantic contribution cannot occur when the agent is pseudo-incorporated (37c).

The contrast between subject and object pseudo-incorporation with respect to the compatibility with bare adverbs is a welcome result in our analysis where agent pseudo-incorporation is made possible by v_inc which semantically combines with its complement and then the agent NP in its specifier. The string in (37c) cannot be generated given that there is no position for the bare adverb above the pseudo-incorporated agent considering that v_inc is introduced above the root phrase.

That said, if bare adverbs attach to the root phrase before the verbalization applies, that raises the question why the bare adverb prevents agent pseudo-incorporation construal in the sentence in (38). Given that the agent NP is a specifier to the verbalizer v_inc, the low attachment of the bare adverb within the root phrase would rule in (38) for the pseudo-incorporation construal, all things being equal.

We already have an account of why this reading is not ruled in. Recall that in explaining the inability of non-specific indefinites to precede bare adverbs, we invoked a prosodic constraint that could be obviated by right adjunction of the bare adverb. We argue that the same constraint is also active in case of agent pseudo-incorporation. The evidence that the prosodic constraint rules out the pseudo-incorporation reading in (38) is shown in (39) where the right adjunction of the bare adverb saves the pseudo-incorporation reading, allowing the adjacency between the incorporated agent and the verb, as opposed to (38) where pseudo-incorporation reading is unavailable.