Expletive There

In the previous post we saw that agreement between determiners and nouns was naturally described in terms of dependencies between lexical items. While the more well-studied case of subject verb agreement in English was treated in terms of agreement over movement dependencies, here we treated agreement in the DP in terms of merge dependencies. This fits much better with the canonical analyses of these constructions, which do not appeal to movement at all.¹ The standard, AGREE-based approach to agreement, which makes use of a separate syntactic operation called AGREE, is something of a hybrid between our dependency-based approach to agreement, and a SPEC-HEAD approach. It is like the SPEC-HEAD approach in that it makes use of a syntactic operation, establishing a direct syntactic dependency between two lexical items that is the same kind of thing that would happen if a movement operation targeted these two lexical items. It is like the present dependency-based approach in that it is envisioned as descending down the tree (along dependencies albeit viewed in terms of constituents) until it finds a goal with relevant morphological features. As such, it is supposed to have locality conditions different from those of movement.

In this post, we investigate the expletive there construction:

• There is no solution to this problem
• There are no solutions to this problem
• There were three piglets in the box
• There will come a time when monkeys rule the earth
• There are times when all you can do is cry

In this construction, the expression which appears in what seems to be the usual position for surface subjects, there, is not the superficial controller of verbal agreement. Instead, the expression that would ‘normally’ be the subject is the controller of agreement. There are three basic ideas in the transformational tradition about how these constructions work, which are (imho) most naturally depicted in terms of dependencies. We will work through them in turn, but we begin with a most basic and hopefully uncontroversial dependency structure.

The high origin account

The basic word dependencies in figure 1, while not exactly analysis neutral (they presuppose a DP analysis of nominal phrases, for example), are common to all three of the analyses we consider. To obtain these analyses from this dependency structure, we will need to add additional dependencies. We begin with an additional dependency between arrived and a, which should capture the fact that DPs require case, which they get from being in contact with tensed items. This is shown in figure 2.

We now decompose arrived into a stem arrive and a past tense head, and have arrived at the refined dependencies underlying Chomsky’s high origin account of there-insertion.

These dependencies give rise to the following lexical items.

\(\textsf{there}\mathrel{::}e\)	\(\textsf{Pst}\mathrel{::}\underline{\bullet v}.{\circ k}.{e\bullet}.s\)
\(\textsf{arrive}\mathrel{::}\bullet d.v\)	\(\textsf{a}\mathrel{::}\bullet n.d.k\)	\(\textsf{man}\mathrel{::}n\)

The dependency between a and Pst is a movement (internal merge) dependency. However, if a man were to overtly move, it would surface in the wrong position (before arrived). Therefore, this movement must be covert - this is indicated by using the feature \(\circ k\). There is realy no reason to order the case movement before or after the expletive selection. Chomsky proposes that the ‘associate’ - the DP which agrees with the T - move covertly to the position occupied by expletive there. So taking this ‘literally’, we would need to first merge the expletive, and second move the DP: Pst would then have the feature bundle \(\underline{\bullet v}.{e\bullet}.{\circ k}.s\). I don’t think it really matters.

We observe that the agreement between tensed verb and noun (which I obscured by using the past tense…) could be mediated over two paths:

1. man – a – arrive – Pst
2. man – a – Pst

The first path arrives at Pst via the merge dependency between a and arrive, whereas the second path arrives directly at Pst via the movement dependency between a and Pst. Only path 2 has the property that its length does not increase as the distance between the DP and tense grows, as in, for example, a raising construction (e.g. There seems to have arrived a man).

The low origin account

We return to our basic dependencies (figure 1), adding once again the case dependency between determiner and tense, as in figure 2. This time, however, we wish to, following Deal (2009), have there move to the canonical subject position. It must, therefore, start out in a lower position, which we assume is dependent on the verb. This is shown in figure 4.

We then again decompose arrived into a head expressing tense, and one expressing the lexical verb, between the two dependencies of there. This creates a complex T–V span expressing the inflected verb, but simplifies the lexicon (instead of having an atomic form arrived we decompose it into the familiar past tense head and a tenseless verb).

The arrive node in this dependency structure is not the usual one - it has too many dependencies! We decompose this node into the usual intransitive arrive and … something else. We do this by decomposing the arrive node between the two dependencies for a.

These dependencies give rise to the following lexical items.

\(\textsf{there}\mathrel{::}e.k\)	\(\textsf{Pst}\mathrel{::}\underline{\bullet x}.{k\bullet}.s\)	\(\textsf{X}\mathrel{::}\underline{\bullet v}.{k\bullet}.{e\bullet}.x\)
\(\textsf{arrive}\mathrel{::}\bullet d.v\)	\(\textsf{a}\mathrel{::}\bullet n.d.k\)	\(\textsf{man}\mathrel{::}n\)

We see that the X lexical item takes a vP, and merges and expletive there, checking the case of the subject in the vP while so doing. It is something like a little-v head. (But of a different kind than those which we have seen previously.) This head provides a locus for the interchange of information between words. The agreement between tensed verb and noun can be mediated over many paths:

1. man – a – arrive – X – Pst
2. man – a – X – Pst
3. man – a – X – there – Pst

All these paths go through X on their way to Pst. Of these three paths, only the third one (making use of there) goes from X to Pst via movement. Thus it is the only one whose length does not increase as the distance between Pst and X grows, as in, again, raising constructions.

The associate internal account

We return to our basic dependencies (figure 1), adding this time a dependency between there and the determiner, as in figure 7. Note that we do not introduce a case checking dependency between the determiner and a tensed head. This will become the associate internal account of there-insertion, a la Basilico 1997.

This time the determiner is more complex than usual, and so we decompose it (after the noun dependency) so as to simplify the morass of dependencies it enters into, as shown below.

Another decomposition of the tensed verb into T and V heads leads to our final dependency structure, the syntactic features underlying which can be reconstructed as per the below.

\(\textsf{there}\mathrel{::}e.k\)	\(\textsf{Pst}\mathrel{::}\underline{\bullet v}.{k\bullet}.s\)	\(\textsf{X}\mathrel{::}\underline{\bullet D}.{e\bullet}.d\)
\(\textsf{arrive}\mathrel{::}\bullet d.v\)	\(\textsf{a}\mathrel{::}\bullet n.D\)	\(\textsf{man}\mathrel{::}n\)

A main problem faced by any associate-internal approach to there-insertion might be called the case-transmission problem, after Chomsky (1986). Here the case requirements of the DP must be satisfied vicariously by there. In the present context, this may be rephrased as: how does the \(k\) feature of the associate DP get transferred to the expletive? Our analysis here does this by divorcing the \(k\) feature from the lexical entry of the determiner. What we have been viewing as determiners are then decomposed throughout the lexicon into big-D and little-d pairs, with the lexical head big-D (features: \(\bullet n.D\)) selecting the nominal complement, and the functional head little-d (features: \(\bullet D.d.k\)) introducing the case requirement. Our associate internal analysis in effect postulates that there is an expletive selecting little-d head.

There are again two basic routes whereby information can pass between N and T.

1. man – a – X – arrive – Pst
2. man – a – X – there – Pst

Again, only the latter, via there, stays constant in length regardless of the distance between DP and T.

---

1. Some people argue that we should appeal to movement in these constructions. Analyzing such constructions using movement seems insightful in some cases. When I am feeling uncharitable, I think that this is a case of ‘when you have a hammer everything looks like a nail.’ When I am feeling charitable, I marvel at the subtle details that this kind of approach reveals. ↩︎