Attributive Comparatives in German

As I have already shown in section 2.4, German is often treated on a par with English in terms of its comparative semantics (cf. Beck et al., 2009; Bhatt & Takahashi, 2011a). However, empirical support for the assumption that German is a language that only has the clausal operator in (2) comes from results of an acquisitional corpus study by Hohaus et al. (2014) already discussed in section 2.4.2 of Chapter 2.

For the current series of experiment, I will focus on German and will assume, accord- ing to the facts presented in section 2.4, that German has only the clausal comparative

operator at its disposal. This means that we discard a Phrasal Analysis for German. We thus do not need to worry about which phrasal operator to use (from Heim 1985 or Kennedy 1997), nor do we need to discuss the implications of such an analysis (which might be simpler than the clausal one) on the processing predictions2_.

Now, acting on the assumption that German only has the clausal operator in (2), and assuming the relational gradable adjective in (4), we arrive at the LF in (5) for the internal derivation and in (6) for the external derivation of our example in (1). One of the movements that takes place is the QR of the indefinite DP (indicated by an arrow in both derivations) from the object position, which is necessary in both cases. The other type of movement that is also present in both trees is wh-movement of a covert how (which can be overt in some languages, like e.g. Russian, cf. Pancheva 2006; Berezovskaya 2014) to create degree abstraction. I also include a bigger version of these trees in the Appendix A on page 188 for better readability.

(5) LF for Int of (1) Maria traf einen besseren Boxer als Peter.

2_{Repeatedly, I refer the interested reader to Schneider (2017) who investigates the processing of English}

(6) LF for Ext of (1) Maria traf einen besseren Boxer als Peter.

Additional structural complexity in (5) comes from internal subjects in DPs (leaving the trace t3) in (5) (cf. Heim & Kratzer 1998, 226-228) . However, since I am assuming this

for the derivation of Int, it is also implicitly assumed for the external derivation, (6), I just do not indicate it there to not complicate the structure unnecessarily. Since this structural assumption applies to both readings, I will be ignoring it forthwith, because it does not affect any of my arguments, predictions and conclusions.

Note that in the derivation of Int there is a small clause (SC) contained in the DP. This small clause is truth-value denoting and provides the right type for degree abstraction to apply and to subsequently supply comp(clausal) with the right type of

argument, namely the characteristic function of the set of degrees (type hd, ti). Another remark is that German is a head-final language when it comes to VPs, that is why the verb comes last in the LFs provided.

The resulting truth conditions for the internal reading are given in (7), for the external reading in (8).

(7) Int:

∃x[meet(M aria, x)]&Max(λd.boxer(x)&µquality(x) ≥ d) >

(8) Ext:

Max(λd.∃x[boxer(x)&meet(x, M aria)&µquality(x) ≥ d]) >

Max(λd0.∃y[boxer(y)&meet(y, P eter)&µquality(y) ≥ d0])

As already discussed for a different example on pp. 35 ff. in Chapter 2 when introducing different phrasal operators, the position of max taking scope over ∃ in both the matrix and the standard clause actually yields the reading where we compare the strongest boxer that Mary met to the strongest boxer that Peter met under Ext. Under int, we compare the maximal degree of strength of an individual whom Maria met and who is a boxer with the maximal degree of strength of Peter who is a boxer. While this is one possible reading that people get for the sentence, it is not the only one, as has already been discussed for computer-example in (14) in Chapter 2. The following reading might also be existent for Ext:

(9) ∃x, y[Max(λd.boxer(x)&meet(Maria, x)&µquality(x) ≥ d) >

Max(λd0.boxer(y)&meet(P eter, y)&µquality(y) ≥ d0)]

This is a reading can be paraphrased as follows: “There is a (specific) boxer Mary met who is stronger than a (specific) boxer Peter met.”. This reading would also be true in a scenario where Maria met a strong boxer and Peter met a weak and a super strong boxer. Importantly, I believe that this complication on the external reading does not change the facts about the differences between the two Logical Forms in (5) for Int vs. (6) for Ext and that the analysis of the indefinite is not central to the ambiguities investigated here. For this reason, I will set this issue aside and concentrate on the differences that are relevant for my predictions in processing.

As already explained, QR of the object indefinite and the wh-movement is needed in both readings. What differs is the QR of the DegP: It applies in the semantic derivations of both readings, but crosses more nodes in the external case. The derivation of the internal reading is less complex on several levels: Firstly, the DegP moves across a sentence boundary (CP) in the external case, but only across a DP-boundary under the internal reading. Hackl et al. (2012) show that longer QR of individual-type quantifiers in object position is costlier than shorter QR. This is the crucial parallel between Hackl et al.’s and the present case. I will elaborate on this in the next subsection. Secondly, the LF of the internal reading has less nodes in total (namely 35 nodes for Int vs. 39 nodes for Ext). In short, the structure that corresponds to Int is less complex in two

senses: it has a shorter QR that happens within the DP and it is the smaller structure overall.3