Fragment answers and movement

Abstract

The nature of fragment answers has been under debate for the past 40 years. Most of the arguments have focused on the mobility and island-(in)sensitivity of the fragments. This paper offers a new empirical domain of investigation: interpretative differences between fragment answers and their full sentence counterparts. I present data regarding an interpretation of superlative expressions that is available only with overt movement but not covert movement, and show that fragment answers allow the reading while their full sentence counterparts do not. Thus I argue that in some cases fragment answers must involve movement in the narrow syntax. Approaches to fragment answers that exclusively involve PF movement of the fragments or in situ fragment answers are challenged.

1 Introduction

The nature of fragment answers as in (1a) and their relation with the full sentence counterparts as in (1b) have been discussed in much research (see Progovac 2013 and Weir 2014 for overviews). Approaches vary whether movement in the narrow syntax is involved or not. Most of the argumentations have focused on whether fragments are sensitive to the constraints that movement in general is sensitive to e.g.: island sensitivity, P-stranding generalization, etc. The interpretative differences between fragment answers and full answers have received much less attention, which this paper takes on to tease the approaches apart.

• (1) What did John eat?

  • a. Apples.

  • b. John ate apples.

The interpretative differences to be discussed in this paper involve superlative constructions. Pancheva and Tomaszewicz (2012) discuss a particular reading in superlative expressions like the best picture of John. Shen (2015) and Tomaszewicz (2015) observe that this reading is only available in English when the focus (John in the best picture of John) is overtly moved out of the NP but not when the movement is covert. This paper shows that this reading is only available in fragment answers but not full answers, thus arguing that the movement involved in fragment answers occurs in the narrow syntax, contra to recent PF movement and in situ approaches.

To make the core contrast more concrete, Szabolcsi (1986) observes that the question in (2) has the following interpretation: ‘Who did you take a better picture of than you took of anyone else?’ Crucially, Shen (2014, 2015) observes that only the fragment answer in (2a) but not the full answer in (2b) can answer the question with the interpretation specified above, i.e. only (2a) but not (2b) can mean ‘the best photo that I took was of John’.

• (2) Who did you take the best picture of?

  • a. John.

  • b. I took the best picture of John.

Section 2 briefly lays out major approaches to fragment answers and their predictions on the interpretations. Section 3 draws on previous research to present an interpretation of superlative expressions like the best picture of John and its asymmetric availability regarding overt and covert movements. Section 4 shows that this reading is only found in fragment answers but not full answers and argues that in some cases movement in narrow syntax must be involved in fragment answers. Section 5 concludes.

2 Approaches to fragment answers

Here I examine three representative approaches to fragment answers, all of which assume that fragment answers underlyingly involve the structure of a full answer.¹ Note that each of these approaches involves intricate variants and arguments which deserve much more discussion than this paper can afford. Instead of trying to lay out a comprehensive picture of the ongoing debate, I sample one recent variant of each approach relevant to the current discussion without committing to their specific implementations or arguments.

2.1 Movement in narrow syntax approach

Merchant (2004) argues that fragment answers involve two operations: movement of the fragment to Spec,FP in the narrow syntax (3b) and deletion of TP in PF (3c).

• (3)

  

Merchant supports this approach by showing that the immobile elements are not licit as fragment answers: embedded CPs that cannot move without their complementizers in English also cannot be fragment answers; weak pronominals that cannot undergo movement in Greek, French, German, and Dutch cannot be fragment answers; in non-P-stranding languages (German, Greek, Yiddish a.o.), fragment answers cannot leave the preposition behind; fragment answers are island sensitive, etc.

2.2 PF movement approach

Weir (2014, 2015) argues that the movement involved in fragment answers is semantically vacuous: NPIs in fragment answers are grammatical in (4a), while NPIs are not licensed when moved over the negation in clefts in (4b) (cf. Merchant 2004 and see Dikken et al. 2000 and Valmala 2007 for the original observations).

• (4)

  

Maintaining the robust correlation between movement and fragment answers discussed by Merchant, Weir proposes that the movement of fragments does not occur in the narrow syntax, but involves an exceptional movement of focus in the PF branch. In (5), while the narrow syntax and the LF of the fragment answer remain unchanged, in the PF any mystery novels moves and the rest of the sentence is deleted. Crucially this PF movement has no interpretative effect.

• (5)

  

2.3 In situ approach

In the in situ approach only one operation is involved in fragment answers: PF deletion. Some of the early implementations can be found in Morgan (1973), Hankamer (1979), Napoli (1982) and this approach is argued for more recently in Bruening (2015), Abe (2016), and Ott and Struckmeier (2016) among many others. As is shown in (6b), the focus stays in situ and the non-focus materials are deleted in the PF.

• (6)

  • a. What did John eat?

  • b. He ate apples.  Narrow Syntax

    He ate apples.  PF

    He ate apples.  LF

2.4 Predictions on interpretation

As is shown above, the debate over the nature of fragment answers largely focuses on whether elements that cannot undergo movements can be fragment answers or not. However, these different approaches also make distinct predictions regarding interpretations of fragment answers. I argue that the interpretative differences between fragment answers and their full sentence counterparts can tease these different approaches apart and locate the movement involved.

The movement-in-narrow-syntax approach argues that the fragments undergo movement in the narrow syntax while the PF movement approach and the in situ approach do not. As a result, if an interpretation is only licensed by overt movement and not covert movement, the movement-in-narrow-syntax approach would predict such an interpretation to be available in fragment answers but not in their full sentence counterparts. The in situ approach on the other hand predicts that this interpretation is not available in either fragment answers or their full sentence counterpart since no movement is involved in fragment answers. Since PF movement in principle cannot affect interpretations, the PF movement approach also predicts this interpretation to be impossible in both full and fragment answers.²

In the next section, I provide a case where fragment answers have an additional interpretation that is only licensed in overt movement, which indicates that at least the fragments in these contexts have undergone overt syntactic movement.

3 The third reading of the largest photo of Ben

In the literature on interpretations of superlatives, the absolute reading (Abs) and the relative reading with NP external focus (Rex) have been discussed by Szabolcsi (1986), Heim (1999), Farkas and Kiss (2000), Sharvit and Stateva (2002) among many others. Recently Pancheva and Tomaszewicz (2012) observe a third reading in Polish and Bulgarian which involves NP internal focus (Rin). The paraphrases of the readings and their availability are shown in Table 1. The Rin is not available in the English sentence Sally took the largest photo of Ben while it is available in the Polish counterpart sentence. This paper focuses on the Rin in English. Scenario 1 in Fig. 1 shows a situation where the Rin is true and other readings are not.³

Fig. 1

Scenario 1

Table 1 Superlative interpretations

The canonical sentence in English in (7) is judged as false in Scenario 1. The polar question in (8) also does not allow the Rin.⁴

• (7) #Sally took the largest photo of Ben.

• (8) Did Sally take the largest photo of Ben?

  # Yes.

However as Shen (2015) and Tomaszewicz (2015) observe, when the focus element for the Rin is overtly moved out of the superlative NP in clefts, pseudoclefts, relative clauses, and free relatives as in (9), the Rin becomes available.⁵

• (9) Overt movement of the focus → Rin available:⁶

  • a. It was Ben that Sally took the largest photo of. (cleft)

  • b. Who Sally took the largest photo of was Ben. (psuedo-cleft)

  • c. The boy that Sally took the largest photo of was Ben. (relative clause)

  • d. I like who Sally took the largest photo of. (free relative)

Shen (2015) and Tomaszewicz (2015) also note that covert movement does not make the Rin available. Note that if the covert movement of the focus could make the Rin available, (7) should be able to have the Rin, contrary to the fact. The same point is shown in the case of QR. Scenario 2 in Table 2 presents a situation where the absolute reading is false and the Rin is true: the largest photo that each photographer took was of a different fruit (John: an apple, Bill: a banana, Mary: a pear), but no one took the largest photos of every fruit. In this scenario, the sentence in (10) with the ∀>∃ scope cannot be judged true. This indicates that covert movement of the focus every fruit out of the NP cannot facilitate the Rin.

• (10) A (different) photographer took the largest photo of every fruit.

  # Rin + ∀>∃: ‘One photographer’s largest photo is a photo of a pear;

  a different photographer’s largest photo is a photo of an apple;

  a third photographer’s largest photo is a photo of a banana.’

Based on the evidence presented above, we reach a generalization regarding the Rin in English: the Rin is available when the focus is overtly moved out of the NP but covert movement of the focus does not facilitate the reading.

Table 2 Scenario 2—John, Bill, and Mary took photos of an apple, a pear, and a banana. The photos are of different sizes as indicated in the table

Before ending this section, I address the question why such a generalization exists. Shen (2015) provides a syntactic account for the contrast between (9) and (7)/(8)/(10). And Tomaszewicz (2015) provides a semantic account for the Rin observed in wh-questions and clefts. Note that the particular account for the generalization is not directly relevant for the argumentation of the present paper. What’s important is the existence of the generalization itself, which the literature agrees on. Here I briefly outline the analysis in Shen (2015).

Following Pancheva and Tomaszewicz (2012), the LF structure of the Rin in (11) requires i. movements of the focus Ben and the Degree Phrase (DegP); ii. that the focus takes scope over the DegP; iii. that the movement of focus precedes the movement of the DegP. See Pancheva and Tomaszewicz (2012) for the motivations for such requirements.

• (11) [Ben₂ [[$_{DegP}$ EST-C]₁ [[1.d [~S [$_{TP2}$ 2,e [$_{TP1}$ Sally [$_{VP}$ took [$_{DP}$ the t₁ large photo [t₂]]]]]]]]]]

When multiple movements occur, the order of these movements and their landing sites are known to be restricted. Richards (2001) shows that cross-linguistically the movement of multiple elements preserves the hierarchical relation of their base-generated positions in multiple wh-movement, scrambling, object shift, and clitic clustering. Richards derives this constraint with an economy principle: Shortest. According to the Shortest principle, the element more local to the attracting head (the higher element) moves first, then the less local element (the lower element) moves and tucks in under the landing site of the higher element, creating a crossing path. Chomsky and Lasnik (1993) use similar economy principles to account for the superiority effects and Bruening (2001) the scope freezing effects in QR.

Shen (2015) assumes that the movement of focus and the DegP also observe this economy principle. Since the base-generated position of the focus Ben is lower than that of the DegP, the movement of the DegP should precede that of the focus, and the focus would then tuck in under the landing site of the DegP. The resulting LF in (12) would not generate the Rin. This explains why the Rin is unavailable when both the focus and the DegP move covertly.

• (12) [[[$_{DegP}$ EST-C]₁ [[Ben₂] [$_{TP2}$ 2,e [1.d [~S [$_{TP1}$ Sally [$_{VP}$ took [$_{DP}$ the t₁ large photo [t₂]]]]]]]]]]

Note that the cases mentioned in Richards (2001) involve only overt movement and the cases discussed in Bruening (2001) involve only covert movement such as QR. Shen (2015) argues that when multiple movement involves one overt movement and one covert movement, the effect of the economy principle is lifted, because the overt movement occurring in the narrow syntax always precedes the covert movement occurring in the LF. In the case of (13), the focus Ben moves in the narrow syntax, then in the LF the DegP moves and tucks in under the landing site of the focus. The LF generated satisfies requirements of the Rin. This explains why the overt movement of the focus facilitates the Rin.

• (13)

  • a. [Ben ₂ [~S [$_{TP2}$ 2,e [$_{TP1}$ Sally [$_{VP}$ took [$_{DP}$ the [$_{DegP}$ EST-C]₁ large photo [t₂]]]]]]]  narrow syntax

  • b. [Ben₂ [[$_{DegP}$ EST-C]₁ [[1.d [~S [$_{TP2}$ 2,e [$_{TP1}$ Sally [$_{VP}$ took [$_{DP}$ the t₁ large photo [t₂]]]]]]]]]]  LF

4 Rin in fragment answers

Returning to fragment answers, to the best of my knowledge, Szabolcsi (1986) was the first to show that wh-questions in English allow the Rin. This is predicted by the generalization mentioned above since wh-questions involve overt movement. Shen (2014, 2015) observes that in Scenario 1 where the Rin is true, the wh-question in (14a) can be truthfully answered with the fragment answer in (14b), but not with its full answer counterpart in (14c). This indicates that the fragment answer allows the Rin while its full answer counterpart does not.

• (14)

  • a. Who did Sally take the largest photo of?

  • b. Ben.

  • c. #Sally took the largest photo of Ben.

The movement-in-narrow-syntax approach in Merchant (2004) predicts this interpretative difference between fragment answers and their full answer counterparts. Since overt movement of the focus is a necessary condition for the Rin in English as we have seen in (9), the availability of the Rin in the fragment answer indicates that it also involves overt movement.

In the PF movement and in situ approaches, an interpretative difference like this can only be accounted for by assuming that the focus undergoes an independent covert movement in fragment answers. But as is already shown, covert focus movement do not license the Rin. If any covert movement did occur in the fragment answer in (14b) to make the Rin available, it is unclear why this operation is not available in the full answer in (14c). Thus the data here pose a challenge for the PF movement and in situ approaches and argue that movement in narrow syntax must occur in some cases of fragment answers.

The contrastive fragment in (15) does not allow the Rin and this is predicted. Given the question-answer congruence conditions (Reich 2007; AnderBois 2011; Barros 2014; Weir 2014 a.o.), the answer only allows the Rin when the question also allows it. Since the polar question in (15) does not allow the Rin, the fragment cannot allow the Rin either.⁷

• (15) Did Sally take the largest photo of Sue?

  #No, Ben.

In the discussion so far, I have been assuming a syntactic identity theory of ellipsis where the ellipsis site in fragment answers is formally identical to the corresponding part in the wh-question.⁸ The facts regarding the Rin discussed above show that the in situ and PF movement approaches to fragment answers cannot be maintained under a syntactic identity theory.

It is worth pointing out that the nature of identity conditions of ellipsis is under debate. Many have suggested weakening the syntactic identity by arguing for either a semantic identity theory or a hybrid of the two.⁹ These approaches suggest that the fragment answers could be derived from sources that are not syntactically identical to the wh-question. Particularly for the current paper, it means that the fragment answer ‘Ben.’, as an answer to ‘Who did Sally take the largest photo of?’, could be derived from sentences other than ‘Sally took the largest photo of Ben.’ I label these other sentences ‘alternative sources.’ The next subsection addresses several potential alternative sources that in-situ or PF movement approach may use to account for the Rin in fragment answers and show that the these alternative sources cannot account for the full range of data without additional assumptions.

4.1 Not alternative sources

One potential way to derive the interpretative difference in (14) without assuming overt movement as a part of fragment answers is to assume that the fragment answer with the Rin is not derived from a canonical full answer as in (16) but from a sentence that already allows the Rin. In (17), fragment answers start with sentences where the focus has undergone overt movement thus the Rin is allowed. The PF deletion occurs to the sentences while sparing the fragment Ben. Since the full sentence answers have the Rin before the deletion and only PF operations are involved in deriving these fragment answers, the Rin should be available in the fragment answers as well.

• (16) Sally took the largest photo of Ben.

• (17) Alternative sources: cleft sentences

  • a. It was Ben that Sally took the largest photo of.

  • b. Who Sally took the largest photo of was Ben.

I argue against the alternative sources by showing that in places where the potential sources in (17) are not possible, fragment answers with the Rin are still available. Note that the alternative sources in (17) involve cleft or pseudo-cleft constructions. A well-known restriction on cleft constructions is the existential presupposition which requires that the property denoted by the cleft sentence is true of some individual. This is shown by the incompatibility of negative quantifiers and cleft sentences in (18) (data from Reeve 2011, see also Postal 1993 and Merchant 2004).

• (18)

  • a. * It was nothing that he drank. (ex. (18a) in Reeve 2011)

  • b. * What he drank was nothing. (modified from ex. (18b) in Reeve 2011)

On the other hand, fragment answers do not show the existential presupposition as is shown in (19) (Merchant 2001).

• (19) Who did you meet?

  No one.

If fragment answers with the Rin are derived from cleft sentences, it is predicted that negative quantifiers like nothing or no one cannot function as fragment answers in the relevant scenario. I will show that this prediction is not borne out: negative quantifiers can function as fragment answers with the Rin.

It is less straight forward to come up with a simple example like (20). Since the Rin involves locating the person depicted in the largest photo that Sally took, as long as Sally took two or more photos of different sizes, the negative answer no one in (20) would be false.¹⁰

• (20)

  • a. Who did Sally take the largest photo of?

  • b. No one.

To circumvent this caveat, I use who else questions in (21c) in a situation where Sally took one unique largest photo which is of Ben as in Scenario 1 in Fig. 1 (inspired by Merchant 2001). The who else question in (21) is more felicitous in scenarios where the Maxim of Quantity is suspended, e.g. in a game show scenario where the blindfolded contestant Bob tries to figure out the situation by asking the host questions. The host must answer truthfully but does not have to be as informative as possible (inspired by Fox 2014). Here the host can truthfully answer the who else question with a negative quantifier. Compare the fragment answer in (21d) and the unacceptable alternative sources in (22), we can conclude that the fragment answer with the Rin in (21d) is not derived from the alternative sources in (22). This suggests that even though the syntactic identity theory is weakened to include alternative sources like cleft/pseudo-cleft sentences, fragment answers with the Rin are still required to involve overt movement.

• (21)

  • a. Bob: Who did Sally take the largest photo of?

  • b. Host: Ben.

  • c. Bob: Who else did Sally take the largest photo of?

  • d. Host: No one.

• (22)

  • a. *It was no one (else) that Sally (also) took the largest photo of.

  • b. *Who Sally (also) took the largest photo of was no one (else).

Two NLLT reviewers pointed out that a predicational clause paraphrasing the Rin in (23) could be another potential source for fragment answers with the Rin. If the full answer starts as (23a) and the focus Ben stays in situ, PF deletion of the rest of the sentence as in (23b) would derive a fragment answer with the Rin.¹¹

• (23) Alternative source: predicational clause

  • a. The largest photo that Sally took was of Ben.

  • b. The largest photo that Sally took was of Ben.  PF deletion

The negative quantifier argument specified above also rules out predicational clauses as the alternative source. If the negative quantifier in (21) were a result of deletion based on a predicational clause like (23), the full answer would have to be (24), which is clearly false according to the scenario (if grammatical at all), unlike the fragment answer in (21).

• (24) #The (other) largest photo that Sally (also) took was of no one.

A final empirical argument against alternative sources comes from reciprocals such as each other. In a scenario where the largest photo that Sally took is a photo of Ben and the largest photo that Ben took is a photo of Sally (see again Scenario 1 in Fig. 1), the fragment each other can be used as a true answer to the question in (25). The reciprocal in the pivot position in cleft sentences in (25b,c) are not acceptable. The predicational sentence in (25d) only allows the interpretation where the largest photo that Sally and Ben collectively took is a photo of themselves together. Thus sentences in (25b–d) cannot be the sources for the fragment answer with the Rin in (25a).¹²

• (25) Who did Sally and Ben take the largest photo of?

  • a. Each other.

  • b. *It was each other that Sally and Ben took the largest photo of.

  • c. *Who Sally and Ben took the largest photo of was each other.

  • d. #The largest photo that Sally and Ben took was of each other.

To conclude, if ellipsis involves syntactic identity, only a movement-in-narrow-syntax approach to fragment answers accounts for the difference in the availability of the Rin reading in fragment answers vs. full sentence answers. While a weakened identity approach involving clefts or predicational sentences can account for some of the Rin cases in fragment answers, such an approach is incompatible with negative fragments or reciprocal fragments. If one loosens the syntactic connection between a fragment answer and the antecedent question even further, it is conceivable that there are source sentences for the fragment which may allow the Rin. Although the current paper cannot exclude that, the lack of restrictiveness of such accounts could be considered problematic for theoretical reasons. As mentioned by a reviewer, the nature and degree of identity in ellipsis is still controversial (see Lipták 2015 for an overview), and this paper is not the place to resolve these issues. What remains, however, is that the distribution of the Rin follows straightforwardly under an syntactic movement account as suggested in Merchant (2004), together with an ellipsis account involving parallel structures of the fragment and the antecedent.

5 Conclusion

This paper utilizes interpretations of superlative expressions as a method to tease apart various approaches to fragment answers. In particular the data from the relative reading with NP internal focus argues against approaches to fragment answers that exclusively involve PF movement or the in situ fragments. Instead, I have provided evidence that at least certain fragment answers involve a derivation with overt syntactic movement.¹³

The superlative facts reported here on top of those from the existing literature suggest an approach with (at least) two mechanisms behind fragment answers: in situ + deletion and narrow syntax movement + deletion. Fragment answers involving immoveable elements and island-insensibility are derived via PF deletion with fragments staying in situ while in superlative cases (maybe among others) fragments move in the syntax before the PF deletion. This paper also offers a promising empirical domain: the interpretative effects of fragment answers. More investigations on interpretative differences between fragments and full sentences might provide novel insights into this topic.