Abstract

I argue that the version of phrase structure theory proposed by Donati and Cecchetto (2011) falls short of accounting for the attested patterns of free relative clauses not only in English but crosslinguistically in general. In particular, I show that free relative clauses can be introduced not only by wh-words like what or where, which is what Donati and Cecchetto predict, but also by wh-phrases like what books or whatever books and their equivalents in other languages, which Donati and Cecchetto explicitly predict not to be possible.

Donati and Cecchetto (D&C) (2011) propose a version of phrase structure theory according to which a lexical item that is “internally merged” (i.e., moved) “can turn a clause into a nominal phrase” (p. 519). In other words, they argue that “there is a type of movement, head movement, which has the property of relabeling the structure it merges with” (p. 552). In particular, if the internally merged “lexical item is a wh-word, a free relative results; if it is an N, a full relative results; if it is a non-wh D, a pseudorelative results” (p. 519).1 Although most of D&C’s examples and arguments are from English, their proposal aims at generality. In this reply, I focus exclusively on how their proposal falls short of accounting for the attested patterns of free relative clauses (henceforth, FRs) in English and crosslinguistically. In particular, I show that FRs can be introduced not only by wh-words like what or where, which is what D&C predict, but also by wh-phrases like what books or whatever books, which D&C explicitly predict not to be possible.

This reply is structured as follows. In section 1, I briefly introduce FRs. In section 2, I sketch D&C’s analysis. In section 3, I present data and arguments against their prediction that FRs cannot be introduced by wh-phrases, showing that this option is available in English and other languages as well. Section 4 concludes.

1 Introducing Free Relative Clauses

FRs are noninterrogative wh-clauses whose distribution resembles that of constituents like nominal (D), prepositional (P), adjectival (Adj), and adverbial (Adv) phrases. The (a) examples of (1)–(5) illustrate FRs, while the (b) examples provide rough paraphrases using DPs, PPs, AdjPs, and AdvPs.

(1)

  • a.

    He read [what Luca read].

  • b.

    He read [DP the stuff Luca read].

(2)

  • a.

    I worked [when the kids were playing].

  • b.

    I worked [PP during the time the kids were playing].

(3)

  • a.

    Sleep [wherever you find a bed].

  • b.

    Sleep [PP in any place you find a bed].

(4)

  • a.

    He can be [however late he wants].

  • b.

    He can be [AdjP as late as he wants].

(5)

  • a.

    I can drive [however fast you can drive].

  • b.

    I can drive [AdvP as fast as you can drive].

The FRs in (1) and (2) are introduced by plain wh-words, that is, wh-words without any morphological enrichment that form a full constituent by themselves without any further material (what, when). I call these FRs plain FRs. The FRs in (3)–(5) are introduced by a wh-word (wherever) or a wh-phrase (however late, however fast) with the suffix -ever. I use the label wh-phrases to refer to phrases containing a wh-word and other lexical material, and the label wh-expressions to refer to both wh-words and wh-phrases. I follow the common usage and call FRs introduced by -ever wh-expressions -ever FRs. As their labels suggest, the bracketed wh-clauses in (1)–(5) have all been assumed to be instances of the same construction—FRs—from as early as Jespersen’s (1909–1949) first description, Bresnan and Grimshaw’s (1978) first syntactic analysis, and Jacobson’s (1995) and Dayal’s (1997) seminal semantic analysis.2 D&C radically depart from this unifying approach, as I summarize in the next section.

2 Donati and Cecchetto (2011) on Free Relative Clauses

It is well-known that a wh-clause like what Luca read is syntactically and semantically ambiguous: it can be an FR and refer to the things Luca read, if embedded in a sentence like (1a), or it can be an embedded wh-interrogative clause and refer to the question about the identity of the things Luca read in a sentence like I wonder what Luca read.

D&C (2011) argue that FRs and wh-interrogative clauses result from two different syntactic processes. The wh-word what is assumed to be a head D in both cases, and in both cases it is internally merged above the head C via head movement. The crucial difference is in the labeling process of the smallest constituent dominating both what and C. This constituent can inherit the D features of what and be a DP, that is, an FR, as shown in (6). The same constituent can instead inherit the features of C and be a CP, that is, an interrogative clause, as shown in (7).3

(6)

graphic

(7)

graphic

D&C also claim that a wh-clause introduced by a DP wh-phrase like what books Luca read in a sentence such as He read what books Luca read can only be a wh-interrogative clause (CP), as shown in (8). The wh-phrase what books has been internally merged above the head via phrasal movement. Therefore, only the head C can transmit its feature to the constituent dominating both C and the DP wh-phrase, producing a CP. This is the case because D&C’s proposal is built on the crucial assumption that DPs (or any other phrases) cannot transmit their syntactic features or labels—only heads can.

(8)

graphic

Finally, D&C (2011:552–555) argue that -ever FRs are not FRs syntactically, but rather headed relative clauses. They sketch an analysis for -ever FRs along the lines of the one in (9) for the -ever FR whatever books Luca reads in He reads whatever books Luca reads (see D&C 2011:554, (111), and related discussion).

(9)

graphic

The nominal books in (9) is base-generated as the object in the relative clause and then internally merged above C via head movement. The resulting constituent is an NP since it inherits the features of N. The NP combines with its D sister whatever via external Merge; that is, whatever does not undergo wh-movement (or any other movement), but is base-generated as the sister of the NP. This is the same analysis D&C argue for with regard to headed relative clauses (see D&C 2011:528, (22b)).

3 Counterexamples to Donati and Cecchetto’s (2011) Prediction about Free Relative Clauses

D&C’s (2011) proposal makes a strong prediction: unlike wh-words, wh-phrases can never introduce FRs. There is compelling evidence that this prediction is not borne out and that FRs can indeed be introduced by wh-phrases. I first provide clear examples of plain FRs that are introduced by wh-phrases in English and other languages (section 3.1); I then show that the arguments that D&C provide to support their claim that -ever FRs are not FRs syntactically do not warrant their conclusion (section 3.2).

3.1 Plain FRs Can Be Introduced by wh-Phrases

D&C exemplify their discussion of plain FRs with English data, although their proposal aims at generality. English alone is enough to provide counterexamples. In varieties of American English, the wh-phrases what + NP and how much + NP can introduce plain FRs, as shown in (10) and (11).5

(10) He read [what books she read].

(11) I drank [how much wine you drank].

D&C do not mention data like (11) in which a plain FR is introduced by the wh-phrase how much + NP. They do make reference to examples like (10) with a plain FR introduced by what + NP. They suggest that what is actually whatever with a silent -ever (see their example (117) and related discussion). Therefore, wh-clauses like the one in (10) are not plain FRs, but -ever FRs, which for D&C are not FRs at all. In section 3.2, I discuss and refute the claim that -ever FRs are not FRs. Here, I show that (10) is a plain FR introduced by a plain wh-phrase, rather than a (covert) -ever wh-phrase. I use Baker’s (1995:216) and Dayal’s (1997:109, (29a–b)) namely test for -ever FRs: while a plain FR (like the definite description that paraphrases it) can be followed by namely and the list of the individuals in the set that the FR is associated to (12), an -ever FR cannot (13).

(12) He read [what she read], namely, Lolita and A Clockwork Orange.

(13) *He read [whatever she read], namely, Lolita and A Clockwork Orange.

Both (10) and (11) pass the namely test, as shown in (14) and (15), respectively. Their -ever counterparts do not, however, as shown in (16) and (17). It follows that the wh-words in (10) and (11) are plain wh-words rather than -ever wh-words. Therefore, plain wh-phrases can introduce FRs, at least in these varieties of American English.6

(14) He read [what books she read], namely, Lolita and A Clockwork Orange.

(15) I drank [how much wine you drank], namely, two glasses.

(16) *He read [whatever books she read], namely, Lolita and A Clockwork Orange.

(17) *I drank [however much wine you drank], namely, two glasses.

I leave a more careful investigation of which plain wh-phrases can introduce plain FRs in which varieties of English for future research. Regardless, the examples above suffice to refute the prediction of D&C’s proposal.7

The same pattern is attested crosslinguistically. In Romanian, a Romance language, all complex wh-phrases can introduce plain FRs.8

(18)

  • Am citit [ce carte/ce cărţi ai citit şi tu].

  • have.1SG read what book/what books have.2SG read also you

  • ‘I read what book(s) you read.’

(19)

  • Am alergat [cât de repede/bine ai alergat şi tu].

  • have.1SG run how.much of fast/well have.2SG run also you

  • ‘I ran as fast/well as you ran.’

(20)

  • Am dormit [câte ore ai dormit şi tu].

  • have.1SG slept how.many hours have.2SG slept also you

  • ‘I slept as many hours as you did.’

Melchor Ocampo Mixtec, an Oto-Manguean language spoken in the Guerrero state of Mexico, exhibits a similar pattern.

(21)

(22)

To sum up, crosslinguistic evidence shows that plain FRs can be introduced by wh-phrases, contra D&C’s prediction.

3.2 -Ever FRs Are Free Relative Clauses

In English, -ever FRs can be introduced by all wh-phrases occurring in wh-interrogative clauses, once enriched with the suffix -ever: whatever/whichever + NP, however + Adj/Adv, however much/many + NP, as shown in (23)–(26).

(23) The internal head and the external head are not part of a movement chain, but are related by [whatever mechanism links an elided constituent and its antecedent in ellipsis cases].9

(24) I can drive [however fast you drive].

(25) He can be [however late he wants].

(26) She can provide [however much financial support is needed].

All the examples in (23)–(26) counter D&C’s prediction that FRs cannot be introduced by wh-phrases, if -ever FRs are indeed FRs syntactically. D&C do not discuss them in the main body of their article, but in an extensive appendix they argue that -ever FRs are not FRs, but some kind of headed relative clause, which they label pseudo free relative (D&C 2011:552) (see (9) above and related discussion). They borrow their conclusion, the label, and five supporting arguments from Battye’s (1989) discussion of the morphosyntactically equivalent construction in Italian. D&C follow Battye in discussing the Italian construction, although they explicitly state that “the analysis for Italian can be extended to English” (D&C 2011:555).

The overall gist of their arguments is that there are differences between -ever FRs and plain FRs showing that they are different syntactic creatures. Below I discuss and refute their five arguments by focusing on English, which is the language that D&C rely on in the main body of their article.

Difference 1. The first difference that D&C discuss is that -ever wh-expressions can have an absolute use; that is, they can be used without introducing a wh-clause (27), while plain wh-expressions cannot (28) (D&C 2011:552–553).

(27) I can eat {whatever is in the fridge}/{whatever}.

(28) I can eat {what is in the fridge}/*{what}.

They interpret this contrast as indicating that -ever wh-expressions are always “quantificational DPs” that are syntactically unrelated to plain wh-expressions: they do not undergo wh-movement and are base-generated in the position in which they are spelled out not only in their absolute use but also, and crucially, in their use in -ever FRs (D&C 2011:555). Historical data do not support this conclusion. In English (and in Italian, Romanian, and Dutch), the absolute use of -ever wh-expressions is attested much later than their use in -ever FRs (see Caponigro and Fălăuş 2018 for examples and references). This would be unexpected if -ever wh-expressions were non-wh quantificational DPs that could optionally be used as heads of headed relative clauses. On the other hand, if they are true wh-expressions, they are expected to introduce wh-clauses. Their absolute use would be a later development in which the -ever wh-phrase dropped its clausal argument.

This later development may have been favored by the semantic properties of -ever wh-phrases, which are crucially different from those of plain wh-expressions (see the namely test above and related discussion). Although the meaning of -ever wh-expressions (and -ever FRs) is an open issue and seems to vary crosslinguistically,10 there is general consensus that an -ever wh-expression behaves like some form of a (possibly modalized) quantifier taking the FR and the matrix clause as its arguments. A plain wh-word in a plain FR, on the other hand, has been argued to be nonquantificational and act as a set restrictor: it applies to the set associated with the FR to return a subset (Jacobson 1995, Dayal 1996, 1997, Caponigro 2003, 2004). An absolute use of an -ever wh-expression with the loss of one of its FR arguments would turn the -ever wh-expression into a monoargumental quantifier—a semantic object for which there is large independent evidence. For instance, whatever in its absolute use would be similar in argument structure to anything or something. On the other hand, what in a hypothetical absolute use could maintain its original meaning as a set restrictor or, maybe, could be simplified to denote a plain set. Regardless, it would not be able to combine with its predicate because of type mismatch. As a set restrictor, it would be roughly similar to an adjective in English without the noun that it restricts: *I read what/interesting. As a plain set, it would behave more or less like a singular count noun in English without a determiner: *I read what/book.

Differences 2 and 3. I discuss the second and third differences together because they touch on the same issue: relative markers. D&C (2011:553) claim that, unlike plain FRs, -ever FRs allow relative markers, just as headed relative clauses do: the complementizer that and relative pronouns. If correct, this pattern would be easily accounted for if -ever FRs were headed relative clauses, rather than FRs. The actual empirical picture looks different, though. D&C provide only two examples with (the Italian morphosyntactic equivalent of) -ever FRs to substantiate their claim. In both cases, the -ever FR is introduced by the (Italian equivalent of the) wh-phrase whatever/whichever + NP and contrasted with a plain FR that is introduced by a wh-word.

Let us start with the complementizer case. (29) shows that the complementizer that can optionally occur between whatever books and the -ever FR that it introduces, while (30) shows that this option is not available for what and the plain FR that it introduces.11

(29) You can read [whatever books (that) are on the table].

(30) You can read [what ’s/*that’s on the table].

However, the behavior of the -ever FR in (29) does not resemble the behavior of a headed relative clause either. A true headed relative clause with a relativized subject like (31) requires the complementizer that in English; it becomes unacceptable without it. (29), instead, is acceptable with or without that.

(31) You can read [any book that’s/*is on the table].

Also, the -ever FR that is syntactically closer to the plain FR in (30)—that is, the -ever FR introduced by the wh-word whatever without an NP—is degraded if followed by the complementizer, as shown in (32).

(32) You can read [whatever is/??that’s on the table].

The examples in (33)–(35) show that sentences with an -ever FR introduced by other wh-words like whoever, wherever, and however disallow the complementizer as well.

(33) I can talk to [whoever is/??that’s on the phone].

(34) He can sleep [wherever (*that) he likes].

(35) I’ll do it [however (*that) you do it].

On the other hand, the corresponding sentences with headed relative clauses replacing -ever FRs exhibit a different pattern, as shown in (36)–(38).

(36) I can talk to [the person that’s/*is on the phone].

(37) He can sleep [in any place (that) he likes].

(38) I’ll do it [in the way (that) you do it].

(36), which has a headed relative clause relativizing a subject, requires the complementizer, while the corresponding sentence with an -ever FR in (33) disallows it. (37) and (38), which have relative clauses relativizing nonsubject constituents, are fully acceptable with or without the complementizer, while the corresponding sentences with -ever FRs in (34) and (35) are unacceptable with the complementizer.

Plain FRs, instead, closely resemble -ever FRs as far as the distribution of the complementizer that is concerned, as shown in (39)–(41).

(39) I can take care of [what ’s/*that’s on the table]. (cf. (33))12

(40) He can sleep [where (*that) he likes]. (cf. (34))

(41) I’ll do it [how (*that) you do it]. (cf. (35))

Interestingly, -ever FRs introduced by wh-phrases, rather than wh-words, allow the complementizer. We already saw an -ever FR introduced by the wh-phrase whatever + NP in (29). (42) shows an example with an -ever FR introduced by the wh-phrase however much + NP.

(42) She can provide [however much financial support is/that’s needed].

At the end of this section, I elaborate more on why -ever FRs introduced by wh-phrases containing a nominal allow an optional complementizer. Still, what is crucial for countering D&C’s claim that -ever FRs are headed relative clauses is that the corresponding sentences with headed relative clauses exhibit a different pattern: they require a complementizer and are unacceptable without one, as shown in (31) and (43).

(43) She can provide [the amount of financial support that’s/*is needed].

Plain FRs introduced by wh-phrases like (10) and (11) differ from both -ever FRs and headed relative clauses in that they do not allow the complementizer at all, as shown in (44) and (45), respectively.

(44) He read [what books (*that) she read].

(45) I drank [how much wine (*that) you drank].

This contrast can be taken as further evidence, in addition to the namely test in section 3.1, against D&C’s claim that FRs like (44) are actually -ever FRs with silent -ever morphemes. It also shows that FRs like (44) and (45) are not headed relative clauses either. At the end of this section, I suggest a way to reconcile this contrast between plain FRs and -ever FRs with the claim that they are both FRs.

Now, let us consider the argument that D&C build using relative pronouns. They point to a contrast like the one between (46) and (47) to argue that all -ever FRs are actually headed relative clauses.

(46) [Whatever books for which she writes reviews] are likely to become bestsellers.13

(47) *[What for which she writes reviews] is likely to become a bestseller.

What (46) shows is that whatever + NP can be followed by a relative pronoun like for which, while the plain wh-word what cannot. Still, the pattern in (46) does not seem to generalize, not even to other instances of whatever + NP, if they are followed by different relative pronouns, as shown in (48)–(52).

(48) I’ll talk to [whatever students (*who) are problematic].

(49) I’ll talk to [whoever (*who) is problematic].

(50) He can sleep [wherever (*where) he likes].

(51) You can do it [however (*in which) you want].

(52) I can handle [however many people (*who) are here for help].

To sum up, -ever FRs disallow relative markers (complementizers or relative pronouns), regardless of the constituent that is relativized. This pattern is close to that of plain FRs, while it contrasts with that of headed relative clauses, which require relative markers when the subject is relativized and optionally allow them in other cases. Therefore, the distribution of relative markers in -ever FRs not only does not support the conclusion that ever-FRs are headed relative clauses—it in fact provides further evidence that -ever FRs are syntactically close to plain FRs.

The only exception seems to be -ever FRs introduced by wh-phrases, like the examples in (29), (42), and (46). The optionality of the complementizer in (29) and (42) may be the result of two different syntactic analyses for the bracketed clause in each example, which in turn may be due to the fact that -ever wh-phrases, unlike plain wh-phrases, allow for an absolute use (see Difference 1 above). When that is present, the bracketed clause is a headed relative clause. The -ever wh-phrase occurs in its absolute use (without an -ever FR) and behaves like the head of a headed relative clause modifying the nominal within the wh-phrase. When that is absent, the bracketed string is a true -ever FR. Its -ever wh-phrase behaves like a true wh-constituent that has been internally merged and prevents a complementizer from cooccurring—a well-known restriction in English.14 -Ever wh-words, instead, are DPs or PPs without further internal structure. In particular, they do not contain any nominal that can be modified by a headed relative clause. As a result, they can only introduce a clause by being internally merged, that is, by forming an -ever FR.

If this suggestion is on the right track, then why is the same dual behavior not observed in -ever wh-phrases containing a nominal that are followed by the relative pronoun who, as in (48) and (52)? I leave this question, together with a more extensive and controlled assessment of the data, to future research. Still, the emerging general picture is clear: -ever FRs do not behave like headed relative clauses as far as relative markers are concerned. The morphosyntactic equivalent of -ever FRs in Italian, with which D&C exemplify their discussion, exhibits the same pattern (see data and discussion in Caponigro and Fălăuş 2018).15

Difference 4. The fourth difference that D&C discuss concerns the fact that -ever wh-clauses that look identical to -ever FRs on the surface are used as clausal adjuncts, as in (53a) and (54a) (D&C 2011:553). Plain FRs do not allow this use, as shown in (53b) and (54b).

(53)

  • a.

    [Whatever you say], I won’t change my mind.

  • b.

    *[What you say], I won’t change my mind.

(54)

  • a.

    [Wherever you go], I’ll be here waiting for you.

  • b.

    *[Where you go], I’ll be here waiting for you.

These -ever wh-clauses exhibit a different syntactic and semantic behavior than -ever FRs (see, e.g., Izvorski 2000, Rawlins 2013). They occur in a dislocated position without playing the usual role of an argument or a standard adjunct of the matrix clause and trigger a concessive interpretation of the whole sentence, which resembles the semantic behavior of clausal adjuncts starting with no matter rather than -ever FRs. D&C take the contrast above as support for their conclusion that -ever FRs are not FRs but headed relative clauses. Still, regular headed relative clauses cannot be used as clausal adjuncts either, as shown in (55) and (56).

(55) *[Any/All the/Those things you say], I won’t change my mind.

(56) *[Anywhere/Everywhere/Those places you go], I’ll be here waiting for you.

In conclusion, clausal-adjunct -ever wh-clauses do not behave like -ever FRs either syntactically or semantically, despite looking identical. Even if a principled explanation of how to derive one from the other is found, it will not support the conclusion that -ever FRs are headed relative clauses because the latter cannot be used as clausal adjuncts.

Difference 5. The last difference that D&C discuss concerns the fact that -ever FRs cannot be infinitival, while D&C claim that plain FRs can. Here, I show that there is no evidence that true plain FRs can be infinitival, while headed relative clauses can be. Therefore, the lack of an infinitival form brings -ever FRs closer to plain FRs and further away from headed relative clauses, contra what D&C argue for. In this case, I need to discuss D&C’s Italian examples since English does not have anything similar. (57) contains an infinitival -ever wh-clause in Italian and the sentence is judged unacceptable (D&C 2011:553, (106); glosses adapted and translation added). (58) is claimed to be an example of an (almost) fully acceptable infinitival plain FR (D&C 2011: 553, (107); glosses and translation adapted).

(57)

  • *Cerco [qualunque studente mandare al mio posto].

  • search.1SG whichever student send.INF to.the my place

  • (‘I am looking for whichever student to send in my place.’)

(58)

  • ?Cerco [quanti mandare al mio posto].

  • search.1SG how.many send.INF to.the my place

  • ‘I am looking for someone to send in my place.’

The bracketed string in (58) is not a plain FR, though; rather, it is what has been called a modal existential wh-construction (Grosu 2004, Šimík 2011) or an existential FR (Caponigro 2003, 2004). Despite differences in their analyses, all these authors agree that this construction is not a DP, but a CP. Šimík (2011) shows convincingly that existential FRs are mainly infinitival across languages, or in the subjunctive form when the infinitival form is not available in the language. Also, existential FRs are more naturally paraphrased with indefinite noun phrases, while plain FRs have been shown to pattern like definite noun phrases. The wh-clause in (58) behaves like an indefinite, as its paraphrase makes clear (‘someone to send in my place’). The set of wh-words that introduce plain FRs is not necessarily the same as the set introducing existential FRs. In Italian, for instance, quando ‘when’ and come ‘how’ can introduce plain FRs only, while di che ‘of what’ can introduce existential FRs only. Although widely attested across languages, existential FRs are not attested in all languages that have plain FRs, English being an example of a language that has one but not the other. On the other hand, the fact that -ever FRs cannot be infinitival does not make them closer to headed relative clauses. Headed relative clauses can be infinitival in Italian, as shown in (59), while I am not aware of any plain FRs in Italian (or in English) that are infinitival.

(59)

  • Cerco [qualche studente da mandare al mio posto].

  • search.1SG some student to send.INF to.the my place

  • ‘I am looking for some students to send in my place.’

In conclusion, the fact that -ever FRs cannot be infinitival does not bring them closer to headed relative clauses. If anything, it makes them resemble plain FRs.

To sum up, the five alleged differences between -ever FRs and plain FRs that D&C present as evidence for their proposal do not hold up under closer scrutiny. They not only do not support D&C’s claim that -ever FRs are headed relative clauses, but in fact further support the standard assumption that -ever FRs are FRs.16

4 Conclusion

D&C’s (2011) proposal on phrase structure theory and (re)labeling through merging makes the strong prediction that languages may have FRs that are introduced by wh-words, but that no language should allow for FRs that are introduced by wh-phrases. I have shown that this prediction is not borne out. Plain FRs can be introduced by wh-phrases, as shown with examples from varieties of English, Romanian, and Melchor Ocampo Mixtec. -Ever FRs are FRs, and can be introduced by wh-phrases as well.

D&C and the literature they rely on may be correct in highlighting an asymmetry in the level of productivity of wh-phrases within FRs: wh-phrases often introduce -ever FRs, while they more rarely introduce plain FRs. At least, this is the pattern found in the relatively small number of languages with both kinds of FRs that have been studied so far. For instance, as shown above, what + NP can introduce plain FRs in some varieties of English, while which + NP cannot introduce plain FRs in any varieties I know of. On the other hand, both whatever + NP and whichever + NP can introduce -ever FRs in all the varieties of English I am familiar with. The same pattern is observed in Romanian and Melchor Ocampo Mixtec. In Nieves Mixtec, instead, the wh-word equivalent to which/what cannot be used to form a wh-phrase introducing plain FRs, but it can be used to introduce -ever FRs, once enriched with the equivalent of -ever (see Caponigro, Torrence, and Cisneros 2013). In English, who is fairly restricted in plain FRs, while its Italian and Spanish equivalents are not. On the other hand, whoever and its counterparts in Italian and Spanish are all fully productive in introducing FRs (see Patterson and Caponigro 2016). If this asymmetry is eventually confirmed by a detailed and typologically balanced study of FRs crosslinguistically, there should be general reasons why this is the case. They cannot be of the kind invoked by D&C, though.

D&C’s proposal does not leave room for any gradience or crosslinguistic variation: true FRs introduced by wh-phrases are not expected to exist in any form in any language because they would violate general and nongradient properties of grammar like the operation of labeling after internal merging. On the other hand, gradience characterizes the pattern of FR formation, as sketched above. The explanation is unlikely to be a syntactic one, because it is not obvious what the syntactic difference between a possible what + NP FR and an impossible which + NP FR would be. Similarly, there is no evidence that -ever FRs are syntactically different from plain FRs, as I argued above. Still, whichever + NP FRs are allowed, while which + NP FRs are not.

A semantic explanation may be more promising. We know independently from interrogative clauses that what + NP and which + NP exhibit semantic differences with respect to each other and to other wh-words. Intuitively, which triggers the presupposition that the set denoted by its NP complement must be contextually salient (“discourse-linked,” according to Pesetsky’s (1987) characterization; see also Heim 1987). Such a meaning component is missing in all other wh-words. What carries no semantic restriction or presuppositional content. All the other wh-words are characterized by some semantic features (who [+human], where [+location], how many [+number], etc.). Finally, the suffix -ever changes the meaning of a wh-expression, as shown by the differences in meaning between plain FRs and -ever FRs in English and across languages. A detailed study of how these differences in lexical meaning interact with the rules of meaning composition that apply to FRs may offer insights into which wh-expressions can occur in FRs.

Notes

1 A first version of this analysis of free relative clauses was presented in Cecchetto and Donati 2010, while Cecchetto and Donati 2015 reprises the version in D&C 2011.

2 See Van Riemsdijk 2006 for an overview of the syntax of FRs and Šimík to appear for an overview of their semantics.

3 D&C are not fully consistent in their labeling of the top node. They use either C or CP depending on the tree (D&C 2011:523, (5)–(6); 528, (22)). In Cecchetto and Donati 2015, they again use either C or CP, as well as either T or TP (pp. 46, (3)–(4); 58, (51a); 95, (16)). I decided to retain the more traditional labels CP and TP for the mother nodes for clarity. Nothing crucial hinges on this assumption as far as my remarks are concerned.

4 In D&C’s (2011) example (111), the constituent right above N after N has internally merged is actually CP, rather than an NP. I assume it to be a typo given D&C’s immediately preceding statement that an -ever item like whatever is a determiner and the noun accompanying it does not form a constituent with it but acts as the “head” of a standard headed relative clause.

5 The judgments are from one native speaker of American English from Washington State, one from Maryland, and one from Georgia. (10) closely resembles Jacobson’s (1995) example (60), I’ll read what(ever) books John read, and two naturally occurring examples I found with a nonsystematic search: We drank what wine we had left (Austin 2012: 223) and King read what books by Gandhi he could get hold of (https://lareviewofbooks.org/article/what-king-learned-from-gandhi/; published on 16 January 2017; accessed on 28 March 2017).

6 An anonymous reviewer suggests that an analogous contrast is revealed by the adverb exactly immediately preceding the wh-expression: He readexactlywhat(*ever) she read, He readexactlywhat(*ever) books she read, and I drankexactlyhow(*ever) much wine you drank.

7Cinque (2017:sec. 4) challenges D&C’s account of plain FRs by discussing evidence from “paucal free relatives” like What beer we found was flat, originally presented in Andrews 1975:75.

8 Thanks to Anamaria Fălăuş for the Romanian data in (18)–(20).

9 Example from D&C 2011:520, main text; bracketing and boldfacing are mine.

10See Dayal 1997, von Fintel 2000, and Condoravdi 2015, among others, for the semantics of -ever FRs in English; and see Caponigro and Fălăuş 2018 for a different semantics for the morphosyntactic equivalents of -ever FRs in Italian and Romanian.

11 See Jacobson 1995:461, (29)–(30), for similar remarks.

12 This plain FR is introduced by the wh-word what and does not fully match the corresponding -ever FR in (33), which is introduced by the wh-word whoever. This is due to the degraded status of plain FRs introduced by who in English, regardless of the presence or absence of the complementizer (see Patterson and Caponigro 2016 for further discussion).

13D&C (2011:553) present an Italian sentence (their (102)) that is structurally similar to (46) as fully acceptable. My consultants find (46) significantly degraded (“??”). The same consultants find [The books for which she writes reviews] are likely to become bestsellers fully acceptable.

14 An anonymous reviewer suggests an analogy between -ever FRs lacking a relative marker and subject contact relative clauses like There’s someone wants to see you in Irish English (Doherty 2000:xii). This possibility would have to be restricted to -ever FRs introduced by whatever + NP and however much/many + NP. They are the only -ever FRs that optionally allow a relative marker, similarly to the optionality that is observed in languages with subject contact relative clauses.

15 An anonymous reviewer points out the contrast in (i) and (ii), inspired by Jacobson (1995:462).

  • (i)

    I will fire [whoever’s signature (*that) appears on this list].

  • (ii)

    *I will fire [anybody’s signature (that) appears on this list].

The -ever FR in (i) is introduced by a wh-phrase (whoever’s signature). The whole sentence is degraded with a complementizer after the wh-phrase, which would make the -ever FR a headed relative clause without any doubt. (ii) looks almost identical to (i) except that whoever is replaced by anybody, a non-wh free choice item that is very close in meaning. (ii) is degraded no matter what, while (i) is acceptable without the complementizer. Also, *I will fire {whoever/anybody}’s signature, with the absolute use of the wh-phrase, is completely unacceptable. This is unexpected if whoever’s signature in (i) behaved like the head of a relative clause. Overall, these data provide further support for the conclusion that -ever FRs cannot be analyzed as headed relative clauses.

16Cinque (2017:sec. 4) too argues against D&C’s proposal that -ever FRs are headed relative clauses. He discusses further crosslinguistic data, including data from Polish (Borsley 1984, Citko 2009) and from Croatian (Gračanin-Yuksek 2008).

Acknowledgments

I am very grateful to Daniel Kane for his invaluable help and to Anamaria Fălăuş and Carson Schütze, as well as to two anonymous reviewers for their insights and suggestions. I am solely responsible for any remaining mistakes.

References

References
Andrews,
Avery
.
1975
.
Studies in the syntax of relative and comparative clauses
.
Doctoral dissertation, MIT, Cambridge, MA
.
Austin,
Paul Britten
.
2012
. 1812.
Napoleon in Moscow
.
Barnsley, UK
:
Frontline Books
.
Baker,
C. L
.
1995
.
English syntax
. 2nd ed.
Cambridge, MA
:
MIT Press
.
Battye,
Adrian
.
1989
.
Free relatives, pseudo-free relatives, and the syntax of CP in Italian
.
Rivista di Linguistica
1
:
219
250
.
Borsley,
Robert D
.
1984
. Free relatives in Polish and English. In
Contrastive linguistics: Prospects and problems
, ed. by
Jacek
Fisiak
,
1
18
.
Berlin
:
Walter de Gruyter
.
Bresnan,
Joan
, and
Jane
Grimshaw
.
1978
.
The syntax of free relatives in English
.
Linguistic Inquiry
9
:
331
391
.
Caponigro,
Ivano
.
2003
.
Free not to ask: On the semantics of free relatives and wh-words cross-linguistically
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Caponigro,
Ivano
.
2004
. The semantic contribution of wh-words and type shifts: Evidence from free relatives cross-linguistically. In
Proceedings of SALT 14
, ed. by
Robert B.
Young
,
38
55
. https://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/95.
Caponigro,
Ivano
, and
Anamaria
Fălăuş
.
2018
.
Free choice free relatives in Italian and Romanian
.
Natural Language and Linguistic Theory
36
:
323
363
.
Caponigro,
Ivano
,
Harold
Torrence
, and
Carlos
Cisneros
.
2013
.
Free relative clauses in two Mixtec languages
.
International Journal of American Linguistics
79
:
41
96
.
Cecchetto,
Carlo
, and
Caterina
Donati
.
2010
.
On labeling: Principle C and head movement
.
Syntax
13
:
241
278
.
Cecchetto,
Carlo
, and
Caterina
Donati
.
2015
.
(Re)labeling
.
Cambridge, MA
:
MIT Press
.
Cinque,
Guglielmo
.
2017
.
On the double-headed analysis of “headless” relative clauses
.
Ms., University of Venice. http://lear.unive.it/jspui/bitstream/11707/6972/1/DoubleHeadedAnalysisOfFreeRCs.pdf (last accessed on 11 November 2017)
.
Citko,
Barbara
.
2008
.
Missing labels
.
Lingua
118
:
907
944
.
Condoravdi,
Cleo
.
2015
. Ignorance, indifference, and individuation with wh-ever. In
Epistemic indefinites
, ed. by
Luis
Alonso-Ovalle
and
Paula
Menéndez-Benito
,
213
243
.
Oxford
:
Oxford University Press
.
Dayal,
Veneeta
.
1996
.
Locality in WH quantification: Questions and relative clauses in Hindi
.
Dordrecht
:
Kluwer
.
Dayal,
Veneeta
.
1997
. Free relatives and ever: Identity and free choice readings. In
Proceedings of SALT 7
, ed. by
Aaron
Lawson
,
99
116
. https://journals.linguisticsociety.org/proceedings/index.php/SALT /issue/view/102.
Doherty,
Cathal
.
2000
.
Clauses without ‘that’: The case for bare sentential complementation in English
.
New York
:
Garland/Routledge
.
Donati,
Caterina
, and
Carlo
Cecchetto
.
2011
.
Relabeling heads: A unified account of relativization structures
.
Linguistic Inquiry
42
:
519
560
.
von Fintel,
Kai
.
2000
. Whatever. In
Proceedings of SALT 10
, ed. by
Brendan
Jackson
and
Tanya
Matthews
,
27
39
. https://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/24.
Gračanin-Yuksek,
Martina
.
2008
.
Free relatives in Croatian: An argument for the Comp account
.
Linguistic Inquiry
39
:
275
294
.
Grosu,
Alexander
.
2004
. The syntax-semantics of modal existential wh-constructions. In
Topics in the syntax and semantics of Balkan languages
, ed. by
Olga M.
Tomic
,
419
449
.
Amsterdam
:
John Benjamins
.
Heim,
Irene
.
1987
. Where does the Definiteness Restriction apply?—Evidence from the definiteness of variables. In
The representation of (in)definiteness
, ed. by
Alice ter
Meulen
and
Eric
Reuland
,
21
42
.
Cambridge, MA
:
MIT Press
.
Izvorski,
Roumyana
.
2000
. Free adjunct free relatives. In
WCCFL 19
, ed. by
Roger
Billerey
and
Brook Danielle
Lillehaugen
,
232
245
.
Somerville, MA
:
Cascadilla Press
.
Jacobson,
Pauline
.
1995
. On the quantificational force of English free relatives. In
Quantification in natural languages
, ed. by
Emmon
Bach
,
Eloise
Jelinek
,
Angelika
Kratzer
, and
Barbara H.
Partee
,
451
486
.
Dordrecht
:
Kluwer
.
Jespersen,
Otto
.
1909–1949
.
A modern English grammar on historical principles
.
Vol. 3
.
London
:
George Allen and Unwin
.
Patterson,
Gary
, and
Ivano
Caponigro
.
2016
.
The puzzling degraded status of who free relative clauses in English
.
English Language and Linguistics
20
:
341
352
.
Pesetsky,
David
.
1987
. Wh-in-situ: Movement and unselective binding. In
The representation of (in)definiteness
, ed. by
Alice ter
Meulen
and
Eric
Reuland
,
98
129
.
Cambridge, MA
:
MIT Press
.
Rawlins,
Kyle
.
2013
.
(Un)conditionals
.
Natural Language Semantics
21
:
111
178
.
Riemsdijk,
Henk van
.
2006
.
Free relatives
.
The Blackwell companion to syntax
, ed. by
Martin
Everaert
and
Henk van
Riemsdijk
,
2
:
338
382
.
Oxford
:
Blackwell
.
Šimík,
Radek
.
2011
.
Modal existential wh-constructions
.
Doctoral dissertation, University of Groningen. Published by LOT – Netherlands Graduate School of Linguistics
.
Šimík,
Radek
.
To appear
. Free relatives. In
Semantics companion
, ed. by
Daniel
Gutzmann
,
Lisa
Matthewson
,
Cécile
Meier
,
Hotze
Rullmann
, and
Thomas Ede
Zimmermann
.
Malden, MA
:
Wiley-Blackwell
. http://ling.auf.net/lingbuzz/003729/current.pdf?_s=v57pC3DVwvijD69s (last accessed on 20 August 2018).