Anti-Pied-Piping

Anti-pied-piping is a widespread but understudied phenomenon where a language targets a proper subpart of the logical focus for focus morphosyntax: for example, focus particle placement or focus movement. We show that anti-pied-piping is attested in over sixty languages from over forty distinct language groups. We present a theory of focus particle syntax/semantics that involves severing the pronounced position of a focus particle and the logical position of its corresponding semantic contribution, which successfully accounts for both anti-pied-piping and pied-piping behavior. Constraints on attested anti-pied-piping behavior and its interaction with movement show that particle placement takes place at particular, punctuated points in the derivation, in a cyclic model of syntactic structure building. We also discuss the relation of particle placement to other processes such as linearization and stress assignment.*

1. Introduction

In many languages, the presence of focus in a sentence triggers a characteristic morphosyntactic response, such as a marked word order via movement or the appearance of a focus particle. For example, in Hungarian (canonically SVO), exhaustive focus triggers movement to a dedicated, immediately preverbal position, as in 1.¹

(1)

Focus-triggered movement in Hungarian (Horvath 1981:117)

 

A language may likewise indicate the presence of focus and an associated semantics using a dedicated particle. For example, additive focus is indicated in Japanese with the particle mo, as in 2. Languages may also use particle placement and movement simultaneously, or use altogether different strategies, as we discuss below.

(2)

Focus-triggered particle placement in Japanese

 

We refer to such morphosyntactic responses to focus as 'MSF' throughout. In both 1 and 2, the constituent targeted for MSF—movement in Hungarian and particle placement in Japanese—is the logically focused constituent, which we annotate with the subscript 'F'. But in some cases, there is a mismatch between the logical focus and the target of MSF, in which case we annotate both with separate subscripts.

Ross 1967 describes a famous type of mismatch termed pied-piping, where MSF targets a constituent properly containing the logical focus. Examples of pied-piping, again from Hungarian and Japanese, are given below. In 3, additional, nonfocused material is moved together with the focused constituent. In 4, the focus particle attaches to a constituent that includes the logical focus, as well as additional, nonfocused material.²

(3)

Pied-piping in Hungarian focus movement (Kenesei 1998b, ex. 13b)

 

(4)

Pied-piping in Japanese focus particle placement³ (based on Kuroda 1965:78)

 

We schematize the syntactic configuration referred to as pied-piping in 5. In this article, we document and investigate the phenomenon of anti-pied-piping, schematized in 6, where a constituent properly contained within the logical focus is marked with a focus particle or targeted for focus movement. Anti-pied-piping can thus be thought of as the inverse of the very well-studied pied-piping pattern.

(5)

Pied-piping

 

(6)

Anti-pied-piping

 

Anti-pied-piping is attested in both Hungarian and Japanese. In the Hungarian example 7, predicate focus results in movement of the object out of the focused verb phrase to the preverbal focus position (see n. 1 above). Similarly, in the Japanese example 8, two whole propositions contrast and license the additive focus particle mo, but the particle appears on the subject within each focus. In both cases, MSF targets a proper subconstituent of the logically focused constituent. It is this type of mismatch that we concern ourselves with in this article.

(7)

Anti-pied-piping in Hungarian focus movement (Kenesei 1998a:77)

 

(8)

Anti-pied-piping in Japanese focus particle placement (Nagano 1951:210)

 

We begin in §2 with a brief introduction to focus semantics, which will establish a methodology for identifying the logically focused constituent and, therefore, mismatches between the target of MSF and the logical focus. Section 3 presents our cross-linguistic survey of anti-pied-piping. We show that anti-pied-piping mismatches of the form in 7 and 8 are attested in over sixty different languages from over forty distinct language groups, as classified by major subfamily or genus (Dryer 1989); a list of all languages discussed here as exhibiting anti-pied-piping is given in the appendix. In addition, we show that the process of anti-pied-piping in many languages must make reference to the linear order of constituents.

In §4, we introduce a new theory for the syntax/semantics of focus particles that allows for anti-pied-piping mismatches. In brief, we propose that many focus particles do not introduce their associated semantics directly, but instead serve as morphosyntactic flags that signal the presence of corresponding abstract operators. In 9a, the entire focus (YP, in gray) is within the sister of the unpronounced operator (op), which associates with focus in a compositional manner. We propose that the particle (prt) may be adjoined to a proper subpart of the focus (XP), as schematized in 9b.⁴

(9)

 

If the particle (prt) is pronounced, we derive particle anti-pied-piping, as in 8. Alternatively, once the particle has been inserted into an anti-pied-piping configuration, as in 9b, the resulting particle phrase (the particle, possibly unpronounced, and its sister) could move, resulting in what we describe as movement anti-pied-piping, as in 7. We show that this theory also naturally extends to cases of pied-piping and captures a number of parallels between the two.

We propose that this particle placement takes place at certain designated, punctuated points during the derivation, in a cyclic spell-out model of grammar (Uriagereka 1999, Chomsky 2000, 2001, among others). This allows particle placement to make reference to some phonological information such as linear order and prosodic information, and to then feed further syntactic operations. Although we concentrate on focus here, our proposal is intended to extend to the morphosyntactic reflexes of other information-structural notions as well.

After we have presented our core proposal, we discuss some potential alternative analyses in §5. Finally, we discuss the question of which constituent is targeted in anti-pied-piping in further detail, and relate this behavior to processes of stress assignment as well as pied-piping in §6. We conclude in §7 with implications of the analysis and a further outlook.

2. Focus as the locus of alternatives

Before diving into the empirical landscape of anti-pied-piping, we first briefly discuss the function of focus in grammar. This section serves an important methodological purpose for our study of anti-pied-piping, as we establish how the interpretation of focus particles and question-answer congruence can be used as diagnostics for the position of focus.

The core function of focus is to highlight a portion of the sentence as standing in contrast to other values in a set of contextually salient alternatives (see e.g. Rooth 1992, Krifka 2008).⁵ Focus-sensitive expressions such as focus particles then quantify over these alternatives. For example, consider the contrast between 10a and 10b. These examples differ only in the placement of focus—realized in English with a pitch accent—but make very different claims about the world.⁶ In example 10a, the theme sandwiches is focused; this claim entails that Alex did not make anything else for Brie. In example 10b, Brie is focused instead, contrasting against other potential beneficiaries; this claim entails that Alex didn't make sandwiches for anyone else. This difference in meaning is reflected in the different felicity patterns of the continuations (i) and (ii) in 10a,b.

(10)

a. Alex only made [sandwiches]_F for Brie.

i. ^✓She didn't make [soup]_F for her.

ii. #She didn't make sandwiches for [Cara]_F.

b. Alex only made sandwiches for [Brie]_F.

i. #She didn't make [soup]_F for her.

ii. ^✓She didn't make sandwiches for [Cara]_F.

In common parlance, focus-sensitive expressions associate with the focused phrase. For example, we may say that the focus particle only associates with sandwiches in 10a and with Brie in 10b.

We indicate the position of logical focus in example sentences with the subscript 'F', commonly referred to as F-marking. However, the position of logical focus is not unambiguously and directly reflected in the linguistic signal. For example, the phonetic realization of example 10a is the same as the first sentence of 11, in that both of these choices of focus result in the most prominent accent being on the object sandwiches. But in 11, the entire VP made sandwiches is focused. The first sentence with only therefore claims that Alex did not do anything for Brie except make sandwiches. The second sentence elaborates on this claim, mentioning washing the car as a particular alternative activity that Alex did not do for Brie.

(11)

Alex only [made sandwiches]_F for Brie. She didn't (also) [wash the car]_F for her.

The surface equivalence of sentences with different positions of focus, as in 10a and 11, makes it challenging to confidently identify the position of focus in a sentence in isolation. As we see from the examples above, however, the position of focus can be elucidated by explicit contexts and continuations that make the extent of contrast between alternatives clear. Consider also the pair of examples with additive also in 12.

(12)

a. Alex made soup for Brie. She also made [sandwiches]_F for her.

b. Alex washed the car for Brie. She also [made sandwiches]_F for her.

The preceding, contrasting propositions serve to identify the extent of contrast between salient alternatives: only the objects in 12a, but the VPs in 12b. The semantics introduced by also presupposes that another alternative is true, in contrast to only, which claims that the other alternatives are false.

Constituent questions and their congruent answers are also useful for identifying the position of focus. Consider the object wh-question in 13 and the predicate wh-question in 14, each with two possible answers. The position of contrast between each set of answers bears focus, roughly corresponding to the material that has been replaced with a wh-word in the question.

(13)

What did Alex make for Brie?

a. She made [sandwiches]_F for her.

b. She made [soup]_F for her.

(14)

What did Alex do for Brie?

a. She [made sandwiches]_F for her.

b. She [washed the car]_F for her.

Again, the utterances in 13a and 14a both result in a pitch accent on sandwiches and cannot be distinguished in isolation, but we can identify their foci by considering the questions they address (13 vs. 14) and the shape of other felicitous answers to those questions.

In the following sections, we continue to indicate the position(s) of logical focus compatible with a particular surface form using F-marking notation. In the interest of space, in most cases we do not include the supporting contexts or continuations that are necessary to verify the choice of F-marking. Most of the data we present comes from work by other scholars; in all such cases, the original, cited sources include such supporting materials or otherwise have sufficiently detailed descriptions that allow us to confidently conclude that the reproduced example indeed has the focusing possibility we report.⁷

Finally, we note that other information-structural notions beyond focus, such as topic or contrast, may also exhibit mismatches between their logical semantic/pragmatic locus and their corresponding morphosyntactic target. We believe that the notions of pied-piping and anti-pied-piping as well as the theoretical proposal we put forward below also extend to other such information-structural categories and their corresponding morphosyntactic reflexes. However, we concentrate on focus here, due to the existence of well-established diagnostics for the logical position of focus, reviewed above, and the fact that focus phenomena are comparatively well described in a wide range of languages.

3. Properties of anti-pied-piping

We now explore the empirical landscape of anti-pied-piping and highlight some of its important properties and points of crosslinguistic variation. We show that anti-pied-piping is widely attested in a range of genetically unrelated and typologically varied languages, with both particle placement (§3.2) and phrasal movement (§3.3). We then discuss the choice of constituent that is targeted for MSF and variation in the obligatoriness of anti-pied-piping (§3.4).

3.1. Yaeyaman

We begin by presenting a detailed and instructive case of anti-pied-piping in Yaeyaman, a Southern Ryukyuan (Japonic) language, from Christopher Davis's work on the Miyara variety. We consider the focus particle du, which in the basic case appears as an enclitic on wh-phrases and on the focused constituent in corresponding answers. This is illustrated by the question-answer pairs in 15–16. Note that the answers in 15b and 16b convey the same proposition, that Chris ate soba. In 15b, as an answer to a subject wh-question, du appears on the subject, while it appears on the object as an answer to an object wh-question in 16b.

(15)

Subject focus (Davis 2014:124)

 

(16)

Object focus (ibid.)

 

What is of particular interest is the behavior of du in utterances with sentence focus and predicate focus (Lambrecht 1994; see also Vydrina 2020:§2.4), such as in the answers to the questions 'What happened?' in 17 and 'What did that woman do?' in 18. In 17b, where the entire sentence constitutes the focus in the answer to the question, the particle du appears on the subject. In 18b, where the predicate 'eat fish' is focused, du appears on the object. The placement of du in 17–18 constitutes cases of anti-pied-piping.⁸

(17)

Sentence focus (Davis 2013:33)

 

(18)

Predicate focus (ibid.)

 

Davis (2013) notes that this anti-pied-piping in 17–18 is obligatory—that is, du cannot instead appear inside of or following the verbal complex—despite the fact that du can encliticize to the verb in cases of narrow focus on the verb. Instead, 'du seems only to be able to occur attached to the leftmost element within its associated focus' (p. 36). Shimoji (2018:96) reports that this description also holds of all of the fifteen other Ryukyuan language varieties he has surveyed. We return to this leftmost effect and the general question of which subpart of the focus is targeted for MSF in anti-pied-piping in §3.4.

3.2. Anti-pied-piping in particle placement

Anti-pied-piping in focus particle placement is readily attested in many other languages. Examples 19–29 illustrate anti-pied-piping in predicate focus in eleven other verb-final languages from distinct language subfamilies or genera. In each of these transitive clauses, a focus particle targets the direct object for attachment (MSF) while semantically associating with the entire predicate VP.⁹

(19)

Burmese¹⁰ (Okell 2002:87)

 

(20)

Imbabura Quechua (Kwon 2013:76)

 

(21)

Ishkashimi (Karvovskaya 2013:81)

 

(22)

Khalkha Mongolian (Jun Jie Lim, p.c.)

 

(23)

Kakabe (Vydrina 2020:518)

 

(24)

Korean (Choe 1996:677)

 

(25)

Masalit (Leffel 2011:30–32)

 

(26)

Telugu (based on Kotani 2008:191)

 

(27)

Tibetan (Erlewine field notes¹¹)

 

(28)

Turkish (Kamali 2011:182)

 

(29)

Qunqi Dargwa (Dmitry Ganenkov, p.c., cited in Forker & Belyaev 2016:249)

 

As noted above, we do not reproduce supporting contexts or continuations that motivate each attested choice of F-marking, but such information is available in the original sources cited.

The focus particles associating with the VP in 19–29 appear between their MSF object and the inflected verb. We might wonder whether anti-pied-piping in such examples is a response to the fact that particle placement directly on the logically focused VP may disrupt the morphology of the verbal complex. The Japanese examples in 30 show that this cannot be the motivation for anti-pied-piping in the general case. When associating with a transitive VP, a focus particle such as additive mo may adjoin to the VP itself, as in 30a, or to the object, as in 30b, the latter being a case of anti-pied-piping parallel to those above. In the former case, 30a, the verbal morphology is indeed disrupted, triggering a process akin to do-support. (In addition to the intended predicate-focus reading, mo in 30a may associate narrowly with the object or the verb, and 30b also allows for narrow focus on the object.)

(30)

Japanese (Ohno 2003:323–24)

 

This optionality of anti-pied-piping in Japanese—despite its obligatoriness in the related and morphosyntactically similar Yaeyaman language in the preceding section—shows that anti-pied-piping cannot be generally described as a kind of repair to satisfy morphological processes, and also that its application is subject to crosslinguistic variation that must be learned.

Focus may also trigger other morphosyntactic reflexes in a clause. In Tundra Yukaghir, when an object is focused with a particle such as leŋ, as in 31a, the subject agreement affix on the verb changes to a dedicated object focus (of) form. In cases of predicate focus, a particle similarly appears on the object, again triggering the object-focus agreement form, as in 31b. See also Nagasaki 2018 for parallel examples in (late nineteenth century) Kolyma Yukaghir.

(31)

Tundra Yukaghir (Matić & Odé 2015:630)

 

Anti-pied-piping in particle placement is not limited to verb-final languages. As seen in examples 32–34, anti-pied-piping in predicate focus is attested in verb-medial languages as well. Schwarz (2009, 2010) also gives examples parallel to 33 in the related Oti-Volta Gur languages Buli, Gurene, and Kɔnni. (Note that example 34 is also compatible with a narrow verb-focus reading (in 34c), which we address in n. 45 below.)

(32)

Awing (Fominyam & Šimík 2017:1039)

 

(33)

Dagbani (Fiedler & Schwarz 2005:120)

 

(34)

Tangale (Hartmann & Zimmermann 2007a:119)

 

Anti-pied-piping in focus particle placement is also attested in verb-initial languages, which we exemplify with Tagalog later in this section.

Readers may note that the anti-pied-piping data just presented fall largely into two categories: OV (head-final) languages with postfocal particles (19–31) and VO (head-initial) languages with prefocal particles (32–34). Readers may rightly wonder whether other combinations are possible: that is, anti-pied-piping in head-final languages with prefocal particles or in head-initial languages with postfocal particles. We imagine that such languages may exist, but have not identified any here, due to the following systematic methodological challenge.

To illustrate the issue, consider example 35 from Konkomba, another Oti-Volta Gur language related to Dagbani. Like the Dagbani example 33 above, 35 is able to express object focus or predicate focus.

(35)

Konkomba (A. Schwarz 2007:123, 126)

 

Before we can determine whether there is a mismatch between the logical focus and focus particle placement, we must identify the syntactic position of the particle. Unlike in the Dagbani 33, the focus particle lá in the Konkomba 35 follows the entire head-initial verb phrase. Therefore, the surface structure in 35 is amenable to either of the parses in 36. Note that we use the label DP (for determiner phrase) for noun phrase projections (Szabolcsi 1983, Abney 1987).

(36)

 

If object focus in 35a and predicate focus in 35b correspond, respectively, to the structures in 36a and 36b, then there is no mismatch of anti-pied-piping or pied-piping. However, if the particle is adjoined to the object as in 36a for both readings in 35, we would describe 35b as a case of anti-pied-piping. Without further work to establish the exact position of the particle, such examples are not sufficiently informative as to whether the particle exhibits anti-pied-piping.¹² In contrast, anti-pied-piping is more immediately identifiable with postfocal particles in head-final structures and prefocal particles in head-initial structures.

Another common pattern of anti-pied-piping involves the subject being marked by a focus particle, with the sentence as a whole being interpreted as the logical focus, as seen in 37–42. We also saw this pattern with Yaeyaman du in 17.

(37)

Even (Matić & Wedgwood 2013:153)

 

(38)

Ishkashimi (Karvovskaya 2013:82)

 

(39)

Kakabe (Vydrina 2020:502)

 

(40)

Korean (Choe 1996:680)

 

(41)

Lak (Victor Friedman, p.c., cited in Forker & Belyaev 2016:251)

 

(42)

Navajo (Perkins 1978:25)

 

The Yukaghir languages also exhibit this same form of sentence-focus anti-pied-piping. Recall that verbal subject agreement morphology is affected by the presence of focus particles, as we saw in 31 above. Similarly, when an intransitive subject is focused with a particle, as in 43a, the agreement morphology on the verb is replaced with an invariant subject focus (sf) suffix. When an entire intransitive clause is focused, as in 43b, its subject bears a focus particle, with the verb again appearing in the subject-focus form.

(43)

Tundra Yukaghir (Matić & Odé 2015:630)

 

Anti-pied-piping in sentence focus is also attested in verb-medial languages, as in 44 and 45.

(44)

Ewe¹³ (Ameka 2010:151)

 

(45)

Konkomba¹⁴ (A. Schwarz 2007:130)

 

Finally, we discuss anti-pied-piping in Tagalog and Latin, which will foreshadow our own proposal. In the Tagalog example 46a, the 'only' particle lang immediately follows the focus, which is an adjunct fronted to initial position. This gives the appearance of lang being an enclitic focus particle. In contrast, lang associates with the verb phrase 'give money' in 46b, but is positioned properly within the predicate, between 'give' and 'money'.

(46)

Tagalog (Kaufman 2005:181)

 

The behavior of lang in 46 is explained in part by recognizing its more general status as a second-position clitic. Second-position clitics in Tagalog follow one phrase or head within the clause, not counting topics (Kroeger 1998, Kaufman 2010). Note that the pronoun ako here is also such a second-position clitic and thus exhibits this same pattern of placement in 46a vs. 46b. Thus we can conclude that the anti-pied-piping manifested by lang in examples like 46b is due to a more general property of second-position clitic placement in Tagalog. Focus particles in Thompson River Salish, which like Tagalog is a predicate-initial language, are also second-position clitics, resulting in expressions parallel to 46; see Koch & Zimmermann 2010:242–43, exx. 14a,b, 17a,b.

Latin too exhibits anti-pied-piping behavior involving the well-known second-position clitic que (see e.g. Zwicky 1977). Here we follow Mitrović and Sauerland (2014) and Szabolcsi (2015) in describing que as an additive focus particle rather than a conjunction. The placement of que in second position within its logical focus leads to examples such as 47.¹⁵

(47)

Latin (Julius Caesar, glossed in Carlson 1983:80)

 

We consider these examples in Tagalog and Latin to be instructive, as anti-pied-piping in these cases can be attributed straightforwardly to a general process of second-position clitic placement, resulting in a mismatch between the position of the particle and the position of its logical interpretation. Of the mismatch exhibited by Latin que, Carlson (1983:73) writes that '[a] much simpler interpretation of -que could be given, though, if we were to somehow "postpone" its semantic effect until a larger unit is encountered in the tree'. The analysis we develop in §4 builds on this intuition and generalizes it to other cases of anti-pied-piping.

3.3. Anti-pied-piping in phrasal movement

Many languages conventionally target focused constituents for movement. Such movement may also exhibit anti-pied-piping, targeting a constituent that is a proper subpart of the logical focus. Although some previous works have described such examples as involving movement that is independent of the focal structure of the sentence (as we review in §5), we instead argue that these patterns are most fruitfully described by paying close attention to the mismatch between the target of movement and the logical focus, and then analyzed as related to the phenomenon of anti-pied-piping in focus particle placement.

The examples in 48–53 illustrate predicate focus with transitive VPs where only the object is moved to a focus position, just as we saw in Hungarian in 7.

(48)

Bura (Hartmann et al. 2008:72)

 

(49)

Garrwa (Mushin 2006:311)

 

(50)

Haitian Creole (Franz Cozier ms., cited in Fanselow & Lenertová 2011:194)

 

(51)

Russian (Fanselow & Lenertová 2011:203)

 

(52)

Somali (Lecarme 1999:284; see also Svolacchia et al. 1995:73–74)

 

(53)

Yoruba (Manfredi 2004, ex. 39a)

 

Fanselow (2004:17–18) describes similar patterns in Czech, Croatian, and Polish as well. We discuss some Czech examples in §6.

The fronted constituent may also host an overt focus particle, as in 54–56. In the German 54, the stranded verb has independently fronted to verb-second position.

(54)

German (Fanselow 2004:17)

 

(55)

Kîîtharaka (Abels & Muriungi 2006:9)

 

(56)

Mandarin (Constant & Gu 2010:28)

 

Focus movement with anti-pied-piping is also attested with sentence focus. Examples 57–61 are all reported as answers to questions such as 'What happened?' or 'What's the matter?', but where only the subject moves to a dedicated focus position. The relevant position is a cleft pivot position in 57, 59, and 60, V2 prefield position in 58, and a left-peripheral position marked by a focus marker in 61.

(57)

French (Sasse 1987:538)

 

(58)

German (Fanselow & Lenertová 2011:181)

 

(59)

Tilapa Otomi (Palancar 2018:261)

 

(60)

Welsh (Mac Cana 1973:93, as glossed in Sasse 1987:539)

 

(61)

Wolof (Robert 1991:125)

 

While such patterns are common, there are also cases of anti-pied-piping with sentence focus leading to focus movement of the object rather than the subject. This is the case in 62 and 63, where there is no overt subject that can be fronted. Bianchi et al. (2016:36, ex. 43) also report this possibility in standard Italian.

(62)

Breton (Jouitteau 2007:178)

 

(63)

Sicilian (Cruschina 2012:71)

 

3.4. Position

Having established the existence of anti-pied-piping, we now turn to the question of which subconstituent of the logical focus is targeted for MSF. We will see that, in many languages, the element targeted for MSF is at or near the left edge of the logical focus, but there is also substantial crosslinguistic variation in the presence or strength of this effect.

First, recall that in Yaeyaman and Ishkashimi, sentence focus is marked by particle placement on the subject, and transitive predicate focus is marked by particle placement on the object. For these two languages, Davis (2013) and Karvovskaya (2013) also show that other possibilities are ungrammatical. These options are schematized in 64.

(64)

Miyara Yaeyaman du (Davis 2013, 2014) and Ishkashimi məs (Karvovskaya 2013)

 

Similar leftmost effects are observed in other languages as well, although with the status of a preference rather than a hard constraint. Example 65 is a case of predicate focus with a ditransitive predicate in Tibetan. For the intended reading, consulted speakers prefer to place the particle yang after the leftmost (goal) argument within the predicate.

(65)

Tibetan (Erlewine field notes)

 

In Japanese, the leftmost requirement is subject to some speaker variation. Ohno (2003:324) reports that for the sentence-focus reading in an SOV sentence, all speakers allow the additive particle mo on the subject, as in 66a, but only some also allow the particle on the object, as in 66b.¹⁶

(66)

Japanese (based on Aoyagi 1998:151, 2006:123)
[Context: At yesterday's party, not only did Hanako dance, but …]

 

Dash and Datta (2020) report optionality in anti-pied-piping particle placement in Hindi-Urdu and Bangla. For instance, both variants of the ditransitive example 67 are described as grammatical for predicate focus, but with a 'weak leftmost preference'; we therefore report the second option with '?'.

(67)

Hindi-Urdu (Dash & Datta 2020, exx. 5, 17)
[Context: During Diwali, Pulkit plans to feed the poor and also distribute gifts to children. However, due to some emergency, he fails to be able to feed the poor.]

 

Next we turn to anti-pied-piping involving phrasal movement. Here, too, leftmost effects have been described in some languages. In the German 68, fronting of the ditransitive's theme allows for the predicate-focus reading in 68a(ii), but fronting of the goal in 68b(ii) does not. The theme is naturally leftmost in the VP's base order.

(68)

German (Fanselow 2004:11)

 

A weaker leftmost preference is reported in Kikuyu. Schwarz (2003:95) notes that for predicate focus with a ditransitive VP with goal-theme base order, goal fronting (as in 69a) 'seems to be slightly preferred' over theme fronting (69b), 'although both seem to be acceptable'.

(69)

Kikuyu (Schwarz 2003:95)

 

In contrast, in San Martín Peras Mixtec (base order: VSO), sentence focus may involve fronting of either the subject or object with apparent optionality.

(70)

San Martín Peras Mixtec (Hedding 2019, exx. 43a,b)

 

We conclude that anti-pied-piping in many languages is subject to a leftmost requirement, whereby MSF must or prefers to target the leftmost subphrase of the logical focus, although there is considerable crosslinguistic variation in the presence or absence and strength of such effects.¹⁷ We discuss further details of this process and its relation to stress placement and the theory of focus projection in §6, after we present our core proposal in §4. For now, it suffices to note that the process of MSF target selection that results in anti-pied-piping must be able to make reference to linearized structures, and ultimately to phrasal stress or its determinants as well. We also note that anti-pied-piping behavior in particle placement and in phrasal movement parallel each other, and these parallels will be strengthened in §6.2. Both of these properties of anti-pied-piping are important features of the proposal we develop here.

3.5. Summary

We have now established a number of facts about anti-pied-piping. Anti-pied-piping is attested in a wide range of languages: in total, we have identified anti-pied-piping in over sixty languages from over forty different subfamilies or genera, which we list in the appendix. Both particle placement and focus fronting allow for anti-pied-piping, with the choice of constituent targeted often subject to a leftmost requirement.

Although the availability of anti-pied-piping is widespread, we do not believe it to be universal. There are, for example, languages where scholars have specifically investigated broad-focus constructions and have shown that possible patterns of anti-pied-piping are not attested. This is the case in two Zapotec languages, both of which are verb-initial and express narrow argument focus by fronting to a preverbal position but where broad focus must be expressed using a verb-initial clause without fronting (Lee 1997:237–38, Bueno Holle 2016:Ch. 5). Just as there is variation in the behavior of languages with anti-pied-piping, so too is its availability a point of potential variation that must be learned.

The existence of anti-pied-piping complicates the syntax/semantics of focus particles. It forces us to divorce the pronounced position of particles and their position of interpretation, just as Carlson (1983) suggests in his discussion of Latin que as noted above. The analysis we develop is also inspired by contemporary theories of pied-piping, allowing us to unify anti-pied-piping in particle placement and in movement and to explain their similarities, as well as to account for parallels between anti-pied-piping and pied-piping, which we present in §6.

4. Proposal

We now present our analysis for the anti-pied-piping patterns presented above. Anti-pied-piping constitutes a serious challenge for the compositional semantics of focus particles, as a central expectation of focus semantics since Jackendoff 1972 and Rooth 1985 has been that the focus be within the focus particle's sister. We therefore begin by putting forward a new and general theory for the compositional semantics and syntactic derivation of focus particle constructions, which will allow for anti-pied-piping in focus particle placement. We then address cases of anti-pied-piping in focus movement, building on the influential proposal of pied-piping as the result of movement targeting a particle-adjoined phrase (see e.g. Tanaka 1999, Horvath 2000, 2007, Watanabe 2006, Cable 2007, 2010a,b). We address the question of how the position of particles is determined in greater detail in §6 below.

4.1. Severing the particle from its semantics

There are broadly two analytic approaches to the semantics of focus particles that adjoin to a subclausal phrase, such as the only in 71.

(71)

Alex made only [sandwiches]_F for Brie.

The first approach, which we call the quantificational particle theory, takes the pronounced only to be a semantically contentful, two-place functor denotation, as in 72. Under this view, only composes with its sister (the α argument, of type σ) to form a quantificational meaning that then composes with its scope (β).¹⁸ The resulting expression presupposes the prejacent proposition β(α) and will be true if and only if all other, alternative meanings to α in the alternative set C, when composed with β, are false.¹⁹

(72)

Two-place only meaning for the quantificational particle prt_only:

 

Under the second approach, which we call the operator-particle theory, the pronounced only in 71 is not itself semantically contentful but is instead simply a morphological flag that signals the presence of a corresponding, unpronounced only operator in the clause (Lee 2004, 2005, Hirsch 2017, Hole 2017, Sun 2021, among others).²⁰ The covert operator adjoins to the clausal spine, above the logical focus and its corresponding particle, resulting in a dissociation between the position of the particle and its associated semantics. The only operator in 73 takes a sister of propositional type with denotation φ (the prejacent), presupposes φ, and returns 'true' if and only if all other alternatives in a set of propositional alternatives C are false.

(73)

One-place only meaning for the operator op_only:

 

In either case, the alternative set C is a contextually determined variable that contains alternatives that are congruent in focus structure to the meaning of the interpreted only's sister. We can ensure the congruence of alternatives by making reference to logical F-marking, as in the alternative semantics of Rooth 1992, 2016, or without reference to F-marking, as in Büring 2015. We continue to discuss examples with only in this section, but these two approaches to its syntax/semantics extend to other types of focus particles as well.

Under the operator-particle approach, example 71 above reflects the syntactic structure in 74: an operator op with the semantics of one-place only in 73 is adjoined to the clausal spine, here taking vP as its sister,²¹ and a corresponding particle prt is adjoined to the focused phrase sandwiches. In English, either the operator or particle can be pronounced as only, but not simultaneously (Hirsch 2017, Quek & Hirsch 2017). If the particle is pronounced, we yield 71 above. If the operator is pronounced instead, we yield the form in 75 with sentential only, which has the equivalent interpretation. See also Hirsch 2017:Ch. 7, Quek & Hirsch 2017, and Bassi et al. 2022 for extensive motivation for the operator-particle theory from the scope-taking behavior of English only.

(74)

 

(75)

 

Some languages allow both the particle and the corresponding operator to be pronounced simultaneously, descriptively in a concord-like relationship, unlike in English. This is attested in Dutch (Barbiers 2010, 2014), German Sign Language (Deutsche Gebärdensprache) (Herrmann 2013:299–300), Vietnamese (Hole 2013, 2017, Erlewine 2017, Sun 2020), and Lavukaleve (see 87 below). See also Bayer 2020:64–66 for discussion of naturally occurring examples in English and German that may be best analyzed as rare cases of simultaneous pronunciation of the operator and particle.

In 74, the particle has adjoined directly to the logical focus, as also schematized in 76. But faithful adjunction to the focused constituent is not the only possibility. The particle could be adjoined to a constituent that properly contains the logical focus, as in 77; this is pied-piping. The particle could also be adjoined to a constituent properly contained within the focus, as in 78, which is the configuration we recognize as anti-pied-piping.

(76)

No mismatch (MSF = F)

 

(77)

Pied-piping (MSF > F)

 

(78)

Anti-pied-piping (MSF < F)

 

We argue that the semantic interpretation of anti-pied-piping requires the operator-particle theory with one-place operator semantics. To see why this is the case, we discuss the interpretation of a predicate-focus structure where prt has adjoined to the object, a proper subpart of the logical focus. For presentational purposes, we discuss a pseudo-English example pronounced as 79a but interpreted as 79b, modeled after the Hungarian example 7. Although this pattern of anti-pied-piping is not attested in English, this example stands in for the many examples of predicate focus discussed above with particle placement on the object.

(79)

a. "Peter read only [Hamlet]_MSF."

b. 'Peter only [read Hamlet]_F.' (He didn't do anything else.)

Under the operator-particle theory, we could posit 80 as the underlying structure for 79. The higher, unpronounced operator op is interpreted with the one-place denotation for only in 73 above.

(80)

Structure for 79 under the operator-particle theory:

 

Notice that the entire focus (VP) is within the sister of the interpreted operator op. This allows op to consider alternatives that vary in their VP meanings, contrasting 'read Hamlet' with other contextually relevant descriptions, like the transitive 'read Macbeth' and 'clean the car', as well as the intransitive 'swim'. The presence of prt serves only to morphologically indicate the presence of the corresponding operator op that is interpreted; prt itself is semantically inert.

In contrast, let us consider how we might attempt to interpret 79 using the quantificational particle theory. Interpreting the pronounced particle in 79a with the two-place semantics for only in 72 above, the constituent [only Hamlet] will yield a quantificational noun phrase meaning. Following quantifier raising (QR; May 1977; see also Heim & Kratzer 1998:Ch. 7), this results in a logical form (LF) representation as in 81, where the trace position is interpreted as a variable (x) and a corresponding λ-binder is adjoined above. For ease of presentation, we illustrate 81 with the agent Peter reconstructed into its base position.

(81)

LF structure for 79 under the quantificational particle theory:

 

Recall that the two-place denotation for only as in 72 considers alternatives to its sister, Hamlet, in the contextually specified variable C and requires that all of those alternatives that are not Hamlet, when composed with its second argument (λx . Peter read x), will be false. For instance, if C includes Macbeth, 81 would require that Peter did not read Macbeth. However, because only alternatives to the object are considered, 81 cannot be used to derive the intended predicate-focus meaning, which contrasts 'read Hamlet' with alternative VP meanings involving other verbs. In other words, if the pronounced particle itself introduces the semantics of only with denotation as in 72, the logical focus must be the particle's sister or a subpart thereof (pied-piping); anti-pied-piping focus association cannot be modeled in this way.²²

The operator-particle theory is also supported by the existence of cases of anti-pied-piping involving multiple particles within a single focus. Eaton (2010a:10) observes that focus particles in Sandawe exhibit anti-pied-piping, as they 'mark the constituent in question as contained within the focus of the sentence'. For instance, example 82a is described as a felicitous sentence-focus answer to 'What happened?', with focus particles appearing on both the subject and object. Example 82b from Jonah 1:5 is an instance of predicate focus, where the theme, goal, and source arguments of the verb all bear particles. The verb can also host focus particles, but only in cases of narrow focus on the verb.

(82)

Sandawe²³

 

Similar multiple particle placement within a single focus is also attested in Kokama-Kokamilla; see Vallejos Yopán 2009:422–23, exx. 25a–c.

Such patterns can be modeled straightforwardly under the operator-particle theory. These languages simply allow for multiple particles to be adjoined within a single focused constituent. A single logical operator then takes the entire focus in its scope. Under the operator-particle theory, particles and operators need not be one to one, although that is often the case. See also Krifka 1991:144 and Lee 2004:47 for discussion of similar structures with multiple focus particles in English and Korean, respectively.

A practical issue that arises in the analysis of focus particle constructions in the operator-particle theory is determining whether a particular overt expression is the realization of an operator or a particle, in our technical sense. Three diagnostics are useful here. First is syntactic position and constituency: operators adjoin to the clausal spine—to a node of propositional type (i.e. type t, at vP or higher; see n. 21)—whereas particles may adjoin to subsentential phrases of arbitrary syntactic category and semantic type, such as noun phrases or prepositional phrases. Second is semantic scope: operators make a semantic contribution, so if they are pronounced, their overt position indicates their scope, whereas particles do not directly indicate the scope of their associated semantics and therefore may lead to scope ambiguities.²⁴ Consider the observation from Taglicht 1984 that English only adjoined to a subsentential constituent may take variable scope, as in 83, but only in a preverbal, sentential adverb position takes fixed scope, as in 84.

(83)

Constituent only has flexible scope (based on Taglicht 1984:150)

 

(84)

Sentential only has fixed scope (ibid.)

 

Such facts receive an immediate explanation under the operator-particle theory. We propose that the two interpretations of 83 reflect two different syntactic structures, given in 85 below, with varying operator positions. Example 83 reflects a realization of 85a or 85b where the particle is pronounced only and its corresponding operator is unpronounced.²⁵ In contrast, examples 84a and 84b reflect the possibility of pronouncing the operator in 85a,b as only, explaining the fact that the position of only in these examples directly reflects its interpreted scope.

(85)

Two structures underlying 83 and 84

 

The third and final diagnostic is the availability of anti-pied-piping: semantic interpretation in the operator-particle theory requires that the logical focus be within the operator's sister at LF, but not necessarily within a particle's sister. Particles therefore may exhibit anti-pied-piping patterns of focus association, but operators do not (except where attributable to independent movements which then reconstruct; see §5.2).

4.2. Particle syntax in the operator-particle theory

We have argued that considerations of compositional semantics necessitate the adoption of the operator-particle theory for focus particles, which allows for anti-pied-piping mismatches. Under this theory, a semantically inert particle is adjoined within the scope of the operator. In this section, we discuss syntactic consequences of particle insertion, accounting for patterns of focus-targeting agreement and movement, before discussing the derivational timing of particle adjunction in §4.4.

We propose that particles are adjoined clitics, as explicitly claimed by Aoyagi (1998), targeting maximal projections. Although particles are semantically inert, they may introduce formal features. Formal features of both the particle (prt) and its sister XP will project to their mother. We refer to the result as a particle phrase and label it prt+XP in the general case. Suppose a particle optionally pronounced as only and bearing the feature [foc] adjoins to a DP sandwiches, as in our English example structure in 74 above. The resulting phrase only sandwiches will project the [foc] feature as well as features projected from sandwiches, such as the category [D] and its φ-features, as in 86.²⁶

(86)

 

The particle and its corresponding operator are then linked by the derivational operation Agree, which allows them to exchange additional formal feature values (see e.g. Chomsky 2000), as proposed in Lee 2004, 2005, Hirsch 2017, and Quek & Hirsch 2017.²⁷ Focus constructions in Lavukaleve (brought to our attention by Isaac Gould, p.c.) offer explicit evidence for this operation. In Lavukaleve, particles encliticize to focused arguments: the subject in 87a and object in 87b. Notice that the form of the particle inflects to reflect the φ-features of the constituent it adjoins to. Lavukaleve then also allows for the pronunciation of another marker in a fixed, postverbal position, which we analyze as the corresponding operator, and which also inflects to reflect the φ-features of the focused constituent.

(87)

Lavukaleve (Terrill 2003:277)

 

Such patterns of φ-agreement arise straightforwardly on our account. As illustrated in 86 above, the particle phrase will bear the formal features both of the particle (e.g. [foc]) and of its host, such as its φ-features. Agree between the particle phrase and its corresponding operator, based on their shared feature ([foc]), then allows for the particle phrase's φ-features to be copied onto the operator.

This approach to particle syntax may also serve to explain interactions between focus and case marking as in Kakataibo (Valle 2014) and Beria (Wolfe & Abdalla Adam 2018). Transitive subjects in each of these languages can be unmarked, but they appear with an optional ergative case marker especially when they are narrowly focused, as well as in cases of sentence focus, constituting a form of anti-pied-piping. (We do not reproduce this data here.) In these languages, targeting a subject for MSF—analyzed as adjoining an unpronounced particle—has the result of affecting the realization of case marking on the subject.

The adjunction of a focus particle may have other consequences for its host, by affecting the structural relationship between the host and its surrounding structure. The syntactic presence of adjoined particles may serve to explain the inability of focused phrases to undergo incorporation or to be visible for external morphological operations (Haiman 1988), or the ability of focus particles to shield nominals from what would otherwise be binding-theoretic violations (Heim 1998:242).

The syntactic visibility of features such as [foc] on the particle phrase is also key to our account for anti-pied-piping in phrasal movement. For concreteness, let us return to our basic English example with the particle phrase only sandwiches. Introducing a higher head that probes for the [foc] feature and moves its goal will result in movement of the particle phrase prt+DP, such as in a cleft.

(88)

It's [_prt+DP prt_only [sandwiches]_F] that Alex made__for Brie.
⇒ It's only sandwiches that Alex made for Brie.

Suppose furthermore that there is also a prt that introduces the [foc] formal feature but is unpronounced. Adjunction of this particle to sandwiches will lead to the appearance of sandwiches moving alone to become the cleft pivot, without an overt particle.

(89)

It's [_prt+DP PRT _∅ [sandwiches]_F] that Alex made __for Brie.
⇒ It's sandwiches that Alex made for Brie.

As we have discussed above in the case of overt focus particles, suppose a particle prt adjoins instead to a proper subpart of the logical focus, as in 90a. Probing for a formal feature introduced by the particle and projected by the particle phrase (e.g. prt+DP) will result in movement of a proper subpart of the logical focus, as in 90b. With nonpronunciation of both the operator and particle and together with independent subject raising, this derives the anti-pied-piping focus movement in the Hungarian example repeated here as 91. Recall that the entire VP with the object Hamlet reconstructed is within the scope of the interpreted operator op (assumed here to be just above vP) and thereby can contrast with other contextually relevant VP meanings such as 'swim'.

(90)

 

(91)

Anti-pied-piping in Hungarian focus movement (Kenesei 1998a:77)

 

Following a suggestion by Cable (2007, 2010b:Ch. 6), we propose that all A′-movement is, by definition, movement of particle phrases.²⁸ Under this proposal, the parallels observed between anti-pied-piping in particle placement and in phrasal movement—for example, in both being subject to leftmost requirements in many languages (discussed in further detail in §6)—fall out immediately: A′-movement such as focus movement is always movement of a particle phrase, although in many cases the relevant particle is unpronounced.²⁹

This view of A′-movement grows out of an influential analysis of pied-piping as particle phrase movement, developed by Tanaka (1999), Horvath (2000, 2007), Watanabe (2006), and Cable (2007, 2010a,b). Not accidentally, then, our proposal for particle syntax also allows for pied-piping mismatches. Particle placement may target a focus-containing phrase, as in 92a. A′-movement targeting the particle phrase then results in focus movement with pied-piping, as in 92b. This derivation followed by independent subject raising yields the Hungarian pied-piping example repeated in 93.

(92)

 

(93)

Pied-piping in Hungarian focus movement (Kenesei 1998b, ex. 13b)

 

In §6.4 below, we present parallels in the fine-grained behavior of pied-piping and anti-pied-piping that further motivate their unification.

4.3. Support from idiom chunks

Additional support for our proposal for anti-pied-piping involving a one-place covert operator comes from the fact that focus particles can appear on a proper subpart of an idiom chunk while taking the entire idiom as its logical focus, in an anti-pied-piping pattern. In example 94 from Japanese, the verb phrase is literally 'eat weeds' but idiomatically means to loiter or waste time along a path. The additive particle mo encliticizes to the object but associates with the predicate in its idiomatic meaning.

(94)

Japanese mo within an idiom chunk (Ohno 2003:248)

 

Under the proposal here, the pronounced particle itself has no semantic contribution and therefore the VP 'eat weeds' can be interpreted as a single constituent with its noncompositional interpretation, all within the scope of a covert additive operator and contrasted against other relevant VP meanings. In contrast, under the quantificational particle theory, the particle's surface sister (here: 'weeds') and its second argument (including 'eat') would be interpreted separately and then composed via the semantics of the particle, which only considers focus alternatives for its sister.

Anti-pied-piping focus movement can also target a subpart of an idiom chunk while retaining idiomatic interpretation. This is observed in the Hungarian example 95, where the VP 'scrape horsehide' retains its idiomatic meaning of sleeping. Similar examples are attested in Dutch (van Riemsdijk & Zwarts 1997 [1974]:18–19), German (Büring 1997:72, ex. 58), Czech (Lenertová & Junghanns 2007:355, ex. 21), and Hausa (Newman 2000:261), as also discussed by Fanselow and Lenertová (2011).

(95)

Hungarian focus fronting of part of an idiom chunk (Kenesei 1998a:85)

 

Under the operator-particle theory, there is no particular semantics that applies specifically to the fronted phrase, but there is instead a corresponding sentential operator that takes the entire focus in its scope. By reconstructing the fronted phrase to its base position, the entire VP 'scrape horsehide' can be interpreted as a unit and thereby idiomatically, within the scope of the covert operator involved in contrastive focus, and contrast as a whole with other VP meanings.

4.4. The timing of particle placement

We now turn to the question of when and how particles are introduced into the derivation. Recall that anti-pied-piping behavior makes reference to the linear order of constituents (§3.4) and furthermore also correlates with stress assignment in some languages, as we elaborate in §6 below. It follows that particle placement must make reference to structures that are linearized and possibly prosodified. But at the same time, particle placement cannot be entirely postsyntactic. As we just proposed, particle placement also feeds further syntactic processes by forming targets for movement and agreement.

Our solution will be to adopt a theory of cyclic structure building where structures are built without access to the phonology, and then undergo a process of spell-out at certain punctuated points in the derivation (Uriagereka 1999, Chomsky 2000, 2001, among others). Spell-out is triggered following the construction of particular, designated structures called phases. When a structure undergoes Spell-out, the pronunciation of its terminal nodes, their relative linear order, and its prosodic representation are calculated (see e.g. Dobashi 2003, 2010, Ishihara 2004, 2007, Fox & Pesetsky 2005, Kratzer & Selkirk 2007, Kahnemuyipour 2009, Sato 2012a). Further derivational steps may build on the output of Spell-out. Following Fox & Pesetsky 2005 and subsequent work, we take Spell-out to target complete phases including their specifiers, in contrast to the description in Chomsky 2001. We concentrate here on the effects of particle placement during Spell-out of the vP phase, which allow us to account for the patterns of anti-pied-piping in sentence focus and predicate focus documented above.³⁰

We propose that particles are adjoined during cyclic Spell-out, as López (2009) and Hartmann (2016) have claimed for abstract information-structural features.³¹ When a phase undergoes Spell-out, first its linear and prosodic representation is calculated. Particle adjunction then takes place via late adjunction (Lebeaux 1988, 1991), targeting a subpart of the structure that has already been built.³² As noted above, particles may adjoin directly to a focused phrase, or to a containing phrase (pied-piping) or to a contained phrase (anti-pied-piping). The determination of particle position can then make reference to the phase's linear and prosodic representation at this stage, allowing for the leftmost effects above.

Concretely, we summarize our proposal for particle placement in 96. We discuss and formalize the relevant notion of 'left-alignment' in §6 below.³³

(96)

Particle placement (preliminary; to be revised in §6): During phasal Spell-out, Late Adjoin the particle to a phrase that {is left-aligned/is preferably left-aligned/overlaps} with the logical focus.

As a final step during Spell-out, we propose that the newly derived particle phrase may optionally move to the edge of the phase. This is necessary in our framework of analysis for allowing the particle phrase to then move out of the current phase, due to a well-established restriction by which only material at the edge of a phase is accessible for further syntactic processes (see e.g. Chomsky 2000, 2001, Fox & Pesetsky 2005). This movement may also be covert; see also n. 27 above.

Another advantage of this late-adjunction approach to particle placement is that it allows us to account for the fact that particles do not disrupt selection between their host and its selector, despite the fact that particle phrases themselves are later syntactically visible for syntactic operations. See Aoyagi 1998:Ch. 2 for discussion of this tension. We see this as a case of derivational opacity: the strictly local structural relationship necessary for selection between, for instance, a transitive verb and its object is satisfied during the construction of the phase, prior to the introduction of particles during phasal Spell-out.

After one phase undergoes Spell-out, the syntactic derivation can continue, building on the result of the lower phase. Therefore, a particle phrase built during a lower phase's Spell-out will be visible for syntactic operations from above, including for movement and agreement, as we have seen. Ultimately, for the derivation to converge, the corresponding operator must be present and Agree with its particle(s), as discussed above, and the logical focus must be in the scope of the operator. (Association is also possible with a focus that has moved out of the scope of the operator, even without reconstruction, for some operators but not others; see Erlewine 2014.)

Further motivation for our proposal comes from opaque interactions between particle placement and other movements. Consider the Ishkashimi example in 97, where the object has scrambled over the subject and the subject bears the focus particle məs. Karvovskaya (2013) reports this sentence as 'marked but somewhat acceptable' with sentence-focus interpretation, which we indicate with '?'. The (relative) acceptability of 97 is puzzling given the strict leftmost requirement of anti-pied-piping particle placement in Ishkashimi (see 64): the particle targets the subject in 97, which is not leftmost within the material that makes up the logical focus.

(97)

Ishkashimi (cf. 38) (Karvovskaya 2013:88)

 

The possibility of 97 with its sentence-focus interpretation supports our proposal for the timing of particle placement in 96. We take the sentence-focus construal to involve focus on the vP event description, which includes the base position of the subject as well as the object, which in turn contains a reflexive that must be bound by the subject. When the vP phase undergoes Spell-out in 98a, the phase is first linearized in its basic SOV order, and then particle placement applies, targeting the leftmost constituent within the focus: the subject. Scrambling of the object then applies for independent reasons, in 98b, fronting the object across the subject to the phase edge, and potentially later to a higher position.

(98)

 

A similar interaction is also reported for Japanese. Consider example 99, where the object is marked with the focus particle dake and has been fronted across the subject. Kotani (2009) shows that examples of this form allow for a predicate-focus interpretation, as indicated in 99. Dash and Datta (2022) show this same interaction to hold in Hindi-Urdu and Bangla as well.

(99)

Japanese (based on Kotani 2009:30)

 

Recall that Japanese allows optional anti-pied-piping in predicate focus (see 30), with particle placement targeting a proper subpart of the VP. On the surface, however, the constituent targeted by particle adjunction in 99 is not a subpart of the logical focus (i.e. the VP).

Such examples are also accounted for straightforwardly under our proposal. We begin by linearizing the vP phase, including the base position of the subject, when it undergoes Spell-out in 100a. Particle adjunction here optionally targets a subpart of the focus, resulting in anti-pied-piping particle placement on the object. Subsequent movement operations, as in 100b, may target the resulting particle phrase.

(100)

 

Notice that we cannot explain such examples by appeal to postsyntactic lowering (see e.g. Embick & Noyer 2001). Such an approach may posit that particles are first adjoined directly to their logical focus but then lower after the end of the syntactic derivation onto its surface host. Contrary to fact, we would then expect movement of an object to a position higher in the clause to bleed particle placement in cases such as 99, or to feed particle placement on the object in the Ishkashimi sentence-focus example with object fronting in 97. In contrast, the theory developed here, which interleaves particle placement and movement operations, derives the attested counterbleeding and counterfeeding patterns above.

4.5. Argument/adjunct asymmetries

The empirical landscape laid out in §3 was restricted, in that the patterns of anti-pied-piping consistently involved core arguments, without adjuncts. While this is in part a function of the sources from which our survey was built, there is also evidence of a systematic asymmetry between arguments and adjuncts in anti-pied-piping.

Consider the Japanese examples in 101 below. Aoyagi (1998) and Ohno (2003:317–18) independently observe that when the focus particle appears on an object, as in 101a, both narrow-focus and anti-pied-piping predicate-focus readings are available, while in 101b, when the focus particle appears on an adverbial, only a narrow-focus reading is available.

(101)

Japanese (Aoyagi 1998:175)

 

Kenesei (1998a) observes a similar asymmetry between arguments and adjuncts in Hungarian focus movement. Although the object can be fronted to the preverbal focus position to express predicate focus, as shown here again in 102a, an adverb such as hangosan 'aloud' in the focus position (as in 102b) is compatible only with narrow focus on the adjunct.

(102)

Hungarian (Kenesei 1998a:77)

 

Hyman and Polinsky (2010) discuss a similar argument/adjunct asymmetry in the interpretation of the immediately after the verb (IAV) focus position in Aghem. With the object in IAV position, narrow object-focus and predicate-focus interpretations are both possible. However, if an adjunct occupies the IAV position, only a narrow adjunct-focus reading is available. If we analyze the IAV position in Aghem as involving movement to a dedicated position, these facts clearly parallel those in Hungarian, with the predicate-focus examples being another instance of anti-pied-piping.

The theory developed here—in which particle placement takes place at punctuated points in the derivation—allows us to account for this asymmetry, based on an approach to the argument/adjunct asymmetry proposed by Lebeaux (1988, 1991) and in much subsequent work. On this approach, adjuncts but not arguments can be introduced into the clause via late adjunction, during cyclic Spell-out (see n. 32). This offers a way of understanding the impossibility of adjunct anti-pied-piping: the adjunct is not yet present in the structure at the point of vP Spell-out, when particles for predicate focus are adjoined. Since these elements are not present, they cannot be targeted for particle adjunction. We leave a fuller investigation of such argument/adjunct contrasts and their consequences for derivational timing for future work.

5. Alternative analyses

We briefly discuss two alternative analyses for anti-pied-piping. The first treats a subpart of the logical focus as carrying a separate, marked information-structural status. The second treats apparent anti-pied-piping as the result of independent movements out of the logical focus. We will see that these approaches could potentially account for a limited set of anti-pied-piping examples, but they fail to explain the widespread possibility of anti-pied-piping and its crosslinguistic tendencies, which our proposal explains.

5.1. Nesting

We first discuss the possibility that anti-pied-piping may involve structures where a subpart of the logical focus independently bears a marked information-structural status. We first observe that it is possible to nest a narrow focus within a larger focus, as discussed in work such as Krifka 1991 and Neeleman & Szendrői 2004 and illustrated in 103. For expository purposes, we annotate the two foci 'F1' and 'F2'.

(103)

Nested foci (based on Krifka 1991:131)
Ted was behaving strangely at last night's party. At one point, he went back to his room to take a nap. He also [drank only [water]_F1 (all night)]_F2.
I wonder if he's feeling okay.

In 103, we can detect the semantic contribution of two overt focus particles, each with their expected semantics: only associates with water (F1), contrasting with other beverages available at the party, while also associates with the entire VP (F2), contrasting with going to take a nap.

Now suppose we have a structure with nested foci akin to 103, but where overt focus morphosyntax applies only to the embedded focus F1 and a focus-sensitive semantics is clearly detectable only in relation to F2. This would lead to the appearance of anti-pied-piping. However, under this approach we expect the syntax/semantics associated with the embedded focus F1 to be independent of the syntax/semantics associated with the larger focus F2, as in 103 above. Instead, in focus particle anti-pied-piping, the choice of overt focus particle (by hypothesis, adjoined to and narrowly associating with F1) correlates with the semantics that associates with the higher focus (F2). The Japanese examples in 104 with predicate-focus interpretation illustrate this correlation between particle form and interpreted semantics.³⁴

(104)

Semantics associated with wide focus correlates with embedded particle choice (Ohno 2003:324, 336)

 

Such correspondences are unexpected if apparent anti-pied-piping is generally attributed to the availability of nested foci, and they instead motivate our own operator-particle proposal, where a syntactic dependency (Agree) ensures a correlation between the choice of overt particle and the choice of semantically contentful but unpronounced operator.

Next we turn to cases of anti-pied-piping in phrasal movement without overt particle placement. Such examples appear to be more amenable to a nesting analysis: as long as a subpart of the focus has some marked information-structural status (as a 'focus' or otherwise) and the language can front such material, it may be moved out of the larger focus. This movement trigger may be a loose pragmatic notion such as 'emphasis', which has been associated with fronting in many languages (see e.g. Frey 2010).³⁵ It is difficult to rule out the possibility, suggested by a referee, that some such pragmatically motivated movement may underlie some of our examples of movement anti-pied-piping in §3.3, especially as the descriptions of the relevant pragmatic notions (see n. 35) are in many cases not precise enough to confidently determine whether they apply narrowly to the fronted constituent or to the larger focus as a whole.

There are, however, at least two considerations that cast doubt on the idea that movement anti-pied-piping is generally due to the fronting of an embedded, independently pragmatically marked constituent. First, under this type of nesting account, the requirement in many languages for anti-pied-piping to target a leftmost subpart of the logical focus (discussed further in §6) goes unexplained; in principle, any subpart of the logical focus may be subject to 'emphasis', depending on the speaker's communicative goals. Second, anti-pied-piping in both particle placement and phrasal movement is generally not described as semantically or pragmatically marked as compared to non-anti-pied-piping constructions,³⁶ and in particular is to our knowledge never described as licensed only in a particular, more complex discourse structure.

Furthermore, in cases where the semantics associated with fronting is clearer and more precisely described, we are able to construct forceful arguments against the nesting account. This is the case for movement to the preverbal focus position in Hungarian (see n. 1). As Surányi (2018:248–49) discusses, the exhaustive identificational semantics associated with focus movement in Hungarian applies to the entire logical focus in anti-pied-piping examples, and not narrowly to the constituent that is moved to the preverbal focus position. This is unexpected under the nesting account, where the observed focus fronting should constitute an independent focus movement of an embedded focus, with its associated conventional semantics, and would not lead to that same semantics instead applying to the larger focus.

We conclude that the availability of information-structural nesting—with a subpart of the larger focus being an independent narrow focus or having some other marked pragmatic status—fails to provide a general explanation for patterns of anti-pied-piping in both particle placement and phrasal movement.

5.2. Movement out of the focus

Next we discuss the possibility that anti-pied-piping mismatches are only apparent, due to independent movements out of the focused constituent that make focus morphosyntax appear to target only a subpart thereof.

This hypothesis is articulated most clearly for focus particle anti-pied-piping in Japanese by Kotani (2008, 2009). Consider the case of anti-pied-piping in predicate focus in 105. Kotani proposes that the particle adjoins directly to its logical focus (VP), but the verb then optionally moves out of the VP to T in order to form the verbal complex. This results in the appearance of object attachment, as schematized in 106.

(105)

Anti-pied-piping in Japanese predicate focus (Ohno 2003:324)

 

(106)

 

If this movement does not take place, the focus particle stays transparently adjacent to its focus, with a process akin to do-support taking place to host the tense affix, as attested in 30a above.

Movement out of the pronounced particle's sister could also explain the anti-pied-piping pattern in the English example 107, where preverbal only—which must associate with its sister or a subpart thereof (Jackendoff 1972, Erlewine 2014) and which we analyze as the realization of the operator (see 74)—associates with the entire proposition.

(107)

 

Assuming that the subject the judge originated within vP and moved out (see n. 21), we may straightforwardly think of this as a case of only associating with the content of its sister vP with the subject reconstructed (Kayne 1998:159, n. 75, Erlewine 2014:82). Krifka (1991:142–43) offers a similar analysis for cases of German predicate focus, where a portion of the VP has moved out of the surface sister of a focus particle.³⁷

However, such cases where anti-pied-piping patterns can be fully attributed to the effects of independently motivated movements are few and far between. For example, returning to Japanese, the approach in 106 above fails to naturally extend to other examples such as cases of anti-pied-piping in sentence focus, which Kotani does not discuss. To explain the word order in examples such as 108, not only the verb but also the object would have to move descriptively to the right, which has no independent motivation in the language.

(108)

Anti-pied-piping in Japanese sentence focus (based on Aoyagi 1998:151, 2006:123)

 

Fanselow (2004:29–35) similarly discusses anti-pied-piping in phrasal movement as involving remnant movement—that is, with some material independently moving out of the logical focus before it undergoes movement—but concludes that such an approach is not feasible, as the necessary movements are otherwise unattested. In contrast, our operator-particle theory does not require an overt particle to take the logical focus as its sister at any point in the derivation, avoiding the need to hypothesize such otherwise unmotivated movements.

This alternative approach to anti-pied-piping mismatches also faces difficulty with the multiple exponence of focus particles. Recall that focus particles can appear multiply in anti-pied-piping, as in 82 above, and that subsentential particles and their corresponding operators can be pronounced simultaneously in some languages, as noted in §4.1 (see 87). Such examples are also modeled straightforwardly in the operator-particle theory.

6. Particle placement and left-alignment

We now return to the question of how languages determine which subconstituent of the focus to target for focus particle placement and/or movement in anti-pied-piping. We saw in §3.4 that anti-pied-piping in many languages exhibits a leftmost effect: MSF often targets the leftmost subphrase of the logical focus, with some variation in the strength of this requirement. In this section, we discuss this aspect of anti-pied-piping in further detail. We motivate a general description for such leftmost requirements that will relate particle position to phrasal stress assignment in stress languages, but that also extends to languages without phrasal stress. We then show how this description naturally explains certain parallels between pied-piping and anti-pied-piping behavior.

6.1. Exceptions to leftmost requirements

Our starting point is the observation that there appears to be a common class of exceptions to the leftmost requirements described above: certain nominals—roughly corresponding to those that are indefinite, given, or less informative—are skipped for the evaluation of 'leftmost'. We illustrate this first with Czech focus fronting: for sentence focus, object fronting is blocked across the subject Linda in 109a, but is allowed across an indefinite subject in 109b. Example 109a can only be interpreted with narrow focus on the object.

(109)

Czech (Radek Šimík, p.c., based on Lenertová & Junghanns 2007:356)
[Q: What's new?]

 

Similar facts are reported for German, where the effect has been studied experimentally by Féry and Drenhaus (2008) and Wierzba and Fanselow (2020). Here we present the results of an acceptability rating study in Féry & Drenhaus 2008. In this experiment, participants were asked to rate audio recordings of question-answer pairs on a 1–6 scale, where 6 is most natural. Example 110 is one representative set of stimuli to test the acceptability of object fronting in a sentence-focus context, with accented words in small caps. Notice that the subject is a pronoun in 110a, an unaccented DP in 110b, and an accented DP in 110c. In all cases, the fronted object was accented. The numbers on the right correspond to average ratings for each condition exemplified by the item at left.

(110)

German (Féry & Drenhaus 2008:24–25)

 

The results reflect a clear and statistically significant difference between these conditions (see Féry & Drenhaus 2008:25–26, n. 10): object fronting is highly degraded across an accented subject (110c) but acceptable across a pronominal subject (110a) or otherwise deaccented subject (110b), which may be interpreted as being given under accommodation.

We observe a similar effect in Japanese. Recall that for sentence focus in Japanese, the additive particle mo could be placed on the subject or object, as we saw in 66 above. In 111, we have modified example 66 so that the subject is a nonspecific indefinite. Speakers then disprefer the placement of the particle on the nonspecific indefinite subject.³⁸

(111)

Japanese (based on 66 above from Aoyagi 1998, 2006)

 

These contrasts show that a certain class of elements are skipped for the evaluation of leftmost requirements. Furthermore, these elements—which are indefinite, given, or less informative—form a natural class in that they avoid phrasal stress in stress languages. This suggests a connection between the process of particle placement and phrasal stress assignment.

6.2. Two stress-based hypotheses

In this section, we consider proposals for particle placement that explicitly make reference to stress information, motivated by the data above.³⁹ We conclude, however, that such approaches are inadequate in the general case, especially considering the possibility of anti-pied-piping in languages without phrasal stress. This sets the stage for our own proposal, in §6.3, which maintains the connection to stress assignment discussed here but also extends to languages without phrasal stress.

Recall that we propose that particle placement takes place during cyclic Spell-out (§4.4), and therefore can make reference to information associated with phonological form (PF), including linear order and prosodic structure. This architecture allows for hypothesis 1 in 112.

(112)

Hypothesis 1 for particle placement given stress information: Particle placement in anti-pied-piping targets the phrase that bears main stress within the focus.

On this approach, cases of apparent optionality would have to be described as fed by independent manipulations in the choice of main stress placement.

This description is conceptually attractive, as it relates anti-pied-piping to the phenomenon of focus projection, which relates the positions of main prosodic prominence and logical focus. (See Arregi 2016 for an overview.) Indeed, a number of previous authors have suggested an explicit comparison or explanatory link between the process of focus projection and anti-pied-piping; we are aware of such discussions in Zsámboki 1995, Choe 2002, Szendrői 2003, Fanselow 2004, F. Schwarz 2007:147–49, Yoshimura 2007, Kotani 2009, and Karvovskaya 2013.

However, one immediate challenge for hypothesis 1 comes from cases where the target of MSF is clearly not the most prominent. Consider the case of predicate focus in German ditransitive clauses, as in 68 above. As noted in Fanselow 2004:23 and Fanselow & Lenertová 2011, both objects in such cases receive pitch accents, but it is the second object (the goal) that bears a more prominent pitch accent. Nonetheless, it is the first (leftmost) object that is targeted for anti-pied-piping, as in 68. Another concrete challenge comes from patterns of anti-pied-piping in Japanese transitive clauses with sentence focus. As discussed in Ishihara 2000, 2001, and Sato 2012b, nuclear stress in Japanese generally targets the immediately preverbal phrase. Nonetheless, as shown in 66 above, MSF may target either the subject or the object, with some speakers in fact preferring subject placement (see n. 16). See also Kahnemuyipour & Megerdoomian 2017 for explicit arguments against tying focus particle position to main stress in Eastern Armenian, although without discussion of anti-pied-piping.

Given these issues, we turn to hypothesis 2 in 113, which does not refer to the main stress and also allows for the observed crosslinguistic variation in the strength of leftmost effects.⁴⁰

(113)

Hypothesis 2 for particle placement given stress information: Particle placement in anti-pied-piping targets {the leftmost/preferably the leftmost/any} accented subphrase of the focus.

An immediate challenge that affects hypothesis 2 (as well as hypothesis 1) comes from the possibility of anti-pied-piping in languages that do not utilize accents or stress, unlike many of the languages discussed thus far. One such example is Hausa, a tone language that lacks accents of the relevant sort and where focus is not generally prosodically marked.⁴¹ Nevertheless, Hausa demonstrates anti-pied-piping behavior similar to that in other languages with stress and/or accent. Sentence focus may be expressed by fronting the subject before the focus marker nèe, as in 114a. However, in certain circumstances, as in 114b, the object is fronted instead; it is further noted that fronting the subject is not an available option in 114b.

(114)

Hausa (Hartmann & Zimmermann 2007b:385)

 

The argument skipped for the evaluation of leftmost in 114b again seems to be of the sort that is skipped in Czech, German, and Japanese above: one that is indefinite, given, or less informative.⁴²

Among other languages that demonstrate anti-pied-piping discussed here, Schwarz (2009) shows that, in the Oti-Volta Gur languages described above (see 33 and 45), there are no prosodic effects of focusing. Focus is also not reflected prosodically in Wolof (see 61) (Rialland & Robert 2001).

We see that anti-pied-piping behavior with leftmost effects—including its familiar exceptions—may be observed in languages such as Hausa that lack phrasal accents and do not prosodically realize focus. This suggests that the crosslinguistically common leftmost effect in anti-pied-piping should not be described as parasitic on surface phonological information such as accent or stress, but should instead rely on a more abstract representation that might feed subsequent phonological processes.

6.3. Proposal: ●-assignment and left-alignment

We put forward a theory that takes hypothesis 2 above to be on the right track—that in languages with phrasal stress, anti-pied-piping appears to target the leftmost accented subphrase of the focus—but without making reference to stress or accents in the general case. We suggest that particle placement is sensitive to the same information that languages with phrasal stress use to determine which elements receive stress and which do not. For expository purposes, we refer to this information as ●-marking (read: bullet), a rule for which is given in 115.

(115)

●-assignment: At phasal Spell-out, assign a ● to each phrase that is not a part of the extended projection that contains the phase head.

Our rule for ●-marking in 115 draws on a long line of work arguing that information relevant for nuclear stress determination is assigned cyclically throughout the derivation (see e.g. Bresnan 1971, 1972, Legate 2003, Adger 2007, Kahnemuyipour 2009, Sato 2012b, Richards 2018). Although these prior works differ in the details of this process, ●-marking and the rule in 115 are meant to be theory-neutral abstractions standing in for whatever its proper characterization turns out to be.⁴³

Importantly for our current purpose, ●-marking is abstract in a way that phrasal stress in and of itself is not. ●-assignment may feed a realization rule like that in 116—giving rise to what we would call a language with phrasal stress—but, crucially, ●-assignment takes place independently of the existence of such a rule in a language. In languages like Hausa, for instance, ●-assignment takes place, but there is no rule like that in 116.

(116)

●-realization: Each element marked with ● is realized with phonological prominence.

One possibility, suggested by a referee, is that ●-marking may instead be relevant for prosodic phrasing in Hausa, following Féry 2013. That is to say, the rule in 116 may not be the only way that ●-marking can be realized and learned. The strong hypothesis would be that ●-marking is always realized in some form, with the inventory of possible reflexes of ●-marking being a question for future research.

We now put forward our general statement for particle placement, in 117, making reference to a ●-relativized notion of left-alignment, in 118.⁴⁴

(117)

Particle placement (revised; subsumes 96): During phasal Spell-out, Late Adjoin the particle to a phrase that {is ●-left-aligned/is preferably ●-left-aligned/overlaps} with the logical focus.

(118)

●-relativized left-alignment: X and Y are ●-left-aligned if the leftmost ●-marked phrase in X and the leftmost ●-marked phrase in Y are left-aligned.

We leave as an open question whether a language's choice between the strong, weak, or free variant of 117 can be predicted by independent properties of the language.

As we have seen earlier in this section, certain types of phrases are crosslinguistically commonly skipped for the determination of particle placement.⁴⁵ We propose that these elements resist ●-marking.

(119)

●-avoidance: Indefinite, given, or less informative elements are not assigned a ●.

This rule explains why these exceptional elements are commonly destressed or deaccented: 119 bleeds 116. However, the calculation of ●-assignment is independent of the presence of a realization rule such as 116, explaining the availability of similar leftmost effects in languages without phrasal stress such as Hausa.

6.4. Pied-piping and anti-pied-piping

Recall that the operator-particle theory we propose also allows for the derivation of pied-piping. In this section, we show that our proposal for particle placement above—motivated by the leftmost effects observed in anti-pied-piping—can also account for observed restrictions on pied-piping. This further supports our account, which views pied-piping and anti-pied-piping as a unified phenomenon.

Pied-piping in many languages also exhibits a form of leftmost requirement. For instance, as shown in 120, pied-piping in English interrogative wh-movement requires the logical focus—the locus of variation across semantic alternatives, that is, the wh-word—to be at the left edge of its pied-piped constituent. The pair in 121 furthermore shows that the restriction is sensitive to linear position, rather than depth of embedding.

(120)

Leftmost requirement in English pied-piping

a. [Whose picture] did you frame __ ?

b. *[A picture of whom] did you frame __ ?

(121)

 

We propose that this behavior can also be explained by our proposal for particle placement in 117 above. In the case of pied-piping, a particle adjoins to a constituent that contains the logical focus (here, the wh-word) and that is ●-left-aligned with the wh-word. Consider 122a,b, which represent the base structures for 120 prior to particle adjunction at vP Spell-out.

(122)

a. you frame [_DP whose picture]

b. you frame [_DP a picture of whom]

In 122a, the constituent marked DP is ●-left-aligned with the wh-word, so an unpronounced particle prt can adjoin to it. Probing for the resulting particle phrase prt+DP leads to what we describe as wh-movement with pied-piping, following the intuition for pied-piping developed in works such as Tanaka 1999, Horvath 2000, 2007, Watanabe 2006, and Cable 2007, 2010a,b. In contrast, in 122b, DP is not ●-left-aligned with the wh-word and therefore cannot host a particle according to the strict formulation of 117.

Our rule for particle placement based on ●-relativized left-alignment also predicts that this leftmost requirement on pied-piping will tolerate certain exceptions, in allowing it to skip material that does not bear stress. This prediction is borne out. For example, in English, a light preposition like to may intervene between the left edge of the pied-piped constituent and the wh-phrase, as in 123, but anything heavier, such as a lexical noun (as in 120b and 121b), may not.

(123)

Not quite leftmost in English wh pied-piping
[To [which]_F student's friend]_MSF did you speak __ ?

Anti-pied-piping with Latin que, discussed in §3.2 above, also tolerates exceptions of precisely this form. Que generally follows one word at the left edge of its logical focus, but skips monosyllabic prepositions.

(124)

Not quite leftmost in Latin que anti-pied-piping (Carlson 1983:73)

 

The fact that these phonologically light elements are ignored for the evaluation of leftmost effects in both pied-piping and anti-pied-piping is a welcome consequence of our proposal for particle placement in 117 and its application to both pied-piping and anti-pied-piping. We expand on our rule in 119 above to propose that these phrases headed by phonologically light prepositions avoid ●-marking. Their presence at the left edge will thus be ignored for the evaluation of ●-left-alignment, allowing for the 'not quite leftmost' pattern of pied-piping in 123 and anti-pied-piping in 124.

Not all instances of pied-piping are subject to this leftmost requirement. For example, Russian does not require interrogative wh-words to be left-aligned within their pied-piped constituents (see 125). Similarly, relative pronouns in English need not be left-aligned (see 126), unlike interrogative wh-words.

(125) No leftmost requirement in Russian pied-piping (Heck 2008:79)

 

(126) No leftmost requirement in English relative pronoun pied-piping (Ross 1967:198)
Reports [_RC [the height of the lettering on the covers of which] the government prescribes __] should be abolished.

Recall that our rule for particle placement allows for parametric variation. English wh-interrogatives require ●-left-alignment between the wh-word and the target for particle placement (and hence, the target for movement), but Russian wh-interrogatives and (certain types of) English relative clauses do not. This distinction parallels the variation observed in anti-pied-piping particle placement in §3.4. The ability of our proposal for particle placement to account for these leftmost effects, including their variation and their exceptions, in a uniform manner in both pied-piping and anti-pied-piping strengthens our view that the two forms of mismatches should be treated together as a unified phenomenon, reflecting different options for particle placement.

7. Discussion and conclusion

The study of focus has featured prominently in linguistic theory, in part because choices of focus placement have consequences for interpretation, morphosyntax, and prosody, naturally leading to questions of grammatical architecture and modularity. In this article, we described anti-pied-piping, a form of mismatch between the morphosyntax and semantics of focus where focus morphosyntax targets a proper subpart of the interpreted position of focus. Anti-pied-piping is quite widely attested crosslinguistically—with examples identified here in over sixty languages from over forty distinct subfamilies or genera—with some notable parallels to well-studied pied-piping behavior. In particular, both mismatches require alignment at the left edge between the logical focus and the corresponding morphosyntactically marked element in many languages.

The details of anti-pied-piping behavior motivate a theory of particle placement in which particles are introduced into the syntactic structure independent of their corresponding semantic operator (the operator-particle theory), at certain punctuated points in the derivation (Spell-out) where PF-branch information such as statements of linear order are accessible to the syntax. The cyclic Spell-out model of grammar (Uriagereka 1999, Chomsky 2000, 2001, among others) adopted here allows us to account for certain opaque interactions between particle placement and scrambling, as well as to unify the anti-pied-piping behavior of overt focus particle placement with anti-pied-piping in focus movement, following previous work on the syntax of pied-piping.

These facts and the resulting theory developed here have important further consequences for the theory of grammar, especially regarding the nature and behavior of A′-movement. Following Tanaka 1999, Horvath 2000, 2007, Watanabe 2006, and Cable 2007, 2010a,b, we take A′-movement to be movement of particle phrases built from a process of particle placement. The theory thus explains aspects of the classic A/A′-distinction in terms of the timing of particle placement (see also Safir 2019): for instance, as particle placement takes place at phasal Spell-out, we predict that a constituent cannot undergo A′-movement until a containing phase is complete and undergoes Spell-out. Furthermore, if movement operations take place as soon as possible, all things being equal, we predict A-movement (movement not contingent on particle placement) to precede A′-movement.

In addition, in the final section we discussed parallels between particle placement position and positions of prosodic prominence, and sketched an approach to both sets of facts that indirectly derives their parallels. Again, as A′-movement is always movement of a particle phrase, we predict that A′-movement can only target phrases that can bear pitch accents in languages with phrasal accents, as has been independently motivated in work such as Cheung 2009 and Branan 2018. We leave the exploration of these and other consequences of the proposal here for future work.