2.1 Basic clause structure

Chuj is a verb-initial language with VSO and VOS as possible basic word orders. Nominal arguments in Mayan languages are cross-referenced with ergative-absolutive alignment through Set A (ergative) and Set B (absolutive) markers on the predicate. The Set A and Set B markers are shown in (5) (Hopkins Reference Hopkins1967, Domingo Pascual Reference Domingo Pascual2007, Buenrostro Reference Buenrostro2009).

1. (5) Ergative and absolutive marking:

Set A markers are also used to mark possessive agreement on nominals:

1. (6) Set A as possessor agreement:

   s-pat   winh  hin-mam

   A3-house cl.masc A1s-father

   ‘my father's house’ (Buenrostro Reference Buenrostro2009: 214)

Classifiers in Chuj cooccur with nominals or appear alone and function as pronouns. See Bielig (Reference Bielig2015) and Royer (Reference Royer2018) for discussion of nominal classifiers in San Mateo Ixtatán Chuj.Footnote ³

(7)  Nominal classifiers in Chuj:

Pre-verbal tense-aspect markers (TAM) in Chuj are shown in (8); see Buenrostro (Reference Buenrostro2007), Carolan (Reference Carolan, Pomerleau and Gendron-Pontbriand2015), Coon and Carolan (Reference Coon and Carolan2017) for details. The majority of the examples discussed below involve perfective aspect. Wherever there is an aspect-related interaction with the uses of wh-words studied here, this is noted explicitly and motivating examples are shown.

(8)  TAM markers:

Examples (9–10) show basic transitive and intransitive sentences. Verbs take a status suffix (intransitive itv -i; transitive tr -V’, -a’ in these examples) when they occur at intonational phrase boundaries or utterance-finally.

1. (9) Basic transitive sentences:

   1. a. Tz-ach-in-chel-a’.

      impf-B2-A1s-hug-tv

      ‘I hug you.’                 (Coon and Carolan Reference Coon and Carolan2017: 7)

   2. b. Ix-∅-in-wa’   ixim   wa'il.

      pfv-B3-A1s-eat cl.grain tortilla

      ‘I ate the tortilla.’

2. (10) Basic intransitive sentences:Footnote ⁴

   1. a. Ix-onh-ulek’-i.

      pfv-B1p-come-itv

      ‘We came.’

   2. b. Ol-∅-wa’   ix.

      prosp-B3-eat cl.fem

      ‘She will eat.’

The full template for a verbal predicate is shown in (11). This template presents the maximal morpheme combination; for example, as noted above, intransitive verbs will lack a Set A (ergative) marker and the status suffix will not appear unless it is at an intonational phrase edge.

1. (11) A template for Chuj verbal predicates: (Coon and Carolan Reference Coon and Carolan2017: 5)

   TAM — Set B — Set A — root — voice — status suffix

As is common in many Mayan languages, $\bar{\rm A} $-extraction of subjects of transitive clauses triggers a change to verbal morphology in that clause. This construction is called Agent Focus (AF) in Mayanist literature (see Stiebels Reference Stiebels2006, Norcliffe Reference Norcliffe, Avelino, Coon and Norcliffe2009, Coon et al. Reference Coon, Pedro and Preminger2014, and references therein). AF verbs can be identified by the lack of a Set A agreement marker and the addition of an AF suffix, -an. We additionally observe the intransitive status suffix on the verb, rather than the transitive. Relevant for this paper, we observe AF in transitive subject wh-questions (12c) and in transitive subject relative clauses (see (14) below).

1. (12) Agent Focus in transitive subject questions:

   1. a. Mach ix-∅-ulek’-i?

      who  pfv-B3-come-itv

      ‘Who came?’ intransitive subject question

   2. b. Mach ix-∅-w-il-a’?

      who pfv-B3-A1s-see-tv

      ‘Who did I see?’ transitive object question

   3. c. Mach ix-in-il-an-i?

      who  pfv-B1s-see-af-itv

      ‘Who saw me?’ transitive subject question

2.2 Headed relative clauses

Since one focus of our study here is free relatives formed with wh-words, we briefly discuss the structure of headed relative clauses in Chuj, which generally do not involve wh-words. Relative clauses in Chuj are simply gapped clauses preceded by the nominal head they modify. Some examples are given in (13). For clarity, the head nouns in the examples in this section are underlined.

1. (13) Headed relative clauses:

   1. a. ix   unin [_RC ix-∅-ulek’-i]

      cl.fem child    pfv-B3-come-itv

      ‘the girl who came’

   2. b. jun (ch'anh)  libro [_RC ix-∅-w-awtej]

      one cl.book book   pfv-B3-A1s-read

      ‘one book that I read’

As with question formation in (12c), transitive subject relativization triggers Agent Focus on the embedded verb, as in (14).

1. (14) Transitive subject relative clause triggers Agent Focus, see (12c):

   winh  unin [_RC ix-∅-man-an  ixim  pastel]

   cl.masc child   pfv-B3-buy-af cl.grain cake

   ‘a boy who bought the cake’

Unlike headed relative clauses in English, relative clauses in Chuj cannot be introduced by an overt complementizer, such as English that. The examples in (15), based on (13) above, show that wh-words cannot be used as relative pronouns in argument relatives.Footnote ⁵

1. (15) Relative clause cannot be introduced by relative pronoun:

   1. a. * ix     unin [_RC mach ix-∅-ulek’-i]

      cl.fem child   who pfv-B3-come-itv

      Intended: ‘the girl who came’   see (13a)

   2. b. * jun (ch'anh)  libro [_RC tas ix-∅-w-awtej]

      one cl.book  book   what pfv-B3-A1s-read

      Intended: ‘one book that I read’   see (13b)

In the case of adjunct relatives, however, wh-words can be used as relative pronouns at the edge of a headed relative clause. This is exemplified in (16) below. This same pattern has been described for Yucatec Maya by Gutiérrez-Bravo (Reference Gutiérrez-Bravo2013): wh-relative pronouns are ungrammatical in argument relativization but possible in adjunct relativization. Example (16a) shows an adjunct wh-word relative pronoun ‘where,’ whereas (16b) shows relativization over the object of the preposition et’ ‘with.’Footnote ⁶ Fronting of the adjunct et’-phrase in (16b) is accompanied by secondary fronting of the wh-word, reversing the order of the wh-word with respect to its pied-piped preposition, as is common in other Mayan languages (Smith-Stark Reference Smith-Stark1988, Aissen Reference Aissen1996, Coon Reference Coon2009, a.o.).

1. (16) Wh-relative pronoun possible in adjunct relativization:

   1. a. Tz-in-kot  t'a   jun lugar [_RC (b'ajtil)  tz-∅-al-chaj  Español].

      impf-B1s-come  prep one place  where  impf-B3-speak-psv  Spanish

      ‘I come from a place where Spanish is spoken.’

   2. b. Ix-∅-w-ilelta   winh  unin [_RC [mach y-et’]   ∅-och  ix  Malin].

      pfv-B3-A1s-meet cl.masc child   who A3s-with B3-help cl.fem Maria

      ‘I met the boy who Maria helped.’ (lit.: ‘the boy with whom Maria helped’)

2.3 Question formation

In this section we present what can be thought of as the canonical use of wh-words, that of constituent question formation. Example (17) shows wh-questions using mach ‘who’ and tas ‘what.’

1. (17)

   1. a. Mach ix-∅-ulek’-i?

      who   pfv-B3–come-itv

      ‘Who came?’

   2. b. Tas   ix-∅-a-man-a’?

      what pfv-B3-A2s-buy-tv

      ‘What did you buy?’

Wh-question formation generally involves the fronting of a wh-phrase to pre-verbal position, leaving a post-verbal gap. (We discuss exceptions to this fronting requirement later in the section.) Recall from the discussion of example (12) above that when the fronted wh-word is a transitive subject, the verb will be in the Agent Focus form.

Some examples of wh-questions with other wh-words are given in (18).

1. (18) Examples of other wh-words:

   1. a. B'ak'in ix-∅-ulek’  ix   Malin?

      when  pfv-B3-come cl.fem Maria

      ‘When did Maria come?’

   2. b. B'ajtil  ix-∅-a-man-a’?

      where  pfv-B3-A2s-buy-tv

      ‘Where did you buy it?’

   3. c. Tasyu'uj  ix-∅-el  ix  Malin?

      why  pfv-B3-leave  cl.fem  Maria

      ‘Why did Maria leave?’

   4. d. Jay-wanh  heb’  winh  unin  ix-∅-ulek’-i?

      how.many-cl.num  pl  cl.masc  child  pfv-B3-come-itv

      ‘How many boys came?’

   5. e. Jay-e’  lapis  ix-∅-a-man-a’?

      how.many-cl.num  pens  pfv-B3-A2s-buy-tv

      ‘How many pens did you buy?’

In this article, we will concentrate on wh-arguments involving the wh-words mach ‘who’ and tas ‘what’ as in (17). Complex wh-phrases akin to the English which/what boy or which/what girl can be formed by adding a nominal domain to mach.Footnote ⁷

1. (19) Mach can take a domain:

   1. a. Mach winh   ix-∅-ulek’-i?

      who  cl.masc pfv-B3-come-itv

      ‘Who_masc came?’

   2. b. Mach winh  unin ix-∅-k-il-a’?

      who  cl.masc child pfv-B3-A1p-see-tv

      ‘Which boy did we see?’

When a plural wh-phrase is constructed, it may optionally be marked with the plural marker -tak, as in (20a). Animate individuals may also be marked by the (animate-only) plural marker heb’, as in (20b). If the wh-word is explicitly marked as plural, the noun must be as well, as in (20c).

1. (20) Mach can be pluralized in two different ways:

   1. a. Mach-tak ix-∅-ulek’-i?

      who-pl  pfv-B3-come-itv

      ‘Who_pl came?’

   2. b. Mach heb’ winh  winak ix-∅-mak’-an cham nok’    tz'i’ chi

      who  pl  cl.masc man  pfv-B3-hit-af  dead  cl.animal dog dem

      ix-∅-el-i?

      pfv-B3-leave-itv

      ‘Which of the men who killed the dog left?’ (Buenrostro Reference Buenrostro2009: 210)

   3. c. Mach-tak *(heb’) winh  unin ix-∅-ulek’-i?

      who-pl  pl    cl.masc child pfv-B3-come-itv

      ‘Which boys came?’

As with mach, an inanimate nominal domain can be added to tas ‘what’ to create a modified wh-phrase, as in (21).

1. (21) Tas can take an inanimate nominal domain:

   Tas  libro-al   ix-∅-y-awtej  ix   Malin?

   what  book-nml  pfv-B3-A3-read cl.fem Maria

   ‘Which book did Maria read?’

Mach can also be used to form inanimate complex wh-phrases, although more often tas is used for this purpose, as in (21). Note, however, that tas cannot be used with an animate domain, as shown in (22b).

1. (22) Mach can take inanimate domain; tas cannot take animate domain:

   1. a. Mach libro-al  ix-∅-y-awtej  ix  Malin?

      what book-nml pfv-B3-A3-read cl.fem Maria

      ‘Which book did Maria read?’

   2. b. *Tas winh ix-∅-ulek’-i?

      what cl.masc pfv-B3-come-itv

      Intended: ‘Who_masc came?’

If a wh-phrase is not fronted, the result is an echo question like (23a).Footnote ⁸ Such a question cannot be embedded under a question-embedding predicate such as ojtak ‘know’, as shown in (23b). Only questions that involve wh-fronting, as in (23c), are ‘true’ questions that can be embedded.

1. (23) Questions without fronting are echo questions and cannot be embedded:

   1. a. Ix-∅-ulek’  mach?

      pfv-B3-come who

      ‘Who came?’ (echo question)  (see 17a)

   2. b. * K-ojtak  [ix-∅-ulek’ mach].

      A1p-know pfv-B3-come who

      Intended: ‘We know who came.’

   3. c. K-ojtak  [mach ix-∅-ulek’-i].

      A1p-know who  pfv-B3-come-itv

      ‘We know who came.’

Question formation involves the optional pied-piping of additional material along with the wh-word to the front of the question. When pied-piping occurs, secondary fronting takes place.

1. (24) Pied-piping with secondary fronting vs. preposition stranding:Footnote ⁹

   1. a. [Mach y-et’(ok)] ix-ach-och-i?

      who  A3-with  pfv-B2p-help-itv

      ‘Who did you help?’

   2. b. Mach ix-ach-och  y-et’(ok)?

      who  pfv-B2p-help A3-with

      ‘Who did you help?’

2. (25)

   1. a. [Tas y-et’(ok)] ix-∅-tajn-i   ix   Malin?

      what A3-with  pfv-B3-play-itv cl.fem Maria

   2. b. Tas  ix-∅-tajn-i  ix  Malin y-et’(ok)?

      what  pfv-B3-play-itv  cl.fem  Maria A3-with

      ‘What did Maria play with?’

Finally, it is important to note that pre-verbal positions in Mayan languages fall into topic and focus categories; see, for example, Aissen (Reference Aissen1992). In Chuj, Bielig (Reference Bielig2015) shows that pre-verbal topics are base-generated high, as in (26) – not triggering the AF extraction marking that appears in (27) – and co-occur with a coreferential post-verbal classifier, if there is an appropriate classifier. (Recall from (7b) above that classifiers can be used as pronouns.) In contrast, pre-verbal foci are fronted from a post-verbal position, triggering AF in (27), with no corresponding post-verbal classifier. Bielig also shows that topics are necessarily higher than the position of focus, and that wh-fronting patterns with focus-fronting, as is common cross-linguistically. See Bielig (Reference Bielig2015) for further discussion on distinguishing topics from foci, and arguments for their distinct derivations.

1. (26) Pre-verbal topic with coreferential classifier pronoun:

   A   ix    Elsa_i, ix-∅-s-xik      te’    k'atitz   *(ix_i).

   top cl.fem   Elsa  pfv-B3s-A3s-chop cl.wood firewood  cl.fem

   ‘As for Elsa_i, she_i cut the firewood.’ (Bielig Reference Bielig2015: 11)

2. (27) Pre-verbal focus with no corresponding post-verbal classifier:

   A   ix    Ana ix-∅-mak’-an nok’    mis (*ix).

   foc cl.fem    Ana pfv-B3s-hit-af cl.animal cat  cl.fem

   ‘It was Ana who hit the cat.’ (Bielig Reference Bielig2015: 16)

2.4 A note on data collection

As mentioned above, our study is based on elicitation with a speaker from San Mateo Ixtatán. Our elicitations were conducted regularly over a period of two years (2014–2016) in Montreal, Canada, with all major contrasts reported in this paper confirmed at least twice in independent sessions. These findings and examples are supplemented by data from existing literature wherever possible. We present homogeneous examples with limited lexical choices to facilitate cross-comparisons between examples and constructions. We believe that this is crucial in order to eliminate any extraneous effects that may influence the contrasts we are interested in.

Following current best practices in semantic fieldwork (e.g., Matthewson Reference Matthewson2004), our data collection involved judgments of felicity/grammaticality in specific contexts, as well as context selection tasks. For example, to test to see whether a free relative admits a singular or plural referent, examples were presented in contexts with either a single relevant individual or with multiple individuals. (This is reported in (98–99) below.) In the interest of space, we do not illustrate all these contexts in the text, but we have included a few of them where they are especially illustrative. See, for example, (62), (76a–c), (95–97), and (101).

3.1 Background: bare wh-indefinites

In addition to their use in interrogative clauses, wh-words are often used cross-linguistically to form indefinites (e.g., Cheng Reference Cheng1991, Postma Reference Postma1994, Haspelmath Reference Haspelmath1997, Bhat Reference Bhat2000, Gärtner Reference Gärtner2009, and references therein).Footnote ¹⁰ We can broadly classify such uses into two categories: indefinites formed of wh-words with additional morphology on them, and indefinites that are bare wh-words. We will show in the following section that Chuj has indefinites of the latter type, which we call bare wh-indefinites. Here we therefore briefly review previous cross-linguistic work on such indefinite uses of bare wh-words.Footnote ¹¹

The use of bare wh-words as indefinites is cross-linguistically quite common. Building on previous literature, Gärtner (Reference Gärtner2009) compiled a list of 62 languages with bare wh-indefinites (what he calls “[i = i]” or “interrogative = indefinite” languages), which he says is based on an aggregate survey of approximately 150 languages. (See his Appendix B for his full list of such languages.) At the same time, it has been noted that the use of bare wh-words as indefinites is limited in these languages in a number of ways.

One common limitation is that, in languages with interrogative wh-fronting, the wh-word must not be fronted. This restriction is observed in Shoshone (Uto-Aztecan) below in (28). The wh-word must be fronted in Shoshone questions (Miller Reference Miller and Goddard1996), as in (28a), whereas wh-indefinites must not be (28b). As we have seen in section 2.3 above, Chuj is also an interrogative wh-fronting language, and we will indeed later observe this in-situ requirement on bare wh-indefinites in Chuj.

(28) Shoshone (Uto-Aztecan) bare wh-indefinites must be in-situ:

Another common restriction is that bare wh-indefinites must be in the scope of a particular licensing environment or operator. A common licensing environment is negation, as illustrated in the Mandarin Chinese contrast shown in (29).

(29) Mandarin Chinese wh-indefinite licensed by negation: (Li Reference Li1992: 127)

Bare wh-indefinites in Mandarin are also licensed in polar questions and in the antecedent of conditionals, leading Huang (Reference Huang1982), Li (Reference Li1992), and Cheng (Reference Cheng1994) to describe bare wh-indefinites as negative polarity items.Footnote ¹² As we will see in the following section, the distribution of bare wh-indefinites in Chuj resembles that in Mandarin Chinese. One immediate difference between Mandarin Chinese and Chuj is, of course, the fact that for interrogative clauses, Mandarin Chinese is a wh-in-situ language whereas Chuj is a wh-fronting language (see section 2.3). While previous work such as Cole and Hermon (Reference Cole and Hermon1998) has claimed that the use of wh-indefinites may correlate with the availability of interrogative wh-in-situ, more recent work by Bruening (Reference Bruening2007) and Aldridge (Reference Aldridge2007) have disputed this alleged connection. The facts we will present from Chuj support the idea that the availability and distribution of bare wh-indefinites is independent of the language's wh interrogative strategy. In particular, we will show that Chuj allows for bare wh-indefinites in licensing conditions very similar to those in Mandarin Chinese, even though they differ in their interrogative wh usage.

3.2 Bare wh-indefinites in Chuj

In this section we describe the limited conditions under which bare wh-words can take an indefinite interpretation. Following the background above, we cover three restrictions on bare wh-indefinites: the behavior of different wh-words and phrases, the in-situ requirement, and licensing environments.

First, we note that bare tas ‘what’ can be freely interpreted as an indefinite in any post-verbal position, as in (30a). In contrast, mach ‘who’ cannot be an indefinite in (30b), although we will see bare mach indefinites in a certain limited set of environments below.Footnote ¹³ An echo question interpretation is also always available.Footnote ¹⁴

(30) Post-verbal ‘what’ but not ‘who’ can be interpreted as a wh-indefinite:

Wh-indefinites must be bare wh-words and cannot take nominal domains. Recall that wh-words in questions can take a domain, as illustrated again in (31) below. In the minimally different example (32), the addition of a nominal domain blocks the indefinite ‘some book’ reading, leaving only the echo question reading available.

(31) ‘What’ tas can take a domain:

1. (32) Indefinite tas cannot take a nominal domain:

   Ix-∅-k-il    tas libro(-al)

   pfv-B3-A1p-see what book-nml

   * ‘We saw some book.’ (see (30a))

   ✓ ‘We saw which book?’ (echo question)

Second, we note that wh-indefinites must be in a post-verbal argument position and cannot be fronted. Compare the intended declarative reading of (33) with an indefinite tas to the grammatical wh-question in (31a) above. We observe that when a wh-word is fronted, only the interrogative reading is available.

1. (33) Fronted wh cannot be indefinite:

   * Tas ix-∅-∅-il-a’.

   what pfv-B3-A2s-see-tv

   Intended: ‘You saw something.’ (see (31a))

Recall that Chuj is an interrogative wh-fronting language. As noted above, a requirement that wh-indefinites be in-situ is common in languages with wh-fronting (Cheng Reference Cheng1991, Bhat Reference Bhat2000, a.o.).

Third, we turn to the special licensing conditions of mach ‘who’ as the indefinite ‘someone.’ Unlike tas which can be interpreted as an indefinite in any post-verbal argument position, a limited set of licensors is necessary for this indefinite reading of mach.

A common licensor of bare wh-indefinites is negation. Negation in Chuj indeed licenses the indefinite use of mach as a narrow-scope ‘someone.’ Negation in Chuj involves two morphemes, surrounding the main predicate. The first morpheme has various realizations, for example surfacing as manh with nonverbal predicates and prospective aspect, as maj with perfective aspect, and as max with imperfective aspect (Coon and Carolan Reference Coon and Carolan2017). The second morpheme is an invariant laj and either attaches to the predicate (the verbal stem itself) or to the first negation morpheme.Footnote ¹⁵

(34) Negation is a licensor of bare mach-indefinites:

Bare wh-indefinites are also licensed in the antecedent of conditional clauses:

1. (35) Conditional licenses bare wh-indefinites:

   Tato tz-∅-∅-il  mach/tas, ∅-∅-al  t'a   hin.

   if    impf-B3-A2s-see who/what B3-A2-say prep B1s

   ‘If you see someone/something, let me know.’ (literally: ‘tell me’)

The last licensors of wh-indefinites are the prospective (future) and progressive aspects, in (36). Here, in addition to tas ‘what,’ mach ‘who’ is also licensed:Footnote ¹⁶

(36) The prospective and progressive aspects license wh-indefinite:

This is in contrast to the perfective, observed above in (30), and the imperfective in (37) below, where tas can take an indefinite interpretation but mach cannot.

(37) Imperfective aspect does not license bare mach-indefinite:

The licensing of bare wh-indefinites by negation, conditionals, and prospective (future-oriented) aspect parallels their licensing environments in Mandarin Chinese, which Giannakidou and Cheng (Reference Giannakidou and Cheng2006), following Lin (Reference Lin1998), describe as all being non-veridical contexts.

Recall from above that tas ‘what’ differs from mach in allowing an indefinite interpretation in any post-verbal position, without a designated licensor, but cannot take a nominal domain as in tas libroal with the intended indefinite interpretation of ‘some book’ (32). We can show this same pattern with mach. Because mach, unlike tas, requires a licensing environment such as negation, we start with the baseline bare wh-indefinite under negation in (38a). The addition of an explicit domain to mach in (38b) again leads to ungrammaticality of the indefinite reading.Footnote ¹⁷

1. (38) Indefinite mach also cannot take a nominal domain:

   1. a. Maj ∅-k-il laj mach/tas.

      neg B3-A1p-see neg who/what

      ‘We didn't see anyone/anything.   (= 34)

   2. b. * Maj ∅-k-il laj mach winh unin.

      neg B3-A1p-see neg who cl.masc child

      Intended: ‘We didn't see any boy.’ (see also (32))

Finally, we note that a wh-indefinite need not be a core argument of the verb, as in the examples we have seen above. This is demonstrated in example (39), where mach is the object of a preposition.

• (39) A wh-indefinite may be the object of a preposition:

• Maj in-och laj y-et’ mach.

  neg B1s-help neg A3-with who

  ‘I didn't help anyone.’

In summary, in this section we presented the distribution of bare wh-indefinites in Chuj. Three types of restrictions were documented, all independently well attested in the distribution of bare wh-indefinites cross-linguistically. First, wh-words differ in their ability to take on an indefinite interpretation: tas ‘what’ can be an indefinite rather freely, mach ‘who’ requires an explicit licensor, and complex wh-phrases with nominal domains cannot be used as indefinites. Second, bare wh-words must be in post-verbal position for their intended indefinite reading. Such a requirement is common in languages with obligatory interrogative wh-fronting. Third, for mach ‘who,’ a class of licensing constructions was documented, including negation, conditionals, and prospective (future) and progressive aspects.

4.1 Free choice yalnhej-wh

In this section we discuss Chuj free choice items formed of wh-words modified with yalnhej. A basic example is given in (40). Here we gloss yalnhej as an unanalyzed unit, but we return to it later in this section.Footnote ¹⁸

1. (40) Free choice item formed of yalnhej and tas ‘what’:

   Ol-∅-w-awtej  yalnhej tas.

   prosp-B3-A1s-read yalnhej what

   ‘I will read anything/whatever.’

The term free choice for these items comes from Vendler (Reference Vendler1962) and Ladusaw (Reference Ladusaw1979) and reflects the intuition that (40) expresses that, no matter what entity is chosen, the speaker will read it. That is to say, the speaker is indifferent to the choice of actual referent. We will translate these items using English wh-ever and any, although the latter also has a use as a negative polarity item in English. The use of wh-words to form free choice items (FCI) is cross-linguistically well-attested; for example, Giannakidou and Cheng (Reference Giannakidou and Cheng2006) present such examples in Greek, Catalan, Spanish, Dutch, Korean, Japanese, and Hindi. See also Dayal (Reference Dayal and Lawson1997 et seq.), Fox (Reference Fox, Sauerland and Stateva2007), Chierchia (Reference Chierchia2013), among others.

The FCI in (40) is in post-verbal argument position, but FCIs are frequently fronted to pre-verbal focus position as in (41). FCIs are also not limited to object position; see (42) for a FCI in subject position. Note also that this FCI in (42) is formed using mach ‘who.’

The domain of yalnhej-wh FCIs can be further restricted by the addition of a nominal domain or a relative clause, as in (43) and (44).

1. (43) FCI restricted by a nominal domain:

   [Yalnhej tas libro-al] tz-∅-∅-awtej.

   yalnhej what book-nml impf-B3-A2s-read

   ‘You read any book.’

2. (44) FCI restricted by a relative clause:

   [Yalnhej mach tz-∅-jaw-i]  ol-in-och  y-et'ok.

   yalnhej who impf-B3-come-itv prosp-B1s-help A3-with

   ‘I will help whoever comes.’

In addition to FCIs formed with tas ‘what’ and mach ‘who,’ FCIs formed of b'aj ‘where’ are also quite natural:

1. (45) A place FCI:

   Yalnhej ba'j tz-∅-∅-al in-b'at-i.

   yalnhej where impf-B3-A2s-say B1s-go-itv

   ‘I go anywhere/wherever you say.’

To summarize, yalnhej can productively combine with a range of wh-words to form a free choice item which can be in pre-verbal focus position or in its post-verbal base position. These FCIs can also take a nominal domain or relative clause.

Now we turn to the structure of these FCIs themselves. There is reason to believe that yalnhej is internally complex and made up of the ability modal yal and the ‘only’ word nhej. Free choice examples are analyzed in this way by Buenrostro (Reference Buenrostro2009: 220), with yal-nhej glossed as ‘able-only.’

1. (46) Yal is an ability modal:

   Tz-∅-yal  w-al-an  kastilla.

   impf-B3-able A1s-speak-sub Spanish

   ‘I can speak Spanish.’ (Buenrostro Reference Buenrostro2009: 142)

2. (47) Nhej is an ‘only’ word:

   A nhej waj Xun tik ko-gana.

   foc only cl.masc Juan dem A3p-like

   ‘We like only this Juan.’ (i.e., not that other Juan)

The idea that yal and nhej should be thought of as separate morphemes is supported by examples such as (48) which also receive a free choice interpretation.

1. (48) Yal and nhej separated:

   Yal ol-∅-w-awtej  nhej tas libro-al.

   able prosp-B3-A1s-read only what book-nml

   ‘I can read any/whichever type of book.’

However, we will argue that in the majority of cases here, where yal-nhej is linearly contiguous, yalnhej forms a nominal constituent with the wh-word and any restricting material. In particular, yal in examples with pre-verbal yalnhej-wh is not a modal predicate yal taking a nhej-wh argument.

We give three arguments for this proposal. First, yalnhej-wh FCIs have the distribution of a nominal constituent: they can be in post-verbal argument position and can be fronted as a unit to pre-verbal focus position, without restriction. This can be observed in the examples above.

Second, items such as pax ‘also’ which normally appear in an immediately post-verbal position cannot split yal and nhej. This would be unexpected under the view that yal here is the regular modal verb.

1. (49) Yal and nhej cannot be split by pax ‘also’:

   1. a. * Yal pax nhej tas libro-al ol-∅-w-awtej.

      able also only what book-nml prosp-B3-A1s-read

   2. b. Ol-∅-w-awtej pax yalnhej tas libro-al.

      prosp-B3-A1s-read also yalnhej what book-nml

      ‘I will also read any BOOK.’

The third and final argument comes from the position of negation. Recall that negation in Chuj surrounds the predicate. Example (50) shows that this is true for the ability modal yal, with the particle laj immediately following yal. In contrast, in the negative cleft in (51), where the yalnhej tas libroal is sentence-initial, negation surrounds the entire fronted FCI (51b), rather than placing laj after yal alone (51a).

1. (50) Negation surrounds the modal yal:

Max   yal laj  in-b'ey   in-ch'ok'ojil.

neg.impf able neg A1s-walk  A1s-alone

‘I cannot walk alone.’    (Buenrostro Reference Buenrostro2009: 230)

1. (51) Yal and nhej cannot be split by negation:

   1. a. * Manh yal (ok)laj nhej tas libro-al ol-∅-w-awtej.

      neg able irr-neg only what book-nml prosp-B3-A1s-read

   2. b. Manh yalnhej tas libro-al ok-laj ol-∅-w-awtej.

      neg yalnhej what book-nml irr-neg prosp-B3-A1s-read

      ‘It's not (just) any book that I read.’ (i.e., I read some special kind of book.)

The conclusion then is that yalnhej in these FCIs is not obviously decomposable into the modal yal and the ‘only’ word nhej. Instead, yalnhej consistently forms a nominal constituent with the wh-phrase. We speculatively conclude that yalnhej is unanalyzable in the synchronic grammar of Chuj, but may be diachronically related to the (now rarer) construction involving the modal yal and a separate ‘only’ nhej, exemplified by (48), which also yields a similar free choice reading.

4.2 Universal masel mach

This variety of Chuj has two common forms of universal quantifiers, masel and masanil. Masel must take a restrictor, as in (52a), whereas masanil can stand on its own as ‘everyone’, as in (52b), or take a nominal domain.

(52) Two forms of universal quantifiers, masanil and masel:

As shown in (53), the former quantifier commonly appears as masel mach ‘every who’ to mean ‘everyone’ (53a), while masanil can also take mach (53b). In this section we present a brief study of these wh-derived universals, masel mach and masanil mach, focusing on masel mach which is more common.

(53) Masel and masanil can take mach ‘who’:

The universal quantifiers have thus far all ranged over the entire set of human individuals. The domain of masel mach can be further restricted by a relative clause or a nominal domain. These nominal domains, as in (55), must be plural, as indicated by the ungrammaticality of removing the plural marker heb’.

1. (54) Masel mach restricted by a relative clause:

   [Masel mach ix-∅-ulek’-i]  ix-in-il-an-i.

   every who pfv-B3-come-itv pfv-B1s-see-af-itv

   ‘Everyone who came saw me.’

2. (55) Masel mach restricted by a plural nominal domain:

   [Masel mach *(heb’) ix     unin]   ix-∅-ulek’-i.

   every who pl    cl.fem girl pfv-B3-come-itv

   ‘All the girls came.’

In the examples above, the universal quantifiers have all been in pre-verbal focus position – as indicated by the use of Agent Focus morpheme in (54) – but they can also be post-verbal, as in (56).

1. (56) Masel mach in post-verbal position:

   Ix-∅-k-il  masel mach (ix-∅-ulek’-i).

   pfv-B3-A1p-see every who (pfv-B3-come-itv)

   ‘We saw everyone (who came).’

Finally, we note that masel is curiously unable to take tas ‘what’ to form a universal quantifier over inanimates, *masel tas, parallel to masel mach. This is shown in (57), while (58) shows that the attested masel mach is limited to animate domains.

1. (57) There is no masel tas:

   * Ix-∅-w-awtej  masel tas juntzan libro tik.

   pfv-B3-A1s-read every what some.pl book dem

   Intended: ‘I read {every one/each} of these books.’

2. (58) Masel mach is limited to animate domains:

   * Ix-∅-w-awtej  masel mach juntzan libro tik.

   pfv-B3-A1s-read every who some.pl book dem

   Intended: ‘I read {every one/each} of these books.’

Instead, inanimate universal quantification involves a simple masanil taking a nominal domain, as in (59).

1. (59) Universals without wh are used instead:

   Ix-∅-w-awtej  masanil juntzan libro tik.

   pfv-B3-A1s-read every some.pl book dem

   ‘I read {every one/each} of these books.’

To our knowledge, the only wh-universal forms that exist are the animate masel mach and masanil mach, with the former being preferred. This exceptional use of mach cannot be subsumed under any independent differences among the wh-words. The only other difference between mach and tas that we have observed is a difference in their licensing as bare indefinites, in section 3 above, where bare tas could be an indefinite in any post-verbal position, whereas bare mach required a particular type of licensing environment.

A possible conclusion at this point would be to say that masel mach and masanil mach now have the status of compounds which are each in the lexicon as syntactic atoms. Evidence from the placement of negation in (60) shows that this is not true. The universal quantifier and mach can be separated by the irrealis marker and second negation marker laj when the universal operator itself is negated:

1. (60) Negation can split ‘every’ and mach:

   1. a. Manh masel ok-laj mach ix-∅-ulek’-i.

      neg every irr-neg who pfv-B3-come-itv

   2. b. Manh masanil ok-laj mach ix-∅-ulek’-i.

      neg every irr-neg who pfv-B3-come-itv

      ‘Not everyone came.’

We conclude that this restriction of universals to mach ‘who’ must be some lexical selectional idiosyncrasy but that these combinations are not compounds. We leave further investigation of these forms for future work.

5.1 Indefinite free relatives

Free relatives with an indefinite meaning are cross-linguistically less common than definite free relatives (Radek Šimík, p.c.). They have been observed in some Indo-European and Semitic languages, and are said to be unavailable in Germanic languages (with the exception of Yiddish; see Caponigro Reference Caponigro2003). Additionally, Caponigro et al. (Reference Caponigro, Torrence and Cisneros2013) document indefinite FRs in two Mixtec languages, Nieves Mixtec and Melchor Ocampo Mixtec. Two examples from Hebrew are given below, with translations reproduced from their sources:

1. (63) Indefinite free relatives in Hebrew:

   1. a. Yesh l-i [_FR im mi le-daber].

      exist to-1s with who to-talk

      ‘I have somebody to talk to.’  (simplified from Caponigro Reference Caponigro2003: 90)

   2. b. Eyn l-i [_FR im mi le-daber].

      not.exist to-1s with who to-talk

      ‘There is nobody I can talk to.’  (Grosu Reference Grosu and Tomić2004: 422)

Although there is considerable cross-linguistic variation within indefinite FRs – see discussion in e.g. Izvorski (Reference Izvorski, Tamanji and Kusumoto1998), Caponigro (Reference Caponigro2003), Grosu (Reference Grosu and Tomić2004), Šimík (Reference Šimík2011) – a generalization is that they must be the internal argument of a verb that expresses existence, often of a ‘have’ or ‘exist’ type. In the Hebrew examples above, the existential verbs yesh (exist) and eyn (not.exist) are used. We can contrast these grammatical cases with the ungrammatical example below, where an existential verb is not used.

1. (64) Hebrew indefinite FR must be the complement of an existential verb:

   * Kani-ti [_FR ma li-kro].

   bought-1s what to-read

   Intended: ‘I bought something to read.’

As we will see, indefinite FRs in Chuj must also be the complement of an existential predicate. For more on such indefinite FRs and the related modal existential wh-constructions cross-linguistically, we refer the reader to Grosu and Landman (Reference Grosu and Landman1998), Izvorski (Reference Izvorski, Tamanji and Kusumoto1998), Caponigro (Reference Caponigro2003), Grosu (Reference Grosu and Tomić2004), Šimík (Reference Šimík2008, Reference Šimík2011, Reference Šimík and Aronoff2017) and references therein.

5.1.2 Licensing predicates

So far we have concentrated on indefinite FRs with the existential predicate ay. In this section we will show that this indefinite FR interpretation is available more generally with predicates that express the existence of their internal argument description.

We begin by taking a brief look at existential predicates in Chuj more generally.Footnote ²² Chuj has three basic existential predicates: the positive predicate ay, its negative counterpart malaj, and a predicate meaning roughly ‘other, distinct, separate’ ch'ok, (73). These predicates can also be used to express possession, as in (74).

1. (73) Existential predicates in Chuj:Footnote ²³

   1. a. Ay jun hu'um sat te’ mexa.

      exist one book A3.surface cl.wood table

      ‘There is a book on the table.’

   2. b. Malaj ch'anh hu'um sat te’ mexa.

      not.exist cl.book book A3.surface cl.wood table

      ‘There is no book on the table.’

   3. c. Ch'ok ch'anh hu'um sat te’ mexa.

      other cl.book book A3.surface cl.wood table

      ‘There is a different book on the table.’

(74) Existential predicates expressing possession:

All three of these one-place existential predicates can take a wh-fronted clause to yield an indefinite FR. Data here is shown for mach ‘who’ but similar facts hold for tas ‘what.’

1. (75) Indefinite FR with different existential predicates:

   1. a. Ay [_FR mach ix-∅-ulek’-i].

      exist who pfv-B3-come-itv

      ‘Someone came.’ (= 65)

   2. b. Malaj [_FR tas w-ojtak].

      not.exist what A1s-know

      ‘I don't know anything.’  (Buenrostro Reference Buenrostro2009: 140)

   3. c. Malaj [_FR mach tz-b'at peresu].

      not.exist who impf-go prisoner

      ‘There is no one who is taken prisoner.’  (Davis Reference Davis2010: 1289)

   4. d. Ch'ok [_FR mach ix-∅-ulek’-i].

      other who pfv-B3-come-itv

      ‘Someone else / Others came.’

In addition to these basic existential predicates, some other verbs that express the existence of their internal argument can license indefinite FRs in that position. This has been shown for indefinite FRs in some other languages as well; see discussion in Grosu (Reference Grosu and Tomić2004). In (76) we demonstrate this with aj-nak ‘be born,’ chax ‘be found,’ and say ‘look for’.

1. (76) Indefinite free relatives with predicates with an existential component:

   1. a. Context: 50 years ago, a boy was born in this village who later became president. So in this place…

      Aj-nak [_FR mach famoso].

      born-stat who famous

      ‘Some famous person was born (here).’

   2. b. Context: My car is broken. I need help so I went to the garage …

      Ix-∅-chax [_FR mach ol-∅-b'o’-an k'en hin-karro].

      prvf-B3-be.found who prosp-B3-fix-af cl.metal A1s-car

      ‘Someone was found who will fix my car.’

   3. c. Ko-say-an [_FR tas ∅-ko-k'ulej].

      A1p-look.for-subFootnote ²⁴ what B3-A1p-do

      ‘We are looking for something to do.’  (Hopkins Reference Hopkins1967: 158)

Finally, we discuss indefinite FRs in the complement of the stative predicate gana. The predicate gana in Chuj is ambiguous between ‘like’ and ‘want’ when it takes a nominal complement (see, e.g., Buenrostro Reference Buenrostro2009).

1. (77) Gana can be ‘like’ or ‘want’:

   Malaj hin-gana  tas.

   neg A1s-want/like what

   ‘I don't want/like anything.’

The predicate gana can take a wh-fronted clause as its complement, interpreted as an indefinite FR, but in such cases only the ‘want’ reading of gana is available:

1. (78) With wh-sister, only the ‘want’ reading survives for gana:

   Hin-gana  [_FR mach tz-∅-b'at-i].

   A1s-want/like who impf-B3-come-itv

   ✓ ‘I want someone to come.’

   * ‘I like someone who comes.’

This too can be explained through the generalization that indefinite FRs are licensed by verbs that directly express existence of their internal argument. As has been widely observed, verbs of desire taking nominal complements underlyingly express a desire to possess the complement. In contemporary terms, such ‘want’ verbs have been analyzed as embedding a silent ‘have’ predicate (e.g., Larson et al. Reference Larson, Den Dikken, Ludlow, Grzankowski and Montagueto appear). To our knowledge, this idea originates with McCawley (Reference McCawley1974), who shows that time adverbials can specifically modify this embedded possession: for example, Bill wants your apartment until June (p. 74) has a reading where Bill's desire is to have the apartment until June, not that his desire extends until June. Returning now to the Chuj example in (78), adopting this decompositional approach to the intensional transitive ‘want’ explains the licensing of indefinite FRs with gana as ‘want’ but not as ‘like’; (78) can thus be thought of as expressing ‘I want there to be people who come.’

5.2 Definite free relatives

Next we turn our attention to definite free relatives. We will show that they are different from indefinite free relatives in two important ways: the distribution of definite FRs within a sentence is not limited in the way that the distribution of indefinite FRs is; instead, they can appear in any syntactic position and as the sister of any verb. In addition, definite FRs allow for nominal domain restrictions, unlike indefinite FRs.

5.3 Definite FRs and FRs with quantifiers as light-headed relatives

Definite free relatives in Chuj appear to simply be a clause with wh-fronting, but we have seen that they have the external distribution of a definite noun phrase. We follow Caponigro (Reference Caponigro and Samiian2002) and Citko (Reference Citko2004) in analyzing definite FRs as involving an unpronounced definite determiner taking the wh-fronting CP as its complement, as in (93). Citko (Reference Citko2004) calls such structures light-headed relatives.Footnote ²⁸

1. (93)

   ‘the person/people who came’ (literally: ‘the [who came]’)

Although some Mayan languages have overt prenominal definite determiners, Chuj does not. Therefore, the D head in (93) may be the general definite determiner in the language, which is unpronounced. Definite FRs can, however, co-occur with the postnominal demonstrative markers tik (proximal) and chi (distal).Footnote ²⁹

1. (94) A definite FR can co-occur with a demonstrative:

   Ix-∅-w-il  [_FR mach ix-∅-ulek’] tik.

   pfv-B3-A1s-see who pfv-B3-come dem

   ‘I saw this person/these people who came.’

Support for the light-headed analysis (93) comes from the fact that various overt, quantificational D heads can also introduce FRs. Examples (95–97) show free relatives as the domains for jantak and tzijtum, two different forms of the word for ‘many,’ as well as juntzan ‘some’ (plural).Footnote ³⁰ Notice that the entire FR may appear pre- or post-verbally and it may include a nominal domain (heb’ winh unin in (96)), just as we saw for definite FRs above. For convenience, we underline the quantifier in each sentence.

1. (95) FRs with quantificational heads, in any syntactic position:

   1. a. [_DP Jantak [_CP mach ix-∅-ulek’-i]] ix-∅-w-il-a’.

      many who pfv-B3-come-itv pfv-B3-A1s-see-tv

   2. b. Ix-∅-w-il  [_DP jantak [_CP mach ix-∅-ulek’-i]].

      pfv-B3-A1s-see many who pfv-B3-come-itv

      ‘I saw the many people who came.’

2. (96)

   1. a. [_DP Tzijtum [_CP mach heb’ winh unin ix-∅-ulek’-i]] ix-∅-w-il-a’.

      many who pl cl.masc child pfv-B3-come-itv pfv-B3-A1s-see-tv

   2. b. Ix-∅-w-il [_DP tzijtum [_CP mach heb’ winh unin ix-∅-ulek’-i]].

      pfv-B3-A1s-see many who pl cl.masc child pfv-B3-come-itv

      ‘I saw the many boys who came.’

3. (97)

   1. a. [_DP Juntzan [_CP mach ix-∅-ulek’-i]]  ix-∅-w-il-a’.

      some.pl who pfv-B3-come-itv pfv-B3-A1s-see-tv

   2. b. Ix-∅-w-il  [_DP juntzan [_CP mach ix-∅-ulek’-i]].

      pfv-B3-A1s-see  some.pl who pfv-B3-come-itv

      ‘I saw these people who came.’

Our translations here are supported by our elicitation work. Such examples were presented in a number of contexts: for example, for (95), (a) where some large number of people came and the speaker saw many of the people who came, as a proportion, but not all, and (b) where many people came and the speaker saw all of them. Our speaker expressed a preference for the (b)-type context being more natural, which is then reflected in our translations here, using the English ‘the many…’. The examples were uniformly rejected in contexts where no people or very few people came or were seen. More careful work is necessary in order to better distinguish such readings, which we leave open for future work.

5.4 Jun free relatives

Having presented the distribution and structural characteristics of indefinite FRs and definite FRs in Chuj, in this section we conclude our discussion of Chuj FRs with the particular characteristics of FRs with jun ‘one.’ We will see that these special FRs seem to be a sort of hybrid which shares some characteristics with the indefinite FRs above and some characteristics with definite FRs.

The word jun in Chuj itself means ‘one.’ The examples below show that jun can be added to both definite and indefinite nominal expressions. The nominal winh unin ixulek'i in the complement of ‘like’ receives a definite interpretation in (98), while the same structure in the complement of the existential verb ay receives an indefinite interpretation in (99). In either case, adding jun fixes the referent as singular, whereas the baselines without jun are underspecified for number. These judgments were obtained by providing our speaker with contexts where there was either just one boy present or more than one boy present; examples (98b) and (99b) were only accepted in contexts with a single boy, whereas the (a) examples were accepted in both types of contexts.

1. (98) Jun can be added to a definite nominal:

   1. a. Ko-gana winh unin ix-∅-ulek’-i.

      A1p-like cl.masc child pfv-B3-come-itv

      ‘We like the boy(s) that came.’ (singular or plural)

   2. b. Hin-gana jun winh unin tik.

      A1s-like one cl.masc child dem

      ‘I like this boy.’ (singular)

2. (99) Jun can be added to an indefinite nominal:

   1. a. Ay winh unin ix-∅-ulek’-i.

      exist cl.masc child pfv-B3-come-itv

      ‘Some boy(s) came.’ (singular or plural)

   2. b. Ay jun winh unin ix-∅-ulek’-i.

      exist one cl.masc child pfv-B3-come-itv

      ‘Some/a/one boy came.’ (singular)

Now we turn to jun free relatives. Our first example is in (100).

1. (100) A jun free relative as the argument of ay:

   Ay jun mach ix-∅-ulek’-i.

   exist one who pfv-B3-come-itv

   ‘Some/a/one person came.’  (literally: ‘There is one who came.’)

   Consultant comment: In Spanish, “hay una persona” (‘there is one person’), but not “hay alguien” (‘there is someone_sg/pl’).

Such examples can be used both when the referent is known to the speaker and when it is not. (The same judgments hold for indefinite FRs without jun.)

(101) Jun FRs optionally express ignorance or unimportance of the referent:

At this point, our FR with jun resembles the indefinite FRs we described in section 5.1 above. Compare (100) with the indefinite FR in (65), repeated here as (102). It seems that the only difference between the two examples is the addition of jun in the former.

1. (102) Indefinite free relative, repeated:

   Ay [_FR mach ix-∅-ulek’-i].

   exist who pfv-B3-come-itv

   ‘Someone came.’  (=65)

There are, however, significant differences between jun FRs and indefinite FRs. Recall that indefinite FRs as in (102) must be the internal argument of a limited set of verbs with an existential meaning (section 5.1.2 above). Jun FRs, by contrast, are not subject to this restriction. Example (103) shows a FR with jun in the object position of ‘see,’ a verb which does not involve existential semantics. Example (104) shows a jun FR as a pre-verbal topic.

1. (103) Jun FR as object of ‘see’:

   Ix-∅-w-il  [jun mach ix-∅-ulek’-i].

   pfv-B3-A1s-see one who pfv-B3-come-itv

   ‘I saw some/a/one person who came.’

2. (104) Jun FR as pre-verbal topic:

   [Jun mach ix-∅-ulek’-i] ix-∅-w-il-a’.

   one who pfv-B3-come-itv pfv-B3-A1s-see-tv

   ‘[Some/a/one person that came]_i, I saw him/her_i.’

Notice also that in both of these examples, the FR with jun is interpreted indefinitely. This is a general property of FRs with jun: unlike adding jun to a lexical noun (see (98–99) above) which does not affect its (in)definiteness, FRs with jun are always interpreted indefinitely. For example, compare example (103) above with example (105) below, which simply differs in the removal of jun. The sentence is grammatical but only with mach ixulek'i interpreted as a definite FR.

1. (105) FR without jun as the object of ‘see’ must be definite:

   Ix-∅-w-il  [_FR mach ix-∅-ulek’-i].

   pfv-B3-A1s-see who pfv-B3-come-itv

   ‘I saw the person/people who came.’ (see (103))

With regard to these distributional facts, then, jun FRs pattern with definite FRs and not with indefinite FRs. Nonetheless, their meaning is indefinite.

In addition – like both indefinite and definite FRs – jun FRs allow for both preposition pied-piping and stranding.

1. (106) Jun FRs optionally pied-pipe with secondary fronting:  (see (92))

   1. a. Ix-∅-w-ilelta  [jun [mach y-et’] h-och ix Malin].

      pfv-B3-A1s-meet one who A3-with B3-help cl.fem Maria

   2. b. Ix-∅-w-ilelta  [jun mach h-och ix Malin y-et'ok].

      pfv-B3-A1s-meet one who B3-help cl.fem Maria A3-with

      ‘I met some/a/one person who Maria helped.’

We propose to model jun FRs as light-headed relatives with the head jun ‘one.’ In this way, jun FRs parallel other cases of FRs restricting a quantifier, such as (97), repeated here:

1. (107)

   1. a. Ix-∅-w-il [_DP juntzan [_CP mach ix-∅-ulek’-i]].

      pfv-B3-A1s-see some.pl who pfv-B3-come-itv

      ‘I saw these people who came.’ (=97)

   2. b. Ix-∅-w-il [_DP jun [_CP mach ix-∅-ulek’-i]].

      pfv-B3-A1s-see one who pfv-B3-come-itv

      ‘I saw some/a/one person who came.’ (=103)

There is, however, another characteristic besides interpretation which unifies jun FRs with indefinite FRs. Recall that indefinite FRs disallow the addition of nominal domains such as ‘boy,’ whereas definite FRs and FRs with quantificational heads allow such restrictors. In this case jun FRs pattern with indefinite FRs, and unlike definite FRs and FRs with quantifiers; see (108). At this stage we do not have an explanation for why jun FRs disallow nominal domains.Footnote ³²

1. (108) Jun FRs disallow nominal domains:

   1. a. * Ix-∅-w-il  [jun mach winh unin ix-∅-ulek’-i].

      pfv-B3-A1s-see one who cl.masc boy pfv-B3-come-itv

      Intended: ‘I saw some/a/one boy who came.’ (see (103))

   2. b. * [Jun mach winh unin ix-∅-ulek’-i] ix-∅-w-il-a’.

      one who cl.masc boy pfv-B3-come-itv pfv-B3-A1s-see-tv

      Intended: ‘[Some/a/one boy that came]_i, I saw him/her_i.’ (see (104))

Jun FRs thus exhibit a combination of the characteristics of indefinite FRs and definite FRs, presented earlier. Like indefinite FRs and bare wh-indefinites, they have an indefinite interpretation and disallow nominal domains. On the other hand, like definite FRs and FRs with quantifiers, they have a free distribution and are not limited to the complement position of certain existential verbs.