The main goal of this article is to argue for the existence and crosslinguistic stability of a phenomenon that I call “negation-licensed commands” (NLCs), in which commands can be interpreted as such only in the presence of negation. The article explores morphosyntactic properties of NLCs in a small number of unrelated languages. It stops short of an analysis of the phenomenon, which will have to await future research.
Commands can be expressed in a number of ways. Crosslinguistically, the most commonly known involve the imperative verb form, the infinitive, and the subjunctive.1 But other syntactic categories can also be used to convey commands.
To the principal!
Hands in the air!
No loud noises!
All the expressions in (1) are interpreted as commands—and if one thinks a command can be conveyed only with functional categories in the extended projection of the verb that contain a modal or some other element with directive force (e.g., Schwager 2006, Kaufmann 2012, Oikonomou 2016), one will have some serious eliding to do to derive them. In a framework in which pragmatics plays the crucial role and the basic semantics is that of a property or a proposition, with the modal meaning originating in the discourse component to which the property or proposition gets mapped (Hausser 1980, Portner 2007, von Fintel and Iatridou 2017), deriving (1a–e) still takes some serious thought. The commands in (1) are not sentences, so again one would have to either consider ellipsis as a possible source or consider nonsentences to be mappable to the hearer’s to-do list (Portner 2007).
This article looks at yet another means of conveying commands. The investigation will delve into a number of languages. Even though it will uncover somewhat different syntactic properties in different languages, one common characteristic will emerge: these commands survive as commands only when they are negated. The absence of negation does not lead to a positive command. It leads either to a completely different (i.e., noncommand) meaning or to ungrammaticality. I will therefore introduce the term negation-licensed commands (NLCs).
This article has three goals: to establish the existence of NLCs, to discuss some of their morphosyntactic properties crosslinguistically, and to show that a number of possible explanations for them cannot be correct. What the correct explanation actually is, the article does not determine.
2 English NLCs (with a Few Greek NLCs Thrown In)
Consider the expressions in (2), all of which can be interpreted as commands.
No talking with your mouth full!
No teasing your sister!
No reading the newspaper in class!
No walking on the grass!
These sentences exhibit several remarkable properties. First and foremost is the fact that negation is obligatory for a command to be conveyed.2
*Reading the newspaper outside of class!
*Playing the piano carefully!
*Staying off the grass!
*Not talking with your mouth full!
*Not teasing your sister!
*Not reading the newspaper in class!
*Never talking with your mouth full!
*Never teasing your sister!
*Never reading the newspaper in class!
The only form of negation that appears to work is the determiner no. This raises the suspicion that the remainder might be a nominal projection—specifically, that it could be a gerund, which is consistent with the presence of the -ing suffix. But if it is a gerund, what sort of gerund is it? Classifications like Abney’s (1987) recognize three types of gerunds in English: ACC-ing (6), POSS-ing (7), and ing-of (8). There is also the PRO-ing gerund, often considered a subcase of ACC-ing (9).3
In this common classification, the only gerund capable of taking the determiner no or the determiner the is the ing-of gerund (10). The ACC-ing and POSS-ing gerunds cannot take the/no ((11)–(12)), nor can the PRO-ing gerund (13).
No singing of any opera impressed me.
The singing of the opera impressed me.
*[The their/them advertising cigars] bothered me.
*I am relying on [the their/them solving the problem].
*[No their/them advertising cigars] bothered me.
*I am relying on [no their/them solving the problem].
(13) *I am relying on the/no solving the problem.
In general, the belief in the literature, at least after Abney 1987, is that if an English gerund is (verbal enough to be) able to assign Case to its object, it cannot take the determiners the/no. Similarly, if a gerund’s modifier is adverbial and not adjectival, it cannot take the/no.
(14) I am relying on [them solving (*of ) the problem].
(15) I am relying on [them carefully/*careful solving the problem].
The POSS-ing gerund also can assign Case to its object, and its modifiers are also adverbial.
(16) [Their advertising cigars] bothered me.
(17) [Their publicly/*public advertising cigars] bothered me.
The object of a PRO-ing gerund also has Case, and the gerund takes adverbs.
(18) Publicly/*Public advertising cigars bothers me.
Finally, the ing-of gerund requires of for its object, an indication that it is incapable of assigning Case, and takes adjectival modifiers.
(19) [The careful/*carefully singing of the opera] impressed me.
With these distinctions in mind, we can look at the NLC and ask which type of gerund it is. We will see that this question does not have a simple answer.
Given that the NLC contains the determiner no, the first gerund we should consider is the only one that can take this type of determiner: the ing-of gerund.
(20) No careless playing of the piano!
(21) *Careful playing of the piano!
As mentioned in footnote 3, the ing-of gerund is considered a “low” nominalization, its head being a noun. So the question arises whether the fact that we observe the NLC with ing-of gerunds is related to their being nouns.4 This presupposes that we know whether we can observe the NLC with nouns. Indeed, we can.5
No crank calls!
A small number of nouns can convey commands without negation.
However, these are few and far between and seem quite formulaic. As far as I can tell, nouns, as a category, need negation to convey a command. In other words, nouns productively form NLCs.
The NLC can be seen with nouns in other languages as well—for example, in Greek.
And the same form of negation appears with a nominalized verb that also displays the NLC phenomenon.6 In cases like (26a), the negation marker oxi must be present for the expression of a command. Plain Treksimata! does not mean ‘Run!’.
no run (nominalized verb)
So nouns do seem to exhibit NLC behavior in English and other languages as well. Possibly, then, the fact that ing-of gerunds can be NLCs is a special case of the fact that nouns in general can exhibit NLC effects. And the fact that the same negative marker must be used with nouns as with deverbal items in Greek NLCs is consistent with this preliminary conclusion.
However, the NLC footprint in English is broader than the ing-of gerund. NLC behavior is not just a property of low nominalizations/nouns. Consider the following examples:
No teasing your sister!
No secretly reading the newspaper in class!
No playing the piano carelessly!
No carelessly playing the piano! (%; some speakers prefer (27c))
These expressions do not have of, so they are not ing-of gerunds. Moreover, they take adverbs, as (27b–d) show, and they cannot take adjectives.
(28) No secretly/*secret reading the newspaper in class!
So we have a gerund that can assign Case to its object and take adverbs but also takes the determiner no (though, crucially, no other determiner). This is not a known category of gerund. From now on, I will use the term no-gerund for such English gerunds.
Does the no-gerund also appear outside NLCs? In section 7, we will see that it appears in another, equally mysterious modal construction in English, which, however, is different from the NLC. But it certainly cannot appear in nonmodal declarative sentences where other gerunds can appear, as object of a preposition or a verb or as a subject.
*I am relying on no teasing your sister.
*I witnessed no teasing your sister.
*No teasing your sister impressed me.
The fact that the no-gerund does not appear in plain declarative sentences is actually quite helpful because it gives us an additional reason to eliminate ellipsis as a possible (or the only) source of the NLC. Ellipsis could in principle have been the source of the NLC phenomenon, if a modal had been deleted. More specifically, one might have wondered whether the NLC behavior of the ing-of gerund was the result of (a somewhat unusual) deletion from a source like either (30) or (31).7
(30) There is no reading of The Guardian allowed →
There isno reading of The Guardian allowed
(31) No reading of The Guardian is allowed → No reading of The Guardian
The first problem that such an account faces is that it does not address the basic question posed by the existence of the NLC: namely, the question of why negation is needed. That is, if deletion is possible in (30) and (31), why would it not be possible in (32)?
So positing deletion as the source of the NLC with the ing-of gerund does not seem promising. Moreover, positing deletion as the source of the NLC with the no-gerund is a nonstarter because the no-gerund cannot appear in declarative sentences. The putative source of deletion is ungrammatical (i.e., in the predeletion sentence, the absence of of is ungrammatical).
*There is no reading The Guardian allowed.
*No reading The Guardian is allowed.
In other words, there are reasons to reject deletion of a modal as the source of the NLC, both for the ing-of gerund and for the no-gerund (though I will return to this possibility in the next section).8
To summarize: We have looked at the NLC phenomenon in English and Greek. We saw that the NLC occurs with the English no-gerund, whose properties do not reduce to those of any previously known English gerund; with the English ing-of gerund; and with nominalized verbs in Greek. We also saw that there is good reason to believe that the NLC can be observed with nouns in both languages.
I will leave English for now, as I try to establish the crosslinguistic stability of the NLC phenomenon. However, I will return to English a few times, to check whether questions raised by NLCs in other languages can be asked about English with any meaningful results.
In this section, I discuss the NLC in Dutch.9 The starting point will be Rooryck and Postma (RP) 2007. RP’s main focus is something other than the NLC, so here I will present only the data and discussion relevant to this article.
Somewhat in passing, as the focus of their paper is different, RP give an example that clearly shows what I have called the NLC.
‘No talking/fidgeting/hanging around/honking/dancing!’
All the expansions in (35) are interpreted as commands. Moreover, the negation is obligatory. In the absence of the negative marker geen, there is no command interpretation. This is the diagnostic profile of the NLC. In addition, the negation in (35) is the negative determiner geen, not the sentential negation niet.
A reader familiar with Dutch might think that gepraat is participial, given the presence of the prefix ge- and the title of RP’s paper. However, despite the presence of ge-, these are not participles but deverbal nouns. For example, for the verb prutsen ‘to mess around/fidget’, the participle is geprutst and the deverbal noun is gepruts. As (35) illustrates, the Dutch NLC contains the deverbal noun.10
RP propose an explanation for a different construction (a hortative); they say this explanation can be extended to (35), though they do not do this themselves. I will lay out their account, attempt this expansion, and point out some questions. Part of my goal will be to see if RP’s account of Dutch can be extended to NLCs in general.
According to RP, Dutch has the following ellipsis process:
Ben jenaar de kapper geweest? are youto the hairdresser’s been
‘Did you go to the hairdresser’s?’
ik hebde verkeerde shampoo gebruikt.
I havethe wrong shampoo used
‘No, I used the wrong shampoo.’
Is ernog wat post voor mij? is therestill some mail for me
‘Is there any mail for me?’
RP’s idea is that the negated deverbal nouns in (35) also involve ellipsis, even though they do not provide the pre-ellipsis sentence. But this hypothesis faces some of the questions that arose in section 2. For example, why would the elided auxiliary have to be a performative modal? Why could it not have been a reportative modal? If that were the case, the sentence would report that someone has an obligation to do something or it would report that someone is forbidden from doing something. This is different from the NLC, which bestows an obligation/prohibition rather than reporting it. In fact, it seems that reportative modals and epistemic modals are not elidable (the pre-ellipsis sentences are fine).
Why does Frank study so much?
Hij moeteen negen halen. he musta nine get
‘He has to get a nine.’
His lights are off.
Hij zalniet thuis zijn. henot at.home be
‘He must not be home.’
A related question is, why would the elided subject necessarily be 2nd person? As (36) shows, the elided subject can refer to variety of persons. In other words, postulating the ellipsis of a performative addressee-oriented modal leaves many questions unanswered.
But let’s assume that it is indeed ellipsis that creates the negated deverbal noun commands. Why should such an ellipsis be possible only under negation? In our terms, why the NLC? To answer this question, RP propose the No-Ambiguity Principle.
*[ . . . AUXMOD. . . ] if AUXMOD is ambiguous.
RP (2007:280) state that the No-Ambiguity Principle “is a particular realization of the Principle of Full Interpretation (FI).” I don’t quite see how it could be a case of FI, as language tolerates plenty of ambiguities, including with modals. But let’s evaluate it, regardless of whether or not it is a special case of FI.11
RP appeal to the No-Ambiguity Principle to capture both the necessary presence of negation and the fact that the elided modal is of existential, not universal, force. The rationale is as follows. RP correctly point out that modals that scope under negation can have either universal or existential force.
She need not leave. Neg > MODUNIV
She cannot leave. Neg > MODEX
Modals that scope over negation have only universal force. There is no deontic modal of existential force that scopes over negation (at least in Dutch or English). RP do not claim to know why (41b) does not exist.
She must not leave. MODUNIV > Neg
???? MODEX > Neg
RP argue that if the elided modal scoped under negation, we would not know whether it was a modal of universal or existential force, and the No-Ambiguity Principle would be violated.
(42) Neg >
But if the elided modal scoped over negation, the No-Ambiguity Principle would be satisfied, as there is only one type of modal that can scope over negation: a modal of universal force (41a). Hence, the elided modal can only be universal.
According to RP, the No-Ambiguity Principle also forces the presence of negation. In the absence of negation, it is not clear whether we are dealing with a universal or existential modal; however, if negation is present, wide scope of a universal modal is the only possible option, as above.
Let us finally consider the structure in [(44)], without negation.
[(44)] [ . . . AUXMOD . . .
In [(44)], if AUX is not lexical, it can be interpreted as MUST or CAN and, hence, is not well-formed with respect to the No-Ambiguity Principle. (RP 2007:281)
However, it is not quite clear how the No-Ambiguity Principle can work as a constraint on ellipsis. Besides the fact that language permits a host of ambiguities, ambiguities created by ellipsis (including antecedent-contained deletion) are well-tolerated. The ellipses in (45) and (46) create the ambiguities in (45a–b) and (46a–b).
(45) John wanted to read the book you did.
John wanted to read the book you read.
John wanted to read the book you wanted to read.
(46) I read every book one day before you did.
I read every book one day before you read it.
I read every book one day before you read every book.
So the idea that the No-Ambiguity Principle guides ellipsis would need more support than it seems to have—unless we formulate it to apply only in cases of ellipsis involving performative modals. But this does not seem ideal.
Furthermore, imperative verb forms are known to be ambiguous between strong (command) and weak (permission/acquiescence) readings.
(47) Open the door! (command)
I want to open the door.
Sure. Open the door. (acquiescence)
So the string Open the door has no problem being underdetermined with respect to its interpretation, in violation of the No-Ambiguity Principle.
Finally, note that the No-Ambiguity Principle does not say that negation is necessary so that a string will be interpreted as a command. It says there can be no ambiguity. Negation appears because it disambiguates. This means that if there is a way to disambiguate without negation, that should be fine too. One such environment is shown in (49), where the permission/acquiescence reading is by far the most felicitous (Wilson and Sperber 1988, Condoravdi and Lauer 2012, Kaufmann 2012, Portner 2012, von Fintel and Iatridou 2017).
I want to open the door.
Sure. Open the door. (acquiescence)
Sure. You may open the door.
(Sure.) #You must open the door.
This means that in a context like (49), one should be able to satisfy the No-Ambiguity Principle by means other than negation. But this prediction is not verified. First, here is the control case, to show that the Dutch imperative does permit the acquiescence interpretation.
Mag ik weg?
may I away
‘May I go away?’
OK, ga (maar) weg.
OK go (MAAR) away
But in this same environment, the affirmative deverbal noun does not work, even though it is clear that the only possible interpretation is that of an existential modal (the presence of maar does not improve the sentence at all).
Mag ik naar de bakker rennen?
may I to the baker run
‘May I run to the bakery?’
(OK.) *(*Maar) geren.
(OK) (MAAR) GE-√ren
Intended: ‘OK. Run.’
In short, for (51B) to comply with the No-Ambiguity Principle, the context should in principle disambiguate, as it does in (49), and negation should not be necessary to reveal which modal force is involved. Yet the utterance without negation is still not possible.
Another way to disambiguate the modal force is to use particles. Grosz (2011, 2014) shows that the German particles bloß and ruhig disambiguate imperatives in favor of universal and existential readings, respectively. Dutch also has such particles (Hedde Zeijlstra, pers. comm.). Maar disambiguates toward permission; nu ‘now’ disambiguates toward commanding.
‘Sit.’ (permission, not command)
‘Sit now!’ (command, not permission)
One might expect that since these particles disambiguate, they should permit the affirmative deverbal noun, as the No-Ambiguity Principle would be satisfied. But this is again not the case.
Intended: ‘Sure. Go ahead and honk.’ (as in (51B))
Intended: ‘Honk!’ (as a command)
So we see that RP’s accounts of ellipsis and the No-Ambiguity Principle do not adequately capture or explain the facts.
But even if RP’s account handled the details of the ellipsis in the Dutch deverbal noun, there is a problem with extending it to English no-gerunds. We have seen that the English no-gerund has properties unlike those of any previously identified gerund and that it cannot appear in a declarative clause. This means that there is no source for an ellipsis whose remnant is the no-gerund. Hence, NLCs cannot in general be the result of the need to disambiguate a remnant of ellipsis.
So RP’s paper does not provide an explanation of the NLC phenomenon. Fortunately, though, it has helped us identify NLCs in Dutch and provided additional support for the belief that the NLC phenomenon is not a quirk of English or Greek.12
4 Next Stop: Russian
On our crosslinguistic search for NLCs, we next come to Russian.13 This language shows the NLC as well, and again also with nominalizations. The following sentences are interpreted as commands, and negation must be present for this effect, the diagnostic of an NLC:
Ni-kak-ogo gulja-nij-a v parke!
NI-which-GEN.SG walk-NMN-GEN.SG in park
‘No walking in the park!’
Ni-kak-ogo pe-nij-a pesen!
NI-which-GEN.SG sing-NMN-GEN.SG song.GEN.PL
‘No singing of (the) songs!’
Let us start with a closer look at the morphology.14 Affixing -nij- to a verb is the regular means of producing a deverbal noun. Both in NLCs and elsewhere, the object of a deverbal noun is in the genitive; that is, the genitive case on the object pesen ‘songs’ in (57) is consistent with the wider behavior of deverbal nouns. The string ni-kak-ogo, which I have translated as ‘no’, is composed as follows. Ni is a negative marker, possibly related to the negative “conjunction,” illustrated here:
ni Vasja, ni Petja
neither Basil nor Peter
Kak is a “quexistential,” a word that has an interrogative or existential use depending on its environment.15 The genitive marker -ogo is obligatory; no other Case will do. The Case on the deverbal noun itself is also necessarily genitive. So ni-kak-ogo displays case, number, and gender concord with the nominal, as would be expected in Russian.
In other words, ni-kak-ogo is like the Dutch negative determiner geen or English no (or German kein), which also incorporate negation and an indefinite existential determiner.
So far, everything in the form of the Russian NLC is accounted for except the necessity of genitive on the negative determiner and the deverbal noun (as just noted, the genitive on the object of the deverbal noun is as expected16). Let’s turn, then, to the genitive on the negative determiner.
Russian is a negative concord language (unlike Dutch or English). As the string ni-kak-ogo is a negative concord item (see Brown 1999:chap. 3), a sentence where it occurs should also display sentential negation. However, no such additional negation is visible in NLCs. To make matters worse, overt negation is not possible, for reasons unknown to me. To stay consistent to the negative concord nature of Russian, I assume that there is a covert negative operator in these cases (for other cases of a covert negative operator licensing negative concord, see Zeijlstra 2008, 2010), though I do not profess to know the consequence of this possibility for the negative concord system of Russian at large.17
OpNEG ni-kak-ogo gulja-nij-a
OpNEG no walking
Postulating a (covert) negative operator explains one more property of (56) and (57), namely, the obligatory genitive on the negative determiner and the deverbal noun: this would then be one more instance of the famous Russian genitive of negation, which appears on indefinites in the scope of negation.
One might object that the genitive of negation is typically optional, while the genitive in the NLC is not.
*Ni-kak-oje pe-nij-e18 pesen!
NEG-which-NOM.SG sing-NMN-NOM.SG song.GEN.PL
Intended: ‘No singing of (the) songs!’
However, Babyonyshev et al. (2001:14), citing Chvany 1975 and Babby 1980, point to cases of obligatory genitive of negation with verbs that assert “existence, nonexistence, or presence at a location.”19
It is not impossible to see NLCs as falling under predicates of “existence, nonexistence, or presence at a location” in that they assert the nonexistence of permission (itself an item with existential force).
I conclude, therefore, that it is possible to analyze the genitive on the negative determiner and deverbal noun in a Russian NLC as a genitive of negation, thereby providing an additional argument for a covert negative operator (i.e., besides the argument regarding the need for a covert negative operator for purposes of negative concord, as outlined earlier).
So far, then, we have seen that Russian, like English and Dutch, has NLCs and that these include a nominalization and a negative determiner. Moreover, as in English and Dutch, not just any negation will do in the Russian NLC. Verbal/Sentential negation doesn’t work—with either genitive or nominative on the deverbal noun.
*Ne pe-nij-a/e pesen!
NEG sing-NMN-GEN/NOM.SG song.GEN.PL
Intended: ‘No singing of (the) songs!’
As for the nominalization contained in the NLC, it behaves like other Russian nominalizations. For example, an adjective precedes a nominalization and an adverb follows (63),21 and the same pattern holds in NLCs (64).
gromk-oje pe-nij-e pesen
loud-NOM.SG sing-NMN-NOM.SG song.GEN.PL
‘loud singing of (the) songs’
pe-nij-e pesen gromk-o
sing-NMN-NOM.SG song.GEN.PL loud-ADV
‘singing of (the) songs loudly’
Ni-kak-ogo gromk-ogo pe-nij-a pesen!
NI-which-GEN.SG loud-GEN.SG sing-NMN-GEN.SG song.GEN.PL
‘No loud singing of (the) songs!’
Ni-kak-ogo pe-nij-a pesen gromk-o!
NI-which-GEN.SG sing-NMN-GEN.SG song.GEN.PL loud-ADV
‘No singing of (the) songs loudly!’
As in Dutch, and as in the English ing-of NLC (though not the NLC based on the English no-gerund), the nominalization can appear in argument position but without the command reading.
Ja ne slyshal [ni-kak-ogo pe-nij-a pesen].
I NEG hear.PSTNI-which-GEN.SG sing-NMN-GEN.SG song.GEN.PL
‘I did not hear any singing of (the) songs.’
To summarize this section, then, we have seen that Russian has the NLC phenomenon, just like English and Dutch. Moreover, the NLCs in the three languages share several semantic and syntactic properties, even though language-specific properties like negative concord also manifest themselves.
So the NLC holds in at least two Germanic languages, in Greek, and in at least one Slavic language. But NLCs are also found in non-Indo-European languages, three of which we turn to next.22
5 Turkish, with a Bit of Hungarian and Hebrew
The NLC phenomenon can also be found in Turkish.23 The Turkish NLC uses the negated existential copula yok and a nominalized clause. The affirmative existential copula var is not permitted. This is exactly the profile of the NLC. Example (66) shows a common use of the negated existential copula. As the translation indicates, there is nothing modal in this sentence; that is, (66) does not mean that milk is forbidden in the fridge.
Dolap-ta süt yok.
fridge-LOC milk YOK
‘There isn’t milk in the fridge.’
The use of yok in the Turkish NLC is illustrated in (67), where it appears to take as its argument a nominalized clause. (I say “appears to” because, given that it is not clear to me what is behind the NLC phenomenon, I do not want to propose any particular syntactic or semantic composition.)
Geç saat-te TV izle-mek yok!
late hour-LOC TV watch-NMNYOK
Literally: ‘There isn’t watching TV late at night!’
‘No watching TV late at night!’
Sınıf-ta sakız çiğne-mek yok!
class-LOC gum chew-NMNYOK
Literally: ‘There isn’t chewing gum in class!’
‘No chewing gum in class!’
Bu kapı-yı aç-mak yok!
this door-ACC open-NMNYOK
Literally: ‘There isn’t opening this door!’
‘Don’t open this door!’
Dolap-ta süt var.
fridge-LOC milk VAR
‘There is milk in the fridge.’
*Yaşlı yolcu-lar-a öncelik ver-mek var!
elderly passenger-PL-DAT priority give-NMNVAR
Intended: ‘Give priority to elderly passengers!’
The contrast between (67a–c) and (69) attests to the presence of NLCs in Turkish as well. The nominalization that is the complement of the negated existential copula yok is quite high, as it contains Case for the object, including the accusative associated with specificity, as in (67c) (this is common for Turkish nominalizations).24
Hungarian exhibits the same syntactic pattern as that associated with the NLC in Turkish.25 That is, the negated existential copula nincs can yield commands (70) and the affirmative existential copula van cannot (71), nor can the absence of any copula (72).
So like the other languages we have seen so far, Hungarian exhibits the NLC phenomenon and its NLCs are constructed with a nominalization of a verbal projection. But like the Turkish NLC, the Hungarian NLC uses the negated existential copula with a nominalized clause, and not a negative determiner on a deverbal noun or gerund.
Hebrew also has the NLC phenomenon, using the same paradigm as Turkish and Hungarian.26 The negated existential copula eyn appears to take as its argument a nominalized clause, while the affirmative copula yeʃ precludes the command interpretation.
Eyn ʃimuʃ be-maxʃevon-im be-mahalax ha-bxina!
EYN use.NMN in-calculator-PL in-course.CS the-exam
‘No using calculators during the exam!’
#Yeʃ haxnasat klav-im la-park!27
YEʃ enter.CAUS.NMN.CS dog-PLDAT.the-park
Intended: ‘Dogs must enter the park!’
To summarize our results so far: the NLC phenomenon occurs in English, Greek, Dutch, Russian, Turkish, Hungarian, and Hebrew.
As a reviewer correctly points out, the NLC is not a “construction.” The morphosyntax involved in English, Dutch, and Russian, where a negative determiner is used, is different from that involved in Turkish, Hungarian, and Hebrew, where a negated copula is used, and Greek, which uses neither a negative determiner nor a negated copula, is different still. At most, what the expressions have in common syntactically in the languages examined (beyond the presence of some form of negation) is the presence of a nominalization. That is, I have not found any cases of NLC with verbs. So the NLC is not a (syntactic) construction but a convergence of properties that can be characterized as I did in section 1: different syntactic expressions that have one common characteristic, namely, that they survive as commands only when they are negated.
6 A Possible Account? (Unfortunately, Not)
Languages like Turkish provide the impetus to explore a particular avenue toward a possible account for NLCs. Consider the Turkish NLC again.
Geç saat-te TV izle-mek yok!
late hour-LOC TV watch-NMNYOK
Literally: ‘There isn’t watching TV late at night!’
‘No watching TV late at night!’
Given the literal meaning of the negated existential copula and the absence of imperative morphology or overt modality, one might think that there is a pragmatic phenomenon going on that can also be observed in English. Consider (76a–b), which can be understood as negative commands like (76c).
In this house, we don’t watch TV at night.
In this house, there’s no TV watching at night.
Don’t watch TV at night!
It is clear how this implicature might come about: if there is no TV watching at night, it must be because it is not allowed. But such a command interpretation is optional, as the possibility of the following continuation indicates:
(77) In this house, we don’t watch TV at night because the power goes off at 8.
In this house, we brush our teeth at 8 p.m.!
Brush your teeth at 8 p.m.!
Could it be that the negated existential copula in Turkish brings about a command in a pragmatic way? That is, is (75) a command because it is interpreted as (76a–b) are when they yield the meaning of (76c)?
The answer is no. In English, the pragmatic inference of a command is optional: it is possible to interpret (76a) simply as a description of facts, as we saw in (77). But this is not possible with the construction used in Turkish.
#Maalesef, gece TV izle-mek yok çünkü akşamları elektrik gid-iyor.
unfortunately night TV watch-NMNYOK because evenings electricity go-IMPF
The meaning of the relevant substring in (79) is still only an NLC, and (79) as a whole has the somewhat incoherent meaning ‘Unfortunately, it is forbidden to watch TV at night because the electricity goes out in the evening’—not ‘Unfortunately, it is not possible to watch TV at night because the electricity goes out in the evening’.
In Turkish, one can express the equivalent of a pragmatically inferred command, as in (76), which can lose the implication of a command reading in context, as in (77); however one cannot do this with the negated existential copula. Instead, simple sentential negation is used.
Bu evde gece hiç TV izlemiyoruz çünkü akşamları elektrik gid-iyor.
this house.LOC night at.all TV we.not.watch because evenings electricity go-IMPF
‘In this house, we do not watch TV at night because the electricity goes off in the evenings.’
Moreover, a pragmatically inferred command does not require negation; see again (78a), which does not. However, the affirmative copula in Turkish does not yield an affirmative command; recall (69). Therefore, as predicted, the equivalent of (78a) with the affirmative copula lacks the command reading and “is a weird sentence. It literally means ‘There is brushing teeth at 8 p.m. in this house’. I cannot imagine this being uttered in any context” (Ömer Demirok, pers.comm.).
Bu evde saat 8’de diş firçala-mak var.
this house.LOC hour 8-LOC tooth brush-NMNVAR
‘In this house, there is some brushing of teeth at 8 p.m.’
In other words, the NLC is not reducible to a pragmatic command, not even in Turkish, where one might at first have thought it might.
7 Another Possible Account? (Unfortunately, Again No)
Since RP brought to mind ellipsis as a possible source for the NLC, and since Turkish raised the topic of (negated) existential sentences, let’s look at another possible source of English no-gerunds.
Kjellmer (1980) describes an interesting construction that also uses the no-gerund (though we will see that it differs from the NLC); see (82a). I will refer to it as the Kjellmer construction for lack of a better term. First, as the paraphrase in (82b) shows, the Kjellmer construction has a modal interpretation.
There was no mistaking that scream. =
It was not possible to mistake that scream.
Second, negation is obligatory. The absence of negation does not result in an interpretation of affirmative modality.
(83) #*There was mistaking that scream.
Third, not just any negation will do. Only the negative determiner can produce the desired result.
*There was not mistaking that scream.
*There was never mistaking that scream.
*There was seldom mistaking that scream.
*There was hardly mistaking that scream.
Finally, the gerund in the Kjellmer construction looks like the no-gerund (which is also used in the NLC), since it can both take a determiner and take an accusative object, as (82a) shows.
Kjellmer himself does not explain the syntactic or semantic properties of the construction in (82a) but observes what it takes to undo the modality. He notes that unlike (82a), (85) lacks a modal interpretation. This sentence asserts that no shouting, merry-making, or waving of flags occurred. It does not say that it was impossible to do these things.
There was no shouting, no merry-making, no waving of flags.
(86) There was shouting, merry-making, and waving of flags.
There wasn’t shouting or merry-making or waving of flags.
There was never shouting or merry-making or waving of flags.
There was seldom shouting or merry-making or waving of flags.
There was hardly shouting or merry-making or waving of flags.
By and large it could be said that the modal gerund can take a direct object, especially an object clause, that it must occur in the singular, that it hardly takes any other adjectival adjunct than no and that it can be turned into a passive. The non-modal gerund, on the other hand, takes a direct object only in special circumstances, and never an object clause, it can be turned into a plural and have various adjuncts other than no and it can hardly be made passive. (Kjellmer 1980:60; my emphasis)
The question now arises whether we can derive the NLC from Kjellmer’s construction by an RP-style ellipsis.
There wasno smoking!
Unfortunately, this path doesn’t seem promising either. First, the no-gerund in the NLC has no verb class restriction and can be intransitive, unlike the Kjellmer construction, which can be modal only with an object, as Kjellmer himself observes in the above quotation; see (89) and (90), respectively.
(89) No smoking! (modal)
There was no smoking. (not modal)28
There was no smoking that cigar. (modal)
Second (and very crucially), the modality is very different in the two cases. In the NLC no-gerund, it is deontic (and performative), while in the Kjellmer construction, it is ability or dynamic (and reportative).
There was no mistaking that scream. =
It was not possible to mistake that scream.
(Not: It was forbidden to mistake that scream.)
There is no killing that rat. =
It is not possible to kill that rat.
(Not: It is forbidden to kill that rat.)
In other words, despite some apparent similarities (e.g., they both appear to use the no-gerund and they are both modal), the NLC cannot be reduced to the Kjellmer construction.
Moreover, we have already seen several languages that have the NLC but I know of the existence of the Kjellmer construction only in English. Even in Turkish, which shows the NLC with the negated existential copula (which would make it close to There is no . . . ), the relevant sentences cannot receive the same interpretation as the Kjellmer construction; they only receive the command reading. Having said this, I should point out that the Kjellmer construction and the role of negation in it is mysterious. But whatever this construction is, it is not an instantiation of the NLC phenomenon, given that it is not a command to begin with.
So we still do not have a good candidate to reduce the NLC to, nor an obvious explanation for it.
In this section, we will look for (and find) the NLC phenomenon in one more language, unrelated to the rest: the Niger-Congo language Buli, spoken in Ghana.29 As expected, Buli has an imperative verb, which can be negated.
Kàn ɲū sīgāārī!
NEG drink cigarettes
‘Don’t smoke cigarettes!’
But Buli also has the NLC. Consider (94), which is an NLC because the absence of negation does not result in an affirmative command. In fact, the resulting strings are ungrammatical as complete sentences (95).
Sīgāārī ɲūkká kà dē!
cigarettes drink.NMNNEG here
‘No smoking here!’
*Sīgāārī ɲūkká dē!
cigarettes drink.NMN here
Intended: ‘Smoke here!’
Sīgāārī ɲūkká ká ɲīɲsā sōkā dōkkú pō!
cigarettes drink.NMNNEG body bath.NMN room inside
‘No smoking in the bathroom!’
Like NLCs in the other languages, Buli NLCs involve nominalizations. So let us take a quick look at some of their properties. First of all, the nominalizing suffix -ka appears on the verb. Buli has no overt Case morphology, so we cannot check the effect of nominalization on the object of the deverbal noun, but the effect of nominalization is apparent from word order: Buli is VO, but when nominalization has taken place, the order is OV.
Asouk dʒù yénní. VO
Asouk burn house.DEF
‘Asouk burned the house.’
*Asouk yénní dʒù. *OV
Asouk house.DEF burn
Intended: ‘Asouk burned the house.’
yénní dʒūká *OVNMN
‘the burning of the house’
*dʒūká yénní VNMNO
Intended: ‘the burning of the house’
Sīgāārī ɲūkká kà-dē!
‘No smoking (here)!’
*ɲūkká sīgāārī kà-dē!
drink.NMN cigarettes NEG-here
So what we have is a nominalization that serves as a command, but only if it is negated. One may wonder about the apparent necessity of a locative (dē ‘here’), or the more complex locative in (96). I do not know the reason for this, but what is important is that a locative is equally necessary in an NLC with a plain noun.
Bāāsā kà *(dē)!
dogs NEG here
‘No dogs here!’
So even though the obligatory presence of a locative remains mysterious, the fact that the same mystery exists with the NLC with nouns shows some consistency.
9 Just One More Attempt
We have seen that if negation is removed from the NLC, the remnant does not express a command. We have also seen that some possible analyses cannot account for the fact that without negation, there is no command reading. In this section, we will look at substituting something else for the negation and see if the result can be a command. That is, we will explore whether there is an X, such that an “X-licensed command” can occur in the same environments in which negation brings about the NLC effect. If such an X exists, we might be able to understand the NLC better by exploring whether and how negation plays a role similar to that of its substitute in forming a command.31 For space reasons, I cannot cast the net too widely here, so I will test only one substitute: a locative predicate, which at first blush appears to yield the desired effect. Consider (102) and its paraphrases in (103). Imagine that I am having a party at my house and I find some guests smoking cigars in the kitchen. I tell them:
(102) Smoking cigars, on the balcony, please!32
If you want to smoke cigars, you must/will do it on the balcony.
If there is to be any cigar-smoking, it must/will happen on the balcony.
A similar phenomenon is seen with nouns, which also occur in the NLC (104a–b). Now imagine a party where many of the guests are likely to bring dogs. Then, one can say (104c) with the meaning of (104d).
Dogs, on the porch!
If you have brought a dog, you must/will put it on the porch.
The same holds for the Dutch deverbal-noun command.
Gepraat, op straat, niet in huis!
talk.NMN on street not in house
‘Talking, on the street, not in the house!’
Getoeter alleen op koninginnedag.
honk.NMN only on Queen’s.Day
‘Honking, only on Queen’s Day.’
Here, we see that the form of the nominalization that appears in the NLC can appear without negation, as long as another constituent is provided that specifies the conditions (location, time) under which the action in the nominalization is permitted. One way to account for that would be to posit a covert permission modal, accompanied by a covert only. Another way would be to invoke a universal deontic modal: the set of situations involving talking/honking/etc. is a subset of the set of situations involving the balcony/Queen’s Day/etc. This would also have the desired result of permitting talking only on the street/honking only on Queen’s Day/etc.
Dogs must be CArried.
[All acceptable situations s: s is a situation in which there is a dog] (s is a situation in which a dog is carried)
Smoking cigars, on the BALcony!
[All acceptable situations s: s is a situation in which there is a smoking of cigars]
(s is a situation in which the smoker is on the balcony)
The tripartite structure of (107a–b) is provided by focus and the deontic modal.33 (108a) also has a focus structure, so an analysis parallel to (107b) might be warranted even though (108a) has no overt modal. After all, its interpretation is clearly modal: ‘Smoking, if it is to happen, must happen on the balcony’. So the conclusion that a covert modal must be postulated in (108a) seems rather unavoidable.
Could it be that the expression stating the conditions under which the action is permitted (as in (108)) satisfies a need of this modal structure that is satisfied by negation in the NLC? For example, does negation contribute the tripartite structure necessary for the (covert) modal, the way the locative (or temporal, etc.) conditions do in (107)?34 Without spelling out the modal structure or the syntax-semantics mapping involved, and keeping in mind that no is a symmetric determiner, we could interpret (109a) as in (109b–c).35
No smoking (here)!
[All acceptable situations s: s is a situation in which one is here] (s is not a smoking situation)
The intersection of the set of smoking situations with the set of acceptable situations is the null set.
So in the absence of negation, the relevant nominalization—or rather, the modal associated with it—will either need to find a different interpretive structure, as in (108), or remain uninterpretable.
In short, from this perspective on the NLC mystery, negation is one way of satisfying the need to “set up” the modal’s tripartite structure.36
Now, does the crosslinguistic picture support this view? Like English and Dutch, Russian seems amenable: the nominalization in question is ungrammatical without negation (110b), but becomes grammatical (and modalized) when another constituent provides the conditions of acceptability (110c).
Ni-kak-ogo gulja-nij-a s sobakoj!
NI-which-GEN.SG walk-NMN-GEN.SG with dog.INSTR
‘No walking with a dog!’
*Gulja-nij-e s sobakoj!
walk-NMN-NOM with dog.INSTR
Intended: ‘Walk with the dog!’
Gulja-nij-e s sobakoj – (tol’ko) vo dvore!
walk-NMN-NOM with dog.INSTR only in yard
‘Walking with a dog, only in the yard!’
As would be expected, the genitive on the determiner and nominalized verb are gone in (110c), as there is no negation to trigger the genitive of negation.
In Hebrew, it is also possible to replace the negation by another constituent that provides conditions of acceptability and yields a grammatical sentence. However, the existential copula must be removed altogether; it is not enough to just replace it with the affirmative one, as in (111c).
Klav-im ba-gina bilvad!
dog-PL in.the-garden only
‘Dogs, in the garden only!’
ʃimuʃ ba-telefon ba-mirpeset bilvad!
use.NMN in.the-phone in.the-balcony only
‘Using phones, on the balcony only!’37
*Yeʃ ʃimuʃ ba-telefon ba-mirpeset bilvad!
YEʃ use.NMN in.the-phone in.the-balcony only
Intended: ‘Using phones, on the balcony only!’
In Hungarian, this is also possible. Unlike in Hebrew, though, the affirmative copula can optionally stay—though its presence affects the word order, as in (112d), or requires a long break, as in (112e).
‘No walking the dog!’
Intended: ‘Walk the dog!’
Kutya-sétál-tat-ás csak a park-ban!
dog-walk-CAUS-NOM only the park-INE
‘Walking a dog, only in the park!’
Csak a park-ban van kutya-sétál-tat-ás!
only the park-INEVAN dog-walk-CAUS-NOM
‘Walking a dog, only in the park!’
Kutya-sétál-tat-ás — csak a park-ban van!
dog-walk-CAUS-NOM — only the park-INEVAN
‘Walking a dog, only in the park!’
But the result of these variations is different in Turkish than in the languages we have looked at so far. In the NLC with plain nouns, the paradigm can be duplicated. That is, removing negation is damaging but replacing negation with a locative condition makes the command resurface.
‘Dogs, to the balcony!’
However, removing negation and introducing the conditions of acceptability results in ungrammaticality (with or without the affirmative copula).
*Sigara iç-mek, balkon-da!
cigarette drink-NMN balcony-LOC
Intended: ‘Smoking, on the balcony!’
*Sigara iç-mek var, balkon-da!
cigarette drink-NMNVAR balcony-LOC
*Balkon-da sigara iç-mek var!
balcony-LOC cigarette drink-NMNVAR
The same result holds in Buli. The equivalent of Smoking, on the balcony! is not possible.
*Sīgāārī ɲūkká vērāndā wá pō.
cigarettes drink-NMN balcony DEM inside
So reducing the NLC to negation, creating a tripartite structure, does not, in an obvious way at least, easily account for the crosslinguistic picture. Yet perhaps future research will find more benefits in this path.
The NLC presents two interrelated questions: Where does the command force come from? Why is negation necessary for the command interpretation?
These two questions contain certain presuppositions. I said that the NLC expresses a “command,” not a “prohibition.” If it’s said to express a prohibition, the necessity of negation is not mysterious, of course: no negation, no prohibition. But where would the directive force come from? The negations used in the languages discussed here do not have a modal or directive meaning in general.38 Nor do the nominalizations involved in the NLC express a command in and of themselves. If they did, they would not need the presence of negation to become commands; they would express affirmative commands all by themselves. So calling NLCs “prohibitives” does not help.
Consider imperatives. The default assumption regarding imperatives is that a directive force is present—semantically or pragmatically, depending on the theory. So in negated imperatives, with all the interesting questions they raise, the directive force scopes over negation, and from that point on, the process is supposed to be compositional. So on the hypothesis that (116a–b) are equivalent, and with the assumption that the negative determiner has no modality, the modal force must be associated with the nominalization, a covert modal, or a mapping to a to-do list. But if that is the case, then why negation is necessary for the modal meaning is mysterious.
Don’t walk on the grass!
No walking on the grass!
Another presupposition hidden in the term “command” is that the directive force, however it is achieved, is strong/universal and that it scopes over negation, the way one would assume happens in (116a), in order to yield the appropriate meaning. This may be the right assumption for (116a) since the directive force does not depend on the presence of negation. However, the same conclusion is not obvious for the NLC in (116b). It could very well be that the NLC contains an element with existential force that scopes under negation. The net effect would be the same, as a universal scoping over negation is equivalent to an existential scoping under it. It might prove fruitful to try this path, though, because—with the exception of Greek and Buli—the languages we looked at might carry the negation-over-existential somewhat on their morphological sleeve: English, Dutch, and Russian use a negative determiner that consists of negation over an existential determiner, while Turkish, Hungarian, and Hebrew use a negated existential copula. However, if the modal existential force is the existential contained in the negative determiner or negated copula, then that brings us very close to saying that the modality or directive force itself is located in the negative determiner and negated copula, a position that seems hard to defend, given the many nonmodal uses of these items.39
So even though the NLC may well prove to be a “negated permission,” as opposed to a “command,” this hypothesis also faces many difficulties, not the least of which is, again, the necessary presence of negation: why should negation be necessary for the permission interpretation to be present? Specifically, if No smoking! is a command, why is Smoking! not a command as well? On the other hand, if No smoking! is a negated permission, why is Smoking! not an affirmative permission? Either way, the obligatory presence of negation remains a mystery. So on the assumption that most would call (116a) a command, I chose to call (116b) a command as well, with the caveat that I cannot exclude that we are dealing with an existential that scopes under negation.
I hope to have shown that the phenomenon of negation-licensed commands exists and has some crosslinguistic stability. Several unrelated languages have nominalizations that become commands in the presence of negation, crucially despite differences in language-particular properties or language-particular morphosyntactic expressions of the NLC. Along the way, I attempted a few analytical paths, mostly inspired by morphosyntactic properties of the NLC in particular languages, but I failed to reduce the NLC to any other phenomenon. So success is quite moderate for now.
1 It is also well-known that many of these verb forms can convey noncommands as well. Note the acquiescence/permission (i) and indifference (ii) uses of the English imperative.
It’s very hot in here. May I open the window?
Sure. Open the window.
Which way should I go?
Go left. Go right. Do whatever you want.
See von Fintel and Iatridou 2017:sec. 3.5 for a crosslinguistic picture of possibilities.
2The NLC phenomenon should not be confused with the known problem of combining negation with the imperative verb. In many languages, such as Greek (i), negation cannot be combined with the imperative verb. Instead, a nonimperative verb must be used—one that typically also has command uses, such as subjunctive or infinitive—to combine with negation (see, e.g., Rivero and Terzi 1995, Han 2000, Zeijlstra 2013).
This is not the same as negation being necessary to express a command (NLC).
Nor should NLCs be confused with prohibitives, where the bare stem combines with a marker that contains both negation and directive force (van der Auwera 2010). I will briefly return to prohibitives in section 10.
3 For Abney (1987) and others, the different types of gerunds are the result of the height of attachment of the nominalizing suffix -ing. The name of the highest nominalization, ACC-ing, is derived from the fact that the subject of the gerund is in the accusative. POSS-ing is derived from the fact that the subject of the gerund looks like a possessor. The term ing-of reflects the belief that this nominalization is the lowest—so low, in fact, that the verb is deprived of its ability to assign Case, which results in the need to insert the (dummy) Case assigner of. Finally, the term PRO-ing, whether or not it refers to a separate type, reflects the belief that the unpronounced subject is PRO.
4 Though if all types of gerunds are DPs (or NPs), this question is not specific to the ing-of gerund. But this is a moot point, as we will soon see.
5 Of course, not every sentential phrase with the no+noun combination has to be an NLC. The following were contributed by Roger Schwarzschild (pers. comm.):
No wonder she’s so happy!
No shit! ≈ I’m not surprised.
6 The negative marker oxi is not sentential negation, nor is it a negative determiner. It is the particle used in negative answers.
Is it raining?
It is also the marker for what is called “constituent negation.”
Idha tin Miranda ala oxi tin Elena.
saw.1SG the Miranda but not the Elena
‘I saw Miranda but not Elena.’
Certain other nominalizations (though not the NLC) can take the negator mi, which otherwise appears in what some call the subjunctive.
i mi apodochi tis pragmatikotitas
the NEG acceptance the reality.GEN
‘the nonacceptance of reality’
But mi does not appear in NLCs.
7 Unlike the no-gerund, the negated ing-of gerund can appear in argument position in a plain declarative sentence, but without a modal interpretation.
I witnessed no teasing of the sister.
8 One might wonder whether the NLC is the result of the presence of a covert modal (rather than the deletion of one). However, neither the obligatory presence of negation nor the particular force of the modal would be better explained than in the deletion account.
9 Many thanks to Gertjan Postma and Hedde Zeijlstra for data and for patient and generous discussion of Dutch.
An initial investigation points to the distinct possibility that what we observe in Dutch holds in German as well. Even so, here I will give only examples from Dutch.
10 While these negated deverbal noun commands are very productive, they are not possible with unaccusatives or verbs that have no argument (Gertjan Postma, pers. comm.).
*Geen geval hier!
no fall here
Intended: ‘No falling here!’
But this restriction is not particular to the command use of these deverbal nouns. It holds for these deverbal nouns in general (Postma 1996). Moreover, like the ing-of NLC, and unlike the no-gerund, these deverbal nouns have (in general) no Case on their object, which must either incorporate or take the preposition van; this is preserved in the NLC.
Geen boekengelees hier!
no book.read here
‘No book-reading here!’
Geen gelees van boeken in mijn tuin!
no read of books in my garden
‘No reading of books in my garden!’
And like the ing-of NLC, but unlike the no-gerund, the negated deverbal noun can be an argument—but, crucially, without the command reading.
Ik hoorde geen gepraat.
I heard no talking
11 To be clear, RP regard the No-Ambiguity Principle as holding of language in general, not just of covert items or ellipsis.
Ambiguous structures do not exist. . . . We will assume that [(39)] is a particular realization of the Principle of Full Interpretation (FI). In the cases in [(40b)] and [(41a)], disambiguation is mediated by the Lexicon. Under the assumption that [(39)] is a general property of Dutch and English, it follows that abstract, i.e. nonlexical AUXMOD is not well-formed unless it is disambiguated. (RP 2007:280)
12 A question that arises is how similar NLCs are to imperatives. For example, one might wonder whether NLCs share with imperatives the ability to express permission/acquiescence or indifference, participate in conditional conjunctions as the first conjunct, and so on (von Fintel and Iatridou 2017 and references therein). As far as I can tell, in English, Greek, and Dutch the NLC can neither yield noncommand readings nor participate in conditional conjunctions. The NLC is only interpreted as a command. In other words, NLCs are properly named: they are commands, not “imperatives,” as they lack the flexibility that imperatives have. Another difference between NLCs and imperatives is that the subject in an NLC can be PROarb, while the subject in an imperative cannot.
No washing oneself in public!
*Don’t wash oneself in public!
This difference is interesting, but it is not clear to me how it necessarily relates to the interpretive variability of imperatives as opposed to the “inflexible” command reading of the NLC.
However, the difference between (i) and (ii) points to a possible insight about a further difference between NLCs and imperatives, suggested by Paul Portner (pers. comm.). According Portner, Pak, and Zanuttini (2019), nonfinite or nominalized clauses are often used for directives that are applied not to a specific interlocutor, but to anyone to whom they can apply (a “generic addressee”). This seems true (though not necessary) for the NLC. Throughout this article, I bring up the possibility of using noun-based NLCs (No dogs!) and other NLCs (No walking on the grass!) on signs, which lack a specific addressee. The NLC can sometimes be seen as reporting a standing prohibition that applies to the addressee (in the absence of addressee-specific markers, as in No teasing your sister!). In Portner, Pak, and Zanuttini’s terms, this suggests absence of the left-peripheral material that encodes the addressee or the speaker-addressee relation. However, as the absence of such material relates to the issue of (un)embeddability of imperatives and commands in Portner, Pak, and Zanuttini 2019, this connection should be explored further because NLCs are unembeddable in the languages discussed.
Finally, a reviewer points out disjunctions like (iii), where imperatives can appear, and suggests that it would be instructive to check whether the NLC can appear there. In fact, it can (iv).
Cover your pipes during the winter or they will burst.
No reading The Guardian in class or you will get an F.
Despite their inherent interest, such disjunctions are not a test for quantificational variability, however, as the first disjunct is always interpreted as a command (Kaufmann 2012).
13 Thanks to David Pesetsky and Sergei Tatevosov for discussion of the Russian data.
14 This decomposition is due to Sergei Tatevosov (pers. comm.).
15 I take this term from Hengeveld, Iatridou, and Roelofsen 2018.
16 So with respect to Case on its object, the Russian deverbal noun is like the Dutch deverbal noun and the English ing-of NLC, and not like the English no-gerund.
17 The marker ni cannot itself be the negative operator, as elsewhere it requires negation to be present in the same sentence (see Brown 1999).
18 This string is ungrammatical with a deverbal noun of any gender. (Nominalizations with -nij are always neuter. So a different gender would not make the sentence grammatical.)
19 Thanks to David Pesetsky for making me aware of obligatory cases of genitive of negation and for discussion of this point.
20 Sergei Tatevosov (pers. comm.) notes that (61b) is acceptable on the reading ‘The doctor was not in town’, with v gorode ‘in town’ focused.
21 This pattern also occurs in nominalizations for the several speakers of English I consulted (including ing-of gerunds).
his deliberate removal of the evidence
his removal of the evidence deliberately
*his deliberately removal of the evidence
The following is another example:
22 One difference between Russian and Dutch/English lies in whether the NLC with nominalizations can be reduced to the NLC with nouns. Though the latter is available in English and Greek (and Dutch as well), it is not available in exactly the same way in Russian. The following construction is ungrammatical on signs:
However, one can utter (i) upon seeing someone with a dog entering a space where dogs are not allowed (Mitya Privoznov, pers. comm.). Why (i) is bad on a sign, I do not know.
23 Many thanks to Ömer Demirok for Turkish data and discussion.
24 Even though nominalization is present, it is unclear whether the NLC with nominalized clauses can be reduced to the NLC with nouns in Turkish, as NLCs are not very productive with nouns. According to Ömer Demirok (pers. comm.), examples like (i) are possible but very restricted. Specifically, (ii) could only be a parent’s comment to a child.
Intended: ‘No dogs!’
25 Many thanks to Dóra Takács for Hungarian data and discussion.
26 Many thanks to Omer Preminger for Hebrew data and discussion.
27 In all three languages, it is only the negated present tense existential copula that delivers NLCs.
The only possible difference I have found between my Hebrew and Hungarian consultants is that for Dóra Takács the presence of the affirmative copula leads to ungrammaticality in Hungarian (see (71)), while Omer Preminger reports that a Hebrew string like (74) can mean ‘The entering of dogs into the park exists’, “though it’s an odd sentence.” (See Ömer Demirok’s identical comment on the equivalent Turkish sentence in section 6.) However, due to the small number of consultants, I do not know how this would generalize. It may be that what is an “odd sentence” for one speaker is an ungrammatical sentence for another. What is clear is that in all these languages, the affirmative copula does not yield affirmative commands.
28 One might rebut that (90a) is modal with the present tense.
There is no whispering in his house.
But this is a case of a pragmatically induced rule, as we saw earlier; and such an interpretation, unlike with the no-gerund, is always optional.
There is no whispering in this house as everybody is by nature loud.
29 Many thanks to Abdul-Razak Sulemana for data and discussion.
30 The plain imperative verb cannot appear in this construction (with or without negation).
*ɲū sígāārī kà-dē!
drink cigarettes NEG-here
*Kàn ɲū sīgāārī kà-dē!
NEG drink cigarettes NEG-here
31 In this context, I should mention that in English, only seems to trigger a phenomenon similar to the NLC.
No eating solid foods! Only drinking liquids!
If only decomposes into negation and another element (e.g., ‘other than’), as in von Fintel and Iatridou 2007, then maybe (i) reduces to the NLC. (Similarly for No more eating solid foods!) However, some of the languages discussed (e.g., Turkish, Hungarian) do not permit their counterpart of only with this effect. For reasons of space, I do not explore this variation here.
32 The comma after smoking represents a small break that is required. Also, the intonation has to have a particular contour. One reviewer’s intuition is that (102) has a different import than, for example, (104) and may be elliptical for Are you smoking cigars? Then, on the balcony. The speakers I consulted, however, do not feel that (102) can only function as a question addressed to known or potential smokers; they feel it can easily be interpreted along the lines suggested in (103).
DOGS must be carried.
[All acceptable situations s: s is a situation in which one is in this place] (s is a situation in which a dog is carried)
[All acceptable situations s: s is a situation in which something is carried] (s is a situation in which a dog is carried)
Smoking ciGARS on the balcony.
[All acceptable situations s: s is a situation in which one is on the balcony] (s is situation in which one smokes cigars)
34 To my knowledge, Partee (1992) was the first to explicitly propose a tripartite structure for negation, though she mentions precursors of the relevant insight. For example, Jackendoff (1972) discusses association with negation, which is also a tripartite partition. Thanks to Roger Schwarzschild for this reminder.
35 Note the paraphrase with an existential modal (acceptable).
36 This discussion somewhat mirrors the observation by Carlson (1989) and others that generic sentences need the overt expression of what Carlson calls the “locus of genericity.”
*John smokes a cigarette.
When John is nervous, he smokes a cigarette.
37 Given the necessary absence of the negated existential copula, one wonders whether these are not pragmatically induced commands, like the ones discussed in section 6.
38 Unlike some prohibitive markers described in van der Auwera 2010, which come from verbs meaning ‘stop’, ‘refrain from’, and so on.
39 One might imagine the following argument: the quantificational force is in the negative determiner or negated copula but the modal itself (sans quantificational force) is elsewhere/lower. This would be some sort of negative existential closure of a modal variable. However, one would still have to explain why the existential closure of the variable would have to be negated. Alternatively, one might propose a covert negative-polarity-item modal. This would capture the obligatory presence of negation. But defending the position that such a covert item exists in so many different languages (and why it would need a nominalization as argument) seems a daunting task and I will not attempt it here.
For discussion of this material, I am very grateful to Ömer Demirok, Kai von Fintel, David Pesetsky, Paul Portner, Gertjan Postma, Omer Preminger, Roger Schwarzschild, Abdul-Razak Sulemana, Dóra Takács, Sergei Tatevosov, Gary Thoms, Vina Tsakali, and Hedde Zeijlstra. Many thanks also to two anonymous LI reviewers. I started this article for the workshop Non-canonical Imperatives (May 2018, Berlin), organized by Despina Oikonomou and Artemis Alexiadou. I thank the organizers and the audience at that workshop, as well as audiences at MIT; the University of Massachusetts, Amherst; and NY-St. Petersburg.