Apprehensional morphology cite

Feature set based on Vuillermet (2018).

Authors: Marine Vuillermet

Acknowledgements: Natalia Chousou-Polydouri, David Inman, David Timothy Perrot

What?

Apprehensional markers are grammatical constructions that conventionally encode both high probability and undesirability, as in the following examples:

(1)	Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru; Vuillermet 2018: 272)
	’Biya	’biya	’biya	’biya!	Kekwa-ka-chana	miya!
	bee	bee	bee	bee	pierce-3A-APPR	2SG.ABS
	‘Bee, bee, bee, bee! Watch out it might sting you!’

(2)	Hup [hupd1244] (Naduhup; Colombia, Brazil; Epps 2008: 631-632)
	a.	ʔam	nɔ́h!
		2SG	fall.APPR
		‘(watch out) you'll fall!’

	b.	náw=yɨʔ	tok-póɡ=hə	cə́c
		good=TEL	pound-EMPH1.IMP=TAG2	INTERJ

			ʔam	tok-pəʔ-yæt-yɨʔ=pɔ̌ɡ!
			2SG	pound-spill-lie-TEL.APPR=EMPH1
		‘Pound (the coca) carefully, darn it! You’ll spill it all out!’

By grammatical constructions, we mean that the construction should not be lexical. Attention-getting particles like watch out! in English or (fai(te)s) attention! ‘(pay) attention’ in French are considered lexical, as they are still very transparently linked to their original lexemes. They are in the imperative form, and can occur in other constructions like in the following subordinate complements: ‘they advised us to watch out for our personal belongings’, or ‘ils nous ont conseillé de faire attention à nos affaires’. Similarly, mind in sentences like mind the gap is still a full lexical verb which can also occur in subordinates ‘old people always carefully mind the gap’. Prepositional phrases like out of fear of X are also not grammaticalized, as they are available with other emotion words (out of shame, out of anger, out of joy, etc.).

The present survey considers dedicated constructions which conventionally encode, rather than pragmatically implicate, an undesirable possibility. For instance, the potential marker might typically encodes a desirable event if the verb is (typically) positive (she might win!), and an undesirable event if the verb is (typically) negative (she might fall!). It is thus not a dedicated apprehensive, and is not considered in our survey. We discuss the distinction below in example (3), and in Appr-01.

Lichtenberk (1995) is a seminal paper on the apprehensional domain, which proposed three distinct functions, namely the apprehensional-epistemics (renamed here apprehensive, following Vuillermet et al forthcoming), the precautioning, and the fear-function. This survey considers an additional fourth function, the timitive. Some languages, especially from Australia, have a single apprehensional construction to encode several functions (e.g. Marithiel, see example 11), while others have a distinct construction for each function (e.g. Ese Ejja, see examples (1, 3) for the Apprehensive vs. (4) for the Precautioning vs. (6) for the Timitive). The four functions are now examined in turn.

Apprehensive

The apprehensional constructions illustrated in (1)-(2) are modals called “apprehensives” that belong to the tense/mood systems of the languages. The speakers consider that the events of stinging, falling and spilling out will very probably happen, and that they are also undesirable, mostly to the addressee in (1) and (2a), but also to the speaker in (2b). While the events in the first three examples are typically undesirable, the undesirability encoded by an apprehensive is still present with more neutral events like ‘go’ (3a) and even typically desirable ones like ‘smile’ (3b).

(3)	Ese Ejja (Vuillermet 2018)
	a.	A’a	María	wowi-jji,	poki-chana!
		PROH	María	tell-PROH	go-APPR
		‘Don’t tell María, (beware of that) she may come along!’

	b.	Marina	swa-chana	mi=’ba-majje!
		Marina	smile-APPR	2SG.ABS=see-TMP.SS
		‘Marine might laugh at seeing you!’ [volunt]

The context of (3a) is that of two friends who want to go into town but would prefer María, the wife of one of them, not to come along. In (3b), swa- ‘smile’ can only be interpreted as ‘smile in an undesirable way’, i.e. ‘laugh at’. These two examples make clear that -chana conventionally encodes undesirability, as they rule out the possibility that the undesirability could only come from pragmatic implications. By contrast, the potential marker might in English encodes either desirability or undesirability, depending on the event itself.

A main feature of the apprehensive function is to always encode the speaker’s perspective: the apprehensive construction is a direct speech act. This perspective (the undesirability and high potentiality of the event) might or might not be shared by the addressee: a sentence like 'watch out you'll fall' in (2a) might be uttered by a worried parent, while their child is confident they are not at risk. The speaker's perspective is a major difference with the closely related precautioning function (see following subsection).

An important characteristic of the apprehensive function is both the frequently reduced availability of grammatical persons, and the impact of these grammatical persons on the pragmatic meaning of the construction (resulting mostly in warning vs. threatening interpretations, see Vuillermet et al (forthcoming) for a detailed discussion). This is investigated with features Appr-03 and Appr-04.

Precautioning

The second function of grammatical apprehensional constructions is called “precautioning”. It is similar to the apprehensive function, but in addition to the apprehensional clause, there is an obligatory second clause, called a “preemptive clause” Evans (1995: 264), which expresses possible actions to avoid the undesirable situation.

In (4), cleaning the rifle (preemptive clause) prevents the game from escaping (precautioning clause). And in (5), taking the umbrella prevents the consequences of the raining.

(4)	Ese Ejja (Vuillermet 2018: 276)
	Owaya	shijja-’bame~’bame-ka-ani	[ijjiakajji	e-’ajjo	kwajejje]
	3ERG	clean-well~RDP-3A-PRS	game	PREC-escape	PREC
	‘He cleans his rifle lest the game escape.’

(5)	He took his umbrella [lest it would rain].

In fact, preemptive clauses are not only present in the precautioning function: in (3a), the Apprehensive clause follows a preemptive clause (‘don't tell Maria, she may come along’). In this case, the preemptive clause and the apprehensive are syntactically independent clauses. In contrast to (3a), the Ese Ejja precautioning clause ‘lest the game escape’ in (4) is subordinated to the preemptive clause, ‘he cleans his rifle’. In cases where syntactic (in)dependence is obvious or easily proven, it is easy to distinguish between the apprehensive and the precautioning functions.

The reality may nevertheless be more complex, as precautioning markers are not always subordinators, and apprehensive markers are not always main clause tense or mood markers: precautioning clauses may have a pragmatic (rather than a syntactic) link to preemptive clauses. For instance, François (2003: 301-302) convincingly argues that in Mwotlap [motl1237] (Austronesian; Vanuatu), the Apprehensive has a strong pragmatic dependence, but no syntactic dependence. Verstraete (2006: 219) also discusses the role of intonation in creating clausal dependencies, typically under-reported in grammatical descriptions. Such cases of dependence not marked in the morphology make the diagnosis of the precautioning vs. apprehensive function difficult.

This is the reason why, in the present survey, we are using the “perspective dependence” criterion of Potts (2007) to distinguish between the apprehensive and the precautioning functions. If, as in (3a), the apprehensional clause (pokichana ‘she may come along’) can only have a modal value – it is the speaker who considers the event undesirable – then the apprehensional is analyzed as having the apprehensive function only. By contrast, if, as in (4), the undesirability can be the perspective of the syntactic subject of the preemptive clause (the hunter cleaning his rifle), then the apprehensional is analyzed as having the precautioning function. If a construction sometimes encodes the perspective of the speaker, and sometimes that of the syntactic subject of the associated preemptive clause, then it has the two functions. See Vuillermet et al (forthcoming), Vuillermet (forthcoming) for a more detailed discussion.

In many languages, the precautioning function is encoded by a negated purpose morpheme/construction, such as (in order) not to or so that not. Example (6) illustrates how Ese Ejja has in fact two constructions that encode the precautioning function: the negated purpose marker e-V-jji pojjiama, and a 'proper' precautioning marker e-V kwajejje.

(6)	Ese Ejja (Vuillermet 2018: 278)
	[...jikio	pia	como	etii	no ves	e’bio=nei	etii=jayojja
	DEM	other	like(Sp)	old	DISC(Sp)	first=ints	old=like

		e-po-jji=pojjiama],	(epoe) aekwá	[e-po	kwajejje].
		PURP-be-PURP=NEG	HESIT	PREC-be	PREC
	(the missionary said, “pupils, may you think about studying) ‘[so as not to be like these elders, like the ancestors, hum, lest you be like them].’ [narr]

Our survey only considers dedicated apprehensional morphology, and thus disregards negated purpose morphemes/constructions, lest every language count as having apprehensional morphology for the precautioning function.

Fear-complementation

The third function is fear-complementation, as in I fear lest he come. Here, the undesirability is encoded lexically by the complemented fear predicate I fear, and also grammatically by the subordinator lest. Unlike the general subordinator that, lest is dedicated, and only appears with fear predicates (*I was happy lest he come).

We are so far not aware of a language that would have a dedicated marker for the fear- complementation function. In English, lest occurs both in the precautioning and the fear-complementation function. In To'abaita [toab1237] (Austronesian; Solomon Island), the morpheme ada occurs in the fear-complementation function (7a) as well as in the apprehensive and precautioning functions (7b-c). In Mojeño Trinitario [trin1278] (Arawakan; Bolivia), the morpheme wichu occurs in the fear-complementation function (8a), as well as in the timitive function (8b).

(7)	To'abaita (Lichtenberk 1995: 295, 301, 296)
	a.	Fear compementation
		Nau	ku	ma’u	‘asia na’a	ada	to’a	na’i	ki
		I	I:FACT	be.afraid	very	APPR/TIM	people	this	PL

			keka	lae	mai	keka	thaungi	kulu.
			they:SEQ	go	hither	they:SEQ	kill	us(incl)
		‘I am scared the people may have come to kill us.’

	b.	Apprehensive function
		Ada	keka	fanga	sui	na’a.
		APPR1	they:SEQ	eat	COMPL	PERF
		‘[I fear] they may have finished eating.’

	c.	Precautioning function
		Ngali-a	kaleko	‘aa’ako	ada	fanu	‘eri	ka
		take-them	clothes	warm	APPR/PREC	place	that	it:SEQ

(8)	Mojeño Trinitario (Rose, p.c.)
	a.	Fear complementation
		mpiko	wichu	sakekoro
		n-piko	wichu	s-a-kekoro
		1SG-be_scared	APPR	3F-IRR-make_fun
		‘‎‎I fear that she will laugh.’

	b.	Apprehensive function
		wichu	jvénopa
		wichu	pi-venopo-a
		APPR	2SG-fall-IRR
		‘‎‎‘‎‎Watch out you might fall.’

Timitive

The fourth apprehensional function is the timitive. It differs from the first three in not having scope over a clause, but over a noun phrase. The morpheme marks an unwanted entity from which some undesirable situation is highly probable.

(9)	Manambu [mana1298] (Sepik; Papua New Guinea; Aikhenvald 2008: 153-154)
	təp-a:k!
	coconut-LK+DAT2
	‘for fear / beware of the coconut!’

(10)	Ese Ejja (Vuillermet 2018: 258)
	Iñawewa	kwaji~kwaji-ani	b’iya=yajjajo.
	dog	run~RDP-PRS	bee=TIM
	‘The dog is running out of fear of the bees.’

As mentioned earlier, some languages like Ese Ejja have three distinct morphemes to encode the apprehensive, precautioning and timitive function (see ex. 1, 3 vs. 4 vs. 6), while other languages like Marithiel [mari1424] (Western Daly; Australia) have a single morpheme for the three functions:

(11)	Marithiel (Green 1989: 80, 170, 58)
	a.	gu-n-ning-pirr-Ø-fang
		3S.REAL-go-1SGO.NSG.NIRR.S-leave-3PL-APPR
		‘(I’m afraid) they might leave me.’

	b.	tharr	guwa-mirrmirr	garri	mitik-a
		thing	3SG.SUBJ.REAL.stand-thunder	3SG.SUBJ.NFEM.rri	extinguish-PST

			watjan	ambi	gu-iwinj-sjang-Ø-fang
			dog	NEG	3NSG.SUBJ.REAL-3NSG-hear-3PL-APPR/PREC
		‘He turned off the “thundering thing” (the generator), lest they not hear the dogs.’

	c.	ambi-ya	guwa-wultharri-ya	gan	duknganan-fang
		NEG-PST	3SG.S.REAL.stand-return-PST	here	policeman-APPR/TIM
		‘He never returned out of fear of the policeman.’

Why?

The still scarce literature on apprehensional morphology mostly focuses on languages from Australia and Papunesia – see Dixon (2011: 292-293) who describes it as a pan-Australian feature and Lichtenberk (1995), the first extensive paper on the topic, illustrated with a few Papunesian languages. However, apprehensional morphology has been found in other macroareas: Dobrushina (2006) is a typological paper on apprehensional morphology that (among others) reports such morphology in Caucasian languages, and Vuillermet (2018) reports its presence in several Amazonian languages. The patchy global distribution of this feature makes it a good candidate for further investigation of its areality in the Americas.

How?

Because of the scarce literature on apprehensional morphology, the consequential variety of terms, and the many functions apprehensional morphology covers, we searched for apprehensional constructions in very different domains of the grammar: NP (especially prepositions), VP (especially mood and modality, imperative/hortative/jussive etc.), complex clauses (especially purpose and complementation), and negation. We made a list of key terms (like ‘‎‎fear', ‘afraid’, ‘beware’, ‘watch out’, ‘look out’ and even ‘snake’, etc), including the most frequent alternative names (like (ad)monitive (Meira 1999), adversative (Olawsky 2006), evitative (Harvey 2001), fear case marker (Dixon 1977), fearitive (Hudson 1983), negative purpose clauses (Vuillermet 2012), preventive (van der Voort 2004), timitive (mood) (Lichtenberk 2008), volitive of fear (de Reuse 1988), and warning (clitic, morpheme) (Derbyshire 1979) . We searched for these key terms in each source, and read the relevant sections looking for grammatical constructions which would match our definition of apprehensional morphology.

Part of the Australian and South American data had been collected in an earlier project.³ David Inman wrote computer code to automatically import the relevant data into ATLAs and they were reviewed to fit the definitions presented in this feature set description.

Features

Appr-01: Does the language have apprehensional morphology?

 { yes | no }

This feature asks whether the language has any dedicated construction that encodes any of the four apprehensional functions. The two key semantic concepts required in the construction are probability and undesirability. A language is coded as <yes> only if there are examples with neutral or (typically) positively valenced verbs or entities, or if the author explicitly describes the undesirability in the text or makes it clear in the translation (e.g. with the systematic presence of 'watch out'-like particles).

yes: Andoke [ando1256] (isolate; Colombia)

Landaburu (1979) calls the apprehensive construction 'énoncé de mise en garde' (warning sentence), which already announces the semantics of high probability and undesirability. The first two examples in Landaburu (1979: 89) are with intrinsically negatively valenced verbs (see 'burn' in (12a)), but the last two examples are with typically neutral verbs, namely ‘‎‎tell' (see (12b)) and ‘‎‎speak', and the event is still undesirable as the translation with attention (‘‎‎watch out') shows.

(12)	a.	ha-du-ə
		you-burn-APPR
		‘‎‎Watch out, you are burning yourself!’
		[Original translation: ‘‎‎attention tu te brûles!’]

	b.	ha-hʌʌ́-ə
		you-tell-APPR
		‘‎‎Watch out, you are telling it!’
		[Original translation: ‘‎‎attention tu racontes!’]

yes: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

See examples (1, 3, 4, 6) above.

yes: English [stan1293] (Indo-European)

The lest morpheme encodes both probability and undesirability: example (13) illustrates its use with the neutral verb 'come'.

(13)	Don't tell him anything lest he come too.

Appr-02: If <yes> to Appr-01, can (at least one of) the apprehensional construction(s) express the apprehensive function?

 { yes | no | NA }

The apprehensive function has to be a direct speech act, i.e. it needs to encode the speaker's perspective (see the definitions in the Apprehensive and Precautioning sections). We thus use the perspective dependence criterion to distinguish between the apprehensive and the precautioning function.

yes: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

See examples (1, 3) above.

yes: Hup [hupd1244] (Naduhup; Colombia, Brazil)

See example (2) above.

no: English [stan1293]

‘‎‎Lest' is a subordinate marker that requires a main clause. Only the very idiomatic lest we forget appears with no main clause.

no: Aguaruna [agua1253] (Chicham; Peru)

The example below shows that the apprehensive is only available in subordinate clauses: ha-ka-i-ha ‘‎‎I might die' is ungrammatical outside a subordinate (Overall 2007: 503).

(14)	[ʃiiha	yu-a-mɨ-i]	[ha-ka-i-ha	tu-sa-mɨ]
	[well	eat-IPFV-2SG-DECL]	[die-INTS-PREC-1SG	say-SBD-2:SS]
	‘you eat well so that you won’t die’ Lit. you are eating well, saying ‘I might die’

Appr-03: If <yes> to Appr-02, which grammatical persons are attested in the subject role (S or A) with the apprehensive?

 { NA | ; separated list of [1, 2, 3] }

Just like the imperative construction in a given language might be limited to a person, apprehensives vary in terms of their availability to grammatical persons.

2: Andoke [ando1256] (isolate; Colombia)

Landaburu (1979: 89; 2000: 282) explicitly mentions that the only available subject is a second person, as illustrated in (12) above.

1;2;3: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

In Ese Ejja, all grammatical persons are available with the Apprehensive -chana.

(15)	a.	1st person
		Miya	kwia-chana!
		2SG.ABS	hit-APPR
		‘Watch out, I might beat you!’

	b.	2nd person
		Shiwi-’io-chana=mi!
		slim-TEL-APPR=SG.2ABS
		‘Watch out, you’ll slim down!’

	c.	3rd person (repeated from (1))
		’Biya	’biya	’biya	’biya!	Kekwa-ka-chana	miya!
		bee	bee	bee	bee	pierce-3A-APPR	2SG.ABS
		‘Bee, bee, bee, bee! Watch out it might sting you!’

1;3: Matsés [mats1244] (Pano-Tacanan; Peru)

Matsés has several apprehensive markers, depending on the grammatical person or on the combination of grammatical persons (for transitive verbs). Only 1st and 3rd person subjects are attested (Fleck 2003: 439ff).

(16)	a.	1st person
		mibi	cues-mane
		2:ABS	hit-APPR:14
		‘‎‎I might accidentally hit you.’[e.g., if you stand behind me as I split firewood]

	b.	3rd person
		chononda
		cho-nunda
		come-APPR:3(>3)
		‘he might come’

The following two morphemes contrast in having different grammatical persons as objects. As Matsés is the only language to our knowledge sensitive to the grammatical person of the object, there is no feature to code this specificity.

(17)	a.	3>1 or 2
		nisi-n	pe-panondac
		snake-ERG	bite-APPR:3>1-2
		‘‎‎Shoot, now maybe a snake will bite me.’ [Said, e.g., as the speaker is going off to hunt on a rainy day.]

	b.	3>3
		nisi-n	pe-nunda
		snake-ERG	bite-APPR:3(>3)
		‘‎‎(Be careful), a snake might bite (e.g., our son).’ *‘ ... bite you/me.’

Appr-04: If the Apprehensive is attested with 1st person subject, then what pragmatics are attested?

 { NA | threat | warning | both }

The most typical context of a threat is with first person subjects acting consciously over a second person, as in ‘‎‎I may hit you (if you don't do X/make X happen)’ (see e.g. Epps 2008: 630ff). However, a threat reading can also be obtained with non-first person subjects, as in ‘‎‎Your child may die (if I don’t receive all the money),’ said by a kidnapper. See Vuillermet et al (forthcoming) for a detailed discussion of the pragmatics of threats.

The conditioning of this feature on first person subjects targets the most typical context where a threat reading is obtained, and the context most likely to be encountered in language descriptions.

warning: Matsés [mats1244] (Pano-Tacanan; Peru)

Matsés is the only language in our sample that encodes warning only. In a sentence like ‘‎‎I might[APPR] hit you’, the only possible context is that of a warning where the (potential) hitter would accidentally hit the hittee, not one where the hitter would no longer stand the hittee's presence or actions and threaten to hit him (Fleck 2003: 439-440 and 2017, p.c.).

(18)	mibi	cues-mane
	2:ABS	hit-APPR:1
	‘‎‎I might accidentally hit you.’ [e.g., if you stand behind me as I split firewood]

threat: Hup [hupd1244] (Naduhup; Colombia, Brazil)

Epps (2008: 631) observes that a “threat is [...] the default interpretation when the subject is in the first person” in her Hup corpus, as illustrated in (19).

(19)	ʔám-ǎn	ʔãh	yɔmɔ̌y	yɔ́k	tán-ã́h!
	2SG-OBJ	1SG	anus	stab.APPR	FUT.CNTR-DECL
	‘I’ll stab you in the anus.’

any: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

With 1st person, the apprehensive reading available is either a threat or a warning, mostly depending on the verb semantics: if with a [-control] verb, then the preferred reading is a warning (‘‎‎I might fall on you’), while a [+control] verb will rather trigger a threat (‘‎‎I will hit you!’, e.g. if you keep behaving like this) (Vuillermet 2018: 273-274).

Appr-05: If <yes> to Appr-02, is the Apprehensive restricted to certain verb types?

 { NA | no | direct control | other }

Two types of undesirable events can be distinguished: the ones which can be prevented altogether, and the ones in which only the consequences can be prevented. For instance, one cannot prevent events of raining or thundering, but one can still protect oneself from their consequences (getting wet). Some languages restrict their apprehensive construction to events which can be avoided altogether.

This restriction though is not easy to investigate, as the use of apprehensional morphology is often not described in detail, examples are scarce, and the context of the examples is not always given. Verbs with no agent, such as “rain”, or “thunder” are the best proxy for evaluating this potential restriction. An inability of such verbs to be used with apprehensive morphology would point to the apprehensive being restricted only to events that can be avoided.

However, even with a verb like 'rain', one could avoid the undesirable event in the right context: e.g. one cannot prevent raining, but one can still prevent raining over one's bike (by covering it with a tarp). In fact, this distinction may point to two subtypes of apprehensives:

• one where only the undesirable event may be avoided, and which one could paraphrase with ‘‎‎make sure X does not happen!’;
• one where either the undesirable event or its consequences can be avoided, and which could be paraphrased ‘‎‎watch out X might happen!’.

While both ‘‎‎watch out it might rain’ and ‘‎‎watch out, it might rain over your bike!’ would be acceptable, only ‘‎‎make sure it does not rain over your bike!’ is possible while ‘‎‎make sure it does not rain!’ is nonsensical. This feature is thus an imperfect proxy for an apprehensive construction expressing fully avoidable undesirable events vs. consequences only. Such a distinction has been discussed for the precautioning function by Lichtenberk (1995: 297-298) (see Appr-08 below).

no: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

Any verb event, including natural phenomena like earthquaking, can occur with the apprehensive -chana.

direct control: Mwotlap [motl1237] (Austronesian; Vanuatu)

The Mwotlap Apprehensive is ungrammatical with [-control] verbs like ‘rain’, ‘thunder’ or ‘happening of an earthquake’ (François 2005).

Appr-06: If <yes> to Appr-01, can (at least one of) the apprehensional construction(s) express the precautioning function?

 { NA | yes | no }

As discussed in the introduction, it is not always easy to distinguish the apprehensive from the precautioning function, since formal cues are not necessarily present. We thus considered the “perspective criterion” to distinguish the two functions: The precautioning function must allow the expression of the perspective of someone else than the speaker, namely that of the subject of the preemptive clause. In order to check for this parameter, it is important to discard preemptive clauses

• with 1st person, as the speaker and the subject of the sentence are merged;
• which are directive speech acts – e.g. do (not) / you should (not) do that (lest…), since they express the perspective of the speaker.

yes: Tariana [tari1256] (Arawakan; Brazil, Colombia)

The apprehensional in Tariana uses the reported speech construction for the precautioning function. It seems restricted to non-1st person subject.

(20)	yaɾumakasi	du-tale	du-a	[hamu-da	du-a-ka]
	clothing	3SG.F-cover	3SG.F-go	[hot-VIS.APPR	3SG.F-say-SUB]
	‘She covered (herself) with clothing, for fear it might be hot.’ [lit. saying “lest it might be hot”]

yes: Aguaruna [agua1253] (Chicham; Peru)

The apprehensional construction in Aguaruna only has the precautioning function (Overall 2007: 503). Note that a reported speech construction is necessary.

(21)	ʃiiha	yu-a-mɨ-i	[ha-ka-i-ha	tu-sa-mɨ]
	well	eat-IMPFV-2SG-DECL	[die-INTS-PREC-1SG	say-SBD-2:SS]
	‘you eat well so that you won’t die’ [lit. you are eating well, saying ‘I might die’]

no: Kotiria [guan1259] (Tucanoan; Brazil, Colombia)

The Kotiria apprehensional marker only has the apprehensive function: it is a modal that only appears in direct speech. Note that the free translation uses ‘‎‎or', which shows how pragmatically linked the apprehensive clause is to the preemptive clause ‘‎‎we have to escape' (Stenzel 2004: 422).

(22)	mʉ'ʉre	chʉri	nia
	~bʉ'ʉ+ré	chʉ́-ri	~di-a
	2SG+OBJ	eat-APPR	say-ASSERT.PERF
	‘‎‎(She urged: “Let's escape right now. They are evil beings.) Or they'll eat you,” she said.’ (Stenzel 2004: 409)

Appr-07: If <yes> to Appr-06, does the precautioning construction require reported speech?

 { NA | yes | no }

yes: Tariana [tari1256] (Arawakan; Brazil, Colombia)

See example (20) and discussion above. Tariana is coded <yes> because it is required for all non-1st persons.

yes: Aguaruna [agua1253] (Chicham; Peru)

See example (21) and discussion above.

no: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

Ese Ejja has a dedicated precautioning marker, which does not include or require a reported speech construction.

(23)	Owaya	shijja-’bame~’bame-ka-ani	[ijjiakajji	e-’ajjo	kwajejje].
	3ERG	clean-well~RDP-3A-PRS	game	PREC-escape	PREC
	‘He cleans his rifle lest the game escape.’

Appr-08: If <yes> to Appr-06, can the precautioning construction be used when one cannot avoid the undesirable event, but only its consequences?

 { NA | yes | no }

This feature is parallel to what is discussed in Appr-05.

Lichtenberk (1995: 297-298) shows how the precautioning function has two distinct subfunctions, depending on the possibility to avoid the undesirable event altogether (the negative purpose subfunction), or only its consequences (the ‘in case’ subfunction). These two subfunctions differ in the causal link between the clauses. This distinction is illustrated with the event of rain in (24) and (25), where lest expresses both subfunctions, and so that not is restricted to the negative purpose.

(24)	The undesirable event itself is avoided:
	a.	He always takes his umbrella lest he get wet.
	b.	He always takes his umbrella so that he does not get wet.

(25)	The undesirable event itself is not avoided, but its consequences:
	a.	He always takes his umbrella lest it rain.
	b.	#He always takes his umbrella so that it does not rain.5

In (24), there is a direct causal relationship between the preemptive and the precautioning clauses, which is not the case in (25).

yes: English [stan1293]

See examples (24-25).

no: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

The precautioning morpheme in Ese Ejja is rare in Vuillermet's corpus, and no 'in case' use is attested, which is unfortunately not a positive proof. This might be the case for many languages.

Appr-09: If <yes> to Appr-01, can (at least one of) the apprehensional construction(s) express fear-complementation?

 { NA | yes | no }

Note: We have only found cases where it is the precautioning marker that extends to this function.

yes: English

The morpheme lest can express fear complementation as in (26).

(26)	I am afraid lest they come.

no: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

Vuillermet (2012) does not document any fear-complementation.

Appr-10: If <yes> to Appr-01, can (at least one of) the apprehensional construction(s) express the timitive function?

 { NA | yes | no }

yes: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

Ese Ejja has a complex postposition dedicated to the timitive function (27).

(27)	Iñawewa	kwaji~kwaji-ani	b’iya=yajjajo.
	dog	run~RDP-PRS	bee=TIM
	‘The dog is running for fear of the bees.’

Appr-11: If <yes> to Appr-10, can the main verb of a timitive construction be [-control]?

 { NA | yes | no }

In some languages, the timitive expresses avoidance (of an undesirable entity, which might cause harm) rather than fear. With [+control] verbs like ‘‎‎hid' in ‘‎‎I hid out of fear of/to avoid the police', the two readings are possible. By contrast, uncontrolled verbs like ‘‎‎shudder' are ungrammatical: ‘‎‎I shuddered out of fear of the wolf' is fine, while ‘‎‎#I shuddered to avoid the wolf' is semantically unacceptable as shuddering cannot prevent someone being harmed by a wolf.

yes: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

(28)	E-sho’i	’oke-’io-naje	pejjepejje=yajjajo.
	NPF-child	go_down-TEL-PST	owl=TIM
	‘The child fell (on the floor) out of fear of the owl.’

no: Marithiel [mari1424] (Western Daly, Australia)

See example (11c).

Appr-12: How many apprehensional morphemes are there and how are they distributed over the four functions of the apprehensional domain?

 { NA | ; separated list of [ A & P & T & F ] }

A = Apprehensive
P = Precautioning
T = Timitive
F = Fear-complementation
& = functions covered by the same morpheme
; = separates different morphemes

A;P;T: Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru)

As has been exemplified above, Ese Ejja has 3 distinct markers to express 3 apprehensional functions: the apprehensive (see examples 1, 3), the precautioning (4), and the timitive (6).

P&F: English [stan1293] (Indo-European)

Lest has the precautioning function (see examples 24-25), as well as the fear-complentation function (see example 26).

A&P&F;T: Trinitario [trin1278] (Arawakan; Bolivia)

Trinitario has 2 apprehensional morphemes, which cover all 4 functions. Examples 8(a-b) above illustrated wichu in the apprehensive and the fear-complementation functions. Example (29) shows wichu in the precautioning function, and example (30) mpisra in the timitive function.

(29)	Precautioning (Rose 2014)
	ene	tepajipo	wichu	tayuka,	ene
	ene	ty-epaji=po	wichu	ta-yu-ko-a,	ene
	and	3-lie.down=PFV	APPR	3NH-bite-ACT-IRR	TAG
	‎‎‘‎‎and he lies down onto the floor lest they bite him, no?’ [Original translation: y se tendió (se echó) para que no le piquen, no?’]

(30)	Timitive (Rose 2018)
	nuchkopo	mpisra	ñi	nima
	n-uchu-ko=po	n-piko-ira	ñi	nima
	1SG-exit-ACT=PFV	1SG-be_scared_of-NMZ	ART.M	1SG-husband
	‘I went out for fear of my husband (drunk, for ex.)’ ‎[original translation: ‘salí por temor de mi esposo (borracho)’]

A&P&F: Harakmbut [amar1274] (isolate; Bolivia)

Harakmbut has 1 apprehensional morpheme, which covers 3 functions. A reported speech construction is required for the fear-complementation and sometimes used for the precautioning function.

(31)	a.	Apprehensive (Tripp 1995: 222)
		O'-kore-apey.
		1SG.APPR-return-APPR
		‘I could come back (if there are troubles).’
		[original translation: ‘Yo podría regresar (si hay problemas).’]

	b.	Precautioning (Van Linden 2019, p.c.)
		mboerek	ö-ïrïng-me	apetpet-a	mbe-arak-apey-a-po
		man	3SG.IND-hide-REC.PST	jaguar-NOM	3>1-kill-APPR-INDIR-DEP
		‘The man hid lest the jaguar kill him.’

	c.	Fear-complementation (Van Linden 2019, p.c.)
		Ken	ö rï-ÿö	öʔ-(ö)ö-po	öʔ-(ö)ö-po
		3SG	Madre.de.Dios-LOC	3SG.IND-bathe-DEP	3SG.IND-bathe-DEP

			Ken-ta	änï	ken	i-tö-mëpuk-me-y
			DIST-LOC	FILLER	DIST	1SG-SOC-fear-REC.PST-1.IND

			ndo.	Änï	mbe-ta-kot-apey	i-nöpöë-po	ndo.
			1SG	FILLER	e-APPL-fall-APPR	1SG-think-DEP	1SG
		‘then she [i.e. An] swam in the Madre de Dios River, she swam, and I was afraid with her, eh, hoping that she would not fall/drown’ [Lit: ‘and I thought: “she mustn't fall/drown” ’ (with ta- expressing endearment) (spontaneous speech; anecdote)]

Derived Features

Appr-12a: How many apprehensional morphemes are present?

 { NA | _number_ }

NA	if Appr-01 is <no>
X+1	where X is the number of <;> symbols in Appr-12

Appr-12b: Which functions are covered by the language’s apprehensional morphemes, regardless of whether they are split among multiple morphemes?

 { ; separated list of [ a, p, t, f ] | NA }

NA	if Appr-01 is <no>
...	a list of all unique functions present in Appr-12

Results

Our survey shows that apprehensional morphology (Appr-01) is certainly not a rara, as it is present in at least 24 % of languages in our sample (the presence of apprehensional morphology is unknown for 6 languages). However, its geographical distribution across macroareas is very uneven. It is present in almost half of the languages in Australia (44%), over a third in Papunesia (at least 36%, 1 language unknown), and almost a third in South America (at least 32%, 2 unknown). By contrast, it is present in less than 1 in 5 in Eurasia (at least 17%, 1 unknown) and North America (at least 15%, 2 unknown), and almost absent in Africa (4%, 1 unknown). Of course, the uneven density of our sample (over 200 languages in the Americas vs about 100 in the rest of the world) calls for confirmation in the less densely sampled macroareas. For instance, Gban [gagu1242] (Mande; Côte d'Ivoire) does not belong to our sample, but Fedotov (to appear) is a detailed report of its apprehensional morphology. It is also interesting to note that South America has twice as much apprehensional morphology as North America. Note however that the geographic distribution is also uneven within these two macroareas: in North America, apprehensional morphology is more common on the Northwest Coast and in Mesoamerica, and in South America, it is more common in Western Amazonia and the Guaporé Mamoré region.

Four features examine the specificities of each of the four apprehensional functions, defined according to our criteria. Apprehensive (Appr-02) is the most frequent function in our worldwide sample, occurring in at least 19% (61 languages, 8 unknown). It is also the most frequent in each macroarea, except for Papunesia (16%) where the Precautioning function is the most frequent (20%). Apprehensive is also the only function attested in our African sample (but see again Gban; Fedotov to appear](Source#cldf:fedotovtoappearapprehensional)). Ten Australian languages have this function (40%), at least 26 in South America (25%, 3 unknown), at least 5 in Eurasia (17%), and at least 4 in Papunesia (16%).

Precautioning (Appr-06) is the second most frequent function in our sample, present in at least 11% of languages (37 verified, 7 unknown). It is present in at least 20% of our Australian and Papunesian samples, and in at least 17% of the South American sample (18 languages, 3 unknown), which again contrasts quite sharply with the 7% of the North American sample (8 languages, 1 unknown).

The much rarer fear-complementation (Appr-09) and timitive (Appr-10) functions occur in respectively 5% and 2% of our sample. Only one language (Kuot [kuot1243]; isolate, Papua New Guinea; Lindström 2002: 11) has an apprehensional morpheme dedicated to the fear-complementation function, a phenomenon which we did not expect to find, since all apprehensional morphemes expressing the fear-complementation function that we knew of also express the precautioning function. The fear-complementation function most frequently occurs in Papunesia (12%), followed by Australia and South America (8%), and Eurasia (7%). It is attested in only one language in North America, and absent in Africa.

The timitive function (Appr-10) is only attested in South America (6%) and Australia (4%). We cannot exclude the possibility that this function has been underreported, since it is a lesser known typological concept, except in the Australianist tradition where cases of “fear-case marker” are in fact overreported (following our definitional criteria) and often express general cause.

It is interesting to note that only seven languages have several apprehensional morphemes (Appr-12a), and they are all in South America. About half of the languages have an apprehensional morpheme dedicated to one apprehensional function only, and the other half have at least one polyfunctional morpheme (Appr-12b).

Contributions

Conceptualization: Marine Vuillermet

Data collection: Marine Vuillermet, David Timothy Perrot

Supervision of data collection: Marine Vuillermet

Computer code: David Inman

References

• Derbyshire, Desmond C. 1979. Hixkaryana syntax. University of London dissertation. (303pp.)

• Reuse, Willem Joseph de. 1988. Siberian Yupik Eskimo: the language and its contacts with Chukchi. Austin: University of Texas dissertation.

• Dixon, R. M. W. 1977. A grammar of Yidiɲ. (Cambridge Studies in Linguistics.) Cambridge: Cambridge University Press.

• Dixon, R. M. W. 2011. The Languages of Australia. Cambridge University Press.

• Dobrushina, Nina. 2006. Grammatičeskie formy i konstrukcii so značeniem opasenija i predostereženija [Grammatical forms and constructions expressing warning and prevention]. Voprosy jazykoznanija 2. 28–67.

• Epps, Patience. 2008. A grammar of Hup. (Mouton Grammar Library, 43.) Berlin, New York: Mouton de Gruyter. xxiii+983pp.

• Evans, Nicholas D. 1995. A Grammar of Kayardild: With Historical-Comparative Notes on Tangkic. (Mouton Grammar Library, 15.) Berlin, New York: Mouton de Gruyter. xxiv+837pp.

• Fedotov, Maksim. toappear. The apprehensional construction in Gban. In Martina Faller and Marine Vuillermet and Eva Schultze-Berndt (eds.), Apprehensional constructions in a cross-linguistic perspective. Berlin: Language Science Press.

• Fleck, David W. 2003. A Grammar of Matses. Houston: Rice University dissertation. (xxi+1257pp.)

• Fleck, David. 2017. personal communication.

• François, Alex. 2003. La sémantique du prédicat en mwotlap (Vanuatu).. (Linguistique de la Société de Linguistique de Paris.) Leuven-Paris: Peeters.

• François, Alexandre. 2005. personal communication.

• Harvey, Mark. 2001. A Grammar of Limilngan: A Language of the Mary River region Northern Territory Australia. (Pacific Linguistics, 516.) Canberra: Research School of Pacific and Asian Studies, Australian National University. xii+209pp.

• Hudson, Joyce. 1983. Grammatical and Semantic Aspects of Fitzroy Valley Kriol. Darwin: Summer Institute of Linguistics.

• Landaburu, Jon. 1979. La Langue des Andoke (Grammaire Colombienne). (Langues et Civilisations a Tradition Orale, 36.) Paris: SELAF. 350pp.

• Landaburu, John. 2000. La lengua Andoque. In González de Pérez, María Stella and Rodríguez de Montes, María Luisa (eds.), Lenguas indígenas de Colombia: una visión descriptiva, 275–288. Instituto Caro y Cuervo, Santafé de Bogotá.

• Lichtenberk, Frantisek. 1995. Apprehensional Epistemics. In Bybee, Joan and Fleischman, Suzanne (eds.), Modality in Grammar and Discourse, 293–327. Amsterdam - Philadelphia: John Benjamins.

• Lichtenberk, Frantisek. 2008. A grammar of Toqabaqita. (Mouton Grammar Library, 42.) Berlin - New York: De Gruyter Mouton.

• Lindström, Eva. 2002. Topics in the Grammar of Kuot. Stockholm University dissertation. (xxii+242pp.)

• Meira, Sérgio. 1999. A Grammar of Tiriyó. Houston: Rice University dissertation. (xviii+708pp.)

• Olawsky, Knut. 2006. A Grammar of Urarina. (Mouton Grammar Library, 37.) Berlin, New York: Mouton de Gruyter. xviii+943pp.

• Overall, Simon. 2007. A Grammar of Aguaruna. Bundoora, Victoria: LaTrobe University dissertation. (xxii+588pp.)

• Potts, Christopher. 2007. The expressive dimension. Theoretical Linguistics 33(2).

• Stenzel, Kristine. 2004. A Reference Grammar of Wanano. University of Colorado at Boulder dissertation. (xxiii+429pp.)

• Verstraete, Jean-Christophe. 2006. The role of mood marking in complex sentences. A case study of Australian languages. \textit{WORD} 57(2-3). 195–236.

• Voort, Hein van der. 2004. A Grammar of Kwaza. (Mouton Grammar Library, 29.) Berlin, New York: Mouton de Gruyter. xxxviii+1026pp.

• Vuillermet, Marine. 2012. A Grammar of Ese Ejja, a Takanan language of the Bolivian Amazon. Université Lumière Lyon 2 dissertation. (xvi+736pp.)

• Vuillermet, Marine. 2018. Grammatical fear morphemes in Ese ejja: making the case for a typological domain?. Studies in Language 42/1. 256–293.

• Vuillermet, Marine, Eva Schultze-Berndt & Martina Faller. n.d. Introduction: Apprehensional constructions in a cross-linguistic perspective. In Faller, Martina and Schultze-Berndt, Eva and Vuillermet, Marine (eds.), Apprehensional constructions in a cross-linguistic perspective.

• Vuillermet, Marine. n.d. A fieldwork questionnaire on apprehensionals. In Faller, Martina and Schultze-Berndt, Eva and Vuillermet, Marine (eds.), Apprehensional constructions in a cross-linguistic perspective.