On the expression of mistaken beliefs in Australian languages

3.1 Marked by a morpheme or collocation of morphemes

A fair number of languages possess a morpheme that either by itself or in a particular construction marks a proposition as a mistaken belief. This morpheme is usually a clitic or free particle (henceforth particle); rarely it is a suffix, and occasionally it is a collocation of morphemes. In the following subsections we discuss these possibilities in order, assuming the categorisation of the source; few grammars, however, mention or explicitly discuss the criteria for identification of the morpheme types. A language may employ more than one of these modes of marking mistaken beliefs; unfortunately, information does not permit postulation of meaning differences among the modes of expression.

3.1.1 A clitic marking ‘mistakenly believe’

This is attested in 11–18 languages (about a fifth to a third of the languages attesting some mode of expression) belonging to three or four genetic families. The languages are widely scattered over the continent, as shown in Figure 2. Examples (6) and (7) are illustrative.

Figure 2:

Location of languages with a possible enclitic marking mistaken belief.

Enindhilyakwa (Gunwinjguan)

(6)

nuw-angkarree-ka	erribaja-bu … wa,	naree-ka	amiyerra- bu
N-run.PST-EMP	away-EMP … XTD	NEG-EMP	further-MB
‘It ran further away, I thought it was going to keep going on, but no.’ (van Egmond 2013: 373)

Kaytetye (Pama-Nyungan)

(7)

re	areynengew- athene	wenherre
3SG	euro-MB	shot
‘He thought it was a euro that he shot (but it wasn’t).’ (Turpin 2000: 114)

Examples in most of the languages reveal that the morpheme is a clitic that attaches to units of various types. Few descriptions, however, say anything about which unit actually hosts the enclitic. In the above examples it seems to be the focus of the mistaken belief. In (6) the distance travelled is the focus of the mistaken belief, and the clitic is attached to the adverbial amiyerra ‘further’; in (7) it is the category of the thing that was shot that is at issue, not the shooting event itself which is evidently presumed to have occurred. However, a focus for the MB is sometimes lacking. For example, in (8) the MB does not focus contrastively on the speaker, but rather concerns the entire issue of their having gone for water. In this instance, the MB clitic appears in Wackernagel’s position in the first clause.

Nyulnyul (Nyulnyulan)

(8)

ŋai- elbe	ŋan-djed	wol-oŋ,	are
ngay- ilbi	nga-ny-jid	wul-ung	arri
1MIN.CRD-MB	1MIN.NOM-PST-go	water-ALL	not
ŋale-djedan
nga-li-jid-an
1MIN.NOM-IRR-go-IMP
‘You mistakenly thought I went for water, but I did not.’ (Nekes and Worms 1953: 519–520)

In some languages the clitic seems to be a dedicated marker of mistaken belief, at least in certain circumstances. This is the case, for instance, in Nyulnyul, Warrwa, Kaytetye, Mparntwe Arrernte, Yankunytjatjara (Pama-Nyungan) and Ngaanyatjarra (Pama-Nyungan), where the clitic seems to be used in no other sense and the mistaken belief sense appears not to be defeasible. Example (7) could not be followed by ‘and it was’, and nor could the final clause of (8) be replaced by the positive nga-ny-jid (1MIN.NOM-PST-go) ‘I went’.

This is likely also the case in various other languages where it is accorded a different gloss, often modal in nature. For example, Kalkatungu (Pama-Nyungan) has what Blake (1979: 48–49) calls a prolocative suffix of uncertain meaning, though the examples he gives suggest it is an MB clitic. In Kugu Nganhcara (Pama-Nyungan) there is a clitic glossed ‘supposedly’, which suggests the MB interpretation, though no clear examples are given in Smith and Johnson (2000: 441). For Karajarri (Pama-Nyungan), Nekes and Worms (1953: 603, 2006: 292) identify a bound form that they refer to as a conjunction ‘that’; examples given are consistent with its having MB as its only meaning, though it is attested only in complement constructions (see Section 3.2.1 below). And Warrongo shows what Tsunoda (2011: 682) refers to as a “counterfactual” clitic that usually has the MB sense; the exceptions warrant further investigation.

In four Pama-Nyungan languages – Arabana-Wangkangurru, Pitta-Pitta,^[2] Yalarnnga and Yandruwandha – the clitic is either the semblative (i.e. similative, ‘like a’) marker (Breen 1984), or is homophonous with such a marker. (We return to this issue in Section 4.4 below.) In Pitta-Pitta the semblative clitic takes the form -widi, as in yurathiyaka-widi (shameless-SEM) ‘as if (they had) no shame, shamelessly’ (Breen 1984: 3). Examples (9) and (10) illustrate use of a clitic with the same phonological form in the expression of mistaken beliefs. (There are examples of -widi with other modal senses, though information is too scant for anything to be said with certainty.)

Pitta-Pitta

(9)

karnta-ya- widi	inpa
go-now-MB	you
‘I thought you were going. (But you aren’t.)’ (Breen 1984: 5)

Pitta-Pitta

(10)

kimpa-layi	itji-ka- widi
alive-EMP	die-did-MB
‘He’s still alive! We thought he’d died.’ (Breen 1984: 6)

The MB construction marked by the clitic does not specify the identity of the holder of the mistaken belief. In (6) and (9) it is the speaker, in (7) it is the third person, the shooter; in (8) the addressee; and in (10) a first person non-singular group. Context must be taken into account to determine (or guess) who the mistaken believer is. However, few descriptions discuss the issue, and for many languages it remains unclear whether or not there are genuine restrictions on who the mistaken believer can be. At least in some languages it is possible to specify the holder of the mistaken belief in a complement construction – see Section 3.2.1.

3.1.2 A (free) particle marking ‘mistakenly believe’

This is the best attested expression type, showing up in 23–27 languages (around a half of the languages attesting an expression), as shown in the map in Figure 3. Belonging to 6–7 families, the languages are again quite widely distributed geographically, though they tend to be more concentrated in the northern half of the continent.

Figure 3:

Location of languages showing a particle expressing mistaken belief.

Examples (11) and (12) illustrate this mode of expression.

Rembarrnga (Arnhem)

(11)

guyala	yarra-many	gurrah	guyala	jeny	yarr-ma-ngœ-gan
MB	1AUG-went	that:way	MB	fish	1AUG>3-get-INF-DAT
‘We went over there, we thought we were going to catch some fish (but we didn’t).’ (Saulwick 2003: 52)

Warrongo

(12)

ngalnga	gogo	goman
MB	language	different
‘I thought it was a different language (but in fact it was not).’ (Tsunoda 2011: 677)

The position of the particle is rarely discussed in grammars, though as in the above examples it commonly occurs initially in the clause representing the mistaken belief.^[3] This is the case also in Gooniyandi, where thaarri ‘MB’ normally precedes the unit it holds in its scope. However, the unit immediately following the particle can be the focus of the mistaken belief, as in (13), where there is contrastive focus is on the speaker and it is presupposed that someone hit someone.

Gooniyandi

(13)

thaarri	nganyi-ngga	gard-looni	ngoorroo-ngga
MB	I-ERG	hit-3SG.ACC.1SG.NOM.CL	that-ERG
yaanya-ngga	gard-bini
other-ERG	hit-3SG.ACC.3SG.NOM.CL
‘I thought I’d hit him, but really it was that other man who hit him.’ (McGregor 1990: 425)

Rarely MB particles occur in some other position in the clause. In (14) it occurs between the initial NP (the Undergoer) and the verb; in (15) it occurs clause finally.

Warray (Gunwinjguan)

(14)

wek	kunj	kat-tirim-pu-n	pin	amala
firewood	MB	1SG.NOM.POT-light-AUX-IRR	but	not
a-kutjkutj	amala	kan-nu
CL-wet	not	POT-burn
‘I mistakenly thought I could light the firewood, but no, it was wet and wouldn’t burn.’ (Harvey 1986: 269)[4]

Kunwinjku (Gunwinjguan)

(15)

dja	ngokko	ø-bal-h-ka-ni	na-wu	mayh
and	already	3PST-away-IMM-carry-IMP	MA-DEM	animal

berrewoneng-ni	yimankek -ni
them.two.OBL-IMP	MB-IMP
‘But already Echidna was bringing back food which she thought was for both of them.’ (Evans 2003: 121)

Possibly in final position the particle serves as a type of afterthought. This interpretation is consistent with the second free translation given for the Warlpiri (Pama-Nyungan) example in (16).

Warlpiri

(16)

yuntardi	nganta
beautiful	MB
‘She thinks she’s beautiful (but).’ Or ‘She’s beautiful, supposedly.’ (Laughren 1982: 141; both free translations Laughren’s)

Examination of the examples available reveals that the holder of the mistaken belief is most often the speaker (as in examples (12)–(14)). However, it can also be a first person group (example (11)) or someone else (examples (15) and (16)). I am unaware of any clear examples where it is the hearer, though this is almost certainly an accidental gap (examples are available for the clitic). Some descriptions state explicitly that context has to be taken into account to determine who the holder of the mistaken belief is, e.g. Evans (1995: 380) on Kayardild (Tangkic) and Laughren (1982: 149) on Warlpiri. Additionally, in some languages the holder of the mistaken belief may be specified by use of a complement construction – we return to this in Section 3.2.2.

In some languages the particle marking a belief as mistaken is dedicated to this use, which may be assumed coded by the particle. This seems to be the situation, for instance, in Gooniyandi, Mangarrayi (Arnhem), Jaru (Pama-Nyungan), Pitjantjatjara (Pama-Nyungan) and Djinang (Pama-Nyungan), where no other sense is attested for the relevant particle. Nor, to the best of my knowledge, is the mistaken sense defeasible in these languages.

Sometimes, however, the particle has other documented modal meanings and/or uses, typically in the domain of the counterfactual (which label is employed in some descriptions instead of mistaken belief). For instance, Gurr-Goni (Maningrida) has a modal particle mundjarra with two main senses in complementary distribution: (i) unrealised intention or expectation when the verb is in the irrealis; (ii) mistaken belief when the verb is in realis, or the clause is verbless (yielding the misclassification sense) (Green 1995: 337). Similarly, Na-Kara has a particle ngamarramung with meanings unfulfilled potential in the irrealis, and mistaken belief when preceding an NP (Eather 1990: 342); there is no mention of what happens with realis verbs. For Bininj Gun-Wok (Gunwinjguan) Evans (2003: 610–611) identifies a range of counterfactual meanings, which he says are associated with different constructions. Kayardild appears to be similar (Evans 1995: 378–379). In circumstances such as these, where the senses are in complementary distribution, there is presumably a single multifunctional modal particle and different constructions.

The different modal senses are not necessarily in complementary distribution in languages such as Ndjébbana (McKay 2000: 300), Kunbarlang (Gunwinjguan) (Kapitonov 2021; Kapitonov and Gentens 2018), and the Pama-Nyungan languages Marra (Heath 1981: 305), Kuku Yalanji (Hershberger and Hershberger 1982; Patz 2002: 113, 198, 220) and Martuthunira (Dench 1994: 168). It is not clear whether or not these are different senses of a single particle, or whether they are associated with different grammatical constructions.

In a few languages it seems that there may be, in addition to the MB particle, a homophonous particle with a modal meaning. Kuku Yalanji, according to Hershberger and Hershberger (1982), has two homophonous particles balu, one an MB particle (with various meanings, as per the previous paragraph), the other a negative particle ‘not want’. In Tsunoda (2011: 677) identifies two distinct but homophonous particles ngalnga, one an MB particle, the other a prohibitive; interestingly, the MB particle almost always hosts the MB clitic, as in (4) above.

Like MB clitics, MB particles also encroach on the semblative domain, as discussed in more detail in Section 4.4 below. In a few languages, mistaken belief is marked by particles dubbed semblative markers. This is the case in Kuuk Thaayorre (Pama-Nyungan): examples (17) and (18) illustrate, respectively, the use of the free particle kar ‘like’ (as Gaby 2017 analyses it) in expressions of similarity and mistaken belief. Similarly, Kunbarlang has the particle yimarne, which Kapitonov and Gentens (2018) consistently gloss as a semblative, and which is also used in expressions of mistaken belief.

Kuuk Thaayorre

(17)

nhul	koo.miing	kar	pam	nhangnip	nhangn
3SG.NOM	face	like	man	father	3SG.GEN
‘He looks like his father.’ (literally, ‘he is like his father in the face’) (Gaby 2017: 328)

Kuuk Thaayorre

(18)

ngul	nganip	yik-r	pul	nganam-u	kar	parr_r
then	father	say-P.PFV	3DU	mother-ERG	MB	child
nhurr	thaa-munthi-rr
2PL	mouth-sink-P.PFV
‘Then dad and mum said “(we thought) you kids had drowned.”’ (Gaby 2017: 449)

3.2 Realised by a complement construction

The second mode of expression of mistaken beliefs involves a complement construction. In some languages a marker (such as described in Section 3.1) may be optionally accompanied by a complement clause; in a few languages a complement clause is obligatory. McGregor (2021: 9–14) discusses the expression of mistaken beliefs in complement constructions; this section is an update.

The following types of matrix clause of thought are attested in mistaken belief complement constructions:

1. with the vague (or polysemous) verb ‘say, do, think’;

2. with the vague (or polysemous) verb ‘hear, think’;

3. with a verb meaning ‘think’;

4. with a verb meaning ‘mistakenly think’;

5. with a nominal meaning ‘mistaken belief’.

Mistaken beliefs are opinions about things, and as such are not (normally) addressed to anyone; the matrix clause never contains an NP specifying an addressee. Correspondingly, a second person category in the complement clause can only be interpreted as a second person in the utterance itself, as illustrated by example (22): the person who might have gone outside must be the addressee of the entire utterance. (See also examples (23) and (24).) The second person category is never according to the person origo of the mistaken thought.

In many languages, however, the matrix clause can include an NP specifying the entity that the mistaken belief is about. This NP is often marked by the dative case, and in some languages is simultaneously cross-referenced by an oblique bound pronominal. The second person bound pronominal enclitic on the matrix verb in (22) cross-references such a participant, though the NP representing it has been ellipsed.

This section begins with a discussion of clitics, particles and suffixes in complement constructions (Sections 3.2.1–3.2.3); collocations of free particles are not attested in complement constructions. It then turns to a range of mistaken belief constructions obligatorily expressed by complement constructions (Sections 3.2.4–3.2.6).

3.3 Mistaken belief as a pragmatic implicature

A few languages express mistaken belief exclusively (in the available examples) in ordinary thought complements using the usual ‘thought’ verb. These include the four Pama-Nyungan languages Walmajarri, Yindjibarndi, Yingkarta and Wangkumara. Only a few examples are provided in the sources, and all of them are previous thoughts of the speaker, as in (43) and (44). In (43) the thought is specified as mistaken by the following clause. In (44) that it is mistaken is clear from the context: it is the final utterance in an exchange concerning the ownership of the meat, where the other party has successfully made a claim on it.

Wangkumara

I would argue that the mistaken belief sense emerges as a pragmatic implicature (McGregor 2021). This implicature can be derived in various ways, none of which are language particular. If we compare a plain clause with that same clause as complement in a thought construction, the latter is a more marked expression, so invokes a more marked interpretation by Levinson’s M principle (Levinson 2000). Thus, if we take ‘it was water’ and contrast it with ‘I thought it was water’ (as in (43)), the latter expresses a more marked meaning than the former. One possible construal is that the proposition was believed (or thought) as stated, but is in fact not true: had it been true, why qualify it as a thought?

Another take on the same example is that ‘it was water’ is a stronger statement on a Horn scale than ‘I thought it was water’. The weaker statement implicates, by Levinson’s Q principle, that the view is no longer held by the speaker. Yet another is that ‘I think it is water’ is stronger than ‘I thought it was water’, and the weaker statement implicates that the view is no longer held.

Example (44) further illustrates the operation of implicatures. Saying that the meat was the speaker’s would have been inappropriate, and contrary to the previous sentence in which the speaker has acknowledged that the meat is the addressee’s and that they will return it to them. By employing the more marked thought complement the speaker acknowledges that they now realise that the meat is not theirs. Framing the proposition as a thought accounts for speaker’s having taken the meat in the first place. They don’t just say ‘it is not mine’, which, though true, explains nothing.

The same implicature can arise in languages that have a means of specifying a belief as mistaken, including English and various other languages (McGregor 2022b). On the Australian continent this is attested in Martuthunira (Dench 1994: 125), which has an MB particle, and Enindhilyakwa, with an MB enclitic, as shown by (45).

Precisely what different nuances of meaning are conveyed by the choice of the dedicated mode of expression and the implicature are uncertain.

Nyulnyul is interesting in this regard. Recall that it has an MB enclitic according to Nekes and Worms (1953, 2006. By the time of my fieldwork in the 1980s, however, there was no memory of the enclitic. Instead, a construction was consistently used with the bound nominal -mungk ‘belief, knowledge’, which takes the same person prefixes as some inalienable body part terms (see McGregor 1995). The nominal status of this word is supported by the fact that it can be used as an attribute in an attributing clause, as in (46); this usage is impossible for a particle. (The morphological potential of the word marks it as a nominal, and distinguishes it from an inflecting verb or preverb.) Such an attributing clause may serve as the matrix clause in a complement construction, as in (47).

Nyulnyul

In the speech of the few remaining speakers of the late twentieth century, the only mode of expressing mistaken beliefs was a complement construction with a verbless attributing matrix clause with this inflecting nominal, as in (48).^[9]

The mistaken belief implicature arises in such examples in the same way as suggested above for verbal matrix clauses with ‘think, believe’. Representing a proposition as a thought implicates a meaning that is more marked than if it had been represented by a simple clause. This may (as per the discussion above) implicate that the proposition is in fact false. This seems to have become a generalised conversational implicature of the complement construction, which by the late twentieth century became entrenched, and a new mistaken belief construction type was emergent. This was not restricted to a first-person thinker and the past tense, as shown by (48). The similar Yidiny construction (Section 3.2.6) appears to have gone a step further in entrenchment, and the nominal has taken on the semantics of mistaken belief: it now codes (in the construction) mistaken belief, which may formerly have been an implicature, as is the case in Nyulnyul.

4.1 The identity of the mistaken believer

Complement constructions expressing mistaken beliefs typically provide specification of the individual who held the belief specified as mistaken: it is typically the “subject” of the matrix clause.

By contrast, the purely morphological expressions (discussed in Section 3.1) do not admit specification of a believer. We have seen that examples from various languages show a range of possibilities: the speaker of the clause, the addressee, a third person, a group including the speaker (‘we’). Examination of the database reveals that the speaker is by far the most common mistaken believer across all languages of the sample, to the extent this can be inferred from the free translations provided in the sources. That is to say, mistaken belief expressions are most commonly used as corrections to the speaker’s own beliefs.

Some descriptions (as mentioned already) explicitly state that the interpretation is contextually engendered; most are silent on this matter. For the bulk of Australian languages, it is not known whether there are any restrictions on who the mistaken thinker can be.

What is encoded by the non-complement expressions is no more than that in some situation someone believed a proposition which turns out to be false (in the speaker’s opinion). Thus, more literal and decontextualised English translations of the examples in Section 3.1 would involve matrix expressions like ‘someone mistakenly believed’, with no specification of the identity of the believer (let alone when or where they held the belief).

In fact, even this component of meaning is too specific. In some instances there is no specific individual in mind, and the mistaken belief is attributed to a generic person in a generic situation: the mistaken belief is attributed, that is, to the average person, to any (normal) member of the society. This interpretation is attested in half a dozen or so languages, including: Gooniyandi ((13) admits the interpretation ‘it was/is thought I’d hit him, but really it was that other man who hit him’); Jaru (as indicated by Tsunoda’s first free translation of example (49), which suggests that in the context anyone might have thought he was dead); Warlpiri (the second free translation of (16) above suggests that the woman is widely believed to be beautiful); Warrongo (Tsunoda 2011: 683); Kayardild (Evans 1995: 381); and Bininj Gun-wok (Evans 2003: 610). This meaning comes close to a semblative one (see further Section 4.4 below).

4.2 Speaker’s uptake

Expressions of mistaken belief are frequently – though not always – followed by a clause stating what (in the speaker’s opinion) the reality actually is or was. This is illustrated in (50), where the second clause specifies what actually made the noise. Significantly, the following clause never asserts the proposition: the MB meaning is not defeasible, and is a part of the encoded meaning of the construction.

In instances such as this, the speaker’s uptake conveys significant and unpredictable information that adds to what is known about the referent world; the same holds also for examples (13), (23) and (30). In the case of mistaken categorisations, the uptake often specifies the real category of the entity, as in examples (3), (27) and (43). In these instances the uptake serves as a replacement of the mistaken belief. In some cases, however, it conveys little in the way of news, merely restating the obvious (examples (8), (28) and (48)).

In some languages a specific morpheme may occur at the beginning of the uptake clause, as shown by (51); occasionally, as in Djambarrpuyngu, it can also occur at the end of that clause (Wilkinson 1991: 686). This morpheme expresses a contrastive meaning which is variously translated ‘but’, ‘but really’, ‘(but) in fact’. The classification of this morpheme differs from language to language. In some sources it is referred to as a conjunction or connective (e.g. Eather 1990; Richards and Hudson 1990: 154). In Kunbarlang the English conjunction but is regularly employed (Kapitonov 2021: 52; see example (31)). In other languages it seems to be a type of modal particle – e.g. Jaru (Tsunoda 1981: 206), Gooniyandi (McGregor 1990: 508), Gurindji (Meakins and McConvell 2021: 552), and Ngardi (Pama-Nyungan) (Ennever 2021: 605). In no description are either the meanings or the uses of the morpheme discussed in detail.

Such a morpheme is attested in just 21–23 languages – i.e., a little over a third of the languages attesting a mode of expressing MB. This includes languages like Walmajarri (Richards and Hudson 1990: 154) that have no dedicated mistaken belief construction. The presence of this morpheme makes explicit that the thought is mistaken. These languages are shown in Figure 4. (Of course, absence of evidence for such a morpheme is not evidence of absence.)

Figure 4:

Languages with an attested ‘but really’ morpheme.

In Mparntwe Arrernte the focus clitic can occur on an NP representing the correct categorisation (no other marker attested), as in (3) above; and in Martuthunira it may be that the emphatic clitic is used in this way.

In Jabirrjabirr (Nyulnyulan), according to Nekes and Worms (1953: 531), alongside -gabad MB are also -gabad … -gal and -ga … -delba. Unfortunately, they provide almost no discussion or exemplification. The one example they do give, (52), suggests that the second component marks what really happened. (The “though” in the gloss line is theirs.)

4.3 Discourse uses of MB expressions

Little is known about the motivation for using MB expressions in discourse, and the meanings that might be intended. Perhaps the most common motivation seems to be to explain behaviour that is unexpected given the interactional context or the referent world and the participants in it.

A number of examples cited above illustrate this. In (25), the misclassification of the birds as water hens is used to explain why the speakers killed the chickens. Examples (26) and (37) come from a description of a drawing in the picture book A boy, a dog, a frog and a friend (Mayer and Mayer 1971) that depicts the boy continuing to dig a hole in which to bury the turtle, which is now clearly alive. The unexpected behaviour of the boy is accounted for by attributing to him the belief that the turtle is dead. And in (38) the behaviour of the police, surprising given that the speaker and cohorts are merely talking together, is accounted for in terms of the belief that they were drinking.

It seems likely that in examples such as (25) the explanation of unexpected behaviour might be in turn employed to excuse that behaviour: one expects that (25) could be employed by the speaker as an excuse for their killing the chickens, not just to explain it. Similarly, the Wangkumara example (44) was clearly used as an excuse in the exchange from which it comes (though in this instance the mistaken belief sense emerges via implicature). Jabirrjabirr example (52) quite likely could have been employed in the same way as an excuse.

By contrast, (23) could potentially be employed as an accusation: that the addressee has failed to perform some action expected of them.

In at least one language, Mparntwe Arrernte, the MB enclitic is sometimes employed in criticisms (Wilkins 1989: 409). This is illustrated by example (53), where the speaker is criticising the hearer for not giving them money.

Wilkins (1989: 426) suggests that the sense of criticism arises as a pragmatic implicature. I agree, though I have a slightly different account of how this meaning is implicated (elaborating somewhat on McGregor 2021: 13). To begin with, the speaker claims that the proposition P (‘you are my friend’) is mistaken in the face of the fact that P is clearly true, and known to be true by both interactants. Their general knowledge thus conflicts with what is stated, namely that P is not true. The speaker flouts the Gricean maxim of quality. This is for a reason. Invoking the counterfactual proposition ¬P the speaker effectively instructs the hearer to infer how it might be that they can claim P to be false. The answer is evident from the interactional context, the situation that the addressee has refused the speaker money, which by ordinary understanding would be the behaviour of people who are not friends. It follows that you are remiss in not giving me money, and the criticism emerges via pragmatic implicature.

This is clearly an effective strategy for indirect criticism, and one would expect it to be widely employed. However, Mparntwe Arrernte is the only Australian language in which this pragmatic implicature is actually attested.

4.4 Mistaken belief and the notion of similarity

We have seen that in some languages there is an affinity between the senses of mistaken belief and similarity. As discussed in Section 4.1, the mistaken belief is sometimes not attributed to anyone in particular, in which case the sense of likeness may emerge, as shown by (54) and (55). In fact, most examples of the MB particle in non-initial position in Warlpiri invoke this sense, and (as is clearly the case in (55)) the preceding material is not included in what is mistakenly believed (see Laughren 1982: 149).

Warlpiri

In these languages the semblative sense can be accounted for as a contextualisation of the coded meaning ‘it was mistakenly believed’: that the proposition is something that could have been believed by someone in the relevant circumstances suggests that what actually happened was like something else (namely what it could have been misconstrued as). The semblative sense is thus not emically distinct from the mistaken belief sense. Glosses like ‘seemingly’ – as per for example Waters (1989: 157) for Djinang – are thus appropriate, provided that they are understood to simultaneously code ‘but in fact not’. It is surely significant that in all instances where this meaning comes to the fore, just a particle or clitic is found, never a complement construction. It can nevertheless be present, albeit backgrounded, in the latter constructions: for example in (23), that something happened that looked to the addressee like the speaker went for water.

I am not suggesting that in all languages showing similarities in the expression of the mistaken belief and semblative senses the relevant clitic or particle codes the MB meaning, with similarity a contextual sense. Recall that in Sections 3.1.1 and 3.1.2 we mentioned languages in which clitics or particles of mistaken belief were (homophonous with) semblative morphemes. I suspect that these languages have a single morpheme that codes a meaning we might gloss ‘like’. The mistaken belief sense can be accounted for in two ways. First, as Kapitonov and Gentens (2018) observe, it can plausibly arise via a Gricean implicature of the coded semblative meaning: what is similar is not identical. Second, as per McGregor (2021: 15), the mistaken belief sense arises when the morpheme holds a clause in its scope, modifying it propositionally. Correspondingly, the semblative sense emerges when the notion of likeness applies not to a proposition but instead to some thing or event in the discourse world. It is not propositions but items of the world of experience that show resemblances. Put in functional grammar terms, for mistaken beliefs the modification is at the interpersonal level; for semblatives it is at the experiential or representational level (Halliday 1985; Hengeveld and Mackenzie 2008).

These remarks show that neither meaning, mistaken belief nor semblative, can be presumed synchronically more fundamental or diachronically prior. Both are equally plausible candidates, and require an evidential basis.

4.5 Puzzles

Tsunoda (2011: 677) makes the astute observation that in Warrongo there is no instance in which the mistaken belief is a negated proposition: there are no instances like ‘I thought he was not going, but in fact he did go’. And indeed, I have come across very few instances of this type in the descriptions. Dixon (1991: 117–119) cites a few examples in Yidiny, though they all involve the privative marker on a nominal, rather than clausal negation. A probable instance is example (56) from Yir Yoront, though the interpretation is unclear due to lack of glossing in the source.

I suspect that there is a reason for the paucity of examples like (56). As is well known, negative clauses are distributionally marked and invoke presupposition of the corresponding positive (e.g. Givón 1984: 324; McGregor 1997: 226–227; Quirk et al. 1985: 808–809). It seems reasonable to hypothesise that the same holds for one’s inner speech: one would think a negative proposition only when there was a presumption that the positive proposition was true. That is to say, one would think ‘she is not here’ in the context that they expected her to be here; otherwise, one would expect the person to entertain a positive proposition such as ‘she has gone’ or ‘she is over there’. But then the presupposition is precisely what would be asserted if the thought was claimed to be false in a mistaken belief construction: saying ‘it was mistakenly believed that she is not here’ asserts that she is here, which is precisely the presupposition of the thinker. The normal expectation in these circumstances would be that one would simply state the proposition: ‘she is here (just as expected)’. It seems pointless to muddy matters by invoking the thought ‘she is not here’, which turns out to be false.

The rub is that to do so, to use the mistaken belief construction is possible, albeit a more marked expression than the simple statement of the positive clause. This must be for a reason. We can account for example (56) as follows. By invoking the mistaken thought ‘you are not here’ the speaker accounts for their own presence on the scene, for why they are now in the presence of the addressee. The speaker thereby excuses themself for imposing on the addressee.

One additional example is significant, (57), the interpretation of which is far from clear. Evans (2003: 611) takes it to be a counterexample to the MB sense of the particle yimankek, and avers that ‘they reckoned’ (given in the translation by a native speaker) is an appropriate free translation. It is common knowledge among the Bininj Gun-wok that the Rainbow Serpent will not hear mosquitoes being slapped, provided it is done loudly. How then can (57) present this as a mistaken belief? I have suggested (McGregor 2021: 13) that an explanation may be possible in terms of a weakening of the second component of meaning of the MB construction (see Section 4.1. above). Specifically, the belief is false not according to the speaker, but rather in relation to general knowledge of the real world – according to which slapping mosquitoes hard would result in being heard.

Bininj Gun-wok

Another explanation is possible. If a Bininj Gun-wok speaker had wanted to express a mistaken belief in this context, one would have expected them to have said ‘They mistakenly believed Ngalyod would hear them if they slapped the mosquitoes hard’. But this could hardly be true: why would they have hit the mosquitoes hard if they believed the rainbow serpent would hear them – and consequently eat them? Perhaps the speaker is negating the entire proposition ‘They mistakenly believed that the rainbow serpent would hear them if they slapped the mosquitoes hard’. This would mean that the negative particle holds the entire sentence in its scope, not just the proposition mistakenly believed. The problem is that this seems to conflict with the scopal relations as suggested by the ordering of the two particles. This is not necessarily an insuperable problem: ordering of elements does not always reflect scoping relations – see also example (56), where the negator appears to be in the matrix clause rather than the complement clause, as predicted by the scoping relation.

It is impossible to decide among the above analyses. My purpose is simply to show that the mistaken belief interpretation need not be ditched out of hand.

4.6 Concluding remarks: mistaken belief as a modal category

We have seen that some Indigenous languages of Australia have a morpheme or construction that codes mistaken belief: this meaning is invariably present and is indefeasible. In such languages, mistaken belief is an epistemic modal category that invokes a dual perspective on a proposition. On the one hand, they indicate that the proposition was believed by someone at some point in time, or could have been believed by someone in the circumstances. On the other hand, they indicate that the proposition is in fact false from the perspective of the present speaker in the here and now. Examples like (53) do not contradict the second component of meaning. Quite the contrary: the implicature relies on the fact that the encoded meaning is contrary to the facts as understood by the speaker. However, as discussed in Section 4.5., it may be that this second requirement also needs to be relaxed somewhat: the proposition might instead be evaluated from the perspective of general knowledge of the world.

As an epistemic modal category mistaken beliefs concern propositions, things that can be true or false, not occurrences of events. Thus I have proposed (Section 3) that in circumstances in which just the category of some entity is at issue, the marker holds a full categorising clause in its scope, not just a nominal. An important fact about mistaken belief constructions is that they represent more than mere thoughts. The thinker does more than simply entertain the thought: they also believe it. One could easily imagine they were living on the moon without mistakenly believing that they were. Hence in contrast with ordinary thought complements mistaken beliefs never comprise just an interjection, and indeed seem never to contain one. This suggests the distinctiveness of expressions coding mistaken belief as constructions separate from reported thoughts, extending on the remarks in Sections 3.2.1. and 3.2.4. Fuller arguments are called for.

The fact that mistaken belief constructions marked by just a morpheme seem generally to be non-specific as to the person holding the mistaken belief underlines their comparability with other modal categories. For instance, both epistemic and deontic senses of must in English admit a similar range of possibilities for the source of the epistemic or obligative meaning: the speaker, some key third person, a general member of the society, even the addressee themself.

It has emerged in the discussion above that there are sometimes similarities in the marking of mistaken belief and other meanings such as similarity (discussed in Section 4.4.) and modal categories such as counterfactuals and irrealis. It is beyond the scope of this paper to explore the similarities and differences among the various modal categories. Suffice it to say that while mistaken belief resembles these categories in terms of the fact that it marks the proposition as false, it conveys an additional meaning not conveyed by either counterfactual or irrealis, namely that someone (might have) believed the proposition.