An Analysis of Colloquial Singapore English lah and Its Interpretation across Speech Acts

Abstract

Previous research has observed that the Colloquial Singapore English particle lah conveys many different, and sometimes contradictory, pragmatic effects. In this paper, I focus specifically on how lah, pronounced in a low falling tone, behaves differently with assertions and directives—although it emphasizes the truth of assertions, it weakens the authoritative force of directives. In addition, it can be used in a non-emphatic way with confirmation-seeking statements. I propose that the particle conveys the not-at-issue or side comment that the lah-marked proposition directly follows from the evidence it is based on, which is interpreted by the addressee as an attempt by the speaker to justify her utterance. The different pragmatic effects of the particle then result from how this not-at-issue comment is interpreted in relation to the speech act of the utterances they mark.

1. Introduction

A wide range of pragmatic functions have been attributed to the Colloquial Singapore English (CSE) particle lah, which has been described as “one of the hardest and most intriguing chestnuts in the field of cross-cultural pragmatics” (Besemeres and Wierzbicka 2003, p. 3). For instance, lah can be used to emphasize utterances, e.g.,:

(1)	A:	Do you want to go? I’m not going lah.1
		(I emphasize to you that I’m not going.)
		(Kwan-Terry 1992, p. 69)

However, lah can also serve to soften the force of utterances:

(2)	A:	Come with us lah.
		(Won’t you come with us?)
		(Ler 2006, p. 156)

Yet another function of lah is to seek confirmation from the addressee:

(3)	(Context: A and B are talking about the celebrity Devon Lee Carlson.)
	B:	i thought she was a model
	A:	then after that like after she got together with the guy right then she became a model
	B:	orh oh because she got famous lah
		(You mean she got famous?)
	A:	ya
	A:	she got got drawn a lot of attention
	(NSC 3009)

This variability is also acknowledged by Gupta, who describes lah as “cover[ing] the full range within the assertive continuum” (Gupta 1992, p. 42).

One way that previous researchers have accounted for lah’s illocutionary effects is by characterizing it directly as a speech act operator (e.g., Kwan-Terry 1978; Besemeres and Wierzbicka 2003). As some of lah’s effects have been observed to correspond with specific intonational contours with which it is pronounced, some researchers (e.g., Kwan-Terry 1978; Bell and Ser 1983; Wong 2004) conclude that there are different variants of the particle, each with its own pragmatic function, that are distinguished by lexical tone. However, lexical variation alone is unable to account for all the differences in lah’s discourse effect, as lah can be pronounced using the same pitch contour—a low falling contour—in the previous examples (1–3).

To avoid the issue of lexical variation, this paper focuses on lah that is pronounced in a low falling contour (henceforth lah₂₁, using Chao’s (1933) 5-point scale for tones). I argue that the speech act of the lah₂₁-marked utterance—whether the utterance is an assertion or directive—is another factor that determines lah’s pragmatic effect. However, rather than analyzing lah₂₁ as a speech act operator or speech act modifier, I propose in this paper to separate its discourse functions from its semantics and characterize it as indicating a not-at-issue or side comment on the lah₂₁-marked proposition p, from which different discourse effects arise based on the addressee’s pragmatic inferences given the utterance context and speech act of the utterance. Specifically, my claim is that lah₂₁ conveys the not-at-issue comment that p follows directly from relevant facts that constitute the speaker’s evidence for p, i.e., the speaker has completely reliable evidence in support of p. While this not-at-issue comment does not affect the at-issue or truth-conditional meaning of the utterance, it is interpreted by the addressee as justification for p, which results in lah₂₁’s pragmatic strengthening of assertions.

With directives, the set of relevant facts in the discourse context are those that would lead the addressee to accept the directive and obey the speaker. Hamblin (1987) notes a distinction made in the literature between a speaker’s social authority, or the social power a speaker of the imperative has over the addressee, and rational authority, or the independent reasons why the addressee should carry out the speaker’s imperative. I argue that when lah₂₁ is used with directives, the speaker’s attempt to provide justification highlights her rational authority, shifting the focus away from her social authority over the addressee. This results in the ‘softening’ effect that has been observed with lah₂₁-marked directives.

Finally, I show that the analysis of lah₂₁ can also account for its non-emphatic interpretation in utterances that have a declarative clause type but are not assertions, such as the confirmation-seeking statement in (3). This function of lah has not been identified in previous literature and poses a problem for analyses of lah as a purely emphatic particle (e.g., Wong 2004). Evidence for this function is drawn from the conversational speech section of Singapore’s National Speech Corpus (NSC), a recently-constructed corpus comprising 1000 hours of spontaneous conversation from Singaporean speakers across different social demographics (Koh et al. 2019), and cited as ‘NSC [conversation ID]’.2 I propose that in such cases, the speaker is using lah₂₁ to indicate explicitly that the lah₂₁-marked confirmatory statement follows directly from the other facts already in common ground in order to support the speaker’s inference. Since the lah₂₁-marked statement is not an assertion, no emphatic effect results from this indication.

The empirical data in this paper can also be regarded in light of the larger debate in the literature on the semantics of imperatives. Similar to Ettinger and Malamud’s (2015) argument in their analysis of the Mandarin particle ba, lah₂₁’s felicity (more generally, lah’s felicity) with both declaratives and imperatives in the previous examples (1–3) argues against a semantics of imperatives in which imperatives have a different semantic type from declaratives, e.g., Portner (2005, 2007), as this would require each clause type to be modified by a different semantic variant of lah. The existence of other particles that can modify both declaratives and imperatives thus provides additional support for a uniform semantics for both clause types, e.g., in Condoravdi and Lauer (2012), Kaufmann (2012), and Starr (2010, 2012).

The paper is structured as follows: Section 2 gives an overview of the discourse particles in CSE. Section 3 explains the theoretical framework in which I will be analyzing assertions and shows how an analysis of lah₂₁ as providing justification for the speaker’s lah₂₁-marked utterance accounts for its pragmatic effect with assertions, while Section 4 extends the analysis of lah₂₁ to directives. Section 5 shows how, by not encoding lah₂₁’s discourse effects directly in its semantics, the analysis can also account for lah₂₁ being used in a non-emphatic way with confirmation-seeking statements. Finally, Section 6 concludes the paper, and suggests directions for future research.

2. Discourse Particles in CSE

Discourse particles in CSE are mostly loans from Cantonese and Hokkien3, and how they are used has been the subject of many previous studies (e.g., Richards and Tay 1977; Kwan-Terry 1978, 1992; Bell and Ser 1983; Loke and Low 1988; Platt and Ho 1989; Gupta 1992; Pakir 1992; Wee 2002, 2004; Besemeres and Wierzbicka 2003; Wong 2004; Ler 2005; Lim 2007; Kim and Wee 2009; Leimgruber 2016). Though syntactically optional, these particles perform key pragmatic functions such as indicating the speaker’s stance or emotional tone (Wee 2004). They also interact with the sentence type or speech act of the utterances in various ways—for example, some particles, e.g., meh below, are only felicitous in questions:

(4)	Meh—indicates skepticism
	A:	You got girlfriend meh?
		(Are you sure you have a girlfriend? I don’t believe it.)
		(Wee 2004, p. 121)

Kim (2014) analyzes the properties of CSE particles as a single syntactic and semantic class, and claims that the additional meanings that they indicate are not conveyed at the level of propositional content. For example, they cannot be challenged by the addressee, unlike adverbials or attitude verbs:

(5)	a.	i.	A:	Your job is stressful meh?
				(Are you sure your job is stressful? I don’t believe it.)
			B:	But it is!/# But you must believe me!4
		ii.	A:	I don’t believe that your job is stressful.
			B:	But it is!/But you must believe me!
				(adapted from Kim 2014, p. 231)
	b.	(Context: Alice, Bob and Cass are talking about a mutual friend Sam, who had gotten into a heated argument with his boss.)
		Bob:	So what did Sam do in the end?
		i.	Alice:	He apologized lor.
				(It should be obvious that Sam apologized in the end.)
			Cass:	No he didn’t! He refused to apologize!/# No it isn’t! It’s not obvious that he had to apologize!
		ii.	Alice:	He apologized, obviously.
			Cass:	No he didn’t! He refused to apologize!/No it isn’t! It’s not obvious that he had to apologize!

Like other CSE particles, the additional meaning of lah₂₁ is, besides being unchallengeable, also unembeddable and unreportable:

(6)	a.	Alicia:	# Bala thinks (that) [the painting is good lah21].
	b.	Alicia:	[Bala thinks (that) the painting is good lah21].

(7)	a.	# Alicia says that [Bala thinks (that) the painting is good lah21].
	b.	[Alicia says that Bala thinks (that) the painting is good lah21].

In (6), lah₂₁ cannot be interpreted as modifying the clause that is embedded under think, but only as modifying the matrix clause—an addressee would attribute the particle only to Alicia and not Bala. Similarly, in (7), lah₂₁ cannot be attributed to Alicia as part of reported speech and can only be attributed to the speaker of (7).

In addition, lah₂₁ cannot be interpreted as modifying only a single conjunct or disjunct in an utterance, or only the consequent of a conditional utterance:

(8)	a.	Alicia:	# It will either rain today or [it will rain tomorrow lah21].5
	b.	Alicia:	[It will either rain today or it will rain tomorrow lah21].

(9)	a.	Alicia:	# It will rain today and [it will rain tomorrow lah21].
	b.	Alicia:	[It will rain today and it will rain tomorrow lah21].

(10)	a.	Alicia:	# If it rains today, [it will also rain tomorrow lah21].
	b.	Alicia:	[If it rains today, it will also rain tomorrow lah21].

In (8–10), Alicia can only use lah₂₁ to emphasize the entire disjunction, conjunction, or conditional in her utterance, and not just the single proposition It will rain tomorrow. For example, Alicia emphasizes to an addressee in (8b) that it would rain on either of the two days, rather than perhaps on other days.

Following Potts’s (2005) analysis of expressives such as damn or bastard as non-propositional content, Kim (2014) proposes that CSE particles indicate the speaker’s attitude towards the marked proposition, some of which include a corresponding instruction to the addressee. For example, the particle meh in (4) indicates explicitly the speaker’s disbelief of the marked proposition, and a request for the addressee to adopt a correlative attitude.

Instead of the speaker’s attitude, I propose that lah₂₁ indicates that the lah₂₁-marked utterance directly follows from the evidence that it is based on. In this way, it fulfills a similar function as epistemic necessity modals such as English must (cf. von Fintel and Gillies 2010), although lah₂₁’s unchallengeability, unembeddability and unreportability indicate that it conveys not-at-issue rather than at-issue content. This not-at-issue content can be further formalized as either propositional CI content in Potts (2005) or as impositions to the common ground in AnderBois et al. (2015), parallel to appositives, although such a step will not significantly affect the analysis of lah₂₁ in the following sections and thus is beyond the scope of this paper.

In the following section, I outline how lah₂₁ functions with assertions, beginning with the model of at-issue assertions that I will be assuming in my analysis.

3. Lah₂₁ with Assertions

3.1. A Discourse Model of At-Issue Assertions

The fundamental intuition that I will be basing my characterization of lah₂₁ on is that when the speaker uses lah₂₁ to emphasize her assertion, the particle conveys additional information that justifies in some way the lah₂₁-marked assertion. I assume, following Gunlogson (2001), Farkas and Bruce (2010), Murray (2014), etc., that when a speaker utters an at-issue assertion, she is merely presenting a proposal to update the common ground that must be accepted by the addressee before it is taken to be common knowledge. In contrast, not-at-issue content in a speaker’s utterances, such as presuppositions or appositives, are non-negotiable additions or restrictions to the common ground (e.g., Murray 2014; AnderBois et al. 2015).6

In addition, following Northrup (2014), I take an assertion to convey not just the update proposal itself, but also an evidential base that contains the set of evidence underwriting the speaker’s epistemic authority. When a speaker asserts a proposition p, the evidential base for the assertion is the set of propositions—speaker beliefs in this case—that together serve as evidence for p, which is always conveyed together with the assertion in discourse. In other words, an assertion’s evidential base constitutes the speaker’s justification for committing to the assertion. Under this analysis, what evidentials do is to indicate the character of such an evidential base—when a speaker asserts the proposition She left using a hearsay evidential, she is indicating that the evidential base that underwrites her assertion is a set of propositions that constitutes hearsay evidence that the person that she is referring to has left (Northrup 2014, p. 39).

More precisely, Northrup (2014) defines evidence as a set of propositions such that the speaker would be unable to commit to her assertion without committing to them as well. Why would the speaker be unable to commit to her assertion otherwise? Davis et al. (2007) suggest an answer to this, basing their pragmatic account on Lewis’s (1976) probabilistic revision of Grice’s (1975) Maxim of Quality, i.e., “Contribute only what you know to be true. Do not say false things. Do not say things for which you lack evidence”:

(11)	Lewisian Quality: “The truthful speaker wants not to assert falsehoods, wherefore he is willing to assert only what he takes to be very probably true. He deems it permissible to assert that A only if P(A) is sufficiently close to 1, where P is the probability function that represents his system of degrees of belief at the time. Assertability goes by subjective probability.”
	(Lewis 1976, p. 133)

Davis et al. (2007) propose that a speaker would only be able to felicitously assert those propositions with subjective probabilities (from the speaker’s perspective) that pass a certain threshold, i.e., that are “sufficiently close to 1” according to Lewis (1976)—how close to 1 is considered sufficient will depend on the discourse situation, such as the discourse participants’ goals, etc. In line with this probabilistic approach, McCready and Ogata (2007) define a proposition q as evidence for a proposition p for an agent a if q increases a’s subjective likelihood of p given that a knows q. Having sufficient evidence, e.g., visual evidence, therefore pushes the speaker’s subjective probability of p past the threshold required to felicitously assert p, while having insufficient evidence keeps the speaker’s subjective probability of p below that threshold.

Besides determining if an assertion is felicitous from the speaker’s own point of view, the evidential base for an assertion also determines if the addressee should accept it as a viable update of common ground. Northrup (2014) and Gunlogson and Carlson (2016) propose that whether the addressee accepts the speaker’s proposal depends on whether he believes her to have sufficient evidence to support her claim. While addressees generally assume that the evidential bases that speakers use to justify their commitments are sufficient, there are situations in which an addressee is less willing to accept a speaker’s assertion by default, e.g.,:

(12)	(Context: Robin is working in a windowless room, apparently with no access to information about the outside world. Another person enters the room.)
	Robin:	(to newcomer) # It’s raining outside.
		(adapted from Gunlogson 2008, p. 118)

In (12), Robin extends a proposal to the newcomer that it is raining outside. However, it would be difficult for the newcomer to accept Robin’s proposal to add the proposition to common ground even though he may also have no knowledge of current weather conditions. (For instance, they might both be working in the basement the entire day.) This is because from the newcomer’s point of view, there seems to be no way that Robin could have found out that it was raining while in a windowless room, so he is unable to imagine how her assessment of the subjective probability of it raining outside can pass what he believes the threshold level for assertion to be. In addition, Robin should have recognized the newcomer’s difficulty, but is still not supplying any evidence for her claim that can convince him to accept it, which is odd given the default assumption that she is a cooperative conversationalist.

In contrast, the same proposal becomes more plausible once Robin reveals sufficient evidence to support her claim:

(13)	(Context: Robin is working in a windowless room, apparently with no access to information about the outside world. Another person enters the room.)
	Robin:	(to newcomer) It’s raining outside. Sam, who just left work a minute ago, called to warn me to take an umbrella when I leave.

In (13), Robin follows up her claim that it is raining with another assertion that a co-worker Sam has called to warn her to take an umbrella when she leaves. She also emphasizes the strength of her evidence—Sam has left work a minute ago and so would likely have first-hand evidence of the weather outside, ensuring that the probability of it raining outside is fairly close to 1, or almost certainly true. This would be far more acceptable to the newcomer as Robin shows both that her proposal to update the common ground with the proposition that it is raining outside is backed by actual evidence, and that the evidence is highly reliable. In the next section, I show that lah₂₁ plays a similar role as the appositive in (13) in highlighting the reliability of the speaker’s evidence for her lah₂₁-marked utterance.

3.2. Lah₂₁’s Emphatic Effect with Assertions

Although the emphatic effect of lah with assertions has been observed multiple times in the literature (e.g., Kwan-Terry 1978; Bell and Ser 1983; Ler 2005, 2006), the recorded examples often do not identify the intonational contour the particle is pronounced with, and more than one may be acceptable according to the discourse context stated. Wong (2004), however, lists a few examples of lah₂₁ specifically, which he describes as the ‘impositional’ lah:

(14)	a.	(Context: B grooms himself meticulously before the mirror for quite some time. A teases him.)
		A:	Nice already lah21.
	b.	(Context: B does not know if a food item can get through the airport customs. A thinks so.)
		A:	Can lah21.
	c.	(Context: B thinks it is A’s last day of work. A says it is not.)
		A:	Not last day lah21.
			(Wong 2004, pp. 763–64, with orthography standardized)

Wong describes the speaker using lah₂₁ in each of the examples in (14) (among others in his paper) as “coercing [the addressee] to accept what to him is a fact, see something the way he sees it, or change her mind about something” (Wong 2004, p. 764).

This characterization of lah₂₁ seems to be too narrow, given the example in the introduction of lah₂₁ used with a confirmation-seeking statement. Indeed, more general characterizations of lah have also been proposed by Besemeres and Wierzbicka (2003) and Ler (2005, 2006)—Besemeres and Wierzbicka (2003) propose that lah indicates the speaker’s expectation that the addressee is able to understand her, while Ler (2005, 2006) proposes using a Relevance-Theoretic framework that lah highlights a shared contextual assumption that the addressee is expected to access and accommodate in order to reach an optimally relevant interpretation of the lah-marked utterance.

However, these characterizations seem to be too general, as they are unable to account for the unacceptability of lah₂₁ in (15):

(15)	(Context: Alicia and Bala are looking at a painting in a museum. This is their first time looking at paintings together, and Bala has no idea what kind of paintings Alicia might like.)
	Bala:	I like this one!
	Alicia:	Yah, I like it too (#lah21)!

Assuming Besemeres and Wierzbicka’s (2003) characterization of lah₂₁, Alicia in (15) should be able to convey using the particle her expectation that Bala is able to understand why she likes the painting, since he has also said he likes it, perhaps to evoke a sense of connection. Similarly, assuming Ler’s (2005, 2006) characterization of the particle, Alicia should be able to use lah₂₁ in (15) to convey that Bala is expected to access and accommodate a shared reason for liking the painting (e.g., because it is painted very well, or because the subject of the painting is impressive), given that he has already said he likes it, in order to derive the optimally relevant interpretation that she approves of his taste in paintings.

While lah₂₁ is inappropriate in (15), the particle becomes acceptable in a slightly different situation where Alicia is correcting Bala’s expectations about her aesthetic preferences after she had frowned at a painting:

(16)	(Context: Alicia and Bala are looking at a painting in a museum. Bala notices Alicia frowning at it.)
	Bala:	I like this one! You don’t like it?
	Alicia:	No, I like it too (lah21)! I just think the frame looks very ugly.

Any successful characterization of lah₂₁ must therefore be flexible enough to account for both its emphatic or corrective function in (14) and (16) and its confirmatory function in (3), while also being precise enough to predict the contrast in acceptability between (15) and (16).

3.3. The Semantics and Pragmatics of lah₂₁

Since it is difficult to pin down the exact discourse context in real-world uses of lah₂₁, (17) is a constructed example, where Ai Ling uses lah₂₁ in its emphatic function to convince Ben that their mutual friend Simon is not attending the party the next day:

(17)	(Context: Ben is under the impression that his friend Simon will be attending a party the next day. He tries to confirm this with a mutual friend Ai Ling.)
	Ben:	Simon is going to the party tomorrow right?
	Ai Ling:	He’s not going lah21. He has a test the day after that.

From discussions with my consultants, as well as from my own intuitions as a native CSE speaker, Ai Ling is indicating in (17) that Ben should believe her as her claim that Simon is not going to the party is definitely true. Why might Ben be reluctant to accept Ai Ling’s claim? As discussed in Section 3.1, one reason could be that Ben was expecting Ai Ling to confirm his impression that Simon will be attending the party, and so does not believe she has sufficient evidence to justify her claim that Simon is not going to the party—either her claim should not have passed her threshold for assertion in the first place or, given his impression to the contrary, it does not pass his more stringent probability threshold for accepting the proposal to update the common ground.

Suppose that, under normal circumstances, Ai Ling believes that it is very likely that Simon will not attend the party because she knows that Simon is a conscientious student who always studies hard for his tests, and so he would be studying for his test that night instead. When Ai Ling asserts “Simon is not going to the party” (without lah₂₁) in this context, the propositions Simon is a conscientious student who always studies hard for his tests and Simon has a test the day after the party form the evidential base that underlies her assertion.7 However, there is still a small chance that Simon will in fact go to the party—for example, he might decide at the last minute that he deserves some fun after studying hard for his test. While Ai Ling may believe that the available evidence raises the probability that Simon is not going to the party sufficiently past the threshold required to felicitously introduce it as a proposal to update the common ground, Ben may require a more stringent threshold (based on an evaluation of his own impression of whether Simon is going to the party, as well as his estimation of what Ai Ling’s threshold is and the available evidence) before he is willing to reject his own impression and accept the update.

I therefore propose that lah₂₁ justifies the speaker’s commitment to the proposition p expressed by the lah₂₁-marked at-issue assertion by indicating that it directly follows from the evidence that it is based on, i.e., there is no possibility that it could be false given the evidence that led the speaker to believe that it is true.8 As examples (6–10) in Section 3.1 illustrate, the proposition that it modifies can only be the maximal proposition expressed by the lah₂₁-marked assertion. By claiming a stronger link between p and the evidence that underlies its assertion, the utterance “p lah₂₁” makes a stronger statement than “p” alone, since “p” can be asserted even if there is a small probability that it is not true, as long as the speaker’s subjective probability of p is “sufficiently close to 1” (Lewis 1976). In (17), Ai Ling is justifying her commitment to the proposition that Simon is not attending the party by claiming, using lah₂₁, that the fact that Simon has a test the next day is a perfect predictor of whether he would attend the party (together with other pieces of relevant evidence that are not explicitly stated but are mutually known, e.g., that Simon is a conscientious student who always studies hard for his tests), i.e., there is no chance Simon would be attending the party given the available evidence.9

I propose that how the addressee interprets lah₂₁ as justifying the speaker’s lah₂₁-marked assertion is as follows:

(18)	a.	The addressee recognizes lah21(p)—that p directly follows from the evidence that it is based on—as not-at-issue or non-truth-conditional content, but assumes that the additional comment is relevant to the conversation.
	b.	The addressee recognizes that lah21(p) is stronger than the bare assertion of p.
	c.	From the discourse context, the addressee recognizes/assumes that the speaker assumes the proposal to add p to common ground to be controversial to him, i.e., that he might not be willing to accept the update of p, and that the speaker knows that he recognizes this.
	d.	The addressee interprets the speaker’s additional comment that p directly follows from the evidence that it is based on as her attempt to use the stronger statement to provide justification for her proposal to add p to common ground.

(18c) accounts for the contrast in acceptability between (15) and (16) (reproduced below as (19) and (20)) noted previously:

(19)	(Context: Alicia and Bala are looking at a painting in a museum. This is their first time looking at paintings together, and Bala has no idea what kind of paintings Alicia might like.)
	Bala:	I like this one!
	Alicia:	Yah, I like it too (#lah21)!

(20)	(Context: Alicia and Bala are looking at a painting in a museum. Bala notices Alicia frowning at it.)
	Bala:	I like this one! You don’t like it?
	Alicia:	No, I like it too (lah21)! I just think the frame looks very ugly.

Alicia’s assertion in (19) is odd with lah₂₁ as there is nothing in the discourse context to suggest that it is controversial and requires justification, while Alicia’s frowning at the painting in (20) leads Bala to guess that she does not like the painting, which renders her use of lah₂₁ to correct the misunderstanding completely natural. This also explains why lah₂₁ is often unacceptable with out-of-the-blue assertions such as in (21), as there is insufficient context to suggest why the assertion might be controversial to the addressee:

(21)	(Context: Ai Ling is looking out the window and talking to Ben.)
	Ai Ling:	It’s a nice day today (#lah21).

Again, a more loaded context such as in (14a) previously (reproduced below as (22)), where B grooming himself meticulously for a long time suggests that he does not think he looks nice enough, makes the use of lah₂₁ at the start of a conversation more acceptable:

(22)	(Context: B grooms himself meticulously before the mirror for quite some time. A teases him.)
	A:	Nice already lah21.

Under this analysis, lah₂₁ is similar to an evidential in that it conveys information about the evidence that underwrites the utterance it marks. However, unlike a conventional evidential that comments on the type of evidence underwriting an utterance (e.g., direct, reportative, etc.), lah₂₁ indicates the complete reliability of such evidence.10 In addition, this evidence does not have to be in common ground, in contrast to the analyses of Besemeres and Wierzbicka (2003), Ler (2005, 2006), and Northrup (2014). In the following example, Alicia can still use lah₂₁ even when her evidence is not shared by Bala:

(23)	(Context: Alicia and Bala are in a museum. Alicia is trying to persuade Bala, who does not know much about art, that the museum has mislabeled the painting in front of them. Bala, however, is reluctant to believe that the museum has made a mistake.)
	Alicia:	It’s painted by Dali lah21, not Magritte! How could they have got it wrong?

Given that a lah₂₁-marked assertion does not require the addressee to recover the exact set of evidence that the speaker is basing her assertion on, the speaker is essentially issuing a promissory note to the addressee that there is reliable evidence backing her assertion, which the addressee may accept on faith if he trusts the speaker, even if he has no idea what that evidence might be. I argue that the mutual knowledge that the speaker is relying on this trust in issuing her lah₂₁-marked assertion contributes to the social meaning of friendliness that lah₂₁ conveys, rather than it being an inherent meaning of the particle.

4. Lah₂₁ with Directives

4.1. A Semantics for Directives

As stated in the introduction, Ettinger and Malamud (2015) assume in their analysis of the Mandarin particle ba for reasons of parsimony that both declaratives and imperatives share the same semantic type, as a different semantic type for each clause type would require each clause type to be modified by a different semantic variant of ba. Since lah is similar to ba in that it can also be used with both declaratives and imperatives, I assume the same in my analysis of lah₂₁ to simplify how the particle combines semantically with the two clause types.11

In this paper, I focus mainly on the variant of imperatives usually termed directives or orders. Recent semantic analyses of imperatives generally share the intuition that a directive imposes in some sense an obligation or preference ranking on the addressee. For example, Condoravdi and Lauer (2011, 2012) propose that directives convey the speaker’s commitment to a preference for possible worlds in which the situation described by the directive is true over other worlds, and that this preference is not in conflict with her other preferences. Kaufmann (2012), on the other hand, formalizes directives as performative modals that, given an evaluation world, rank possible worlds in which the situation described by the directive is true higher than other worlds, where the ranking is determined by relevant criteria in the discourse context such as the speaker’s preferences.12

While the acceptance of an assertion depends to a large extent on the addressee’s assessment of the evidence that the assertion is based on, whether a directive is accepted by the addressee seems to be determined by the authority of the speaker, more specifically the speaker’s social authority over the addressee, using Hamblin’s (1987) terms. For example, Condoravdi and Lauer propose that directives carry a presupposition that the speaker has authority over the addressee, such that the addressee is “socially or institutionally obligated” (Condoravdi and Lauer 2011, p. 13) to accept the directive. Similarly, Kaufmann (2012) analyzes felicitous directives that are accepted by the addressee as directives where the addressee accepts the speaker’s presumption of social authority, while Starr (2012) simply defines the speaker as an authority over the addressee with respect to a directive if it is accepted by the addressee.

However, Hamblin (1987) notes that other variants of imperatives, such as advice or suggestions, may be accepted by the addressee even if the speaker has no social authority over him. Instead, there are independent reasons to justify why the addressee might benefit from obeying the speaker’s imperative, which he terms the speaker’s rational authority. He lists the following imperatives as examples:

(24)	a.	Turn left at the lights and go straight on past the Town Hall.
	b.	Take three of these immediately and try to get a good night’s rest.
	c.	If I were you I wouldn’t tell Julia.
	d.	You could ask him to phone the recommendation through.
	(Hamblin 1987, p. 11)

In each of the examples in (24), the speaker may not have sufficient social authority over the addressee to compel him to carry out the imperative. Instead, the addressee has to decide whether to accept the speaker’s advice or suggestion depending on whether he believes that the speaker is basing it on reliable information, e.g., knowledge of the city layout in (24a) or specific knowledge about Julia’s personality in (24c), such that carrying it out will make him better off.

In Northrup’s (2014) evidential base framework, a speaker’s authority is indicated by the utterance’s evidential base. While the evidential base for an assertion constitutes the speaker’s epistemic authority, comprising propositions that justify the speaker’s commitment to that assertion, the evidential base for an imperative constitutes the speaker’s deontic authority, comprising propositions that justify why the speaker can issue an order to the addressee. Extending Northrup’s analysis, I argue that the evidential base for an imperative contains all propositions that the speaker believes justify the imperative as a valid one that the addressee should accept. This would therefore include not just the fact of the speaker’s social authority over the addressee (if she has any), but also additional propositions that relate to a speaker’s rational authority.

In the following section, I show how a speaker can use lah₂₁ to highlight these additional propositions in an attempt, just as with assertions, to get the addressee to accept her directive. However, this shifts the focus away from the speaker’s social authority, which results in lah₂₁’s ‘softening’ effect with directives.

4.2. Lah₂₁’s De-Emphasis of Speaker Authority in Directives

To reiterate the puzzle with lah₂₁ that this paper attempts to solve, the particle that strengthens declaratives can also be used to ‘soften’ directives and weaken their authoritative force, e.g.,:

(25)	a.	A:	(to B) Come with us. (Direct command)
	b.	A:	(to B) Come with us lah21. (Persuading tone)
	(Ler 2005, p. 297, with my addition of lah’s pitch contour)

One observation that is fundamental to the analysis of (25b) is that, related to lah₂₁’s function of weakening the authoritative force of directives, it also reduces the authority the speaker needs to utter the directive felicitously in the first place. For example, a friend of the addressee would be able to direct (25b) to him in an attempt to get the addressee to go with her group, but it would likely be odd or rude for her to use the bare directive as a command in (25a) to try and achieve the same aim. Since a friend would not have any authority over the addressee, it would be difficult to get the addressee to accept the obligation that is imposed upon him by the command in (25a) without any additional motivating factors.

However, this is exactly the problem that I propose lah₂₁ is used to address—in a situation where the speaker does not have the requisite authority to compel the addressee to accept her imposition of deontic obligations, she may use lah₂₁ in an attempt to justify her imposition to the addressee based on other facts in the discourse context. As argued in the previous section, the evidential base for imperatives includes both the speaker’s authority over the addressee (if any) and independent reasons that justify why the addressee might benefit from obeying the speaker’s imperative. Following the analysis in Section 3, (25b) therefore indicates, in addition to the directive itself, that the obligation or preference ranking for the addressee to go with the speaker’s group directly follows from these independent reasons, since the speaker has no authority over the addressee in this example.13

In the context of (25b), these independent reasons might be the fact that the addressee may wish to make the speaker happy by obeying her since she is his friend, or that the speaker as the addressee’s friend would likely bring him somewhere he would enjoy. The speaker of (25b) is therefore indicating that these facts constitute very strong reasons for the addressee to go with the speaker’s group, establishing the speaker’s rational authority (cf. Hamblin 1987) and providing additional justification for why the addressee should accept her imposition despite her lack of social authority over him.

In contrast, if we assume that the speaker utters (25b) in a discourse context where she does have social authority over the addressee, the additional comment that lah₂₁ indicates is that the obligation or preference ranking expressed by the speaker’s imperative directly follows from her social authority over the addressee and the fact that she wants the addressee to go with her and her group. As discussed in the previous section, this is essentially what the bare directive in (25a) already indicates, since the speaker’s authority over the addressee would allow her to compel him to accept her directive. Moreover, a speaker would not have needed to justify her directive in the first place if she has sufficient social authority over the addressee to compel him to accept her directive, even if there are additional reasons in the discourse context that would motivate the addressee to obey the speaker. In such a situation, lah₂₁ would be redundant, and its use would violate Grice’s (1975) Maxim of Quantity. I argue that an addressee would therefore pragmatically infer from a lah₂₁-marked directive that the speaker either does not have the social authority to unilaterally impose obligations onto him or that, despite having authority over him, she is deliberately treating him as a social equal by making an effort to justify her directive to him. The former accounts for Besemeres and Wierzbicka’s (2003) observation that lah-marked utterances can seem to be ‘pleading’, while the latter accounts for the social meaning of friendliness or solidarity that previous researchers (e.g., Richards and Tay 1977; Kwan-Terry 1978; Bell and Ser 1983) have observed is conveyed by the particle, without the need to build such social meanings directly into the semantics of the particle.

Parallel with the addressee’s interpretation of lah₂₁ with assertions in (18), I propose that how the addressee interprets lah₂₁ with directives is as follows:

(26)	a.	The addressee recognizes lah21(p!)—that the preference ranking p! directly follows from the evidence that it is based on—as not-at-issue or non-truth-conditional content, but assumes that the additional comment is relevant to the conversation.
	b.	The addressee recognizes that the additional comment conveyed by lah21(p!) would be redundant if the directive’s evidential base only contains the fact that the speaker has social authority over the addressee and that she wants him to carry out the directive, and so assumes that the speaker intends to use lah21 to highlight other relevant propositions in the directive’s evidential base as independent reasons why he should accept the directive.
	c.	From the discourse context, the addressee can ascertain the degree of social authority the speaker has over himself, and recognizes that the speaker knows that he can ascertain this.
	d.	The addressee interprets the speaker’s additional comment that the preference ranking p! directly follows from the evidence that it is based on as her attempt to use the other relevant facts that have been highlighted to provide justification in order to convince him to accept the directive.
	e.	In a situation where the speaker does not have sufficient social authority to unilaterally impose obligations onto him, the addressee infers that the speaker is trying to convince him to accept the directive in spite of her lack of social authority. If the speaker does have sufficient social authority over him, the addressee infers that she is deliberately treating him as a social equal by making an effort to justify her directive to him.

As with a lah₂₁-marked assertion, a lah₂₁-marked directive does not require the addressee to recover the exact set of reasons that the speaker believes justifies her directive. If he has no idea what the set of reasons might be, the addressee may still trust that there is some rational justification behind the speaker’s directive given their relationship as friends, etc., and accept the directive. Again, I argue that this additional pragmatic reasoning by the addressee also contributes to the social meaning of friendliness and solidarity that lah₂₁ conveys, without it having to be part of the particle’s semantics.

5. Lah₂₁ with Confirmation-Seeking Statements

One advantage of separating lah₂₁’s semantics from its pragmatic effect of emphasis is that the analysis can accommodate situations where the speaker is not using the particle to emphasize the truth of her claim. As stated in the introduction, lah₂₁ can also be used to seek confirmation from the addressee:

(27)	(Context: A and B are talking about the celebrity Devon Lee Carlson.)
	B:	i thought she was a model
	A:	then after that like after she got together with the guy right then she became a model
	B:	orh oh because she got famous lah21
	A:	ya
	A:	she got got drawn a lot of attention
	(NSC 3009)

In (27), B had mistakenly thought that Carlson was always a model, but A corrected him by saying that she only became a model after she started dating her boyfriend Jesse Rutherford (the frontman of a popular band). B then concluded, based on A’s correction, that Carlson became a model as a result of the fame she garnered by dating Rutherford, and sought confirmation from A.

(28) is another example, where B confirms with A that there are no classes during recess week in A’s school:

(28)	(Context: A is telling B about her meeting with a friend to study during recess week, where there are no scheduled classes.)
	B:	orh what time
	A:	uh OTOT14 I guess we going to study
	A:	cause next week is recess week
	B:	oh so oh ya so the whole week you all no school lah21
	B:	is test or exam ah coming up
	A:	ya midterms
	(NSC 3038)

Unlike with assertions or directives, lah₂₁ in (27) and (28) does not indicate that the speaker (B) is certain of the truth of their utterances, or that they are trying to get the addressee to accept their utterances despite resistance. Instead, the lah₂₁-marked utterances in (27) and (28) are tentative inferences by the speaker that require confirmation from the addressee, which analyses of lah₂₁ as a particle that specifically indicates the speaker’s strong epistemic commitment (e.g., Wong 2004) would be unable to account for.

In the literature, these confirmation-seeking utterances have been analyzed as falling declarative questions (Gunlogson 2001, 2008). (29c) is another example in Standard American English15:

(29)	(Context: Robin is sitting, as before, in a windowless computer room when another person enters. The newcomer is wearing a wet raincoat and boots.)
	a.	Robin:	Is it raining?
	b.	Robin:	It’s raining?
	c.	Robin:	(I see that/So) It’s raining.
(Gunlogson 2001, p. 109; 2008, p. 104, with orthography standardized)

Gunlogson (2001, 2008) notes that falling declarative questions appear in situations where the speaker has sufficient evidence to venture an inference, but leaves her own commitment to her inference contingent on the addressee’s confirmation, as the addressee has significantly stronger or first-hand evidence regarding the issue. Because of this contingency, falling declarative questions require a ‘yes/no’ response from the addressee, whereas a response of ‘okay’ or silence, both of which are fine with assertions, are inappropriate. In support of this analysis, the confirmation-seeking utterances in (27) and (28) pattern with falling declarative questions in this regard:

(30)	(Context: A and B are talking about the celebrity Devon Lee Carlson.)
	B:	i thought she was a model
	A:	then after that like after she got together with the guy right then she became a model
	B:	orh oh because she got famous lah21
	A:	{ya/no/#okay/#(silence)}

(31)	(Context: A is telling B about her meeting with a friend to study during recess week, where there are no scheduled classes.)
	B:	orh what time
	A:	uh OTOT I guess we going to study
	A:	cause next week is recess week
	B:	oh so oh ya so the whole week you all no school lah21
	A:	{ya/no/#okay/#(silence)}

(32)	(Context: Robin is sitting in a windowless computer room when another person enters. The newcomer is wearing a wet raincoat and boots.)
	Robin:	(I see that/So) It’s raining.
	Newcomer:	{Yes it is./No it isn’t./#Okay./#(silence)}

However, lah₂₁ cannot be used with neutral or biased polar interrogatives (Richards and Tay 1977; Wee 2004), suggesting that it is only acceptable with falling declarative questions because of its clause type16:

(33)	(Context: A and B are talking about the celebrity Devon Lee Carlson.)
	B:	i thought she was a model
	A:	then after that like after she got together with the guy right then she became a model
	a.	B:	oh because she got famous (lah21)
	b.	B:	oh is it because she got famous (#lah21)
	c.	B:	oh isn’t it because she got famous (#lah21)

(34)	(Context: A is telling B about her meeting with a friend to study during recess week, where there are no scheduled classes.)
	B:	orh what time
	A:	uh OTOT I guess we going to study
	A:	cause next week is recess week
	a.	B:	oh so oh ya so the whole week you all no school (lah21)
	b.	B:	oh so oh ya so is it the whole week you all no school (#lah21)
	c.	B:	oh so oh ya so isn’t it the whole week you all no school (#lah21)

I propose that lah₂₁ does not have an emphatic effect in (27) and (28) because the lah₂₁-marked utterance, while being a declarative, is recognized by the addressee as not being an assertion. Given this context, the addressee interprets the additional comment conveyed by lah₂₁—that the proposition described by the lah₂₁-marked falling declarative question follows directly from the other facts already in common ground—as the speaker marking the proposition explicitly as a strong inference and justifying it given the available evidence for the addressee’s confirmation:

(35)	a.	The addressee recognizes lah21(p)—that p directly follows from the evidence that it is based on—as not-at-issue or non-truth-conditional content, but assumes that the additional comment is relevant to the conversation.
	b.	The addressee recognizes that lah21(p) is stronger than the bare assertion of p.
	c.	From the discourse context, the addressee recognizes that he has significantly stronger or first-hand evidence regarding p compared to the speaker, and that the speaker also knows that he recognizes this. He therefore interprets the speaker’s utterance not as an assertion of p, but as a statement that seeks to confirm that p is true.
	d.	The addressee interprets the speaker’s additional comment that p directly follows from the evidence that it is based on as her attempt to use the stronger statement to justify her inference given the available evidence for his confirmation.

To illustrate the process using (28), the speaker B concludes that A does not have school the following week from A’s comment that she is meeting up with a friend to study, with the meeting time being flexible, i.e., including times where they would usually have to go for classes, as well as her explicitly stating that the following week is “recess week”, which is commonly used to describe a period just before final examinations when there are no official classes (although there might still be ad-hoc classes scheduled). A’s comments thus constitute the evidential base for the proposition A does not have school the whole week, and the utterance “the whole week you all no school lah₂₁” indicates that B’s conclusion that A does not have school the following week directly follows from the evidence of A’s previous comments.

From A’s perspective, she recognizes that she has significantly stronger or first-hand evidence regarding whether she has school the following week or not compared to B, and that this is mutually known, so as a rational interlocutor, B would not be using lah₂₁ to convince A of this claim. However, there is another possible interpretation of lah₂₁’s not-at-issue comment given the discourse context, if A takes B’s statement as a falling declarative question and not a conventional at-issue assertion. A would therefore interpret B’s lah₂₁-marked utterance instead as a statement that seeks to confirm whether it is true that she has no school the following week, with the not-at-issue comment attempting to justify B’s inference as the strongest, most obvious or most reasonable one given A’s previous comments.

6. Conclusions

In this paper, I have proposed that the semantics of the CSE particle lah₂₁ interacts differently with the speech act of the utterances it marks. Instead of being directly encoded by the particle, I argue that lah₂₁’s observed illocutionary effects of emphasis, attenuation and confirmation result from the addressee’s pragmatic inferences based on the particle’s semantics.

To reiterate, I characterize lah₂₁ as conveying the not-at-issue comment that the lah₂₁-marked proposition directly follows from the evidence that it is based on, which is then interpreted by the addressee as an attempt by the speaker to justify her utterance. Although this has an emphatic effect with assertions, it weakens directives’ authoritative force because the attempt itself implies a shift in focus away from the speaker’s authority. With confirmation-seeking statements, the addressee’s recognition that the lah₂₁-marked declarative is not an assertion blocks the emphatic interpretation of the particle, and in such cases, lah₂₁ marks the declarative explicitly as a strong inference for the addressee’s confirmation instead. Several social meanings that have been observed with lah₂₁-marked utterances, such as friendliness or solidarity, can also be derived from pragmatic inferences given lah₂₁’s semantic meaning.

If the analysis in this paper is correct, one direction for further research is to extend the analysis to other intonational variants of lah such as lah₅₁. As mentioned in the introduction, some researchers (e.g., Kwan-Terry 1978; Bell and Ser 1983; Wong 2004) argue that these intonational variants are also semantically distinct, albeit with related meanings. However, another approach adopted by Davis (2009) for the rising and falling intonational variants of the Japanese particle yo is to assume one semantic meaning for the particle, while the rising and falling intonational contours themselves are analyzed as distinct morphemes that contribute additional semantic meaning. Future research can thus explore which approach will work best for lah using the semantic meaning of lah₂₁ proposed in this paper as a stepping-stone.

Finally, cross-linguistic comparisons with emphatic particles in other languages may also be fruitful. For example, Uyeno (1971) observes that yo can be used both to convey a sense of emphasis in an utterance, but also to soften the imperative tone of directives. Exploring the extent to which the analysis of lah₂₁ in this paper can be used to account for the behavior of these other particles, and the ways in which the particles differ, may also reveal much about the semantics and pragmatics of emphatic particles in general.