European Portuguese lá: Use-Conditional Meaning and Pragmaticalization

Abstract

This study focusses on non-adverbial uses of lá in European Portuguese, whose exact meaning contribution still remains an open research question. Applying a multidimensional semantics framework, the central claim is that non-adverbial uses of lá represent use-conditional items. Passing the standard tests suggested in the literature, they thus do not contribute to the truth conditions of an utterance but specify particular use conditions. It is argued that they are felicitously used if a speaker wants to convey illocutionary modification, which pragmatically leads to mitigation or reinforcement effects. Diachronically, substantiated by historical data from the Corpus do Português, use-conditional lá is argued to be a product of a pragmaticalization process that led to so-called pragmatic fission at some point, i.e., to the polysemy of two synchronically available configurations of a truth-conditional and a use-conditional lá.

1. Introduction

In European Portuguese, lá is primarily known as an adverbial that serves to designate a location relative to the speaker’s origo, i.e., as a spatial deictic expression, as illustrated in the examples in (1).

(1)	a.	Foi para a cidade e por lá ficou.1 ‘He went downtown and there he stayed.’
	b.	Compraram um terreno e lá edificaram a sua casa. ‘They bought a plot of land and built their house there.’

In particular, it signals a location which, according to the standard norm of European Portuguese, is proximate to neither the speaker nor the addressee (Costa and Martins 2010, p. 61). In addition to this basic function of lá, there are non-adverbial uses (cf. e.g., Duarte 1989), as exemplified in (2), which have seen growing attention over the past years.

(2)	[À mesa, diz a mãe para o filho:] (Pereira 2009, p. 87) ‘[At the table, says the mother to the son:]’
	a.	Acaba lá de comer a sopa! ‘(Just) eat the soup!’
	b.	Dá-me lá a colher que eu ajudo-te. ‘(Just) give me the spoon, I will help you.’

While it is clear that in these examples, lá is not meant to indicate a spatial location but something else, its exact meaning contribution to the entire utterance remains an open research question. Previous descriptions are mainly based on detailed paraphrases of individual examples to cover that lá adds certain nuances to the overall meaning of an utterance, which is commonly verified by checking the effect of its omission, i.e., by means of minimal pairs. The general assumption is that non-adverbial lá carries a highly context-sensitive, speaker-oriented meaning. Although not stated explicitly, this suggests the status of a conversational implicature, which is to be resolved taking into account the individual context that lá is uttered in. In this vein, Pereira (2009) assigns lá in (2-a) with the function of affecting the directive speech act (’Eat the soup!’) to signal impatience. Elsewhere, lá has been described as a marker of emphasis, a marker of cognitive distance and as a so-called emotional deictic expression (cf. Marques and Duarte 2017). Meisnitzer (2012) and (Meisnitzer and Müller 2024) assign non-adverbial lá with the status of a modal particle, very much in the tradition of Franco (1988, 1991), which means that they assume a conventionalized meaning entry, which goes against the assumption of a conversational implicature. However, lá is not paid detailed attention to in these studies.

What is needed for a better understanding of the meaning of lá is an elaborated account of its functional spectrum that can be encountered in synchrony, as well as a careful delimitation of its conventional semantics and its pragmatic potential. Therefore, the goal of this contribution is to describe its meaning building on the framework of multidimensional semantics as discussed in Gutzmann (2013, 2015), a successor of Potts (2005)’s work on the logics of conventional implicatures. As we will see, this approach provides a range of insights that serve to describe the conventional semantics of non-adverbial lá in a unified framework based on use conditions. In addition, this framework provides a way to account for the diachronic dimension underlying lá, suggesting quite a regular pragmaticalization path, which has not been studied yet to my knowledge.

The article is organized as follows. In Section 2, I briefly introduce the main characteristics of the framework of mutidimensional semantics featuring the fundamental distinction between truth-conditional and use-conditional meaning. Applying standard tests from the literature, I then identify a truth-conditional lá_tc, as well as a use-conditional lá_uc in Section 3. Building on the assumptions developed by Davis and Gutzmann (2015) concerning the semantics of pragmaticalization of use-conditional items, I extend the analysis to the diachronic dimension in Section 4 to investigate the historical development of lá_uc out of lá_tc. The Section 5 concludes the study.

2. Hybrid Semantics: Truth- and Use-Conditional Meaning

The analysis builds on the mutlidimensional semantics framework in the sense of Gutzmann (2013, 2015) and the crucial idea that the conventional meaning of a linguistic expression disposes of a truth-conditional, as well as a use-conditional dimension. This approach is heavily influenced by work on expressivity by Cruse (1986); Kaplan (1999); (Recanati 2004, p. 447); and Potts (2005, 2007), that also assume multidimensionality (cf. e.g., “two modes of expression” in Kaplan 1999). Accordingly, there is both a descriptive/propositional/truth-conditional as well as an expressive dimension in the semantics of a linguistic expression, which explains the term multidimensional semantics. Basically, the truth-conditional dimension concerns descriptive meaning contributing to truth values in the traditional sense, whereas expressive meaning is about non-truth-conditional, subjective, speaker-oriented, evaluative meaning. Expressive items may be used felicously and infelicitously, which means that they do not add anything to the truth-conditions of an expression. The prime example for expressives are slurs (3).

(3)	Example discussed in (Gutzmann 2013, p. 8).
	a.	Lessing was a German.
	b.	Lessing was a Kraut.

Both examples display the same descriptive, truth-conditional meaning, which is traditionally understood to define the conditions that need to be fulfilled so that a sentence can be rendered true. In most general terms, this is in line with Wittgenstein (1922)’s Bildtheorie des Satzes and the idea that understanding a sentence means to know what the case is if it is true. Accordingly, speakers have to know the conditions that (3-a) and (3-b) impose on the world for judging it true or false. Still, the minimal pair clearly displays a contrast in meaning, which goes beyond the truth-conditional dimension but is captured by the expressive component contained in the slur Kraut, which might be paraphrased more or less as “Generally, I do not like Germans“. If a speaker chooses to linguistically commit to this negative evaluation conventionally denoted by the slur Kraut, they utter (3-b) felicitously, regardless of the question of whether it is a true or false statement.

(4)	Multidimensional denotations for (3-b) (modified example based on Davis and Gutzmann 2015, p. 200).
	a.	Mapping to a set of worlds “Lessing was a Kraut” is true if Lessing was a German.
	b.	Mapping to a set of contexts “Lessing was a Kraut” is felicitously used if the speaker has a negative attitude towards Germans.

The multidimensional system gives rise to four possible combinations concerning the truth and use conditions of a sentence, i.e., $\langle 1,✓\rangle$, $\langle 1,↯\rangle$, $\langle 0,✓\rangle$, $\langle 0,↯\rangle$. While expressives in the narrow sense are mostly associated with negative or positive evaluations, Gutzmann (2015)’s notion of use-conditional meaning covers a broader range of kinds of subjective, speaker-oriented meaning, which turn out to share many properties with expressives (cf. also the discussion in Amaral 2018, p. 330f.), such as particles (e.g., German modal particles), certain intonation patterns, or certain morphosyntactic constructions of mirativity or topicalization. The multidimensionality of these examples can be illustrated informally by the fraction notation shown in (5).

(5)	Examples of use-conditional content complementing truth conditions and the fraction notation (Gutzmann 2013).
	a.	Conventional meaning of a linguistic expression = $\frac{\mathrm{use}\mathrm{conditions}}{\mathrm{truth}\mathrm{conditions}}$.
	b.	Lessing was a Kraut = $\frac{\mathrm{Generally},\mathrm{I}\mathrm{do}\mathrm{not}\mathrm{like}\mathrm{Germans}}{\mathrm{Lessing}\mathrm{was}\mathrm{a}\mathrm{German}}$
	c.	Hans-i (German) = $\frac{\mathrm{Familiar}\mathrm{relationship}\mathrm{between}\mathrm{the}\mathrm{speaker}\mathrm{and}\mathrm{Hans}}{\mathrm{Hans}}$.
	d.	John, Mary loves. = $\frac{\mathrm{The}\mathrm{speaker}’\mathrm{s}\mathrm{mental}\mathrm{representation}\mathrm{of}\mathrm{John}\mathrm{is}\mathrm{active}}{\mathrm{Mary}\mathrm{loves}\mathrm{John}}$.
	e.	Carl did finish their book. = $\frac{\mathrm{It}\mathrm{is}\mathrm{true}\mathrm{that}\mathrm{Carl}\mathrm{finished}\mathrm{their}\mathrm{book}}{\mathrm{Carl}\mathrm{finished}\mathrm{their}\mathrm{book}}$.
	f.	How tall Michael is! = $\frac{\mathrm{It}\mathrm{is}\mathrm{unexpected}\mathrm{how}\mathrm{tall}\mathrm{Michael}\mathrm{is}}{Ø}$.
	g.	Ouch! = $\frac{\mathrm{I}’\mathrm{m}\mathrm{in}\mathrm{pain}}{Ø}$.

In (5-c), it is the social meaning of a nickname that can be used felicitously or infelicitously without changing its reference; (5-d) suggests linking the information structure to use conditions based on the idea that the marked word order does not describe but expresses different things, e.g., that the speaker’s mental representation of John is active in the respective topicalization. In (5-e), the idea is that the verum focus associated with “did” may be used (in)felicitously without interfering with the sentence’s truth value. Finally, (5-f) shows that there are also expressions that completely lack a truth-conditional component, meaning that they are not mixed but purely use-conditional items, which also applies to (5-g).

In the tradition of Potts (2005)’s seminal book, there has been a lot of subsequent work on the main properties of expressives that led to the development of linguistic tests for expressive items. Partly, these have been controversially discussed (cf. overview in Amaral 2018, p. 327ff. and the second issue of volume 33 of the journal Theoretical Linguistics). Still, they provide a solid base to test for expressivity and, in fact, have been shown to apply to use-conditional meanings similarly (cf. Gutzmann 2013, p. 31ff.). Among the main properties of use-conditional items are independence, nondisplaceability, immediacy, descriptive ineffability, and perspective dependence. I will turn back to these in more detail in the analysis of lá_uc in Section 3.2.

3. Two Semantic Configurations of European Portuguese lá

In this section, European Portuguese lá is described based on the multidimensional semantics framework introduced above. I argue that there are two configurations, namely a purely truth-conditional lá_tc and a purely use-conditional lá_uc. In Section 3.3, I comment on whether there might actually also be a third type: a mixed type denoting both truth- and use-conditional content at the same time. In fact, there is a candidate, namely a particular use of lá which in the literature is often referred to as a marker of meta-linguistic negation (cf. Martins 2014, 2021; Schwenter 2016). Although the detailed study of metalinguistic negation with lá within a multidimensional framework remains for future work, I develop the hypothesis in Section 3.3 that it represents just another pragmatic outcome of lá_uc as discussed in the following.

3.1. The Truth-Conditional Configuration: lá_tc with Deictic Meaning

The truth-conditional lá_tc arguably constitutes the basic meaning of lá. The term basic intends to account for that (i) it is commonly mentioned in the first place in lists of uses of lá, as, e.g., in the Dicionário da Língua Portuguesa by the Academia das Ciências de Lisboa (cf. footnote 1), and (ii) non-adverbial lá is commonly understood as derived; i.e., it has a somewhat secondary meaning in the previous literature, as we will see below.

(6)	a.	Foi para a cidade e por lá ficou. ‘He went downtown and there he stayed.’
	b.	Compraram um terreno e lá edificaram a sua casa. ‘They bought a plot of land and built their house there.’

In these examples, lá_tc functions as an adverbial to designate a location relative to the speaker origo. In particular, it signals a location which, according to the standard norm, is proximate to neither the speaker nor the addressee (Costa and Martins 2010, p. 61). Close synonyms of lá are aí and ali, featuring subtle differences regarding the parameters of proximity between speaker and addressee.2

(7)	Spatical deictic expressions (Costa and Martins 2010, p. 61).
	a.	aqui/cá ‘here’ [+ close to speaker].
	b.	aí ‘there’ [– close to speaker, + close to addressee].
	c.	ali/lá ‘there’ [– close to speaker and addressee].

In (6), lá_tc clearly contributes to the conditions that render the utterances true, which are evaluated based on the agreement with the extralinguistic world’s facts. If in (6-a), they went downtown but then returned, (6-a) is false. Likewise, the location of the construction of the house is crucial for deriving the truth value of the whole expression in (6-b). Applying the fraction notation of the framework of hybrid semantics introduced in the previous section, we can informally describe the meaning of lá_tc by means of the formula indicated in (8), which shows a non-occupied use-conditional dimension.3

(8)	Multidimensional semantics of látc in the fraction notation.
	a.	látc = $\frac{Ø}{\left(\mathrm{over}\right)\mathrm{there}}$.
	b.	Foi para a cidade e por lá ficou. = $\frac{Ø}{\mathrm{she}\mathrm{went}\mathrm{downtown}\mathrm{and}\mathrm{there}\mathrm{she}\mathrm{stayed}}$.

3.2. The Use-Conditional Configuration: lá_uc and Illocutionary Modification

Although non adverbial-uses of lá have not been identified as expressing use-conditional content in previous work, their functions have been described in several studies. A first crucial aspect concerns diamesic variation, namely that there is a restriction to oral, informal registers, an observation already made by Ali ([1930] 1971) and maintained in recent studies. As for the meaning contribution, (Duarte 2010, p. 180) speaks of marqueurs d’aiguillage communicatif (’communicative referral markers’) but also alludes to that they may count as modal particles in the tradition of Franco (1991) (cf. also Meisnitzer 2012; Meisnitzer and Müller 2024). Further, she qualifies their type of meaning as either of a conventional or a generalized conversational implicature in the sense of Grice (1975). The list of functions associated with lá contains distance in time/uncertainty (la distance dans le temps/l’incertitude), reinforcement of directives (le renforcement de l’injonction), attenuation, and fulfillment of expectations (l’accomplissement des attentes). In Marques and Duarte (2015), this list is reduced to the main functions of mitigation and reinforcement (strengthening) featuring specific interactions with particular speech acts and potentially further pragmatic factors.

The present analysis builds on these assumptions, though shifting from pragmatics to semantics, claiming that lá_uc carries the conventional use-conditional meaning of illocutionary modification, as illustrated in the fraction in (9).

(9)	látc = $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$

I follow (Gutzmann 2015, p. 168) and the distinction of sentence mood on the semantic level (a reflection of syntactic sentence type) and illocutionary force on the pragmatic level, as discussed in (Grewendorf and Zaefferer 1991, p. 270):4

(10)	sentence mood	illocutionary force
	declarative	assertion
	interrogative	question
	imperative	directive
	exclamative	exclamation

In the tradition of speech act theory, illocutionary force thus concerns the pragmatic level. This means that lá_uc operates right at the semantics–pragmatics interface in that it carries the conventional semantics of modifying on the pragmatic level. In line with (Hengeveld 2004, p. 1192), illocutionary modification does not however concern the identification of a certain speech act. Put differently, lá_uc does not conventionally contribute to the creation of indirect speech acts. Instead, illocutionary modification is part of a general strategy serving to “reinforce or mitigate the force of the speech act”. In particular, this means that assertions, questions, directives, and exclamations can be reinforced or mitigated in order to manipulate their degree of assertivity, directivity, etc. Although the concepts of mitigation and reinforcement might be objected to for their pre-theoretical and intuitive nature (cf. Schwenter 2003, p. 1026), they are broadly used in research, enabling onomasiological comparisons in a cross-linguistic perspective (cf. also the discussion in Favaro 2023, p. 184). In this vein, there has been extensive work on the range of linguistic strategies associated with the mitigation (sometimes also labelled as attenuation, downtoning) and reinforcement (also boosting, strengthening) of speech acts (Holmes 1984; Sbisa 2001; Schneider 2010). According to (Holmes 1984, p. 348), speakers choose to attenuate or boost certain speech acts either to express their “attitude to the content of the proposition” or to express their attitude “to the addressee in the context of utterance”, serving the higher goal of increasing either solidarity or social distance between speaker and hearer, very much in the spirit of politeness theory (cf. Brown and Levinson 1978). Note that this comes quite close to descriptions of expressivity, where “a speaker can establish a relation of intimacy with a hearer merely by choosing one lexical item rather than another in the course of a conversation” (Cruse 1986, p. 285). Although cross-linguistic research on illocutionary modification has often focussed rather on strategies of mitigation than on reinforcement, perhaps due to the “negative-politeness cultures“ prevalent in the anglophone academic community (cf. Holmes 1984, p. 348), both are symmetrical and “of the same nature, since both are to be explained as variations in the illocutionary effects of speech acts“ (Sbisa 2001, p. 1811). Put differently, they represent two sides of a coin, two pragmatic outcomes that trace back to the basic semantic instruction of illocutionary modification.

In terms of descriptive data for lá_uc, Pereira (2009) provides an elaborated list of examples together with descriptions of the situations and paraphrases of the meaning effect as verified by minimal pairs manipulating the presence of lá. Starting with imperatives (11), it is illustrated that lá_uc’s illocutionary modification may swing towards both reinforcement and mitigation.

(11)	láuc in imperatives (Examples from Pereira 2009, p. 87).
	a.	[À mesa, diz a mãe para o filho:]
		(i)	Acaba lá de comer a sopa! ‘(Just) eat the soup!’ ∼ Reinforcement of directive speech act, signaling impatience.
		(ii)	Dá-me lá a colher que eu ajudo-te. ‘(Just) pass me the spoon, I will help you.’ ∼ Mitigation to turn an order into a request.
		(iii)	Ó filho, vê lá se comes a sopa para seres um homem. ‘Son, make sure you eat enough soup to be a man’ ∼ Reinforcement (with a slight threat.)

Since one could think of contexts leading to opposite effect, respectively, it becomes clear that it is a pragmatic issue how lá_uc’s illocutionary modification is exactly to be resolved. In this regard, (Marques and Duarte 2015, p. 123) point out that intonation also impacts whether there is an effect of attenuation or reinforcement in directives (without identifying typical intonational patterns). Furthermore, they identify politeness strategies in the sense of Leech (1983) as an underlying motivation for using lá_uc in directives. This is in line with the general observation that politeness is often mentioned as one (but not the only) reason for linguistic attenuation, as it contributes to efficiency in terms of achieving communicative goals (cf. Briz and Albelda 2013, p. 292). Using the fraction notation again, (11-a-i) can be understood as shown in (12). The perfectly grammatical and acceptable Acaba de comer a sopa!, conveying a directive speech act, is combined with the illocutionary modifier lá_uc. This composition outputs a multidimensional expression that is felicitously used if and only if the speaker wants to manipulate the degree of directivity. In the given context, it is most likely pragmatically resolved as an impatient reinforcement.

(12)	Acaba lá de comer a sopa! = $\frac{Ø}{\mathrm{acaba}\mathrm{de}\mathrm{comer}\mathrm{a}\mathrm{sopa}!}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$ = $\frac{\mathrm{illocutionary}\mathrm{modification}}{\mathrm{acaba}\mathrm{de}\mathrm{comer}\mathrm{a}\mathrm{sopa}!}$

Withregard to declaratives, Marques and Duarte (2015) refer to mitigation (they use the term attenuation) that leads to the effect of epistemic distance or uncertainty and potentially to a negative evaluation of the underlying proposition for reasons of self-protection. In the same direction, Marques and Duarte (2017) refer to the concept of emotional deixis as derived based on the original spatial deictic item (cf. also Lopes 2019 for emotional deixis as a broader category, e.g., also to be found with personal deictics).

(13)	láuc in declaratives (Pereira 2009, p. 91).
	a.	Tem estado muito doente, mas ontem lá se sentou à mesa. ‘He’s been very ill, but yesterday he sat at the table.’
	b.	O João tem estado muito doente, mas ontem, a muito custo ou com muito esforço, conseguiu tomar a sua refeição à mesa, o que indicia que o seu estado de saúde já não é tão grave como antes. ‘João has been very ill, but yesterday, with great difficulty or effort, he managed to eat their meal at the table, which indicates that their state of health is no longer as serious as before.’
	c.	Só acabarei este trabalho lá para Setembro. ‘I will not finish this job until (approximately) September.’

Pereira (2009) provides (13-b) as a paraphrase of (13-a), indicating that lá contributes to a contrast, since a very sick person is commonly not expected to sit down at a table. The mitigation effect of lá_uc thus serves to weaken the assertion of p tem estado muito doente ultimamente. In terms of (Favaro 2023, p. 132), it “reduce[s] the level of assertivity”, i.e., the degree of the speaker’s commitment to what is conveyed by p (cf. also Shapiro 2020 on assertion and commitment). Mitigating the seriousness of the message, there is a flavor of “let us not worry too much, it will be alright”. While there might be a way to express the same meaning in (13-a) via certain intonational contours, it is lexically conveyed by lá_uc in (13-b).

(14)

Tem estado muito doente, mas ontem lá se sentou à mesa. = $\frac{Ø}{\mathrm{Tem}\mathrm{estado}\mathrm{muito}\mathrm{doente},\mathrm{mas}\mathrm{ontem}\mathrm{sentou}\text{-}\mathrm{se}à\mathrm{mesa}.}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$ = $\frac{\mathrm{illocutionary}\mathrm{modification}}{\mathrm{Tem}\mathrm{estado}\mathrm{muito}\mathrm{doente},\mathrm{mas}\mathrm{ontem}\mathrm{sentou}\text{-}\mathrm{se}à\mathrm{mesa}.}$

Another note on this example concerns the presence of mas, which clearly acts as a trigger of a conventional implicature in the classical sense of Grice (1975), also indicating some kind of contrast between the two propositions p and q. Although leading to a related meaning effect, mas and lá_uc act separately.

Concerning (13-c), Pereira (2009) argues that lá is not unequivocally disjoined from a local (i.e., temporal) interpretation, which underlines the polysemy of lá and the associated high degree of ambiguity obtained in many examples also mentioned by (Marques and Duarte 2015, p. 121). However, Pereira ascribes lá in (13-c) a modalizing meaning that conveys the speaker’s uncertainty about the date, which thus turns to be perceived as imprecise, resulting again in a reduced level of assertivity. Crucially, this effect is lost when omitting lá_uc. In fact, Marques and Duarte (2017) claim that this mitigation effect in declaratives oftentimes coincides with indefinite NPs or other linguistic markers of vagueness, such as hesitation, creating epistemic distance, both illustrated in (15).

(15)	E depois e depois [hesitação], pronto, uns acontecimentos lá, que [hesitação] não gosto muito de coisas trocadas […]. ‘And then there was [hesitation], a few events there, [hesitation], I do not really like things being changed.’

In questions, the data suggest that lá_uc’s use condition of illocutionary modification uniquely creates reinforcement that leads, e.g., to exclamative-like questions (16-a). Note that this might suggest that lá_uc has the potential to create indirect speech acts, e.g., to turn a question into an order (16-b). However, omitting lá_uc in the examples in (16) verifies that this is not the case, as indirect speech acts are also easily created without it. This means that lá_uc bears the potential to contribute to the creation of an indirect speech act together with a range of other factors, e.g. intonation. However, its function is not to specify but to mitigate or reinforce the illocutionary force. Regarding the missing mitigation, it remains unclear if these actually match with questions at all (similarly to exclamatives, cf. below), which might explain the absence of examples.

(16)	láuc in questions (Pereira 2009, p. 93).
	a.	Então, como vai lá isso?! ‘So, how’s it going?’
	b.	Podes lá parar com isso?! ‘Will you stop it?!’
	c.	E que tem lá isso? ‘What is up with that?’

Last but not least, Pereira (2009) discusses exclamatives and idiomatic expressions. Since exclamatives do not carry truth-conditional meaning per definitionem, they obviously represent prime examples of expressive/use-conditional meaning. In this sense, lá_uc acts again as an item of reinforcement, as the idea of mitigating an exclamation seems to be somewhat contradictory. Considering (17-b), Pereira (2009) ascribes lá_uc the emotional expression of surprise and admiration for what is perceived.

(17)	Exclamatives (Pereira 2009, p. 89).
	a.	Vá lá, podia ter sido pior! ‘Come on, it could have been worse!’
	b.	Eh lá! Que catástrofe! ‘Hey! What a catastrophe!’
		(i)	*Lá! Que catástrofe! (constructed by LM)
		(ii)	Eh! Que catástrofe! (constructed by LM)

Pereira further analyzes eh as reinforcing lá (instead of the other way round). Given that lá alone does not represent a well-formed exclamation though (17-b-i), which eh does (17-b-ii), it rather seems that it is eh that sets up the exclamation, which then is boosted by lá_uc, as illustrated in (18).

(18)	Eh lá! = $\frac{\mathrm{emotional}\mathrm{involvement}}{Ø}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$ = $\frac{\mathrm{emotional}\mathrm{involvement}+\mathrm{illocutionary}\mathrm{modification}}{Ø}$

Finally, there are idiomatic constructions that typically host lá. They share that they pass lá to highly subjective, speaker-oriented constructions, e.g., seja quem for (19-b), again with a function of illocutionary modification, as shown in (20). Given that these are easily amenable to expressivity per se, the conventionalization of lá_uc at some point in diachrony is plausible and should be reconstructable based on the account discussed here and particularly in Section 4, which I leave for future study. Concerning the type of illocutionary modification, they can be seen as declaratives whose degree of assertivity is boosted.

(19)	Idiomatic constructions (Pereira 2009, p. 94).
	a.	Já lá vai o tempo em a família se reunia ao serão. ‘Gone are the days when the family would get together in the evening.’
	b.	Seja lá quem for deve ser castigado pelo acto que cometeu ‘Whoever it is must be punished for the act they have committed. ’
	c.	Seja lá como for deve ser castigado pelo acto que cometeu. ‘Whatever the case, he should be punished for the act he committed.’
	d.	Sabe lá Deus se chegamos amanhã. ‘God knows if we’ll arrive tomorrow.’
	e.	Não penses mais no que te aconteceu. O que lá vai lá vai. ‘Do not think about what happened to you anymore. What goes around comes around.’

(20)	Seja lá quem for = $\frac{Ø}{\mathrm{seja}\mathrm{quem}\mathrm{for}}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$ = $\frac{\mathrm{illocutionary}\mathrm{modification}}{\mathrm{seja}\mathrm{quem}\mathrm{for}}$

Summing up, lá_uc is felicitously used if a speaker wants to express illocutionary modification that is pragmatically resolved in terms of mitigation or reinforcement. This holds across several sentence moods, as summarized in

Table 1. Illocutionary modification with lá_uc and sentence mood.

	Mitigation	Reinforcement
imperative	✔	✔
declarative	✔	✔
interrogative	?	✔
exclamative	◼	✔

Question mark means “no example encountered” and black square means “contradictory”.

. This generalization is restricted though to the extent that mitigation turns out to be somewhat contradictory to the meaning of the exclamative speech act. It remains furthermore unclear if interrogatives can be mitigated by means of lá_uc.

Testing for Common Properties of Use-Conditional Items

There are some typical properties of expressives suggested by (Potts 2007, p. 166f.) that help identify truth- and use-conditional meaning and that have been discussed a lot subsequently (cf. e.g., Gutzmann 2013, p. 31 and the overview provided in Amaral 2018). In line with the claim that non-adverbial uses of lá carry use-conditional meaning, they should pass the typical tests listed in (21).

(21)	Properties of expressives proposed by (Potts 2007) (brackets indicating partly problematic criteria, cf. Gutzmann 2013; Amaral 2018).
	a.	Independence.
	b.	Nondisplaceability.
	c.	Perspective dependence.
	d.	(Descriptive ineffability).
	e.	Immediacy.
	f.	(Repeatability).

To start with, the property of independence covers that expressive content contributes a meaning dimension separate from the regular truth-conditional content, meaning that the descriptive, truth-conditional dimension imposes truth conditions, whereas the use-conditional dimension imposes felicity conditions that render an utterance felicitous or infelicitous instead of true or false. This can be tested by comparing a sentence with and without a purported use-conditional item and thus reflects the procedure in Pereira (2009). Omitting lá_uc should not affect the descriptive meaning of a sentence. Related to this, it should not be possible to negate lá_uc by ordinary negation (22-a-i), to directly deny it in dialogue (22-a-ii), nor to target it in an interrogative (22-a-iii). A way to object to content imposed by lá_uc is exemplified in (22-a-iv), where the só que… construction (‘it is just that …’) provides an indirect way to refuse lá_uc. In line with Longhin-Thomazi (2003), só que represents a conjunctional periphrasis that adds to a preceding statement a new circumstance that breaks with some of the assumptions shared by the speakers, leading to meaning effects of refusal, surprise or counter argument. In (22-b), where the preceding directive lacks lá_uc, the só que construction is highly marked if not odd, given that it is unclear what exactly the objection is supposed to refer to.5

(22)	a.	A: Acaba lá de comer a sopa! (with intensification reading). ‘(Just) finish your soup!’
		(i)	B: #Não, mas já vou fazê-lo. ‘No, but I will do it.’
		(ii)	B: #Já vou fazê-lo, mas não lá. ‘I will do it right away, but not there.’
		(iii)	B: #Como/onde é que devo fazê-lo? Lá? ‘Where am I supposed to do it?’
		(iv)	B: Está bem, só que vou fazê-lo com calma. ‘Alright, but I’m going to do it slowly.’.
	b.	A: Acaba {ø} de comer a sopa! B: ??Está bem, só que vou fazê-lo com calma. ‘A: (Just) finish your soup! B: Alright, but I’m going to do it slowly.’

The property of nondisplaceability states that expressives/use-conditional items predicate something of the utterance of the situation, meaning that they cannot be shifted away from the utterance situation. While descriptive, truth-conditional content can be easily dissociated from the situation of the actual utterance, e.g., by means of past or conditional operators, expressives are bound to the utterer and the time and place of the utterance (cf. Cruse 1986, p. 272). Yule (2006, p. 9) compares this immediacy of expressives to animal language, exemplified by a dog whose expressive interjection “Grrr” means “Grrr right now” and not “Grrr last night, over in the park”. The property of nondisplaceability also proves to be a good fit for the examples in (23), as the listed imperatives are strictly tied to the speakers origo. Trying to shift lá_uc in a directive speech act to the past (23-a) or to enter it in the scope of a universal quantifier and the subjunctive mood (23-b) results in oddity. A repair strategy might be to force a reading, i.e., where lá_tc adverbially modifies the location of the eating event.

(23)	a.	Eu lembro-me de quando me ligaste. Foi ontem ao meio dia quando eu pedi para o meu filho acabar [*láuc/látc] a sopa. ‘I remember when you called me. It was yesterday at noon when I asked my son to finish their soup.’
	b.	Sempre que peço ao meu filho que acabe [*láuc/látc] de comer a sopa, ele come tudinho. ‘Whenever I ask my son to finish their soup, he eats it all.’
	c.	Se alguém estivesse muito doente mas [*láuc/látc] se sentasse à mesa, era surpreendente. ‘If someone was very ill but sat down at the table, it was surprising.’

The property of perspective independence is quite similar to the nondisplaceability condition and states that expressive or use-conditional content is evaluated from a particular perspective. In general, the perspective is the speaker’s. Perspectival shifts, e.g., initiated by reported speech or attitude reporting verbs, as in (24-a) and (24-b), clash with the strong association of use-conditional content with the immediate speaker’s origo, although there can be deviations if conditions are right (cf. Amaral 2018, p. 329). In fact, (24-c) may mark such an exception, since an anonymous reviewer points out that it is acceptable in the lá_uc reading. This does not exclude the possibility of a nonshifted interpretation though, where lá_uc is still attributed to the speaker’s perspective. Concerning future studies, this suggests investigating if sentence mood impacts the shiftability of lá_uc, i.e., if it is shiftable indeed when occurring in declaratives.

(24)	a.	A tua tia é muito severa, sabias? Pede ao nosso filho que acabe [*láuc/látc] de comer a sopa. ‘Your aunt is very strict, you know. She asks our son to finish their soup.’
	b.	[Mãe: Acaba lá de comer a sopa! Filho: Não percebi, o rádio está muito alto.] Pai: A tua mãe quer que acabes [*láuc/látc] de comer a tua sopa. ‘Finish your soup! Son: I do not understand, the radio is too loud. Dad: Your mother wants you to finish your soup.’
	c.	O João tem estado muito doente, mas o Pedro disse que ontem [láuc/látc] se sentou à mesa. ‘João has been very ill, but Pedro said he sat at the table yesterday.’

The property of descriptive ineffability states that “speakers are never fully satisfied when they paraphrase expressive content using descriptive, i.e., non-expressive, terms” (Gutzmann 2013, p. 42). In fact, this criterion has received a lot of criticism. An important issue raised by Geurts (2007, p. 210) is that there are also truth-conditional items with meaning content “suffering” from descriptive ineffability, such as “the” or “because” (cf. also Amaral 2018, p. 328); i.e., this cannot be unique to expressive/use-conditional meaning. In fact, Pereira (2009) actually does provide paraphrases for some of the examples discussed above. However, it is easy to notice that these ponderous explanations, complemented by further clarifications concerning the context, confirm that they are not to be understood as semantically equivalent to what is conveyed by lá_uc.

The last criterion, the one of immediacy, covers that expressives/use-conditional items, “like performatives […], achieve their intended act simply by being uttered“ (Gutzmann 2013, p. 44). Put differently, while truth-conditional meaning can be revised in certain ways, use-conditional meaning can hardly be taken back, and its effect certainly evolves once the underlying use-conditional item has been uttered. As alluded to by (Potts 2007, p. 180), immediacy is related to the non-deniability in discourse discussed above and also applies to lá_uc.

In sum, the discussed data support the claim that non-adverbial lá can be understood in terms of the use-conditional item lá_uc.

3.3. A Note on Meta-Linguistic Negation

As mentioned above, here is a brief note on certain uses of lá that are excluded from the analysis.

(25)	(Marques and Duarte 2015, p. 118).
	a.	Sei lá. ‘I do not know.’
	b.	Quero lá saber. ‘I do not want to know.’

Regarding lá’s meaning contribution in these examples, (Marques and Duarte 2015, p. 118) allude to its negating function, which is also conveyed by the English translations ‘I do not know’ and ‘I do not want to know’, complemented furthermore by an axiological distance between the speaker and content, i.e., again by illocutionary modification as discussed in the previous subsection. In fact, this might actually suggest a status of a mixed item in terms of Gutzmann (2015), featuring both truth-conditional (negation) and use-conditional content (illocutionary modification):

(26)	Sei lá = $\frac{Ø}{\mathrm{sei}}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{\neg }$ = $\frac{\mathrm{illocutionary}\mathrm{modification}}{\neg sei}$

In the sense of (Martins 2014, p. 638), these uses are described as giving rise to metalinguistic negation on merely pragmatic grounds in the tradition of Horn (1989):

(27)	A: Estás um pouco preocupado?—B: Estou lá um pouco preocupado, estou morto de preocupação. ‘A: Are you a little worried?—B: I’m not a little worried, I am worried sick’

(Horn 1989, p. 363) defines metalinguistic negation as an objection of a previous utterance by not focusing on its truth conditions but on its assertability. In this vein, lá in (27) does not negate the truth conditions of the previous utterance but objects its understatement, as it prepares an utterance that is actually meant to reinforce it. According to this view, lá does clearly not carry a negation operator in its truth-conditional meaning component. Instead, the negating effect is a product of a pragmatic inference that furthermore does not operate on the descriptive meaning. Conceiving of this implicature as a conventional one, metalinguistic negation might be understood as an own use-conditional meaning component as shown in (28).

(28)	estou lá um pouco preocupado = $\frac{Ø}{\mathrm{estou}\mathrm{um}\mathrm{pouco}\mathrm{preocupado}}$ + $\frac{\mathrm{metalinguistic}\mathrm{negation}}{Ø}$ = $\frac{\mathrm{metalinguistic}\mathrm{negation}}{\mathrm{estou}\mathrm{um}\mathrm{pouco}\mathrm{preocupado}}$

Another option though would be to understand metalinguistic negation as just another pragmatic outcome of illocutionary modification (29). Based on the conception of emotional deixis, lá accordingly creates a very strong mitigation effect of subjective distance. This could be understood as leading to a tremendously low degree of assertivity/commitment to what is said, creating thus metalinguistic negation pragmatically.

(29)	estou lá um pouco preocupado = $\frac{Ø}{\mathrm{estou}\mathrm{um}\mathrm{pouco}\mathrm{preocupado}}$ + $\frac{\mathrm{illocutionary}\mathrm{modification}}{Ø}$ = $\frac{\mathrm{illocutionary}\mathrm{modification}}{\mathrm{estou}\mathrm{um}\mathrm{pouco}\mathrm{preocupado}}$

Assuming (29) would mean that there is nothing conventional about metalinguistic negation (the conventional underlying meaning would be again illocutionary modification). In fact, this is supported by (30), where an interpretation of ‘God does not know’ is clearly ruled out. This suggests a merely pragmatic nature of metalinguistic negation indeed, derived on the grounds of the semantically stored illocutionary modification.

(30)	Sabe lá Deus se chegamos amanhã. (Pereira 2009, p. 94) ‘God [knows/*does not know] if we’ll arrive tomorrow.’

In sum, it seems plausible to treat metalinguistic negation as an implicature derived from lá_uc’s use-conditional meaning of illocutionary modification. Since this assumption calls for more substantiating research though, e.g., also concerning the exact type of implicature, I leave the intriguing desideratum of a proper multidimensional treatment of metalinguistic negation for future research.6

4. Lá in Diachrony

4.1. A Case of Pragmaticalization: From lá_tc to lá_uc

In this section, the study of lá is approached from a diachronic perspective, based on the working hypothesis that lá_uc historically evolved from lá_tc. In fact, non-adverbial uses of lá have also been discussed in the previous literature as a result of a derivation process. Though not explicitly stated, they seem to be understood as a product of a synchronic, ad hoc creation made in conversation. Put differently, the assumption is that there is only one lexical entry for lá such that lá_uc is conceived of as a result of an implicature-based inference. What drives this process is assumed to be an inferred transfer of lá_tc’s spatial distance to the domain of subjective distance, an operation whose result is also known under the general label of emotional deixis (cf. Lakoff 1974; Potts and Schwarz 2010; Aguiar and Barbosa 2023, cf. also Lopes 2019 and the discussion of further Portuguese examples of emotional deictics, like cá (‘here’) and aí (‘there’)). In a broader research context, assuming one and only lexical entry for lá relates to the monosemy vs. polysemy (or minimalist vs. maximalist) problem (cf. e.g., Hansen 1998, p. 85 and Waltereit 2006, p. 8 in the context of discourse/modal particles). This minimalist view seems to be particularly in line with the high degree of ambiguity obtained in many examples of lá, which oftentimes makes it also hard to distinguish lá_uc from lá_tc in linguistic analysis for native speakers (cf. Marques and Duarte 2015, p. 121). On the other hand, there is a great body of research on the diachronic rise of discourse/modal particles (cf. e.g., Detges and Waltereit 2007) assuming a historic process leading up to synchronic polysemy, i.e., a reflection of diachronic change. A guiding principle in this process has been identified in terms of the subjectification thesis and the related concept of pragmaticalization (Traugott 1989, cf. also Diewald 2011).7 In (Traugott 1989, pp. 34–35), three general tendencies of subjectification are listed. Among these are the typical process from external (description of situation) to internal (evaluative/perceptual/cognitive) and the development of expressing “the speaker’s subjective belief state/attitude toward the propositon“, that qualify to also apply to a diachronic formation of lá_uc. Within multidimensional semantics, pragmaticalization processes are assumed to typically evolve from descriptive/truth-conditional to expressive/use-conditional content (cf. Davis and Gutzmann 2015; Amaral 2018, p. 331), which the present analysis is built on. Accordingly, the process of pragmaticalization is understood as a diachronic type shift from truth to use conditions, illustrated in (31) with examples in English and German.

(31)	Pragmaticalization as a type shift (Davis and Gutzmann 2015, p. 203).
	a.	A : $\langle \sigma ,t\rangle$ > Aex : $\langle \sigma ,u\rangle$
	b.	Descriptive nouns > expressives (Davis and Gutzmann 2015, p. 198).
		(i)	Boor ‘countryman, farmer’ > ‘crude person’ (Engl.).
		(ii)	Wip ‘woman’ > weib ‘woman.PEJ’ (Germ.).
	c.	Adverbs/adjectives > modal particles (Germ., Davis and Gutzmann 2015, p. 198).
		(i)	Eben ‘flat’ > ‘just, exactly’.
		(ii)	Schon ‘already’ > ‘somewhat’.

As a fundamental building block in the approach, a two-step pattern is assumed. This means that there is an intermediate stage in the derivation of boor _exp or eben_mp, quite similar to bridging contexts in the overlap model known from grammaticalization theory (cf. e.g., Heine 2003). The diamond operator (⋄) in (32) therefore indicates a mixed use-conditional item that contains both a truth- and a use-conditional meaning dimension that mediates between the truth-conditional origin and the purely use-conditional destination:

(32)	Two-step pragmaticalization.
	a.	A > A ⋄ Aex > Aex
	b.	Boor: $\langle e,t\rangle$ > boor $\langle e,t\rangle$ ⋄ boorex : $\langle e,u\rangle$ > boorex $\langle e,u\rangle$.

The intermediate stage is typically associated with ambiguous contexts oscillating between an old and a new meaning, before the new one gets stored as conventionalized. In more detail, this process is understood as follows. A purely descriptive/truth-conditional item starts to generate expressive meaning at some point by means of a conversational implicature that is generated in particular contexts and situations (Stage I). Once there is enough of a regularity, the conversational implicature turns into a conventionalized one, stored now as conventional use-conditional content that complements the truth-conditional dimension (Stage II). Finally, the truth-conditional meaning component may be lost, creating a purely use-conditional item, as is the case with, e.g., “boor” (Stage III). However, in case of what the authors call pragmatic fission (33), the third stage may also split into a purely truth-conditional item restored along with the use-conditional item, creating lexical ambiguity.

(33)	Pragmatic fission (Davis and Gutzmann 2015, p. 204). $A>A\diamond A_{\mathrm{ex}}>\left\{\begin{array}{c}A\hfill \\ A_{\mathrm{ex}}\hfill \end{array}\right.$

Applied to the present study, this is a promising scenario to understand the process that lá has passed through diachronically and that we keep as a working hypothesis for the corpus study discussed in the next section.

(34)	Possible pragmaticalization path of lá. $lá>{lá}_{\mathtt{T}\mathtt{C}}\diamond {lá}_{\mathtt{U}\mathtt{C}}>\left\{\begin{array}{c}{lá}_{\mathtt{T}\mathtt{C}}\hfill \\ {lá}_{\mathtt{U}\mathtt{C}}\hfill \end{array}\right.$

4.2. Corpus Study

For the corpus query, the historical section of the Corpus do Português was consulted Davies (2006). Based on the working hypothesis that lá’s pragmaticalization path can be described with (34), I particularly hypothesized the following stages.

• Stage I:
  The original morpheme lá carries only truth-conditional content and serves to modify the location of a predicate. A conversational implicature of illocutionary modification may be generated in certain contexts, leading to mitigation or reinforcement effects as discussed in Section 3.2.
• Stage II:
  In certain contexts, the implicature of illocutionary modification is conventionalized, leading to a hybrid, i.e., two-dimensional expression.
• Stage III:
  The hybrid expression splits into two homophonous expressions, one encoding original truth-conditional content and another encoding use-conditional content.

The related research questions that the query aims at can be formulated as follows.

• What are the time frames corresponding to the stages formulated above?
• What are typical onset contexts for (some of) the stages, i.e., potential linguistic properties promoting the development?

Regarding the corpus queries, the procedure was as follows. First, a general overview of the frequencies of lá in diachrony was generated, which is illustrated in

Table 2. Total occurences of lá in the Corpus do Português per century (Davies 2006).

	16th	17th	18th	19th	20th
Tokens	1644	977	305	10,499	16,170
Size	4,435,031	3,407,741	2,234,951	10,008,622	20,777,725
Tokens/size (*1000)	0.37	0.29	0.03	0.17	1.05

. It shows the raw frequencies of the lexical item lá per century. Given that the subcorpora for each century are not balanced in terms of their size, I calculated ratios by dividing the number of tokens by the corpus size in order to compare them.8 The subcorpus for the 20th century is furthermore annotated according to diatopic variation, based on the distinction of European (EP) and Brazilian Portuguese (BP), as well as for register, based on the distinction of academic, newspaper and fictional texts, as well as spoken language. This means that the subcorpora for the 16th–19th century contain mixed data in terms of variation, which impedes filtering exclusively for EP data.9 In order to maintain the comparability across the columns, the query for the 20th century was not restricted either. In this vein, the data presented in Table 2 provide a general overview of the most important trends in the life cycle of lá in Portuguese.10

Before describing the findings, another note of clarification is needed. The search of instances of lá is based on the spelling lá. This is not trivial, as many occurrences in older texts are still spelled la, i.e., without an acute accent, due to the diachronic evolution of the feminine definite article of Lt. illa(m) > la > a (Azevedo 2005, p. 158). I refrained from including the spelling la in the query due to the problem of the massive quantity of definite articles in the data. In this sense, the query is based on a sample. This should not be problematic, since corpus data always represent samples.11 As can be seen in Table 2, the sample is big enough for the purposes of the study, allowing for generalizations.

Considering the data presented in Table 2, there is a decline to be noted leading from 0.37 in the 16th to a low point at 0.03 in the 18th century, before the numbers rise again, until they explode in the 20th century (1.05). This suggests an important turning point around the 18th/19th century, serving as a preparatory stage for the upcoming rise of lá. Potentially, it was not until then that speakers started to conventionally add a use-conditional meaning component to lá (Stage II). Going in that direction, the oldest examples suggesting use-conditional uses were of interest in the next step. Therefore, the search focussed on lá occurring in singular imperatives, both in formal and informal ones (e.g., tome lá!/toma lá! ’(please) take!’). The imperative construction was chosen for two reasons. First, this again meant sampling, i.e., filtering for the sentence mood of the imperative to improve the overview of the data. Second, imperatives are described to be the central sentence mood in terms of frequencies that lá_uc occurs in (Marques and Duarte 2015, p. 123), which suggests a high degree of pragmaticalization. Accordingly, imperatives should be amenable to implicatures of illocutionary modification from very early on already. For the formal imperatives, the search was based on the keyword [VS* lá] to find the most frequent constructions, complemented by manual searches for the informal imperative of these. Note that both forms are equivalent with other ones: the informal toma coheres with the third person singular indicative present tense and the formal tome with the third person singular subjunctive present tense, meaning that qualitative inspections of relevant examples are needed.

An overview of the results of the described query is shown in

Table 3. Singular imperative + lá, CDP.

	16th	17th	18th	19th	20th	EP	BP	ACA	NEWS	FICT	ORAL
VEJA LÁ		1	1	62	94	86	8		7	44	43
VÊ LÁ			1	83	85	67	18	1	1	59	24
DIGA LÁ				36	51	45	6		3	23	25
OLHE LÁ				37	40	23	17		2	20	18
OLHA LÁ		1		47	62	48	14		5	33	24
DEIXE LÁ			1	42	32	29	3			21	11
DEIXA LÁ				45	43	37	6	1	4	29	9
ANDE LÁ			1	32	6	2	4			6
ANDA LÁ	1	2		38	24	21	3		1	20	3
TOME LÁ				30	9	7	2		1	6	2
TOMA LÀ	1			36	23	21	2		5	15	3
ESPERE LÁ				17	14	14			1		13
ESPERA LÁ				10	11	11				5	6
OUÇA LÁ				7	18	18			1	1	16
OUVE LÁ				20	23	20	3			19	14
CONTE LÁ				7	14	13	1			4	10
CONTA LÁ	1			10	16	14	2			15	1
DIGA-ME LÁ				4	11	11				1	10
DIZ-ME LÁ					5	5				3	2
DESCULPE LÁ					9	9				1	8
DESCULPA LÁ					2	2			1	1
EH LÁ				5	3	2	1			3

. Similarly to Table 2, the 18th/19th century seems to be an important turning point, as there is a rise to be registered from the 19th century onwards. Frequency has actually been shown to be an important catalyst in pragmaticalization (Bybee 2003). With an increase with respect to both type and token frequency, conditions are good for a new function to be consolidated in the language system or, to put it in individual-cognitive terms, conditions are good for a new function to be entrenched, before lá_uc receives its own lexical entry at a later stage. Furthermore, the data suggest a strong focus on oral and fictional registers in the 20th century, thus confirming previous studies. The striking role of fictional data in this regard can be explained by a likely high amount of staged orality, i.e., of imitations of spoken language or language of proximity in the sense of Koch and Oesterreicher (1985).

Another crucial insight revealed when comparing the two tables lies in that the data are compatible with the idea of a gradual pragmaticalization of lá over the centuries. While the general frequencies of lá developed from 0.37 to 0.03 to 1.05 (a curve like a checkmark), this trend is clearly not reflected in Table 3 (explosion from the 18th to the 19th centuries). Crucially, this suggests ruling out the possibility that lá_tc has been available ever since as a synchronically available, secondary meaning component in the sense of the minimalist view discussed above. Otherwise, the data in Table 3 should much more resemble the trend depicted in Table 2. Instead, the use-conditional meaning component of lá must have been added in the course of the time as a result of a process of meaning expansion.

As a next step, I checked each of the examples occurring prior to the 19th century for uses of lá_uc, i.e., featuring illocutionary modification or at least a flavor thereof, suggesting either a status of Stage I, i.e., of a conversational implicature of illocutionary modification (35), or of Stage II, i.e., of a conventional implicature of illocutionary modification (36). Examples hosting clear uses of lá_tc are not listed here, but they can easily be accessed online using the platform.

(35)	Examples compatible with Stage I (conversational implicature of illocutionary modification).
	a.	Manuel Bernardes, “Nova floresta”, 1710. Eis aí te fez Cristo digno de levares a sua cruz e segui-lo; vê lá nao faças pé atrás e acudas mais pela tua carne que pela tua alma e te furte o diabo o reino da glória. ‘Behold, Christ has made you worthy to take up your cross and follow him. Do not stand back and fight more for your flesh than for your soul, and let the devil rob you of the kingdom of glory.’
	b.	Francisco Manuel de Melo, “Apolo”, 17th century. Mulher, olha lá como amaldiçoas: não toques no campanário! ‘Woman, look how you curse: do not touch the bell tower!’
	c.	Candido Lusitanoazendo, “Carta a um filólogo de Espanha”, 1750. Como v.m. o tem, veja lá se gosta daquelle timbre das armas, e timbre daquelles, por quem chora o Tejo…. ‘Since you have it, see if you like that timbre of the arms, and timbre of those for whom the Tagus weeps….’
	d.	Manuel da Costa, “Arte de Furtar”, 1645. se algum naõ tiver isto por factivel, veja lá naõ lhe provêm, que lhe succedeo a elle. ‘If anyone does not take this for granted, look at what happened to him.’

In each of these examples, lá is used together with predicates amenable to a locative semantic role, i.e., to local specification. In (35-a) and (35-b), ver and olhar (’look, see’) are easily amenable to a specification of where to look. Given that the examples furthermore connect with the idea of a subjective meaning component, the presence of a conversational implicature seems to be plausible. In (35-a), the reinforcement effect goes hand in hand with the seriousness of the religious task of not standing back. In (35-b), there is expressivity triggered by almadiçoar, also suggesting a reinforcing effect. In (35-c), the opposite function seems to in place, i.e., a mitigating effect, suggested by the socially much higher addressee of vossa mercê (‘your grace’). Example (35-d) is a bit out of competition, as it is part of a monologue, i.e., there is no direct addressee in “if somebody […]”, and it is embedded under a conditional operator se, which might be problematic with expressive/use-conditional items as discussed in Section 3. In fact, (Sambrana 2021, p. 94) analyzes this very example as a first occurrence of veja lá as a discourse marker, which suggests that it not necessarily about illocutionary modification but more about discourse organization.

Concerning the candidates for Stage II listed in (36), there are now predicates involved that do not actually call for a locative semantic role, like tomar and contar (‘(to) take, (to) tell’). As indicated by a reviewer, there is actually also a locative component contained in the semantics of tomar, as there is a movement of an item away from the speaker origo towards a goal. However, the goal of the location is exactly defined by the receiver, so there is no need for further specification, which marks the contrast with olhar and ver as discussed above in (35).

(36)	Examples compatible with Stage II (conventional implicature of illocutionary modification).
	a.	Gil Vicente, Farsa do juiz da beira, 1525. Vai-se e vem um Escudeiro com um seu Moço e diz: Toma lá esse sombreiro. ‘A squire comes along with a young man and says, ‘Take that parasol.’’
	b.	Gil Vicente, Auto da barca do purgatório, 1518. Senhor tartarugo[,] digo que mentis como bestigo, salvanor. Fala em tua menencória e não fales em passar e conta lá outra história porque em festa de tal glória não hás ninguém de levar. ‘Mr. Tortoise, I say you lie like a beast, savior. Speak in your menagerie and do not talk about passing and tell another story because you have no one to take to such a glorious feast.’

Given that these are nevertheless combined with lá seems now to be possible because of the conventionalization of the implicature of illocutionary modification, which had been previously generated as a conversational one. Although declaratives were not explicitly searched, there is one in (36-b), featuring a clearly evaluating function, i.e., reducing the degree of assertivity as discussed in Section 3.2. Future studies may thus follow up on whether lá_uc with declaratives developed after lá_uc was conventionalized with imperatives indeed. That would mean that at first, conversational implicatures (Stage I) were generated with imperatives, before in Stage II, where the conventionalization already enabled a broader range of sentence mood. Related to this, the first instances of the purely exclamative eh lá are also indicated in Table 3. They trace back to the 19th century and also suggest an emergence posterior to imperatives.

It may seem a bit counterintuitive that the discussed examples for Stages I and II do not perfectly mirror their actual chronological occurrences. In this vein, Gil Vicente appears as an innovative promoter, as he is the only author that already very early on uses lá with rather advanced functions in terms of its pragmaticalization path. Since processes of change are multilayered developments that may be promoted to different degrees in different speech communities at different times before conventionalization in the system is accomplished, this does not directly contradict the discussed assumptions. In particular, the change process from conventionalizing a conversational implicature is not to be understood as an abrupt and clearcut process that speakers suddenly obey. As noted by a reviewer, the leading role of Vicente’s theater plays may be particularly due to the contained passages featuring informality and language close to orality and the fact that the number of theater plays in the corpus decreases over the centuries. In fact, a more balanced corpus in terms of register would certainly improve the whole picture, particularly regarding the frequencies and distributions of innovative uses to see if the conventionalization proceeded to different extents in different registers at different times. In (37), examples from the 19th century are listed. They suggest a status of either Stage II or even III. Note that in the historic corpus-based methodology, it is hardly possible to identify when exactly the transfer from Stage II to III must have taken place. Using synchronic experimental methods, this could be approached by testing to what extent the typical properties of expressive/use-conditional items discussed in Section 3.2 apply to lá at a given stage. Given that there is no operationalization suggested in Davis and Gutzmann (2015) either, I leave it for further research to identify if there are other ways to distinguish Stage II and III based on diachronic corpus data. In both examples in (37), it is obvious though that there is not a potential but a certain reinforcement effect associated with lá and thus a strong subjective note of involvement by the speaker, emphasized by their impatience and joy. In sum, the 19th clearly started to host a great number of uses of pertaining to at least Stage II, whereas prior occurrences were still highly sporadic.

(37)	Stage II (and III?).
	a.	Eça de Queirós, “O Crime do Padre Amaro”, 1875. E eu por mim o que fazia, para calar toda essa gente, era casar-me já. Eu bem sei que tu não morres por ele, bem sei. Deixa lá! Isso vem depois. O João é bom rapaz, vai ter o emprego. ‘And what I’d do, to shut all those people up, is get married now. I know you will not die for him, I know that. Never mind! That’ll come later. João is a good boy, he’ll get the job.’
	b.	Almeida Garrett, “Falar verdade e mentir”, 1845. Vamos fala, conta-me lá como isso foi, quero saber tudo por meudo. ‘Come on, tell me how it happened, I want to know everything.’
	c.	Camilo Castelo Branco, “Maria Moisés”, 1876. Vá, mano, conte lá a história. ‘Come on, man, tell the story.’

4.3. Discussion

The results of the corpus analysis turn out to be compatible with the hypotheses formulated above, crucially with the claim that lá passed through a diachronic process of pragmaticalization of lá_tc > lá_uc. With respect to the first research question of the corpus study, the data show first instances of conversational and also conventionalized implicatures from the 15th century onwards and suggest a stronger conventionalization from the 19th century onwards. According to the discussed account, Stage III must have been approached at around this time. Concerning the transition from Stage II to III, it was not possible to identify a specific period using the current method, mainly due to limitations of the historical corpus approach. Regarding the second research question, the amenability of predicates to the semantic role of location turned out to serve as an indicator for the degree of pragmaticalization at a certain stage. In this sense, the analysis suggests that, at first, predicates prone to call for a locative semantic role allowed for the generation of conversational implicatures of illocutionary modification before there was an extension to predicates less amenable to specifying a location.

5. Conclusions

In this contribution, I analyzed European Portuguese lá based on a multidimensional semantic framework in the sense of Gutzmann (2013, 2015). Crucially, I showed a synchronic split into two conventionalized meaning configurations leading up to polysemy, i.e., to a truth-conditional and a use-conditional lá. While lá_tc is well known to contribute to the descriptive meaning of an utterance by adverbially modifying a location, lá_uc expresses use-conditions (or felicity conditions) that apply to particular contexts (instead of worlds). From a synchronic point of view, this claim was substantiated based on a range of diagnostic tests for use-conditional meaning discussed in the literature. Concerning the diachronic dimension, data from the Corpus do Português were presented, supporting the assumption of a two-step pragmaticalization path of lá > lá ⋄ lá_uc > lá_tc/lá_uc and resulting in pragmatic fission in the sense of Davis and Gutzmann (2015).

Regarding previous research on lá, I would like to highlight the following three achievements. First, it was possible to reduce a great part of the inventory of functions assumed for non-adverbial lá to the one of use-conditional meaning of illocutionary modification, which also holds across sentence moods. As discussed in Section 3.3, the analysis might also be compatible with metalinguistic negation (in the sense of Martins 2014), which, however, remains to be investigated in detail. As pointed out by a reviewer, another function of lá not treated in the analysis refers to it occurring in particular information structural constructions, e.g., in topicalizations. Again, it remains to be studied if examples like Lá de Jorge nada temas (‘With respect to Jorge, you do not fear anything.’) are also compatible with the analysis or not. Second, the semantic meaning contribution of lá_uc was highlighted and delimited from the pragmatic potential associated with it. Third, it was possible to explain the formation of lá_uc based on an pragmaticalization approach featuring the idea of a stepwise conventionalization.

Besides these achievements, there is a range of open questions remaining for future work. Next to the ones mentioned above, another one certainly pertains to whether there are the same synchronic effects and diachronic processes to be observed with lá in other varieties of Portuguese, as well as with other spatial deictics such as cá (‘here’) and aí (‘there’).