First, we focus on the non-subject status of jo. Catalan is a null subject language, and as such it allows subjects not to be pronounced, and be syntactically represented by a null pronoun pro. Alternatively, the subject may also be overt. The two possibilities of example (2a) are illustrated in (5a, b).5

The status of jo↑ in jo–constructions is different from subject jo because it may show up even in a left peripheral position of an impersonal existential sentence, such as (6a). In this example it is the null subject in the subordinate clause, whose first person feature is expressed in the verbal morphology, that resumes jo↑, as shown in (6b).

Jo↑ in jo–constructions may also co-occur with an overt first person subject (7). However, recall from section 1, that the first jo↑ has comma intonation, while the second jo has an integrated intonation:

The examples in (7) remind us of the so-called “double subject” construction (Duarte 1995; Kato 1999; Costa et al. 2004; Costa 2010; a.o.), which has been postulated in Brazilian Portuguese. In (8), we show that a resumed subject may occur with left-peripheral DPs – nominals, pronouns or quantifiers – that have an integrated intonation with the clause.6

It should be noted that in sharp contrast with the examples in (8), double subject constructions in Brazilian Portuguese are rare when a first person pronoun is resumed (Costa et al. 2004). However, examples are easily found in oral speech (9a), and even on the internet (9b), from a post at a site dedicated to complaints.

This construction has never been separately studied, since the literature on Brazilian Portuguese focuses on the double subject construction in (8). However, notice that the examples in (9) (see also footnote 1, ex. (ic)) need a special intonation on eu↑ ‘I’ in initial position, differently from what happens in double subject constructions. The Brazilian Portuguese examples in (10), as well as the Catalan sentences in (11), show that eu↑ / jo↑ may occur without a double subject, from which we conclude that the strong pronoun heads a different construction.

At the time of characterizing the core properties of the jo–construction, it must also be pointed out that jo↑ differs also from so-called “quirky subjects” (Sigurdsson 1992). In Romance, quirky subjects are dative (12a) or locative (12b) (Belletti & Rizzi 1988; Masullo 1993; Fernández-Soriano 1999; Rigau 1999), and it is this dative or locative subject that is assumed to satisfy the EPP feature of Tense. A sentence that contains both the dative and the locative subjects is ungrammatical (12c), thus suggesting that the dative and locative compete for the same subject position. The dative clitic subject em can be doubled with a mi (12d) (Demonte 1995), and it can even combine with jo↑ (12e), which has a default nominative case, thus supporting once more the conclusion that the peripheral first person strong pronoun is not a subject.

Before we close this section we may also consider the extent to which jo↑ could be conceived an expletive subject or topic similar to third person strong pronouns used for purely pragmatic purposes. Thus, the pronoun ele in European Portuguese is hypothesized as being an expletive element that lexicalizes Force Phrase and, interestingly, can precede an overt first person singular pronoun.

Similarly, the pronoun ell in Majorcan Catalan and the pronoun ello in Dominican Spanish have been claimed to be used as exclamative particles and discourse markers, respectively.

These expletive items share with jo↑ the fact that they are not obligatory but differ in that jo↑ (i) requires a resumptive item in the clause, (ii) is excluded with meteorological verbs, (iii) and does not precede assertions, but rather declarations, as argued in Section 3.

In this section we show that jo↑ in the jo–construction is a (kind of) hanging topic, since on the one hand it fulfills the properties attributed to hanging topics in general, but on the other hand it is restricted to first person. The jo–construction shares syntactic properties with hanging topics, but not necessarily their interpretive properties.

Hanging topics have several distinguishing properties that have been pointed out by Cinque (1997: 96; see also Cinque 1990).

The properties in (15a, c) apply to jo–construction directly. The jo–construction also complies with line (15b). In fact, hanging topics normally cannot be iterated but can be combined with left-dislocated topics, and they are always higher (Benincà & Poletto 2004; López 2016; Catasso 2022). Consider the Catalan example in (16), where en Joan is a hanging topic and de llibres is a left-dislocated topic.7

Likewise, jo↑ precedes so-called scene setting adverbs and left-dislocated topic constituents, if they are present, as illustrated in (17).8

Concerning resumption (15d), the literature (Cinque 1990; Benincà & Poletto 2004; López 2016; a.o.) points out that hanging topics require a resumptive item in the clause. Likewise, the jo–construction requires an element in the sentence that resumes the first person in the left periphery. Besides resumption by a pronoun, which may also have a specifier a mi (18a) or a PP per mi (dubbed an attitude-holder or opinion-holder; Bosse et al. 2012; Bylinina 2017) (18b), the expression of this resumption can be in the form of a null subject pro and be instantiated by means of morphological marking on the verb (-o in marxo ‘I leave’) (18c). See also the examples in (1), (6) and (7) above.

If there is neither a verbal first person agreement morphology nor an overt first person pronoun, the jo–construction is discarded. See the minimal pairs in (19) and (20).

Importantly, number agreement is not crucial, as we can also have plural first person resumption.

There is an apparent exception to that restriction, however. Unlike clitic pronouns (e.g., the dative em in (22a)), possessive determiners bearing first person features cannot serve as the resumptive element in the clause, no matter whether the possessive appears in preverbal or in postverbal position, (22b, c).

As pointed out by a reviewer the difference illustrated in these examples is that whereas (22a) is interpreted as a statement about the speaker, (22b, c) are statements about the speaker’s computer. However, regrettably, this semantic difference does not explain the ungrammaticality of (22b, c).

We consider that the impossibility of (22b, c) is due to the fact that meu is embedded in a DP whose phi-features do not agree in person with jo↑. Therefore, we postulate that the ungrammaticality of these examples arises because of the A-over-A principle (Chomsky 1964), which states that if a rule ambiguously refers to A in a structure of the form of (23), the rule must apply to the higher, more inclusive, node (marked with a box).

This principle rules out meu as a potential pronoun to resume jo↑. Since the highest DP projection bears third person, according to the A-over-A principle the embedded DP bearing the first person pronoun is not available to fulfill the requirement for jo↑ resumption. See the simplified structure in (24).9

Crucially, it must be emphasized once again that jo↑ is a kind of hanging topic – that is, its peculiarity is related to being first person singular, whereas hanging topics are tacitly assumed in the literature to be third person DPs (Cinque 1997; López 2016; a.o.) and, hence, are not subject to the A-over-A principle.10

Concerning case-matching (15e), it has been claimed in the literature that hanging topics do not generally agree in case (see also López 2016). These constituents take a default case marking, which in Catalan (like in Italian, Portuguese and Spanish) is nominative (Kato 1999; Schütze 2001; Villa-García 2023).11 This notwithstanding, notice that the strong pronoun with default case can be coindexed not only with a default nominative subject (as in (7), (9), (10b)), but also with a pronoun in accusative or dative case inside the sentence, as illustrated in (10a), (11a), (12e) and in (25).

Concerning islandhood (15f), it has been claimed that hanging topics are not sensitive to islands (Cinque 1997; Villa-Garcia 2023). In other words, unlike what happens in the case of sequences containing movement, this restriction does not apply to the jo–construction in strong islands such as the Coordinate Structure Constraint, as illustrated in (26) (or the Complex DP Constraint, illustrated in (10a) and (11a)). In fact, the contrast in (26) shows the possibility of a resumptive pronoun occurring in a jo–construction (26a), as opposed to a Left Dislocation construction (26b), in which there is leftward movement.

In sum, in this section we have argued that jo↑ in the jo–construction introduces a (kind of) hanging topic. While jo↑ obeys the properties in (15), it should be highlighted that, unlike regular hanging topics, the jo–construction only introduces a first person strong pronoun at the highest left periphery position, resumed by an item (a first person null or overt strong pronoun or a weak pronoun), inside the co-occurring sentence.

We here show that sentences preceded by jo↑ are distinct from sentences containing predicates of personal taste (as well as psychological predicates and verbs of propositional attitude). We show that, while the experiencer of the jo–construction can have multiple roles, as speaker (jo↑ ‘I’) and judge (the resumptive overt subject jo ‘I’, the opinion-holder PP per mi ‘as for me’, or the double dative a mi ‘to me’), and must be animate or sentient, the judge of predicates of personal taste (and the like) does not have to be.

First, consider the incompatibility of jo↑ with so-called predicates of personal taste (e.g., fun, tasty; Lasersohn 2005: 643, adapted from ex. (1)).12

Example (27b) shows that opinion-holders (i.e., ‘judge’ PPs) can be second and third person nominal expressions. Thus, while the expressions per mi / per tu / per als visitants / per a l’ajuntament introduce an individual parameter (the judge) with respect to which the truth of the sentence is evaluated and relativized, jo↑ is not a mere instantiation of the judge. The contribution of jo↑ is higher than the levels where the truth of p is evaluated. Furthermore, (27c) shows explicitly that, even though one might infer that it is the speaker of the utterance ‘Roller coasters are fun’ in a specific context who finds roller coasters fun, an explicit instantiation of the first person inside the sentence (i.e., the clitic em ‘me’) is required to make the sequence well-formed, as illustrated in (27d).

A similar paradigm is illustrated in (28) with an adjectival predicate of the ‘tasty’ class. The ill-formedness of (28c) is due to the absence of a first person element to resume jo↑, a problem solved in (28d) by a first person affix on the verb and in (28e) by a first person clitic affixed to a psychological verb.

The conclusion is that jo↑, as an instantiation of the speaker, can combine with verbs of personal taste only when inside the sentence there is a resumptive item instantiating a first person judge.

In Section 2.2 we argued that jo↑ is a kind of hanging topic. Here we show that hanging topics differ from sentence topics (the overt subject pronoun jo ‘I’, the opinion-holder per mi ‘as for me’, and the quirky subject em ‘me’ of psychological verbs) in that sentence topics precede assertions, correspond to topics of declarative sentences (used as assertions), and introduce a relation between an argument and a proposition relative to a context (Reinhart 1981). Consider the discourse in (29):

By uttering (29B) the speaker makes an assertion from which the proposition that they do not know whether there are any beers left in the fridge should be inferred. In (29B) jo is a sentence topic that precedes a new assertion: a proposition to be added to the context set of a given discourse (Stalnaker 1978). In this sense sentence topics combine with assertions that add information about what is the matter of current interest or concern and can be analyzed in terms of truth conditions.

The use of jo↑ instead of jo in (29B) would be inappropriate, since this subjective statement would not satisfy the informative requirements of the conversational exchange initiated by the question in (29A). Therefore, the hypothesis we postulate is that by means of the jo–construction the speaker is performing an act of declaration, although of a specific kind – as we argue, which may be hidden.

Most commonly, declarations differ from assertions in that they have no truth value (Krifka 2024b: 56). To illustrate this claim, consider the conversational exchanges in (30) and (31). While (30A) introduces an assertion with respect to which a reaction such as That’s not true is acceptable, (31A) does not allow a similar reaction.13

These examples show that while (30A) can be falsified, (31A) cannot. Against (31A) the addressee can only express disagreement (31B’), by means of which the interlocutor’s declaration is rejected. Thus, (30A) differs significantly from (31A). The jo–construction introduces a declaration speech act, which directly changes a situation once it has been uttered. This is possible because the speaker creates a new fact by their utterance, and crucially the contribution of jo↑ is at a level of representation higher than the levels where the truth of p is evaluated.

The hypothesis that the jo–construction serves the speaker to perform an act of declaration conveys the prediction that the speaker does not make any public commitment to the truth of p, since declarations differ from assertions in that they do not convey the social commitment characteristic of assertions (Krifka 2024b). Evidence supporting this claim is provided by the incompatibility of jo↑ with expressions that appear to modify the commitment to the truth of p, such as: veritablement ‘truly’, vertaderament ‘truthfully’, de veritat ‘for real’, positivament ‘positively’, sens dubte ‘no doubt’, amb certesa ‘with certainty’. Consider the data in (32), which exemplify that what the speaker adds to the common ground by means of the jo–construction is not a contribution to the truth-conditions of the sentence.

Likewise, the jo–construction is incompatible with exclamatives, the reason being that exclamatives introduce a commitment to a psychological state s (of surprise or annoyance) with respect to p (Villalba 2024).15 Notice that the exclamative in (33a) contains a first person judge PP. Therefore, the ill-formedness of (33b) must be due to the incompatibility between jo↑ and exclamatives.

Two caveats are in order here. First, from a crosslinguistic perspective it is unclear which modifiers are precisely commitment modifiers. Thus, while German sicherlich ‘certainly’ is analyzed as a judgment modifier in Krifka (2023), the data in (32) apply equally to Catalan amb certesa / certament ‘certainly’, thus suggesting that these expressions are commitment modifiers: ‘it is certain that p’.

Second, while according to Krifka (2024b) declarations are claimed not to allow for epistemic and evidential modifiers, (34) shows that the Catalan jo–construction is compatible with judgment (epistemic, evidential, evaluative) modifiers.

To sum up, in this section we have argued that jo–constructions differ from assertions in that they cannot be evaluated for truth and they resemble declarations in not allowing commitment modifiers. However, in contrast to Krifka’s (2024a; 2024b) analysis of declarations, jo–constructions admit judgment modifiers; jo–constructions are incompatible with certainty operators but are compatible with operators that express the speaker’s uncertainty towards p, which suggests that in their act of declaration the speaker declares that p is a fact relative to the worlds they know.

Semantically speaking, although most commonly no explicit performative is expressed by means of the jo–construction, jo↑ always introduces a declaration speech act at the time of the utterance act itself (Krifka 2023; 2024a; 2024b), no matter whether a performative predicate declare is overt, as illustrated in (35), or covert.

The act of uttering (35) is an act of declaration because by means of it the speaker does not merely say something but enforces an index change at the time of uttering the sentence (Szabolcsi 1982); the speaker declares p and has the authority to perform that change. Most importantly, this is the case whenever jo↑ precedes an utterance, even in the absence of an explicit performative, as illustrated in all the examples presented so far.

Jo↑ is the instantiation of a hanging topic projection whose head encodes an indexical feature [spk] (for speaker) that constrains the interpretation of the co-occurring statement as a declaration of the speaker’s subjective point of view. Therefore, as expected, in a jo–construction jo↑ is compatible with and must precede speaker-oriented modifiers (Schreiber 1972), whereas in regular assertions the adverb must precede the subject. This property follows from the hypothesis that speaker-oriented adverbs modify the speech act layer, whereas the contribution of jo↑ is at a higher hanging topic projection. Consider the minimal pair in (36).

In (36b) the strong pronoun jo is not any longer a hanging topic but corresponds to an overt subject (see footnote 3).

The examples in (36) and (37) constitute an additional argument in support of our hypothesis that jo↑ at the syntax-pragmatics interface sits in a syntactic position different from the one assigned to judge expressions (nominatives (36b), datives (37b) and opinion-holders (37c)).

Before we close Section 3, let us present an interpretive puzzle.

The previous sections support the hypothesis that sentence-initial jo↑ is an experiencer non-argumental expression that introduces a subjective declaration speech act compatible with judgment modifiers. These properties, together with the fact that jo↑ does not contribute to the truth conditions of p, appear to suggest that the strong pronoun jo↑ contributes to not-at-issue meaning (Potts 2005). Example (38), repeated from (1a), can only be interpreted as a declaration by the speaker that they believe they will not come, thus showing that jo↑ is out of the scope of negation. The same applies to the neg-raising construction in (38b).

Can jo↑ be questioned? The answer is negative as well. The jo–construction can only precede declarative sentences that are declarations, which excludes requests and commands. Accordingly, the possibility of finding examples such as (39) is discarded.

Likewise, the sequence in (40) shows that jo↑ cannot co-occur and combine with a command.16

All together these properties suggest that jo↑ contributes to (or constrains) the not-at-issue meaning.

However, we have repeatedly shown that jo↑ requires a resumptive first person pronoun within the sentence, supporting the hypothesis that the representation of the speaker is part of syntax. Furthermore, the previous discussion supports that jo↑ is part of ‘what is said’ (Grice 1975), a claim to which we add that jo↑ does not contribute to implicatures (its meaning cannot be cancelled without introducing contradiction, and its meaning is independent of the utterance in which it occurs), which suggests that jo↑ constrains the at-issue content of the sentence.

The solution to this puzzle is given in Section 4 where we present an analysis of the jo–construction at the syntax-pragmatics interface that consists in postulating a Hanging Topic Phrase with a [spk] head, and a Speech Act Phrase with a declare operator head.

To sum up this section, firstly we have argued that the jo–construction instantiates a kind of subjective statement headed by a non-argumental experiencer. Secondly, we have argued that the strong pronoun jo↑ instantiates the experiencer role of speaker, to be distinguished from the experiencer role of judge that is instantiated either by means of nominative and dative expressions or by judge PPs. Thirdly, we have argued that jo↑ indicates that the speaker is performing an act of subjective declaration compatible with judgment modifiers, neither an assertion nor a default (objective) declaration. Finally, we have shown that the jo–construction introduces a semantic puzzle, for jo↑ appears to contribute to / constrain the not-at-issue content as well as to the at-issue meaning.

We have already shown that the jo–construction, being an instantiation of a subjective sentence, may contain various coindexed instances of an affected experiencer (Bosse et al. 2012) at different layers. Consider the data in (44), which illustrates that when jo↑ is explicitly stated, the opinion-holder PP and the judge must be first person too.20

The syntactic structure we postulate for these examples is given in (45):

This structure combines a hanging topic with various sentential topics. The lowest a mi_i corresponds to the expression of the judge, when TP contains a psychological verb or a verb of propositional attitude. JP holds the constituent with respect to which the truth of the proposition is relativized (Lasersohn 2005; Stephenson 2007; among others). This constituent can take either the form of nominative or dative case. The intermediate per mi_i corresponds to the expression of the opinion-holder, usually introduced by means of the preposition per ‘as for’ (see the term judge-PP, Bylinina 2017; Perspective Phrase, Dayal 2023). The highest jo↑ corresponds to the expression of the speaker, which is the trigger of an act of subjective declaration. The structure in (45) requires that the speaker and the judge must agree in person, but not in case and number.

The structures in (43) and (45) also make explicit that speakers and judges are part of ‘what is said’ and because of this they must be represented in syntax. However, recall (Section 3.4) that jo↑ appears not to be part of the at-issue meaning, since it cannot be questioned or negated, and it does not contribute to the ‘truth-conditional content’. On the other hand, jo↑ imposes first person resumption with some constituent within the sentence, expressed either in the JP layer, in the TP layer, or on both, and this is precisely its contribution to the at-issue content. Therefore, the puzzle we presented in Section 3.4 is solved at the time we represent the jo–construction in syntax, since jo↑ in (43) and (45) – as a hanging topic – sits in a projection higher than ActP, while it is resumed by another first person element within JP or TP.

Let us now consider how the structure in (45) is altered at the time the hanging topic jo↑ interacts with judgement modifiers, as described in Section 3.2. Consider first the evidential and epistemic modifiers in (46), which are represented in the structure in (47).21

In the next section we consider the External Merge position for speech act modifiers that may co-occur with the jo–construction.

Consider the speaker-oriented adverbs in (48), represented in the structure in (49).

As pointed out in Section 3.3 the ill-formedness of sequences in which the hanging topic follows the speaker-oriented adverb is accounted for by postulating that: (i) jo↑ merges in a position higher than ActP, and (ii) speaker-oriented adverbs merge in ActP.

We are now left with the meaning associated with the structure in (43). We hold with Krifka (2024a; 2024b) that an act of declaration uttered in a context c conveys the performative update of c with the proposition expressed by the declarative sentence it is based on. The speaker of a declaration speech act changes the indices of the context (Szabolcsi 1982) and causes a common ground update at the very moment the utterance is produced, but crucially the speaker is not involved in a social commitment nor makes the addressee accept the truth of that p. This means that in declaration speech acts the speaker is the only one responsible for the performative update and does not trigger an informative update of c on the side of the addressee. That is, declarations (in Krifka’s sense) are assumed to be simpler than assertions (see (41)) and, accordingly, they lack ComP in their formal structure (see (42)). Besides, the structure in (43) differs from the one in (42) by including a reference to the speaker above ActP and a reference to the judge below ActP, in JP, which makes it a subjective declaration.

The distinction between declarations (that only involve a performative update) and assertions (that involve both informative and performative updates) accounts for the meaning differences observed between (30A) and (31A) of Section 3.2, repeated here for convenience as (50) and (52), with their corresponding structures in (51) and (53).

The example in (50), an assertion, conveys a performative update followed by a perlocutionary effect on the addressee, by means of which (s)he is informed that the proposition ‘I will not come’ is true and restricts an input context c to a new output context set c’. In (52), a subjective declaration, jo↑ hosts an indexical reference to the speaker, who directly serves to indicate that the utterance conveys a performative update. Its meaning is such that it changes indices of the context c so that the proposition Em sembla que no vindré holds at a new index i’, once the declaration has been uttered. Thus, the jo–construction introduces a declaration, as represented in (53), by which the speaker believes p and there are no side effects on the hearer such that he also believes p.