2. Passive and topic constructions in English

Syntactic (canonical) passives in a language such as English or French surface in two different forms. They are either long, as illustrated by (1a) from English and (2a) from French, or short, as illustrated by (1b) and (2b) from English and French respectively. The difference lies in whether they use a by-phrase or not:

1. (1)

   1. a. The car was repaired by John.

   2. b. The car was repaired.

2. (2)

   1. a. La voiture a été réparée par Jean.

      ‘The car was repaired by John.’

   2. b. La voiture a été réparée.

      ‘The car was repaired.’

One property of passive constructions is that their surface theme subject is understood as the logical object of the verb. This is clear from the fact that the object position left empty after the object has been displaced cannot be filled by another theme NP. Examples such as (3a) and (3b) are therefore unintelligible or ungrammatical since they violate the principle of Full Interpretation (Chomsky Reference Chomsky1986) or the Case Filter (Chomsky Reference Chomsky1981).

1. (3)

   1. a. * The car was repaired the bus (by John)

   2. b. * La voiture a été réparée l'autobus (par Jean)

      ‘The car was repaired the bus (by John).’

A second property of passive constructions is that their verb is morphologically inflected with a passive participle form -en/-ed in English (e.g., stol-en, writt-en, repair-ed, and so on) and é in French (e.g., lav-é ‘washed’, répar-é ‘repaired’, donn-é ‘given’, and so on). A third property, related to the second one, is that they have an auxiliary, be in English and être ‘to be’ in French, that accompanies the passive participle suffix on the verb, thus displaying a periphrastic passive form. In many languages, however, an auxiliary is not used, and the passive form is signalled by morphology alone. In other words, they have what Keenan and Dryer (Reference Keenan and Dryer2007: 7) call “strict morphological passives”. A further property is that they may have an optional agent by-phrase. Whenever this agent by-phrase is not present, as in short passives, an agent argument is assumed to be implicitly present (Baker et al. Reference Baker, Johnson and Roberts1989), an assumption that is claimed to be supported by the fact that it can control the null subject of a rationale clause as in The ship was sunk to collect insurance and it can occur with a subject-oriented adverb such as deliberately as in The ship was deliberately sunk.

The analysis of passive within the framework of transformational generative grammar (e.g., Chomsky Reference Chomsky1981, Burzio Reference Burzio1986, among others) argues that the displaced theme object in examples like (1) and (2) moves from its canonical object position to a base-generated empty subject position. This movement is necessitated by the fact that the passive verb cannot assign a theta role to its subject and an accusative case to its object. Movement of the object thus takes place in accordance with the Case Filter and the Theta Criterion (Chomsky Reference Chomsky1981, Burzio Reference Burzio1986, Jaeggli Reference Jaeggli1986, Baker et al. Reference Baker, Johnson and Roberts1989) as the moved object NP receives nominative case in the derived subject position but no additional theta role.

As for topic constructions in a language like English, they are also derived by movement of a clause-internal constituent but to a non-argument position on the left periphery of the clause, not the structural subject position. The moved constituent is linked to a gap (____) in the comment clause, as illustrated in (4):

1. (4)

   1. a. These books, I bought ___ for my daughter

   2. b. Him, we see ____ everyday at school

In an example like (4a), the fronted object NP these books is the topic, and it names an individual (person, object, etc.) already present in the universe of discourse (Kiss Reference Kiss and Karimi2003: 36). The clause I bought ____ for my daughter in that sentence is the comment, and it asserts something about the topic these books. The topic NP is linked to the gap (the position from which it has moved) inside the comment. In Rizzi's (Reference Rizzi and Haegeman1997) Cartographic framework, which we adopt here, when a constituent is topicalized, it targets the specifier of a Topic Phrase (TopP), one of several functional projections on the left periphery. Movement in this case is triggered by a feature (e.g., [+Top]) on the head of TopP and it is in this respect similar to other types of A-bar movement (e.g., wh-movement), which are also triggered by some feature on the head of some functional projection. Thus, when a wh-phrase moves to form a wh-question such as What will you tell them? it targets the specifier of Focus Phrase (FocP) on the left periphery, and movement is triggered by the [+Foc] feature on its head. The similarity between topic and wh-movement is also illustrated among other things by the fact that they both display similar reconstruction effects for Binding Principle C (Chomsky Reference Chomsky1981), as shown in the following:

1. (5)

   1. a. * Mary_i she_i loves ___

   2. b. * Who_i does she_i love ___ ?

Both these sentences end up violating Binding Principle C when the fronted constituents (R-expressions) are moved back into their original base-generated object position.

With this brief overview of some of the basic properties and derivation of passive and topic constructions in English in mind, let us now turn to consider some examples of transitive OV constructions in creoles. We will, however, focus mainly on such constructions in Mauritian Creole.

3. Transitive OV constructions in creoles

As illustrated below, Mauritian Creole, like other creoles, has sentences which, at the surface, resemble English or French short passives.Footnote ⁶ The initial argument, just like those in (1b) and (2b), is understood as the semantic object of the verb. Such sentences, as noted earlier, have been identified as passive in several studies (see Corne Reference Corne1977, Reference Corne and Muysken1981; Winford Reference Winford1993; LaCharité and Wellington Reference LaCharité and Wellington1999; Veenstra Reference Veenstra, Alexiadou, Anagnostopoulou and Everaert2004; DeGraff Reference DeGraff, Patrick and Holm2007; and Velupillai Reference Velupillai2015).Footnote ⁷

A brief examination of such examples reveals that, unlike their English and French counterparts, they do not have an auxiliary (the equivalent of the English auxiliary be or French être ‘to be’) nor a passive participle suffix on the verb.Footnote ⁸ Additionally, they almost never have a by-phrase.Footnote ⁹ Constructions with a by-phrase, like those in (7), are ungrammatical or rare, and usually restricted to acrolectal varieties (see Bernabé Reference Bernabé1983).Footnote ¹⁰

This being so, the only property of English or French short passives that these creole transitive OV constructions possess is a change in the position of the object NP. It precedes rather than follows the verb (and its auxiliary). This change is what lies behind the assertion that creoles have passive (see Winford Reference Winford1993, LaCharité and Wellington Reference LaCharité and Wellington1999, and DeGraff Reference DeGraff, Patrick and Holm2007, among others), and it provides much of the justification for the passive analysis proposed in LaCharité and Wellington (Reference LaCharité and Wellington1999) for OV constructions such as (6f–h) in Jamaican Creole. Nevertheless, in the absence of any other kind of clear indicator that the displaced object is in subject position, the possibility of an alternative topic analysis cannot be ruled out. First, we briefly review LaCharité and Wellington's (Reference LaCharité and Wellington1999) analysis.

3.1 A passive analysis of transitive OV constructions

Working within Chomsky's (Reference Chomsky1981) framework, LaCharité and Wellington (Reference LaCharité and Wellington1999) argue that constructions like (6f–h) in Jamaican Creole are passive. Their central assumption is that these constructions have a phonologically null, but syntactically active, argumental passive morpheme (the equivalent of English passive -en) under INFL. Following previous analyses of passives (see Chomsky Reference Chomsky1981, Burzio Reference Burzio1986, Jaeggli Reference Jaeggli1986, and Baker et al. Reference Baker, Johnson and Roberts1989), they argue that this abstract argument (PASSIVE) ‘absorbs’ the verb's accusative case and its external theta role. Consequently, the object NP moves into the structural subject position, a case-marked non-theta position, in compliance with the Case Filter (Chomsky and Lasnik Reference Chomsky and Lasnik1977) and the Theta Criterion (Chomsky Reference Chomsky1981, Burzio Reference Burzio1986). A sentence like (6g) is therefore represented as in (8):

This account looks theoretically plausible and is consistent with the principle of Uniformity (Chomsky Reference Chomsky and Kenstowicz2001: 2).Footnote ¹¹ Creoles are after all natural languages, and the difference between them and their lexifiers concerning passive lies in the phonological form of the passive argumental morpheme, covert in creoles and overt in their lexifiers.

Although the analysis of LaCharité and Wellington (Reference LaCharité and Wellington1999) seems theoretically well grounded and extendable to similar OV constructions in other creoles, it does nevertheless face some empirical difficulties, particularly when we consider examples similar to (6f–h) in Mauritian Creole. The passive analysis seems particularly problematic when we consider semantic restrictions on the displaced theme object and syntactic restrictions on its position relative to other constituents (e.g., fronted wh-phrases and negative adverbial phrases) and its distribution (ability to coordinate with unaccusatives). Such difficulties, as we show, do not arise under the alternative topic analysis. Rather, the semantic properties and syntactic restrictions follow naturally.

3.2 A topic analysis of transitive OV constructions

Following the Cartographic guidelines in Rizzi (Reference Rizzi and Haegeman1997) and drawing on data mainly from Mauritian Creole, we propose that the displaced theme object in examples like (6 a–b) occupies the specifier of a topic phrase (TopP) and is linked to a topic-gap (trace) in the canonical object position, an instance of topicalization (Ross Reference Ross1967), as illustrated by (9):

The theme object is attracted to the specifier position of TopP by the [+Top] feature on its head. Under this analysis, a structure like (9) is transferred to the interfaces for semantic and phonological interpretation. Importantly, its interpretation at the semantics-discourse interface involves locating some discourse-linked antecedent. Such a search, according to Reinhart (Reference Reinhart1981), can be construed as an instruction to the hearer to access their mental representation concerning the individual/entity in question, and then add the salient information to the open clause (FinP). A pragmatic ‘aboutness’ relation (Reinhart Reference Reinhart1981) then holds between the topic NP and the open clause (the comment).Footnote ¹² The comment is then ‘about’ the topic phrase as it asserts something about it.

A further assumption concerning the structure in (9) is that the subject position is not empty, as it would be in the case of passive in the underlying structure. Rather, it is filled with an impersonal null pronoun (represented as pro_imp) functioning as the external argument of an active transitive verb in the open/comment clause. Finite clauses with an unexpressed impersonal subject and a transitive verb are common in Mauritian Creole, both as independent and complement clauses, and they have a structure similar to that of the open clause in (9) but with the object remaining in situ, as shown in (10):

The following examples illustrate matrix and embedded sentences with unexpressed impersonal subject:

The unexpressed subject in such finite sentences is a non-referential/non-specific pronoun like the indefinite pronoun ‘someone’ or the impersonal/arbitrary indefinite third person plural pronoun ‘they’ in English. And, as shown by the glosses in (11b) and (12b), it can also be glossed as an arbitrary indefinite NP, namely ‘people’. In this respect, the unexpressed subject in these sentences is similar in its interpretation to the unexpressed subject of impersonal constructions in other languages (e.g., Spanish (Jaeggli Reference Jaeggli1986), Italian (Cinque Reference Cinque1988), Irish (McCloskey Reference McCloskey2007), Lithuanian (Šereikaite Reference Šereikaite2017), and so on). It is generally construed as some unspecified individual or collective of individuals whose identity is unknown (Siewierska Reference Siewierska2010). Additionally, it can have an existential, or, in Cinque's (Reference Cinque1988) term, quasi-existential force, which can be attributed to the presence of episodic tense or aspect marking, signalling a specific time reference or a quasi-generic force (Cinque Reference Cinque1988), as in (11b) and (12b), usually in a habitual context, which, in these sentences, is established by the presence of the temporal adverb avan ‘before’. These are properties that the unexpressed impersonal subject in (11) and (12) shares with impersonal or arbitrary subjects in other languages such as Italian, French, German, Irish, and so on (see McCloskey Reference McCloskey2007 and references cited there). Furthermore, since it is restricted to the subject position of finite clauses, it is identifiable with the impersonal subject pronouns in some of the Germanic languages (e.g., men in Dutch), which, according to Fenger (Reference Fenger2018), are also restricted to a nominative case position, but not with the English impersonal pronoun one, which is not restricted in its distribution and has only an inclusive generic reading in contrast to the multiple (existential and generic) reading that the Dutch impersonal men has. Given this distributional and interpretational similarity to the impersonal subjects found in some of the Germanic languages (not English), the unexpressed subject in (10) may also be said to share their structural makeup, which, according to Fenger, is deficient (i.e., lacking a functional layer unlike English one), thus lending itself to multiple (i.e., generic and existential) interpretations.

As an impersonal pronoun, the unexpressed subject in examples such as (11) and (12) cannot be anaphorically linked to a referential specific (definite) personal pronoun. Thus, example (13a) is ungrammatical if the unexpressed impersonal subject of the embedded clause is linked to (or bound by) the overt referential personal subject pronoun zott ‘they’ in the main clause. Example (13b) is likewise ungrammatical if the unexpressed impersonal subject in the main clause is linked to the overt referential personal subject zott ‘they’ in the embedded clause. Thus, neither a backward nor a forward anaphoric relation appears possible between an unexpressed non-referential/impersonal subject pronoun and an overt referential personal pronoun even though they share person and number features:

This restriction follows naturally since the anaphoric relation is between an unexpressed subject pronoun that is impersonal, non-referential, and indefinite and one that is personal, referential, and definite. Such a restriction, as pointed out by McCloskey (Reference McCloskey2007: 834–5), also seems to hold between impersonal arbitrary and referential subject pronouns in other languages (e.g., Italian, Irish, and German).

Another observation on the unexpressed subject is that it is not only impersonal but also understood as human. Thus, the meaning of a sentence like (14a) is that a person, not an animal (e.g., a cat), left a (dead) mouse by the dog and the meaning of (14b) is that an implicit human agent, not some non-human or inanimate (e.g., a machine), turned the water off:Footnote ¹³

Interestingly, an inanimate reading of the unexpressed subject is impossible even when an appropriate context is created to facilitate it. Example (15b), for example, is not felicitous when compared to (15a) since the unexpressed subject cannot be construed as referring to cars, vans, and so on:

This difference highlights the fact that in such impersonal constructions like (11) and (12), a human animate reading of the unexpressed subject persists even when an inanimate causer is pragmatically available in the preceding context. The unexpressed impersonal subject of the constructions in (11) and (12) is therefore identifiable as a human agent, which is thought to be a general characteristic of impersonal constructions (Blevins Reference Blevins2003, Siewierska Reference Siewierska2010, Šereikaite Reference Šereikaite2017, Legate et al. Reference Legate, Akkus, Šereikaité and Ringe2020, among others) – but see McCloskey (Reference McCloskey2007) for other possibilities in the interpretation of Irish arbitrary subjects.Footnote ¹⁴

A further observation is that the impersonal agent subject, although unexpressed, is syntactically active, unlike the agent in passive, which is syntactically deleted but semantically present as an implicit argument. Looking back at the structure in (9) and (10), the assumption is that the subject is not syntactically missing but simply unexpressed. Using diagnostic tests for the syntactic presence of a subject in subjectless impersonal constructions (see McCloskey Reference McCloskey2007, Legate et al. Reference Legate, Akkus, Šereikaité and Ringe2020, among others), we show that an unexpressed impersonal subject is indeed syntactically present in structures like (9) or (10).

One indication of its presence is that an adjunct by-phrase, as observed earlier, is impossible. Thus (16a) and (16b) are ungrammatical:

This restriction follows naturally, it may be suggested, from the assumption that an impersonal unexpressed subject is already present in these examples and has been assigned the verb's external theta role. The insertion of a by-phrase thus results in a violation of the principle of Full Interpretation (Chomsky Reference Chomsky1986). Additionally, examples with an agent by-phrase, like those in (16), also involve a conflict of features, as the unexpressed subject is non-referential while the agent in the by-phrase is referential. However, since by-phrases can also be missing in languages which have passives, its impossibility in examples like (16) can only suggest, rather than confirm, that an unexpressed subject is syntactically present.Footnote ¹⁵

A stronger argument in support of this claim is the possibility of anaphor binding. The reciprocal len-e-lot ‘each other’ in Mauritian Creole, like reciprocals in other languages, requires a plural antecedent, as illustrated by the contrast in (17):

Examples like (18) show that a reciprocal is possible in impersonal sentences like (11):

Since reciprocals require binding (Chomsky Reference Chomsky1981), we conclude that they are bound by an unexpressed subject pronoun inside the bracketed clause in (18).

The syntactic presence of an impersonal subject in (11) or a structure like (9) or (10) is also suggested by remnant NPs (such as ‘who’, ‘what’, and so on) following sluicing. According to Ross (Reference Ross1967), Merchant (Reference Merchant2001), and others, a remnant NP or DP in examples like (19) results from sluicing, an operation whereby everything in a wh-interrogative, except the wh-phrase, is elided, as illustrated by the strike through notation:

1. (19)

   1. a. Someone called this morning, but I don't know who called this morning

   2. b. John hit someone but no one knows who John hit

The remnant wh-phrase ‘who’ is understood as subject in the indirect question in (19a) and object in the indirect question in (19b). In both cases, the remnant NP is linked to the indefinite, non-specific pronoun ‘someone’ in the first conjunct. Now, as (20) illustrates, sluicing is also possible in Mauritian Creole:

Applying sluicing to constructions with impersonal subject, like those in (11), provides further support for the claim that an unexpressed subject is syntactically present in their underlying structure, and not deleted or suppressed as would be the case if they were passive. The remnant wh-phrase kisennla in (21)–(23) confirms the presence of a syntactic subject.

Interestingly, in all these examples, a remnant by-phrase is impossible. Thus, speaker B cannot use par kisennla ‘by who (m)’ instead of kisennla ‘who’ in (21). Likewise, replacing the remnant NP in (22) and (23) with a by-phrase, as shown in (24), is also impossible:

The contrast between (21)–(23) on the one hand and (24) on the other is due to a voice mismatch (Legate et al. Reference Legate, Akkus, Šereikaité and Ringe2020). The impossibility of (24 a–b) with a passive remnant by-phrase suggests that constructions like (11) are active, not passive, and they have a syntactically projected unexpressed indefinite subject, understood as ‘someone’ or ‘they’.

As for the displaced theme object in (9), we argue that it occupies a specifier position on the periphery of the clause and is linked to the gap (trace) in the comment clause. Under a topicalization analysis, we take it to have moved there from its canonical object position. In what follows, we discuss semantic and syntactic restrictions on the displaced theme object and argue that they can be straightforwardly accounted for under a topic analysis.

4. Semantic and syntactic restrictions on the theme object

One observation on transitive OV constructions like (6 a–b) in Mauritian Creole is that they are generally judged grammatical if the initial argument is specific (old information known to the hearer), as shown in (25), and ungrammatical if non-specific (new information unknown to the hearer), as shown in (26) and (27). Specificity is signalled by the postposed definite determiner la ‘the’, while non-specificity by the indefinite determiner enn ‘a/an’ or null (zero) determiner.Footnote ¹⁶

Such a restriction, however, does not apply to the subject of transitive or unergative sentences. It can be either specific or non-specific, as illustrated by (28):

Nor does it apply to the subject of unaccusatives, since it too can be either specific or non-specific, as illustrated by (29):

If the initial argument (the theme object) in examples like (25)–(27) were in subject position, as expected under a passive analysis, the contrast between (25) on the one hand and (26)–(27) on the other would be surprising, for, as (28) and (29) make clear, non-specific (indefinite) NPs, including theme ones, are not barred from subject position. By contrast, in a topic analysis, the difference between (25) and (26)–(27) is expected because, as illustrated by the difference between (30a–b) and (30c–d), non-specific (indefinite) NPs, unlike specific NPs, are generally excluded from a topic position (see Kiss Reference Kiss2002, Reference Kiss and Karimi2003):Footnote ¹⁷

The fact that the displaced theme object in (26), (27) and (30a–b) cannot be non-specific (indefinite) seems to follow directly from a general discourse requirement that topic phrases be specific (in the sense of Enç Reference Enç1991).Footnote ¹⁸ That is, they must be discourse-linked to some unique discourse antecedent or, in the words of Aissen (Reference Aissen1992: 50), to ‘some identifiable participant in discourse’. What follows naturally from such a requirement are two things: firstly, as pointed out by Prince (Reference Prince1981) and Kiss (Reference Kiss2002, Reference Kiss and Karimi2003), the topic constituent cannot be non-specific and secondly, it cannot be non-referential (e.g., a pleonastic element like ‘there’).Footnote ¹⁹

A second semantic restriction on the displaced theme object in transitive OV constructions like those in (6a, b) is that it cannot be animate. The following are judged ungrammatical.Footnote ²⁰

It is interesting to note that a similar restriction holds on displaced animate theme objects in Caribbean English Creoles transitive OV constructions (see Alleyne Reference Alleyne and Alleyne1987, Winford Reference Winford1993). Compare, for example, (6f–h) above with the examples in (32), also from Jamaican Creole:

Under a passive analysis, the ungrammaticality/marginality of the transitive OV constructions in (31) and (36) is unexpected because specific animate NPs (including proper nouns) are not excluded from the structural subject position. This is clear from transitive constructions like (33), unaccusative constructions like (34), and unergatives constructions like (35):

If the fronted theme object in (31) moves from its canonical object position to the structural subject position, as per the analysis of passive in Chomsky (Reference Chomsky1981), Burzio (Reference Burzio1986), and others, it is predicted that, much like the unaccusatives in (34), the OV constructions in (31) should also be grammatical. This prediction is not borne out, as the ungrammaticality of (31) demonstrates.

However, examples like (31) are problematic not only for a passive analysis but also for a topic analysis that posits that the fronted object lies in the periphery of the clause (more specifically, in a topic position), given that animate proper nouns and NPs, being specific, are not barred from such a position. And, indeed, as the examples in (36) show, they are not, particularly when the topic phrase is a hanging topic linked to a resumptive pronoun inside the comment clause.

What the contrast between (31) and (36) suggests then is that animate NPs, when topicalized, cannot be gap-linked, as evidenced by the ungrammaticality of (31), but must instead be reprised by a resumptive pronoun, as shown in (36).Footnote ²¹ Why should that be? A possible and plausible answer is that, given the strong association between animacy and agency and the availability of an unexpressed subject, animate fronted objects in OV constructions like (31) tend to be read as the agent of the verb rather than its theme. The relationship between animacy and agency is also invoked by Alleyne (Reference Alleyne and Alleyne1987) in connection with examples like (32) in Jamaican Creole. Only the presence of a resumptive pronoun coreferential to the fronted object, as shown in (36), or contrastive focus on the topic NP (see also footnote 20), it seems, can ensure that this ‘theme as agent’ misinterpretation (or processing error) is avoided.

It is interesting to note that a ‘theme as agent’ reading in such OV transitive constructions persists even when the context discourages it, as in the following:

What is clear from (37) is that even the availability of a non-human or inanimate causer, either through linguistic coreferencing, as in (37b), or pragmatic anaphora, as in (37a) and (37c), does not stop the topic animate object polisie la ‘the policeman’ from being incorrectly interpreted/processed as agent of the verb morde ‘bite’. Thus, the reading of the OV construction inside a construction like (37b) is not ‘the policeman was bitten’ or ‘the policeman, someone bit (him/her)’ but rather the incorrect interpretation ‘the policeman bit (someone)’. The facts relating to the impossibility of fronted animate objects in OV constructions like (31) can therefore be captured relatively straightforwardly under a topic analysis.

Turning now to a third restriction on the fronted theme object in transitive OV constructions like (6a–b), it appears that these objects cannot be pronouns either, whether they are animate or inanimate, which is puzzling since personal pronouns are by definition definite (Roberts Reference Roberts, Reimer and Bezuidenhout2005).Footnote ²²

Unlike the third person (singular and plural), the first and second person singular in Mauritian Creole display different case forms, depending on their distribution. Thus, they have the nominative form mo ‘I’ and to ‘you’ for first and second person singular respectively when they appear in the subject position of a finite clause, as shown in (39a–b) and the corresponding first and second person singular accusative form mwa ‘me’ and twa ‘you’ when they appear as the object of a verb, as shown in (39a–b) or a preposition, as shown in (39c), and as topics, as shown in (39d):

With the fronted object pronouns displaying accusative (not nominative) form, the examples in (38) are clearly problematic for a passive analysis. Surprisingly, they are also problematic even when these pronouns are in the nominative, as shown in (40). A passive analysis incorrectly predicts that the examples in (40) should be grammatical:

The nominative form is expected on a pronoun when it appears in the subject position of a finite clause, as shown in (39a–b) above, an unaccusative finite clause, as shown in (41), or unergative finite clause, as shown in (42):

Under a topic analysis, the ungrammaticality of the examples in (40) follows directly from the fact that such structures involve conflicting cases: thus, the fronted pronoun has the nominative form (mo ‘I’, to ‘you’) in its derived (topic) position while the gap to which it is linked has the accusative form. It also follows under this analysis that a nominative pronoun, being phonologically weak/unstressed, is naturally excluded from an edge/peripheral (topic) position, which typically hosts phonologically strong/stressed constituents. The ungrammaticality of (38), on the other hand, is unexpected under a topic analysis: the fronted object pronouns have the accusative (phonologically strong) form, which, as is clear from (39d), is the case form expected on topic pronouns. This is further illustrated by the hanging topics in (43) and the contrastive focused constituents in (44):

Why, then, are the examples in (38), with a fronted accusative object pronoun, ungrammatical?Footnote ²³ It is apparent from the difference between (38) on the one hand and (43) and (44) on the other that a fronted object pronoun, unless contrastively focused, cannot be linked to a gap inside the clause. Instead, it must be linked to a resumptive pronoun, as shown in (43). Given that (38a–b) pattern with (31) in that neither allows a fronted animate object (pronoun or NP) to be gap-linked (unless it is contrastively focused), we could account for the ungrammaticality of the former in the same way that we explained the ungrammaticality of the latter: that is, in the absence of a resumptive pronoun inside the comment clause or contrastive focus, a fronted animate object is misinterpreted/misanalysed as the agent of the transitive verb, a possibility enhanced by the availability of an indefinite null subject and a universally strong association between animacy and agency. The ungrammaticality of (38c–d), with a fronted inanimate pronoun remains unexplained under the proposed ‘theme as agent’ misanalysis approach, as the association between animacy and agency cannot be invoked in these cases. One solution is to appeal to the phonological form of the fronted pronouns in these examples. The third person singular pronoun li ‘he/she/it’ and the third person plural zott ‘they’, as noted earlier, are phonologically weak forms and are therefore naturally excluded from peripheral positions that generally tend to host phonologically strong forms. A topic analysis, unlike a passive analysis, can thus account for the impossibility of fronted object pronouns (nominative and accusative) in transitive OV constructions in Mauritian Creole.

It is interesting to note that such misinterpretation or incorrect processing (i.e., theme as agent), as discussed above, does not occur in the presence of an intervener (e.g., a wh-phrase or a clause), as shown in (45):

This would suggest that the strong association between animacy and agency that encourages the ‘theme as agent’ misinterpretation only holds when a gap-linked non-focused animate topic constituent and the unexpressed indefinite subject in the OV construction are adjacent.

In summary, it seems that the displaced theme object in transitive OV constructions in Mauritian Creole cannot be non-specific or animate or pronominal, and this is the case unless reprised by a resumptive pronoun or contrastively focused.

5. Further evidence in support of a topic analysis of transitive OV constructions

In what follows, we discuss some additional evidence that further supports the topic analysis of the transitive OV constructions in Mauritian Creole.

The first piece of evidence comes from the position of the displaced theme object relative to a fronted wh-phrase. As illustrated by (46) and (47), whenever a wh-phrase is fronted in a transitive OV construction, the displaced object appears to its left, as shown in (46b) and (47b), not to its right, as illustrated by the ungrammaticality of (46c) and (47c).Footnote ²⁴

Under a passive analysis, we do not expect (46c) and (47c) to be ungrammatical, since the fronted object occupies the subject position and the fronted wh-phrase, to its left, occupies the specifier of FocP, as per the template posited in Rizzi (Reference Rizzi and Haegeman1997). They should therefore be just as grammatical as the unaccusative in (48b), where the object has moved into the subject position:

And yet, (46c) and (47c) are ungrammatical, unlike (48b). Under a topic analysis, however, their ungrammaticality is not unexpected. As a topic phrase, the theme object in these examples appears higher than (or to the left of) the wh-phrase. This is because topic constituents target the specifier of TopP, while wh-phrases target the specifier of FocP, following Cartographic syntax (Rizzi Reference Rizzi and Haegeman1997). These sentences are therefore ungrammatical because they either violate Relativised Minimality (Rizzi Reference Rizzi1990) or require an additional FocP, an option that violates Cartographic guidelines (Rizzi Reference Rizzi and Haegeman1997) where only one FocP is allowed in the periphery of a clause.Footnote ²⁵ Of course, no such problem arises with the unaccusative (48b) since the theme object appears in subject position and the wh-phrase in the specifier of FocP.

A second syntactic restriction that also supports the topic analysis relates to the position of the theme object in relation to a negative adverb of frequency like zamen ‘never’. As illustrated by (49) and (50), a theme object must be to the left of this type of adverb for a transitive OV construction to be well formed:

If negative adverbs occupy the specifier of FocP (Rizzi Reference Rizzi and Haegeman1997), we then expect zamen ‘never’ to precede the NP in subject position as in the transitive, unergative, and unaccusative constructions in (51) where the agent or theme is unambiguously in subject position:

This being so, a passive analysis of transitive OV constructions incorrectly leads us to expect (49c) and (50c) to be well-formed. A topic analysis, in contrast, correctly rules them out because, with the theme object in the specifier of TopP, the negative adverb has moved across it to the specifier of a higher projection, violating locality (Relativized Minimality (Rizzi Reference Rizzi1990)) as illustrated by (49c) and (50c). Alternatively, as was the case with (46c) and (47c), a second FocP is required, which is inconsistent with Cartographic guidelines (Rizzi Reference Rizzi and Haegeman1997).

A further restriction that also supports a topic analysis of transitive OV constructions relates to coordination. As illustrated by (52) and (53), it is possible to coordinate two such constructions, with and without an overt subject, as in (52), or two unaccusatives, as in (53), but not a transitive OV construction and an unaccusative, as shown in (54).

The ungrammaticality of (54a) and (54b) follows directly from the principle that requires conjuncts to belong to the same category (e.g., PP and PP or TP and TP) (Ross Reference Ross1967). Under the proposed topic analysis, these constructions involve the coordination of a TopP (the transitive OV construction in the first conjunct) and TP (the unaccusative in the second conjunct), two categorially dissimilar constituents, and are therefore correctly ruled out. Under a passive analysis, in contrast, these constructions involve coordination of two TPs and are therefore incorrectly predicted to be well formed. As expected, the topic analysis also correctly rules out the coordination of a transitive construction (a TP) and an OV construction (a TopP), as illustrated by (55):

The ungrammaticality of constructions such as (54) and (55) therefore provides additional support for the claim that the OV transitive constructions under discussion are topic, not passive, constructions. The topic analysis of OV constructions is therefore syntactically supported not only by facts relating to word order but also by facts relating to coordination.

6. Implication of a topic analysis of transitive OV constructions in Mauritian Creole

The analysis of transitive OV constructions in Mauritian Creole as topics rather than passives suggests that passivization, an instance of A-movement, is an option that has not been exploited in this language, and possibly in some other creoles too (e.g., West African Pidgin English, Tok Pisin, Bislama, among a few others), while other instances of A-movement (e.g., movement to subject position in unaccusative and raising constructions) appear to have been.Footnote ²⁶ The proposal that Mauritian Creole has topic rather passive constructions can be taken to support the traditional intuitions of many creolists that creoles, like many non-creole languages, lack canonical passives (see Siewierska Reference Siewierska2010, Keenan and Dryer Reference Keenan and Dryer2007). Interestingly, native speakers of creoles generally tend to reject passive structures. McWhorter and Good (Reference McWhorter and Good2012: 164) note for instance that their Saramaccan consultants translated English-type passives as active sentences with the generic third person plural subject de ‘they’, as shown in (56):Footnote ²⁷

Likewise, Veenstra (Reference Veenstra, Alexiadou, Anagnostopoulou and Everaert2004: 277) notes that not all speakers of Saramaccan accept sentences with a fronted object as passive unless they were contextualized (i.e., they included aspectual information focusing on the endpoint of an event denoted by the verb). There are also similar caveats from contributors to the chapter on passives in the APiCS databases (Michaelis et al. Reference Michaelis, Maurer, Haspelmath and Huber2013). Thus, in her contribution to these databases, Escure (Reference Escure2013) notes in connection to Belizean passives that “since only the context provides the correct interpretation (passive or active), it is in fact doubtful that Belizean Creole has a passive construction”. Altogether, it seems that there are some doubts as to whether passive (in which the object appears in subject position) is available in these languages.

Creoles, including Mauritian Creole, thus tend to resort to other well-known strategies for expressing a passive meaning. These include suppressing the logical subject, or using an impersonal subject, or using a verb similar to ‘get’ as in John got arrested. All three are used in Mauritian Creole although the last one is limited to just a few verbs like gany bate ‘got beaten’, gany kriye ‘got shouted at’ but not * gany tuye ‘got killed’ (see Winford Reference Winford1993: 142 for similar structures in Caribbean English creoles).