A brief introduction to Comparative syntax across grammars and structures, the second issue of Probus 36 dedicated to Mary A. Kato on her 90th anniversary

Comparative syntax across grammars and structures, issue 2 of Probus 36, completes the series initiated with The interplay between theory and observation: studies in syntax and semantics, issue 1 of Probus 36. This series is dedicated to Professor Mary Aizawa Kato, whose biography can be found in the publicly available Preface to Probus 36.1.

Kayne (2013), in his influential paper entitled Comparative Syntax, attributes a significant role in the qualitative improvement of our knowledge and understanding of human language grammar in the last six or seven decades (taking 2024 as reference), to the successful development of the field of inquiry of formal comparative syntax. As he puts it:

Close cross-linguistic comparison is used as a tool to unearth properties of the language faculty that might otherwise have been difficult to bring to the fore. Explaining why a property found in language A is not found in language B often sheds light on the syntax of both languages. (Richard Kayne, in the presentation of the series Oxford Studies to Comparative Syntax)^[1]

As an illustration of Kayne’s final assertion in the above quotation, let us consider the paradigms in (1) and (2) below, which feature data from two closely related, and geographically contiguous, Romance languages, namely European Portuguese and Spanish. While both languages allow TP Ellipsis (TPE), illustrated by the (a) examples, only European Portuguese also allows VP Ellipsis (VPE), as shown by the grammaticality contrast between (1b) and (2b).^[2] Why should a superficially insignificant difference (the absence/presence of the verb after the polarity word também/también), with no impact on interpretation, have such a strong effect in Spanish, but not in European Portuguese grammar? As a starting point, note that the observation that (1b), thus VPE, is not available in Spanish – that is, it is not a grammatical option, as (2b) demonstrates – was prompted by the comparison with another language, and might well have gone unnoticed were Spanish the strict focus of attention. Due to work carried on in comparative syntax, the puzzling contrast between Spanish and Portuguese (which could as well be between Spanish and English)^[3] can be put in a larger context. Goldberg (2005) has observed that VPE (which she calls V-stranding VP Ellipsis) is typologically rare, which is in contrast with the widespread availability of TPE. The question is now why different types of ellipsis occurring in similar structures (e.g. coordination structures) display dissimilar distribution across languages, even though ellipsis, lato sensu, appears to be universally available in human language syntax.

Kayne (2013) establishes as a central aim of comparative syntax going beyond “observational adequacy” (i.e. the full and straight picture of the relevant facts to address a particular research question) with the goal to explain why things are as observed, relative to what we know about the human language faculty (also referred to as Universal Grammar/UG). Advancing from observational to “explanatory adequacy” requires taking comparison to the step of “descriptive adequacy”, unveiling generalization supporting correlations not trivially accessible to observation.

In comparative syntax, as in syntax in general, one can and must also aim at explanatory adequacy, above and beyond observational and descriptive adequacy. In the case of comparative syntax, we can try to understand, in general UG terms, why a given cross-linguistic correlation should hold in the first place. (…) It is to be noted that both for comparative syntax and for syntax in general there is no suggestion in any of the preceding discussion that descriptive adequacy must be met in a fully prior way to explanatory adequacy or that descriptive adequacy must fully wait until observational adequacy is met. In practice one must aim at all three simultaneously, and work simultaneously on developing more and more observations, generalizations and explanations (Kayne 2013, 137).^[4]

Going back to the VPE/TPE illustration, extending the comparative observation across the Romance languages family (besides Spanish and European Portuguese) permits to perceive certain non-trivial correlations. The languages that allow VPE (European Portuguese, Galician), on the one hand, and the languages that bar it (Spanish, Catalan, Italian, French), on the other, display different patterns of polar answering system. In the languages with VPE the verb by itself can express affirmative confirmation or denial; in fact, bare verb answers constitute the unmarked pattern of polar affirmative answer in European Portuguese and Galician.^[5] Moreover, these same languages display enclisis in finite clauses, whereas languages that do not allow VPE and bare verb affirmative answers also exclude enclisis in finite clauses.^[6] These descriptive correlations/generalizations, established by looking across different structures, besides different languages, can be explained under the hypothesis that the abstract functional polarity-encoding head Sigma (Σ) – which, in UG terms, is part of the functional architecture of the clause – is subject to parametric variation regarding its ±V-relatedness. Only languages with V-related Σ, hence with verb movement to Σ, license VPE, allow bare verb affirmative answers (with the verb playing the role of a polarity word) and derive enclisis in finite clauses (with the verb moving higher – to Σ – than the clitic – in the T projection). This hypothesis can be tested against new empirical observations and descriptions, not limited to Romance (see Martins (1994, 2016, forthcoming), for Romance (and English); Costa et al. (2012), for Capeverdean; Lipták (2013), for Hungarian; see also Cyrino and Matos (2005) and Rouveret (2012), for alternative approaches to VPE). As for TPE, it has been proposed by different authors that it is licensed by ΣP under the condition that its Spec, its head or both be filled with polarity particles, such as também/también ‘also’ in (1a/2a) above (e.g. Matos (1992), for Portuguese; López (1999) and Vicente (2006, 2010, for Spanish; Busquet (2006), for Catalan; Morris (2009), for French; Lipták (2013) for Hungarian). Crucially, licensing of TPE does not require that Σ and the verb merge together and this seems to be the reason why TPE displays more relaxed licensing requirements and a broader availability across languages.

Three of the five articles in the current issue, whose theme is “comparative syntax across grammars and structures”, deal with silent constituents. Ana Maria Martins and Jairo Nunes, as well as Andrés Saab analyze null subjects as deleted constituents, which undergo ellipsis under specific licensing conditions (cf. Perlmutter (1971), revived by Sheehan (2006, 2016, Saab (2008, 2016, Duguine 2013, among others, as part of the new takes on the former pro-drop parameter).^[7] The article by Pilar Barbosa addresses null objects, extending her previous work on null subjects (Barbosa 2019) inspired by Tomioka’s (2003) concept of minimal n. The fourth article, by Maria Lobo, explores a formal comparative approach to different types of gerund clauses in European Portuguese; the cross-structural variation observed language-internally shows that the availability/activation of specific functional categories may vary not only across languages but across structures as well (see e.g. Thrainsson (1996), Bobaljik and Thráinsson (1998), Fukui (2006), Rizzi (1997, 2004, among others). The fifth article, by Maria Eugenia Duarte, deals with the challenges of handling in the best possible way the limited access to observation in comparative historical syntax, specifically addressing syntactic change in the pronominal clitic system of Brazilian Portuguese – with European Portuguese as term of comparison.

In their article, Martins and Nunes examine how the predictions of Holmberg’s (2005) and Holmberg et al.’s (2009) theories of null subject licensing fare against the comparative empirical evidence offered by Brazilian and European Portuguese, the former a partial, the latter a consistent null subject language (NSL). They conclude that Holmberg’s (and coauthors’) proposal that T has a D(efiniteness) feature in consistent NSLs but lacks it in partial NSLs does not account for the Brazilian and European Portuguese data. Alternatively, they argue for an approach that drops the hypothetical D-feature and operates only with abstract ϕ-features (which need not bear a transparent relation with agreement morphology) and case. Martins and Nunes’ approach retains from Holmberg (and coauthors) the insight that the licensing of null subjects depends on the interaction between the features of T and the features of subject pronouns (the former requiring proper valuation to allow subject ellipsis). The degree of ϕ-feature (under)specification of subject pronouns is the source of cross-linguistic variation as well as language-internal microvariation regarding the licensing of subject ellipsis. Moreover, variation in the availability of null subjects across different structures (namely, finite, infinitival, participial and gerund clauses) is the effect of the diverse feature composition of T. This allows, for example, to derive the fact that BP displays three degrees of acceptability of a null subject in finite clauses, from totally acceptable (nós ‘we’) to marginally acceptable (eu ‘I’, vocês ‘you(PL)’, eles/elas ‘they(MASC/FEM)' to excluded (você ‘you(SG)’, a gente ‘we’ and ele/ela ‘he/she’), as well as the fact that these differences disappear in gerund clauses, where subjects ellipsis is generally allowed.

The paper by Andrés Saab approaches the complex cross-linguistic variation and commonly accepted language typology relative to the null subject property within the general theory of ellipsis developed in Saab (2008, 2016, 2020). The author maintains from the original intuition behind the pro-drop parameter the idea that (rich/impoverished) agreement is in fact behind the relevant typological distinctions. But crucially, as he puts forward, abstract/syntactic Agree (Chomsky 2000, 2001), though an important ingredient, is not enough to allow a language to be an NSL. For subject ellipsis to be an option, in addition to syntactic Agree, T⁰ and agreement must be properly dissociated at PF. Agreement dissociation is necessary because “only an expanded agreement morpheme adjoined to the T⁰ node can serve as a licit antecedent for ellipsis of a subject D⁰ head at PF”. The interplay between the dichotomies presence/absence of abstract Agree and presence/absence of agreement dissociation at PF then derives, at large, the full typological spectrum of languages relative to the null subject property. Languages without abstract Agree are radical NSL (e.g. Japanese and Chinese); languages with abstract Agree but without agreement dissociation are non NSL (e.g. English and Standard German); languages with abstract Agree but restricted agreement dissociation are partial NSL (e.g. Brazilian Portuguese); languages with abstract Agree and unrestricted agreement dissociation are (with some further qualification) consistent NSL (e.g. European Portuguese and Spanish).

An apparent exception to the regular behavior of a consistent NSL like European Portuguese is provided by the semantically first person plural pronoun a gente (‘we’, originated from ‘the people’), whose grammaticalization path and categorial status have been extensively debated in the literature (see, for example, Costa and Pereira 2013; Lopes 1999; Lopes and Brocardo 2016; Taylor 2009, and references therein). As the example in (3) shows, the subject a gente cannot be deleted in European Portuguese, whereas, in the same context, ele ‘he’, ela ‘she’, você ‘you(SG)’ can, although these pronouns are exactly like a gente ‘we’ in being associated with verbal forms without overt agreement morphology.^[8]

Martins and Nunes’ article and and Saab’s article account in different ways for the peculiar behavior of a gente in European Portuguese. Under Martins and Nunes approach, the pronouns ele ‘he’, ela ‘she’, você ‘you(SG)’ have specified values for person and number in European Portuguese, therefore can be deleted after they value the ϕ-features of T, whereas a gente is fully underspecified. Having no values for person and number, it cannot be deleted even in a consistent NSL like European Portuguese, since it is unable to value the ϕ-features of T.^[9] Saab proposes a different explanation, consistent with his theory of Ellipsis. He considers that a gente is a full DP, not a pronominal head; hence it cannot undergo ellipsis in European Portuguese because a consistent NSL only elides heads (in morphology), while radical NSLs elide phrases (in syntax). Discussion of the two theoretically well motivated explanations might be enriched by bringing into consideration novel observations and descriptive insights. Insular dialects of European Portuguese appear to provide relevant data. These dialects display a construction where a gente co-occurs with impersonal/indefinite se in subject position, as illustrated in (4) and (5).^[10]

Interestingly, a gente can be, and often is,^[11] null in this construction, as exemplified in (6) to (9). As Henrique Pestana (2023) points out, the subject clearly has a referential, definite interpretation in sentences (6) to (9). How easily the data of the Portuguese dialect of Madeira Island fit in the theories put forward by Saab, on the one hand, and Martins and Nunes, on the other, remains to be seen. Be it as it may, microvariation relative to a gente and its resistance to/allowance of subject ellipsis in Portuguese (hopefully bringing in the picture the different African varieties of Portuguese) is undoubtedly of theoretical interest as it may contribute to improve our understanding of (facets of) subject ellipsis in natural language.

Pilar Barbosa offers a critical thorough review of the extensive literature on null objects (and null arguments in general), with a focus on the properties of null objects in European Portuguese (EP) and Castillian Spanish. While European Portuguese has null objects interpreted as definite or indefinite, Castillian Spanish only allows indefinite null objects, which require a bare plural nominal or a bare mass noun as antecedent. Barbosa adopts from Raposo (1998, 2004 the idea that the contrast between European Portuguese and Castillian Spanish relative to the licensing of null objects is related to the properties/distribution of bare nominals in each language, and argues for Ruda’s (2017) hypothesis that a base-generated rootless nP proform underlies the derivation of both definite and indefinite null objects. Besides, Barbosa retains from Tomioka (2003) the proposal that the relevant null arguments denote properties, and their semantic differences result from general (independently needed) type-shifting operations – whose precise formal implementation Barbosa details for each of the languages under scrutiny (Existential Closure yields an indefinite interpretation; otherwise, a specific/definite interpretation arises, contextually retrieved). A central component of Barbosa’s investigation is the discussion on how the fact that definite null objects and epithets share similarities (they can, but need not, be A′-bound; they cannot be A-bound) may be made into an argument supporting the base-generated null nominal approach to object drop (in alternative to argument ellipsis and other approaches).

The comparative syntax research presented in Maria Lobo’s article explores syntactic variation between closely related grammatical structures within a particular language, with the goal to identify the theoretically motivated factors that simultaneously allow and limit such variation. Lobo systematically compares for a defined set of properties different types of gerund clauses in European Portuguese, namely peripheral adverbial, central adverbial, predicative and como (‘as’) gerund clauses. After testing and discussing each construction in detail, Lobo proposes that the syntactic differences between them can be derived by the interplay of four factors: (i) their more/less expanded functional architecture (only adverbial gerund clauses are CPs; predicative gerund clauses lack the C‒T domain); (ii) their structurally higher (‘peripheral’) or lower (internal/‘central’) locus of merge; (iii) the existence of an Agree/matching relation between embedded T features and matrix T features when matrix T c-commands embedded T (which arises in central, but not in peripheral, adverbial gerund clauses); (iv) the blocking of the Agree relation by a head intervening between matrix and c-commanded embedded T (the como Predicative head of como gerund clauses).

The article by Eugenia Duarte starts from the observation that Brazilian Portuguese blatantly diverges from European Portuguese regarding third person object pronouns, as colloquial BP generally replaced accusative and dative third person clitics with either null objects (reanalyzing a strategy also possible in EP)^[12] or non clitic forms of third person pronouns. The author’s aim is to trace the change back to the moment and circumstances of its emergence, testing the hypothesis that third person clitics were never acquired by the large population of African slaves (mainly speakers of Bantu languages) who acquired Portuguese as a second language (L2), thus originating a change later diffused outside the African-descendants communities.^[13] To this aim, a diachronic investigation of theater plays written by Brazilian and Portuguese authors from the eighteenth to the late twentieth century is undertaken and a quantitative and qualitative description of the data presented. Comparison between the BP and EP corpora shows a broad similarity in the use of third person clitics on both sides of the Atlantic until the middle of the twentieth century, when their rate of occurrence in the plays written in Brazil sharply drops and the contrast between BP and EP pops up crystal clear. This chronology is unexpected given Duarte’s hypothesis that the change arose as the effect of L2 acquisition of Portuguese by native speakers of Bantu languages, in colonial Brazil. However, the abrupt fall of frequency of third person clitics in the middle twentieth century theater plays may be interpreted as indicating a significant delay between the starting point of the change and its appearance in written sources. Thus, putting together, in addition to the quantitative corpus data, elements of demographic history, literacy history and the still clear gap between colloquial and formal/written language in Brazil, Duarte makes a case for her hypothesis of syntactic change prompted by L2 acquisition, notwithstanding the impossibility of attesting it close to its initial stage.

Duarte’s research, which closes the series of 11 articles dedicated to Professor Mary Aizawa Kato, thus agrees with the programmatic claim of Meisel (2024), who in an article of the same series published in Probus 36.1, states: “L2 learners are (…) the most likely agents of fundamental syntactic change.” (Meisel 2024: 31); “Non-native grammars can indeed differ from native ones in core properties, and L2 learners are therefore possible agents of such changes, provided they play a significant role in language transmission. This need not imply that they are a socially dominant group within a speech community” (Meisel 2024: 48). Closed the cycle and our task as editors, we wish to thank all the authors for their scientific contributions and for their willingness to join this tribute initiative to “express recognition and gratitude for all the influence and inspiration Mary has been providing to so many scholars in linguistics” (from the Preface to Probus 36.1).