Introduction: towards a diachronic typology of the middle voice

1 Preliminaries

This special issue finds its root in a thematic workshop that we have organized at the 53rd meeting of the Societas Linguistica Europaea, held online in September 2020. The workshop originated from the observation that, in spite of the typological work that has been devoted to the middle voice over the past decades, a systematic diachronic typology of this domain is still lacking, so that we only have an impressionistic understanding of how middle voice systems (henceforth, MVSs) emerge and develop in the world’s languages. This is particularly unfortunate, since, as we discuss in Section 3, diachronic considerations are of paramount importance to achieve a better understanding of the nature of the middle voice. The goal of the workshop was therefore to bring together scholars working on the middle and related phenomena in a typological and diachronic perspective, in order to provide a forum for discussion of new data tackling the core question of how middle voice markers emerge cross-linguistically. Our expectations where happily met by the high-quality contributions that were presented at the workshop and by the stimulating scholarly discussion that followed, which prompted us to put together this special issue.

2 The middle voice in linguistic typology

The identification of the middle as a distinct type of verbal voice was first proposed in antiquity to describe the opposition between different sets of verbal endings in ancient Indo-European languages such as Ancient Greek and Latin (Zúñiga and Kittilä 2019: 169–171; see also Inglese forthcoming for a more detailed discussion). For this reason, research on the middle voice has long remained within the boundaries of Indo-European linguistics (Benedetti 2016; Inglese 2020: 9–14), where the middle voice has been traditionally described as the voice category concerned with the involvement of the subject in the event expressed by the verb (e.g. Benveniste 1966; Gonda 1961). This definition of the middle was essentially carried over to general linguistics (Lyons 1968: 373) and it also served as the basis for early descriptions of comparable phenomena in non-Indo-European languages, including for example 19th missionary descriptions of Bantu languages (see Dom et al. 2023).

It is only starting from the 80s that the middle voice has entered the research agenda of linguistic typology, thanks to the groundbreaking cross-linguistic studies on reflexive and middle voice markers by Geniušienė (1987), Klaiman (1991), and especially Kemmer (1993). Drawing from the analysis of a 32-language sample, Kemmer argued that middle markers cross-linguistically occur on a well-defined set of situation types, which subsume both valency-related functions such as reflexive and passive, and also other semantic classes of verbs, including e.g. verbs of motion and experiencer verbs.

In spite of Kemmer’s influential work, the middle voice has remained an elusive notion in typological studies. This is partly due to the lack of an explicit comparative concept of what counts as a middle marker in Kemmer’s work, but also to a long-lasting descriptive tradition to use the term middle to refer to quite disparate phenomena, to the extent that “middles […] represent a major terminological problem area” (Zúñiga and Kittilä 2019: 151; see also Dixon and Aikhenvald 2000: 11–12). A general consensus exists regarding the fact that middles are inherently polyfunctional constructions (Klaiman 1991: Chapter 2; Kulikov 2013: 256–257; Zúñiga and Kittilä 2019: 176), but even more recent cross-linguistic studies operate with distinct definitions of middle voice: Bahrt (2021: 74) uses the term middle to refer to syncretic intransitivizers, whereas in Inglese (2022a), middle markers are defined as those markers that, besides intransitivizing functions, also show a lexically obligatory distribution with verbs that can only occur with middle marking (see also Inglese and Verstraete 2023). In Inglese (2022a), these two macro-classes of middle verbs go under the name of oppositional and non-oppositional middles (corresponding to Klaiman 1991: 106 alternating vs. non-alternating). Typical functions of oppositional middles include valency changing operations such as anticausative, passive, antipassive, reflexive, reciprocal, impersonal, facilitative, and involuntary agent constructions (Bahrt 2021; Dixon and Aikhenvald 2000; Kulikov 2010; Zúñiga and Kittilä 2019). Non-oppositional middles show a consistent distribution across languages, as they tend to occur with a specific set of verb classes: grooming verbs, verbs of position and of change in body posture, verbs of (non-)translational motions, verbs of speech, experiencer verbs, and spontaneous events.

Besides the non-trivial issue of how to identify and compare middle voice constructions across languages, another key problem is how to explain the observed recurrent similarities that middle markers display cross-linguistically. In this respect, Kemmer reached the conclusion that the middle domain cross-linguistically constitutes a “coherent but relatively diffused category that comprises a set of loosely linked semantic subdomains” (1993: 238). In particular, she famously proposed that situation types expressed by middle markers can all be linked, more or less directly, to a core semantic property that can be described as low degree of elaboration of events. In a nutshell, Kemmer’s idea is that situation types that receive middle marking are characterized by involving two participants that are, however, not fully physically and conceptually distinguishable (also Næss 2007: 22–24, 27–29).

The suitability of the notion of elaboration of events as the overarching functional motivation for middle marking has more recently been called into question on at least two grounds. The first objection comes from synchronic data: once a wider language sample is taken into account, cross-linguistic data does not point to a particular connection of middle constructions and reflexives and/or grooming situations (Inglese 2022a: 521), which instead constitute the core of middle marking in Kemmer’s account. The second objection is diachronic in nature, and we briefly outline it in Section 3.

3 The diachronic typology of middle voice markers

Diachronic studies of middle markers remain few, to the effect that nowadays there is no comprehensive diachronic typology of MVSs. Existing studies focus almost exclusively either on the lexical sources of valency-related functions or on the historical connections among these (Bahrt 2021: Chapter 7; Zúñiga and Kittilä 2019: Chapter 8). Dedicated studies have appeared on passives (Haspelmath 1990; Wiemer 2011), reflexives (Evseeva and Salaberri 2018; König and Siemund 2000; Schladt 2000), reciprocals (Heine and Miyashita 2008), and antipassives (Sansò 2017). What is lacking in most of these studies, however, is a careful consideration of how oppositional, i.e. grammatical, functions of middle markers historically relate to non-oppositional, i.e. lexical ones.

The mainstream view on the development of middle markers holds that (i) reflexives are the preferred source of middle markers across languages (ii) there is a unidirectional path from oppositional to non-oppositional functions (Haspelmath 2003; Kaufmann 2007; Kemmer 1993). It must be said that historical evidence typically adduced in support of these claims mostly comes from better-known modern Indo-European languages of Europe (that is, Germanic, Romance, and Balto-Slavic languages) (cf. Kemmer 1993: Chapter 5). In a recent survey on the diachrony of middle marking in a 129-languages sample, Inglese (2023) found that neither of these assumptions is borne out by actual data. In the first place, middle markers originate out of a much richer pool of sources, including lexical verbs of various types, non-reflexive valency changing markers, markers of plurality, markers of spontaneous events, nominalizers and verbalizers, and even aspectual markers. In the second place, there is evidence that middle markers may originate as markers of non-oppositional middles and only develop oppositional functions at a later stage, thus countering the purported unidirectionality of the oppositional > non-oppositional shift (e.g. Kaufmann 2007: 1688). These and other findings suggest that a reassessment of the diachronic typology of the middle voice domain is in order, so as to explore the full range of possible sources and mechanisms of language change that may lead to the emergence of middle voice systems.

Diachronic considerations may also play a key role in reshaping current explanations of the polyfunctionality of middle markers. In recent years, typologists working in a source-oriented perspective have strongly advocated the need to integrate a diachronic component in the explanation of cross-linguistic regularities (e.g. Bybee 2006; Cristofaro 2010, 2019, 2021; Sansò 2018; but see also Haspelmath 2019 for a critical discussion). In the case of the middle voice, a comparable approach has already been suggested by Haspelmath (1995: 373), who wondered whether the occurrence of middle marking across a variety of contexts may not be due to a unitary semantics that all these participate in, but rather to independent historical developments that led to the extension of middle marking from one individual context to another. In this respect, the notion of middle voice would not be dissimilar from the controversial typological notion of irrealis (Sansò 2020). In both cases, recurrent polyfunctionality clusters find their ultimate motivation in the diachronic processes that account for their emergence in the first place and not in a single overarching semantic component.

While persuasive, Haspelmath’s suggestion has only recently been followed up in the study of individual languages (Holvoet 2020 on Baltic languages; Inglese 2020: 243 on Hittite; Post and Modi 2021 on Macro-Tani languages), and has not yet been tested on a large scale, partly following from the lack of solid empirical evidence concerning the origin and development of middle markers cross-linguistically.

4 Contributions to this special issue

The five papers collected in this special issue offer in-depth case studies on middle markers in languages from various areas of the world, and all contribute to elucidating hitherto unexplored aspects of the diachronic typology of the middle voice.

Carlota de Benito Moreno offers a thorough corpus study concerning the origin of non-oppositional middle verbs marked with se in Spanish. Thanks to the careful analysis of historical corpus data, de Benito Moreno compellingly shows that the obligatorification of reflexive marking with intransitive verbs that only allowed optional reflexive marking in origin, e.g. reir(se) ‘laugh’, can, at least in some cases, be explained as the result of the analogical extension of se-marking from semantically similar verbs, with which se occurred to signal a valency-related operation, as is the case of conversive experiencer verbs, e.g. alegrar ‘please’ → alegrar-se ‘be pleased (of)’. This finding further enriches our knowledge of how non-oppositional middles may historically come about.

The existence of multiple middle markers within the same language is addressed in two papers (see Inglese 2022b for a typological overview). Matthew Zaslansky traces the history of middle marking in the Turkic language family. Turkic languages offer a particularly interesting piece of evidence because in this family two middle suffixes that have a clearly distinct etymology in origin, that is, ‘reciprocal’ -(I)š and ‘reflexive’ -(I)n, progressively converged in their range of usages. This resulted in the creation of a bipartite middle voice system in which the two suffixes coexist in a situation of paradigmatic competition and become increasingly associated to non-oppositional (or deponent) verbs. Sebastian Dom, Leora Bar-el, Ponsiano Sawaka Kanijo, and Malin Petzell focus instead on Kagulu, a Bantu language from Tanzania. Recent studies dedicated to the function of verbal affixes in the Bantu languages have pointed out how some of these in fact behave as middle markers in individual languages, albeit with a narrower set of functions when compared to the better-known middle of ancient Indo-European languages. The authors contribute to this ongoing debate by offering a thorough description of detransitivizing suffixes in Kagulu and their origin, showing that only a sub-set of said affixes in fact qualify as middle marker. The findings of these two papers show that middle affixes can stably coexist over time, against the communis opinio that in two-form middle systems, one marker takes over in time, and also show that non-oppositional middles do not merely represent idiosyncratic lexicalizations but may also constitute a productive class.

The issue of the directionality in the development of middle voice systems is tackled by Guglielmo Inglese and Jean-Christophe Verstraete. The authors discuss evidence from two unrelated language families, that is, Anatolian (Indo-European) and Paman (Pama-Nyungan), and argue that in both cases, constructions that can rightfully be described as middle voice markers were originally confined to non-oppositional verbs on a lexical basis, and only subsequently developed oppositional valency-related functions typical of middle markers. These results show that the unidirectionality of the oppositional > non-oppositional shift cannot be upheld.

The paper by Karaj and Sansò deals with Malayic languages, and discusses the case of prefixes derived from Proto-Malayic *(mb)AR-, whose multifunctionality patterns include functions that are typically encoded by middle markers across languages but, crucially, they are also used as denominal verb formatives, i.e. as strategies used to derive intransitive verbs with various meanings from nominal bases (e.g. ‘have, contain, use N’). Karaj and Sansò provide a diachronic explanation for such an unusual combination of functions and show that the denominal verbalizing function of reflexes of *(mb)AR- is primary in diachronic terms. Moreover, they sketch a diachronic scenario through which a denominal verb formative has extended to cover a portion of the functional space usually encoded by middle markers across languages. This diachronic scenario revolves around the semantic effects brought about by an aspectual component of the denominal verb formative.