Towards a diachronic typology of individual person markers

2.1 Independent pronouns and bound indexes

Siewierska (2004: § 2.1; § 3.4.1) provides a thorough typological introduction to the differences between independent pronouns on the one hand, and bound indexes on the other hand. For a variety of reasons, independent pronouns are typically treated as being the primary type of person markers. That is to say, starting with their morphophonologically free status, independent pronouns are also virtually universal across the languages of the world (i.e., much more so than bound indexes), and, if the paradigms of independent and dependent forms differ, independent paradigms tend to be richer, typically displaying more semantic distinctions (e.g., clusivity, number, or gender).

Also taking a diachronic perspective, the traditional view suggests that independent forms are primary, with bound forms being reduced versions of the independent forms (i.e., the commonly cited scenario for the evolution of verbal person markers in Givón [1976], also termed Givón’s Cycle in Seržant, this volume).

However, Mithun (2013) provides a compelling argument that especially a diachronic perspective may also yield a very different insight. By providing a case study of Mohawk (Iroquoian, northeastern North America), she shows that it may be the bound verbal indexes that represent a semantically differentiated and functionally consistent person paradigm, while independent forms are the opposite, displaying fewer semantic distinctions and occurring in a scattered set of different discourse functions.

There is also robust cross-linguistic evidence that either whole paradigms or individual person forms have originated in the direction of bound > independent, rather than in the grammaticalization-conforming direction of independent > bound forms (Siewierska 2004: § 7.1.3; cf. also Mushin and Simpson 2008).

Besides developments between the two types of paradigms (independent vs. bound), it is obvious that there must additionally be other kinds of diachronic changes differentially affecting the two types. As DeLancey (this volume) proposes for the Trans-Himalayan language family, bound 2sg forms may experience a significant pressure for avoidance not experienced by their free 2sg pronoun ‘counterparts’, due to the bound forms’ obligatory morphological status.

There must also be clear differences in source constructions from which independent pronouns and bound indexes originate. A case in point are complex nominal expressions, which may give rise to independent pronouns, such as Spanish Vuestra Merced ‘your grace’ > usted 2sg.hon (Siewierska 2004: 224). Clearly, such a nominal expression would not be able to give rise to a bound person marker, at least not immediately without first becoming an independent pronoun.

Nonetheless, there are also interesting parallels in sources that can give rise both to independent pronouns and bound indexes. For example, items expressing spatial deixis can make it into both, the paradigm of independent pronouns (consider demonstrative origins of independent pronouns, for all three persons, see the following section) as well as the paradigm of bound person indexes, where cislocative (venitive) markers are known to develop into SAP object indexes (Konnerth 2015; Mithun 1996). Of course, the morphosyntactic statuses of demonstrative pronouns and cislocative markers are different and so are the actual source constructions and diachronic pathways, but it is still interesting to see the analogous metaphorical correspondences.

2.2 SAP vs. 3rd person

Person marker paradigms in languages often consist of SAP forms and 3rd person forms. Clearly, any typological approach needs to keep all person forms in mind. At the same time, our diachronic approach agrees with previous literature (see, e.g., Siewierska 2004: 5–8) that there are substantial differences between all SAP forms on the one hand and third person forms on the other hand.

First, on a more general cross-linguistic level, there are patterns of diachronic development that either affect SAP forms or third persons. A good example are various pathways that lead to so-called Hierarchical Alignment or Inverse Alignment in verbal person indexation (Gildea and Zúñiga 2016), which may involve differential morphosyntactic treatment of SAP’s as opposed to 3rd person participants, although more specific configurations are also common. Another example is that zero forms in bound verbal person indexes more commonly arise for third person rather than SAP forms (Bickel et al. 2015; Cristofaro, this volume).

Cross-linguistic tendencies for differences in diachronic development of SAP vs. third person forms also correlate with different motivations and triggers for change. SAP forms experience sociopragmatic pressures, both in the form of reference avoidance as well as in special highlighting for reasons of seeking empathy (Section 4.1). Such sociopragmatically motivated pressures do not apply in the case of third person forms, which refer to participants outside the sphere of you-and-me. Loss or renewal of third person forms are more commonly tied to language contact, and other typical internal variation and change developments also found in grammatical domains other than person.

Second, considering particular developments in single languages, it is common to find attestation of more idiosyncratic changes that affect all SAP forms yet not third person forms, or vice versa. For example, Ciucci (this volume) shows how the realis/irrealis distinction in Chamacoco was first neutralized for SAP forms only. Another example is that object indexes may only ever develop for SAP’s but not for third persons (Konnerth 2015), i.e. a special case of the above-mentioned greater likelihood to find zero expression in third person rather than SAP’s.

Despite the differences, there are also some intriguing similarities we find in the diachronic developments of SAP and third person forms. For example, both types of person forms may originate from the same conceptual source category. This is particularly evident in forms originating in spatial deixis, such as cislocative markers for SAP object forms (Gildea and Zúñiga 2016; Konnerth 2015; Mithun 1996) or the commonly attested case of demonstrative elements turning into regular third person pronouns.

2.3 Person scenario markers, including inverse markers

While clear 1sg forms, for example, may be identified across all bound and independent paradigms of some languages, other languages may have person markers that are only used as verbal bound indexes occurring with (di)transitive verbs. Specifically, we have in mind markers that do not refer to a single person (group) but to one or a set of person ‘scenario(s)’ where arguments act on one another. An example is (1) from Yakkha (Trans-Himalayan/Sino-Tibetan), where -nen is a person scenario marker that occurs whenever a first person acts on a second person (conventionally represented as ‘1→2’).

Another example is the cislocative suffix -(i)m in Nez Perce, which has conventionalized as a person scenario marker for 2→1 (Gildea and Zúñiga 2016; see also Mithun 1996; Rude 1985) in (2).

The fact that innovative forms for particular person scenarios develop shows that there are diachronic pressures that act at that level of specificity. In the literature, we find discussion on (certain types of) such forms mostly within the framework of inverse marking and hierarchical indexation. Heath’s (1991, 1998) discussion of a wide variety of changes occurring in person forms in SAP→SAP scenarios also provides fascinating evidence for the need to take person scenarios seriously. There are other, particular proposals that fall under this line of research. For example, Mithun (1996: 428–429), citing data from Northern Iroquoian languages, suggests that an innovative, cislocative-derived 2→1 marker specifically in imperative forms finds a plausible explanation in the sociopragmatics of this specific speech act. DeLancey (2018a) finds a common pattern of innovative 1→2 markers in Trans-Himalayan, as well as argues that there are recurrent cases of 2→1 marking to merge with 3→1 in this language family, both of which he explains with reference to sociopragmatics (cf. Section 4.1). Note that the latter merger of 2→1 and 3→1 may result in unified first person object marking. This clearly shows that pure person markers sometimes develop from person scenario markers as well as crucially based on person scenario specific motivations.

The diachronic connection between pure person markers and person scenario markers also surfaces in other developments. A case in point is the recently discussed innovation of inverse marking (specifically, the combination of 3→SAP plus 2→1 marking) from antipassive origins via a reconstructed first person object marking stage (Konnerth 2021).

All of the just-mentioned cases of person scenario markers are innovations within the person paradigms where they are found. This is typical, and as a result, person scenario markers commonly have a morphosyntactic status that is in some way or another non-paradigmatic (e.g., occupying a different morphological slot). However, this is not anything unique to person scenario markers, as pure person markers can likewise be innovated from extra-paradigmatic material and remain non-paradigmatic. An example is a second person prefix preliminarily reconstructed as *tV- by Pons (to appear) for Proto-Trans-Himalayan (see also DeLancey 2014). In many modern languages, a reflex of this prefix has become a conventionalized index for second person, remaining in prefixal position among otherwise suffixal verbal person indexes.

3.1 Internal sources

Innovative person markers may derive from category-internal sources either by being ‘re-purposed’ from already existing person markers through person shift, or else through fusion of already existing person markers.

3.2 External sources

The main external sources for person markers attested across languages include impersonal and valency-decreasing constructions, as well as demonstratives/spatial deictic expressions, nominal expressions, and intensifiers/reflexives. As will be shown below, there are recurrent pathways of change through which each source type is preferentially reinterpreted as a specific person marker, based on the specific conditions that foster such a process of reinterpretation. The motivations for these pathways of change will be discussed in more detail in Section 4.

3.3 Fusion of person markers and other elements

New person markers may develop through fusion of original person markers with elements they were bound to. For example, Siewierska (2004: § 7.1.4) documents the innovation of person markers based on conjugated verbal forms, apparently typically conjugated copulas or other auxiliaries, although other verbs may also enter this diachronic path. The result may be new independent pronouns and specifically new emphatic forms. Moreover, the result may also be new verbal person indexes, when larger periphrastic verbal constructions fuse. This latter pathway to new verbal person indexes is also discussed by Seržant (this volume). Third, Siewierska also mentions the innovation of possessive person markers resulting from fusion and reanalysis of conjugated forms of possessive verbs ‘to have’ or ‘to own’.

Elsewhere, Siewierska also mentions another case of innovation through fusion. In Teribe, a Chibchan language of Costa Rica, we find the innovation of specific case forms of independent pronouns through fusion of person marker plus adposition (Siewierska 2004: 48, based on Quesada 2000: 46).

One also comes across hypotheses that new person markers may derive from fusion with information structure markers (e.g. in different Trans-Himalayan languages, see Benedict 1995; DeLancey 2018b: 79, this volume), and, perhaps similarly, with a “morpheme of identification” in Yaouré, a Mande language of the Ivory Coast (Siewierska 2004: 69, discussing data from Hopkins 1986: 196).

4.1 Sociopragmatics

Sociopragmatic considerations (in the sense of DeLancey 2018a) appear to be an essential factor of change in the realm of person marking involving the SAP’s. Avoiding direct reference to specific participants is the reason why impersonal and valency-decreasing constructions are recruited to refer to speech act participants in the first place, but is possibly at work also in other scenarios.

In a pioneering paper, Heath (1991) has drawn attention to the fact that person markers that refer to speech act participants frequently have shapes that are unpredictable from the rest of the paradigm. In many Australian languages of the Northern Territory where affixes index the subject and the object at the same time, for instance, the combinations 1→2 and 2→1 (I saw you; you saw me) display unusual features. While in configurations involving 3rd persons and speech act participants (3→1; 1→3; 3→2; 2→3) the verb complex is regularly marked as inverse if the object ranks higher than the subject (i.e. in 3→1 and 3→2 scenarios), the same does not hold true when both speech act participants are involved (i.e. in 1→2 and 2→1 configurations). In the latter cases, we also find rampant morphological irregularity, including special allomorphs for 1st or 2nd person, unanalyzable portmanteau morphs, zero forms for one or the other participant or for both, cases of number and person shift, or combinations thereof. An instructive case is that of Ngandi. A partial paradigm of subject/object combinations in Ngandi is provided in (8):

The prefixes in (8) can be decomposed into smaller morphemes quite straightforwardly: in scenarios involving 3rd person and SAP’s, for instance, ŋa- is an exponent of 1st person singular, ṇi- is an exponent of 3rd person singular masculine (losing retroflexion when appearing non-initially), etc. On the contrary, the 1sg→2pl and 1pl→2 sg/pl prefix gura- exemplifies the kinds of irregularities listed above that characterize combinations involving SAP’s. The form can be analyzed as gur-na-, where na- is possibly the 2du/pl morpheme (as in ṇarguni-, 3sg→2du/pl): given that in all other configurations 1st person morphemes precede other person morphemes, we shall conclude that gur- is an exponent of 1st person. A look at the regular combinations in (8), however, shows that exponents of 1st person are ŋa-/ŋar-, and thus gur- (perhaps decomposable into gu-r, with -r- as plural) must be taken as a special allomorph of 1st persons limited to 1→2 forms. Another interesting case is 2sg →1sg ñunu-, with an unexpected shape with respect to the expected *ŋanu-. In other languages, mood and reality status morphs are recruited in 1↔2 configurations: this is the case of Jawoyn (Western Gunwinyguan) wal-, 1→2, which is the same as 3rd person irrealis wal-. In this case, an irrealis (polite, non-factual) form unmarked for person has been taken as a substitute for person morpheme combinations involving speech act participants irrespective of the reality status of the clause (Heath 1991: 83).

These irregularities are ultimately motivated in terms of the “pragmatic delicacy and dangerousness of using first and second person pronouns (particularly singulars), and the particular delicacy of combining them in a noun phrase or sentence in a manner overtly specifying their relationship to each other” (Heath 1991: 78). Similar phenomena are attested across Native American languages, as discussed by Heath (1998), and in Tibeto-Burman languages (DeLancey 2018a). As DeLancey (2018a: 354) argues, avoiding reference to the agent is the motivation behind all deviant 1→2 and 2→1 configurations: these scenarios are particularly likely to be sensitive to face issues, since “a 1→2 scenario is likely be a promise or a threat […], and 2→1 is the realm of requests, demands and accusations” (DeLancey 2018a: 354), and avoiding reference to the agent is notoriously one of the commonest politeness strategies across languages (Brown and Levinson 1987).

Sociopragmatic motivations for changes in the person marking domain are also discussed in contributions by DeLancey (this volume) and Ozerov and Konnerth (this volume).

Note, however, that the opposite of reference avoidance, which we may term ‘reference emphasis’ can similarly be a politeness strategy. Mithun (2013: 302) states that the overt use of the first person free pronoun in Mohawk, which is rare and only occurs in less than 10% of sentences involving a first person participant, is in fact a politeness strategy: “<This is just my opinion or experience or feeling, not necessarily shared by you or others.>”

Another interesting phenomenon is person shift connected with specific expressive needs in narrations. As discussed by Margetts (2015), in many languages 3rd person referents are indexed by 1st or 2nd person pronouns in events that are highlighted for some reason. These person shifts function as attention-getting devices. In Saliba-Logea (Oceanic), for instance, characters in narration can be indexed by a second person index (ku-) instead of the 3rd person one (ye-) to mark the narrative peak (e.g. sudden and unexpected events). Although 2-to-3 and 1-to-3 shifts in person marking are not described in Helmbrecht’s (2015) survey (and are similarly not attested in Bates’ survey, this volume), it is a fact that cases of 1/3 and 2/3 homophony in singular paradigms are widely attested (see Cysouw 2003: 41–48), and at least some of these cases might be explainable as the long term result of expressive shifts such as those described by Margetts (2015).

4.2 Areal effects

Areal effects are likely to shape person markers paradigms. As is well-known, the use of a 2nd person plural pronoun for polite 2nd singular appears to be a Standard Average European phenomenon (cf. Helmbrecht 2005). Cysouw (2003) mentions various cases in which the configuration of the person paradigm may depend on areal contact. A case in point is the cross-linguistically quite rare 1/3 non-singular homophony, i.e. the use of the same form for 1st and 3rd person plural, which is attested in a handful of only distantly genetically related languages on the south-eastern tip of Papua New Guinea (Cysouw 2003: 134–135). Authier (this volume) shows that in East Caucasian there is a tendency to retain clusivity (or to renew it after loss) that is likely to depend on areal effects in an area where this distinction remains predominant.

5.1 Stability vs. susceptibility for change

Some person markers are reported to be diachronically more stable than others. Personal pronouns, for instance, “are on all the lists of relatively stable lexemes” (Nichols 2003: 292), but cases of restructuring due to paradigmatic pressures are also widespread across languages. Nichols (2003: 292–295) discusses the cases of restructuring that can be reconstructed in the development of Nakh-Daghestanian languages: these cases appear to be driven by the principle of resonance, whereby forms that are phonologically rather different from one another tend to develop similar phonological traits (e.g. in terms of rhyme or syllabic structure).

Nichols (1995) includes the inclusive/exclusive opposition in her list of diachronically stable morphosyntactic features, and discusses whether such stability is indicative of areal or genetic tendencies (see also Bickel and Nichols 2005). While the lack of clusivity oppositions in Indo-European languages might speak in favor of genetic inheritance of this specific trait (especially if we consider the rampant presence of such distinctions in the Caucasus), the universality of these distinctions in Australia might be indicative of areal effects. Moreover, clusivity distinctions display a clinal distribution (almost 100% in Australia, around 50% in the Americas, near zero in Europe and Africa), which shows that “if there were any appreciable tendency for [them] to be spontaneously innovated, [their] worldwide frequency would be more even” (Nichols 2003: 304). In other words, inclusive/exclusive oppositions seem to be generally inherited, with a residual possibility to be created ex nihilo under areal influence. Wichmann and Holman’s (2009) measures of diachronic stability of inclusive/exclusive oppositions confirm a high degree of stability for this feature, with no significant differences between independent and bound pronouns.

Authier (this volume) discusses the history of East Caucasian inclusive and exclusive indexes and shows that while inclusive indexes are more prone to be maintained or renewed (possibly due to areal influence), their exclusive counterparts may be lost due to avoidance strategies. The result of these dynamics is that clusivity distinctions may disappear as the result of the loss of one member of the inclusive/exclusive pair, resulting in reinterpretation of the surviving member as a general 1st person plural marker.

Seržant (this volume) also turns to the question of stability, but particularly in light of the possible loss of a whole indexing system. Counter to the well-known account by Givón (1976), which assumes loss of verbal person indexation followed by new grammaticalization from independent forms, Seržant argues against a universal bias for reduction/attrition of pronominal indexes, finding it very rare in the world’s languages.

5.2 Optionality and variation

Synchronic variation in the use of person markers may be indicative of on-going change. The process through which on in French has become a 1st person plural independent pronoun has been gradual: on was used since Old French as a stylistically marked substitute for other personal pronouns, but its use as a 1st person plural pronoun “was [not] general in the Paris area before the nineteenth century” (Coveney 2000: 450).^[2] Similarly, the 1st person plural interpretation of the reflexive impersonal construction in Italian (see Section 3.2) is already available in texts from the 14th and 15th century, but it is not before the end of the 18th century that this use gains ground, even though it has not become obligatory in contemporary Italian (Giacalone Ramat and Sansò 2011: 220–223). In Sango (Ubangi), the optional use of álà (2nd person plural) as a polite 2nd person singular form, fostered by contact with French, is determined by social and demographic factors and is likely to survive in the foreseeable future (Heine and Song 2011: 609). In varieties of European Portuguese, a gente may trigger 1st person plural agreement, which is not possible in Standard European Portuguese: this means that in these varieties its process of reinterpretation as a 1st person plural pronoun is far more advanced than in the standard variety (Posio and Vilkuna 2013: 207).

All these facts show that changes leading to the emergence of new person markers follow the usual steps of grammatical change in terms of becoming obligatory after rather long periods of free variation. Looking at individual and discourse variation in the use of person indexes is thus crucial to capture the mechanisms of change in this realm. Ozerov and Konnerth (this volume) show by means of a synchronic analysis of Anal Naga discourse that the inclusive form in this language may be shifting towards a 1st person singular index, with contexts in which it is used to express shared responsibility and involvement acting as potential bridging contexts in the sense of Heine (2002).