From reflexive clitic to marker of human impersonal subjects: the Resian reflexive sa

3.1 Reflexive clitics in Resian

The Resian system of pronominal clitics is largely identical to that in standard Slovenian. However, there is one curious difference that is of great relevance for the analysis of reflexives in impersonal constructions. It concerns the accusative form of the reflexive clitic where we find two variants: refl se/sa (Steenwijk 1992: 118–119). The entire paradigm of the Resian reflexive pronoun is given in Table 2.

Although the variation in the accusative form of the personal pronouns – it is characteristic also for 1st-person pronominal clitic: me/ma ‘I.acc’ (Steenwijk 1992: 118–119) – has already been noticed by Baudouin de Courtenay (1875: 52–54), its origin is still unclear. In Common Slavic, the accusative ending consisted of the nasal vowel *ę (probably [ɛ̰] or [æ̰] in IPA notation). This vowel, according to Baudouin de Courtenay (1875: 54), has various reflexes in Resian, among them e and a. In more recent descriptions of Resian historical phonology, authors either mention only e as the reflex of Common Slavic *ę or at least do not provide a discussion of other variants (e.g., Vermeer 1987: 246, Šekli 2015: 204–205). Therefore, it seems that the origin of the variation in the accusative form cannot be determined with certainty. The most likely scenario is that the outcome of the Common Slavic nasal vowel differed depending on the phonological environment. This is a plausible explanation because Resian has undergone a number of complex vowel changes as a result of which the quality of a vowel is to some extent depending on the quality of the vowel in the subsequent syllable.^[10] Considering that in Slavic, clitics together with their host formed a phonological word,^[11] it is reasonable to assume that clitic personal pronouns (including the reflexive) were affected by various of these assimilation processes. Due to the significant changes affecting the rules determining clitic placement in Resian, the original distribution of sa and se would, however, have become opaque (Steenwijk 2023: 72).

Be this as it may, the distribution of the two reflexive variants in the main resource investigated in this study suggests that they have received a new function. As is shown in Section 3.5, the variant sa seems to have developed into a subject marker with an arbitrary human referent. Before this issue as well as the rules governing the placement of the reflexive clitics in impersonal constructions (Section 3.4) are addressed in detail, it will first be shown that Resian reflexive impersonals exhibit the same properties as the corresponding constructions in standard Slovenian and Polish (Sections 3.2 and 3.3).

3.2 Reflexive impersonal constructions shared with standard Slovenian

Like standard Slovenian, Resian allows the formation of reflexive impersonals from both transitive and intransitive verbs. Regarding verbal agreement, the characteristics of these constructions do not differ in the present tense. We always find the finite verb in the 3rd person singular. However, in the past participle I have come across agreement patterns deviating from those observed in standard Slovenian. In Example 13a taken from The little prince, the participle počarnjali ‘blush’ in the second of the two coordinated reflexive-impersonal clauses unexpectedly takes the plural ending and not the ending of the neuter singular (cf. Section 2). If this is not simply an error, it could be interpreted in at least two ways. First, we could be dealing with an instance of agreement ad sensum. As has been pointed out in Section 2, depending on their environment (e.g., verb semantics, presence of adverbs), reflexive impersonals can have the reading of a subject with the meaning ‘many/few people’ (cf. Rivero and Milojević Sheppard 2003: 123–131 based on Chierchia 1995). Therefore, one could argue that it is these quantifying semantics that trigger agreement according to sense, i.e., agreement in the plural. An instance of agreement ad sensum with a quantifier can be found in The little prince. In Example 13b, the plural subject clitic ni ‘they’ doubles the quantifier phrase malö njyh ‘few of them’ where the quantifier malö is neuter singular (cf. Wandl Forthcoming). A parallel phenomenon could be invoked to explain the plural form in Example 13a.

As an alternative, influence from Italian could be considered. In Italian, reflexive impersonals show curious agreement patterns (cf. Maiden and Robustelli 2013: 125–126). When introducing predicative adjectives, indefinite reflexives require the copula verb to take 3rd-person singular endings while the adjective appears in the masculine plural (cf. 14a). Moreover, past participles of intransitive verbs take plural endings if they appear with the auxiliary essere ‘be’ also outside reflexive impersonals.^[12] On the other hand, if an intransitive verb normally has avere ‘have’ as its auxiliary, the past participle appears in the masculine singular form. This can be seen in Examples 14b and 14c where arrivati and dormito appear in the masculine plural and singular forms, respectively. With transitive verbs, the patterns are even more complex (Maiden and Robustelli 2013: 125–126). However, since they involve the same forms of the predicative adjectives and participles, it is not necessary to discuss them separately here. What is important is that the plural agreement in počarnjali (Example 13a) could be interpreted as the result of language contact with Italian.

With regard to Example 13a, it is further interesting to note that in Slovenian, verbs like Resian počarnjati ‘blush’ are identified as unaccusative by Ilc and Marvin (2016: 149–150). This type of intransitives is not acceptable in reflexive impersonal constructions unlike a generic or habitual character of the expressed event is implied by the sentence or indicated by an adverbial (e.g., Ilc and Marvin 2016: 154 and Lenardič 2020). In Example 13a, the causal construction (ko… to) or perhaps the adverb mej ‘never’ even though it appears in a different clause could be interpreted as indicative of a habitual character. Accordingly, it must remain unclear whether Resian allows forming reflexive impersonals from unaccusative verbs at this point.

The second type of a deviating agreement pattern in Resian comes from Steenwijk’s (1992: 187) fieldwork data. In Example 15 (repeated as Example 31d below), the past participle has masculine endings instead of the expected neuter ones (cf. also Example 41 which is, however, ambiguous). Here, it is not possible to invoke agreement ad sensum as a possible explanation considering that a maškira, according to Steenwijk (1992: 279), is a ‘certain female dress for Carneval’. One possibility to account for this form is to refer to predicative constructions. As will be shown further below in this section, predicative adjectives, participles, and pronouns usually take the masculine rather than the neuter form in Resian reflexive impersonals. Therefore, one could argue that the masculine ending in Example 15 results from analogical adaptation to predicative adjectives. On the other hand, one could again argue that the masculine form is the result of influence from Italian (cf. Example 14c).

Finally, it should be kept in mind that differences in the agreement patterns between Steenwijk’s (1992) data and the data extracted from The little prince could also reflect inter-dialectal or even inter-speaker variation. Clarifying this, would be possible only based on a sufficiently large corpus of spoken Resian including data from different speakers. Unfortunately, such a resource does not currently exist.

As mentioned in Section 2, reflexive impersonals derived from intransitive verbs occur in all Slavic languages. Therefore, it does not come as a surprise that we also find them in Resian. Examples 16a and 16b contain two unergative verbs (sednut ‘sit down, take a seat’ and jtyt ‘go, walk’). In the latter case, the verb is embedded into a modal construction. Example 16c contains a verbum sentiendi with a complement clause containing a modal construction. The latter example is further interesting because it shows that in coordinated clauses with reflexive impersonals the reflexive clitic is repeated.

Of greater interest for this paper are reflexive impersonals derived from transitive verbs. In The little prince, I found transitive reflexives with their objects expressed by nouns and/or object clitics. Example 17a contains the verb ‘see’ which has wsake sjorte ‘all kinds.acc.pl’ specified by riči ‘things.gen.pl’ as its direct object. In Example 17b, the direct object is expressed by the clitic jo ‘she.acc’. Additionally, this example contains the object clitic jin ‘they.dat’. Examples 17c and 17d are interesting because they show that in Resian, relative pronouns can act as objects of reflexive impersonals. In both cases, the relative clause is introduced by the uninflected relative marker ki and contains the object clitic jin ‘they.dat’. However, the function of this clitic is different in the two clauses. In Example 17c, it represents an indirect object while the relative marker takes the role of the direct object; in Example 17d, on the other hand, jin constitutes a resumptive pronoun in the dative plural which accompanies ki while jïmë ‘name.acc.sg.n’ acts as the direct object of the clause.

The final example I want to discuss with regard to argument structure is 18. It contains the only instance of a genitive of negation in a clause with a reflexive impersonal that I have come across. Strictly speaking, the form saminče ‘seeds’ is ambiguous because the ending -e expresses both the feminine genitive singular and accusative plural cases. Negation alone is insufficient to determine the case of the object because in Resian, the genitive of negation is not consequently realized (Steenwijk 1992: 181). However, Considering that saminča is a collective noun (cf. Steenwijk 2005: 121), one expects a singular form here.

Example 18 is interesting for yet another reason. It shows that direct objects can be clitic-doubled in reflexive impersonals. Clitic doubling is a phenomenon that, within Slavic, is best known from Bulgarian, Macedonian, and the Torlak dialects of Serbian (cf., for instance, the contributions on Slavic languages in Kallulli and Tasmowski 2008; Escher 2021 on the Timok dialect of Torlak). However, it also occurs in Western Slovenian dialects (e.g., Krošelj 2013; Marušič and Rok 2013; Marušič and Žaucer 2009). In Example 18, the direct object saminče is doubled by the object clitic je.^[13]

Like in standard Slovenian, the subject of reflexive impersonals can antecede reflexive pronouns. Nevertheless, in The little prince, I have come across only one instance which demonstrates this. Example 19 contains a verb phrase consisting of rivät ‘finish’ and se wmyt ‘wash oneself’. The latter verb is a reflexive verb derived from wmyt ‘wash’ by adding se (cf. Example 1a). The addition of the tonic reflexive pronoun sabe is rather unexpected since it is not necessary to express reflexive meaning here. Neither can the appearance of both reflexive variants (clitic and tonic) be interpreted as an instance of clitic doubling since the clitic is not doubling the tonic pronoun, but rather the other way around. Therefore, I suggest that the function of the tonic variant is related to contrast. It is a characteristic of Slavic languages with pairs of tonic and clitic personal pronouns to express contrast or emphasis by means of the tonic variant (cf. Steenwijk 1992: 119 and Šekli 2010 on the function of the tonic personal pronouns in Resian). Therefore, the tonic pronoun sabe in Example 19 could serve the purpose of contrasting the object of the subordinate clause (‘oneself’) with the object of the main clause (‘planet’).

Considering the special function of the reflexive in Example 19 as well as the fact that it occurs together with a reflexive verb, one may doubt whether it provides as clear evidence for the appearance of reflexive pronouns as antecedents of reflexive impersonals as 5a does for standard Slovenian. To verify whether reflexive impersonals can indeed antecede reflexive pronouns, I therefore construed a parallel sentence in Resian which was judged acceptable by a native speaker. The sentence is given in 20. Apart from the vocabulary, it differs from the standard Slovenian example in 5a in that the verb ‘think’ in the expression ‘think of’, in Resian, is formed with the dative reflexive clitic si.^[14] Moreover, it requires the preposition za + acc instead of na + loc. In the accusative, Slovenian personal pronouns have a specific form that is used with prepositions. It is usually referred to as “bound” and, in script, is written together with the preposition (e.g., Herrity 2016: 128–130). In the Resian orthography developed by Steenwijk, personal pronouns in the accusative are separated from prepositions by hyphens (Steenwijk 1994: 17), viz., Resian za-se ‘for oneself’.

Like in standard Slovenian, Resian impersonal reflexives can antecede reciprocal constructions. This is shown by Example 21, where the reciprocal construction ‘one another / each other’ is dependent upon the infinitive ǵat ‘put’. The latter is embedded into a modal construction with the verb ‘have’ (here: ‘have to’) modified by the particle ba, which, according to Steenwijk (1992: 186), has an attenuative meaning.

Resian reflexive impersonals can also act as controllers of passive complements. Since The little prince does not contain an example proving this, I again, constructed a sentence corresponding to the standard Slovenian Example 7) and had it judged acceptable by a native speaker. The sentence is given in 22. Note that the modal verb ćë ‘want.prs.3sg’ could also be replaced by the construction ma wojo ‘have.prs.3sg will.acc.sg.f’.

Finally, reflexive impersonals in Resian, can also be formed from copular constructions. Examples are given in 23. In the first two examples the predicative expressions are realized by adjectives (Examples 23a and 23b). Since adjectives in Resian (and Slavic generally) are targets of gender agreement, these examples are of special interest for determining the properties of subjects in reflexive impersonals. As mentioned in Section 2, native speakers of Slovenian seem to vary with regard to their acceptance of masculine and neuter adjective forms in impersonal predicative constructions. This variation seems to be characteristic for Resian, too. As can be seen, in Example 23a the adjective tožan appears in the masculine form while risnö in Example 23b is neuter.

Overall, the number of reflexive impersonals with adjectival predicates in The little prince is four. Of these, one example contains an indeclinable adjective (or passive participle) which makes it meaningless for determining number and gender features. Additionally, there are two examples with the masculine form of the pronoun sam ‘self; alone’ (cf. Example 23c). Of these six examples, only Example 23b contains a neuter adjective in the predicate. However, since in Slavic the most common way to form an adverb from an adjective is by simply adding the neuter nominative/accusative endings, the form risnö could perhaps be explained as an adverb. Be this as it may, the limited number of examples does not allow drawing any firm conclusions about the distribution of masculine and neuter forms in The little prince.

Example 24 shows that the complement in an impersonal predicative construction does not have to be a form that is target to agreement. Here, a phrase consisting of the preposition za ‘for’ and an infinitive appears in the predicate. This construction constitutes a structural loan of the Romance final construction represented by Italian per ‘for’ + inf.

To conclude, the investigated data provide clear evidence that in Resian, reflexive impersonals have the same properties as in standard Slovenian. They can be formed from both intransitive and transitive verbs. In case of the latter, they require the object to be in the accusative or when negated in the genitive. Moreover, the subject of reflexive impersonals can antecede reflexive (possessive) pronouns and reciprocal constructions, and it can control passive participles. Additionally, reflexive impersonals derived from the copular verb ‘be’ seem to occur in Resian while native speaker judgments on corresponding constructions in standard Slovenian vary (cf. Section 2). Therefore, the discussed data show that if the interpretation of the reflexive clitic as a subject pronoun is assumed for standard Slovenian, it would also have to be applied to Resian. However, Resian provides additional evidence potentially strengthening this analysis. This evidence will be presented in the following two sections (Section 3.3 and 3.4).

3.3 Reflexive impersonal constructions specific to Resian

Of interest for the analysis of the status of the reflexive sa/se in reflexive impersonals are further constructions with reflexive verbs and optative constructions with imperatives and the particle da. They will be discussed consecutively in following two sections (Sections 3.3.1 and 3.3.2).

3.4 The place of impersonal reflexive clitics in the clause

In the previous sections I have first, provided evidence that in Resian, reflexive impersonals show the same syntactic properties as they do in standard Slovenian (Section 3.2), and second, discussed two constructions specific to Resian which also could be interpreted as supporting the analysis of impersonal reflexive pronouns as subject clitics (Section 3.3). This Section is now concerned with the placement of reflexive clitics in impersonal constructions.

While the standard Slovenian system of pronominal clitics consists of object clitics in the genitive, dative, and accusative cases (including the corresponding reflexive clitics), Resian has additionally developed a set of subject clitics (Benacchio 2002; Šekli 2010; Skubic 1997). Moreover, clitics in Resian are not placed after the first phrase in the clause as they usually are in standard Slovenian (cf. Franks and King 2000: 31–48; Golden 2003; Marušič 2008), but are proclitic to the verb (see below). However, most interesting for this paper is that in Resian, the domains of subject and object clitics differ from each other. This provides us with the unique opportunity (within Slavic) to investigate whether impersonal reflexives in their placement group together with subject clitics or with object clitics.

The Resian clitic system has yet to be properly described. In the literature, it is merely recognized that pronominal clitics are proclitic to the verb (Benacchio 2002; Skubic 1997). The exact slots that the individual clitics occupy in the clause as well as their order, so far, has not been studied in detail. Therefore, in the following I cite material from a separate study based on the same resources as the present article that I am currently preparing (Wandl Submitted).

As mentioned above, Resian clitics are oriented towards the verb. Subject clitics always precede the verb. The same is true for object clitics unless they appear in imperative clauses with morphological imperatives. In this case, they follow the verb. However, the verbal forms towards which subject and object clitics are oriented are different. Subject clitics always precede the finite verb in the clause regardless of whether it is a full verb or an auxiliary. Object clitics, on the other hand, precede their lexical head verb. Since in Resian finite verbal forms in unmarked word order precede non-finite forms, subject and object clitics do not occur adjacent to each other whenever a verbal complex contains more than one verb.^[22] Accordingly, subject and object clitics do not form a clitic cluster. Especially telling in this regard are periphrastic constructions with tonic auxiliaries because here it cannot be argued that subject and object clitics form a cluster together with the clitic auxiliary. Thus, in Example 35, the subject clitic an ‘he’ precedes the tonic negative auxiliary nï ‘not be’ whereas the object clitic mi ‘I.dat’ precedes its lexical head verb which is realized as the past participle rišpundäl ‘answer’.

Etymologically, Slavic reflexive clitics are object clitics (cf. Section 3.1). Accordingly, we find them to be placed before their lexical head verb in Resian.^[23] As can be seen in Example 36, subject clitics and reflexive clitics are placed according to the same pattern as in Example 35, irrespective of whether the auxiliary is tonic (cf. bil in Example 36a) or itself clitic (cf. jë in Example 36b).

If we now turn to the placement of reflexive clitics in impersonal constructions in Resian, we see that here the reflexives are placed like subject clitics and not like object clitics. In Example 37a taken from The little prince, the impersonal reflexive sa precedes the tonic auxiliary nï ‘not be’ exactly like the subject clitic an ‘he’ in Example 35. Example 37b comes from Steenwijk’s (1992: 187) fieldwork data. It shows that the clitic placement observed in the examples from The little prince cannot be attributed to influence of the French original. Again, sa precedes the finite verb tëšë. The latter presents the 3rd-person imperfect form of the verb tët ‘want’. However, as can be seen in the translation provided by Steenwijk (1992: 187), the original meaning is significantly bleached so that the verbal form is at least approaching the function of an auxiliary expressing counterfactual meaning (cf. Steenwijk 1992: 182–183 and Pila 2023 on the semantics of the former imperfect in Resian). It is further interesting to note that sa, in this example, also precedes the attenuative particle ba. At least in The little prince, this position is otherwise only occupied by subject clitics and never by object clitics (cf. also Example 21). The placement of the object clitic jin ‘they.dat’ in Example 37b again corresponds to the placement of the object clitic mi ‘I.dat’ in Example 35.

However, it should be noted that the The little prince also contains two instances of the reflexive clitic preceding auxiliaries in other constructions. In Example 38a, se appears before the perfect auxiliary jë and in Example 38b before the future auxiliary bo. In both instances, the reflexive belongs to an inherently reflexive verb.

While this variation deserves further attention,^[24] the fact that in the 39 other sentences which contain se and past- and future-tense auxiliaries (including two instances of the auxiliary jë ‘be.prs.3sg’ and four instances of nï neg.be.prs.3sg) the former always follows the latter suggests that it is this order which constitutes the unmarked pattern in non-impersonal constructions. Further support for this assumption can be found in the placement of object clitics – the type of clitics that the reflexive clitic originally belongs to – which do not occur before auxiliaries when acting as verbal arguments.

Other examples that are instructive with regard to the placement of impersonal reflexives are those containing modal verbs. These are quite numerous in The little prince, which does not come as a surprise considering that generalizing expressions such as ‘one/people should/can/must etc.’ are rather common. The reason why modal constructions are interesting for the analysis of impersonal reflexives is due to the function of modal verbs. Because they merely modify the meaning of the embedded verbs, modal verbs share their subject with the latter. As with the periphrastic constructions in Example 37, one could, therefore, argue that there is no reason why the impersonal reflexive should not be placed before the lexical verb, i.e., in the slot assigned to object clitics, the type of clitics that reflexives belong to etymologically. Accordingly, it could be interpreted as indicative of the subject analysis of impersonal reflexives that sa always precedes the modal verbs in Example 39. The latter includes the modal verbs ‘can’ (Example 39a), ‘have to / must’ (Examples 39b and 39c), and ‘want’ (Example 39d). Examples 39a and 39b further show that object clitics are placed before the lexical verb, and Example 39c confirms the same slot for non-impersonal reflexives.

Other multi-verb constructions are less suggestive with regard to the placement of impersonal reflexives. This is because here, it could be explained by reference to object clitics. Remember that object clitics, in Resian, are proclitic to their lexical head. If a matrix verb has a strong lexical meaning, it could be argued that the placement of the impersonal reflexive before the matrix verb is due to the fact that it is an object clitic. This argumentation seems impossible for periphrastic constructions with auxiliaries which merely have grammatical meaning (cf. especially Example 37a). Against the background of Resian clitic placement rules, the position of impersonal reflexives, therefore, suggests a relationship between these clitics and subject clitics.

3.5 An emerging subject pronoun with an arbitrary human referent

When introducing the paradigm of the reflexive pronoun in Section 3.1, it was pointed out that Resian has two phonologically different clitic variants acc sa/se. It has further been said that these variants most probably originated as the result of a context-sensitive phonological change whose conditioning factors have been obscured. However, in The little prince we find a clear functional distribution of the two reflexive variants that seems indicative of the status of the reflexive clitic in impersonal constructions.

It has already been noticed by Steenwijk (1992: 187) that in reflexive impersonals with reflexive verbs the variant sa always precedes the variant se. The single example of this kind from The little prince confirms this observation (Example 31a). However, in this text the distribution of the two variants goes even further. As can be seen in Table 3 showing the distribution of sa and se across reflexive constructions of all types (cf. Section 2) in the Resian version of The little prince, the former almost exclusively appears in impersonal constructions while the latter, apart from two exceptions, is limited to other reflexive constructions.

The distribution presented in Table 3 clearly speaks for a reinterpretation of sa as a marker of impersonal subjects that might, at present, still be an ongoing process. That the process is in fact not yet completed even in the dialect of the translator could be suggested by the four counterexamples. However, at least, three of these examples allow for alternative interpretations.

In the counterexample given in 40a, the reflexive se appears in an impersonal construction which is at odds with the overall distribution presented in Table 3. The relevant phrase consists of a phasal verb (‘finish’) and a reflexive verb (‘wash oneself’). Accordingly, it contains two reflexive clitics and one could argue that the use of se as a marker of an impersonal subject is related to the occurrence of the same clitic in the embedded infinitive. Similarly, the use of sa as a marker of reciprocAudioVolumeMuteity in Example 40b could be due to the fact that the following verb samajat begins with the identical letters.^[25]

As regards the third counterexample, no such explanation is available. Example 41 is, however, ambiguous. The combination sa čül could either be interpreted as a reflexive impersonal or as a reflexive passive. This is because the nominative and accusative forms of the noun phrase dan drügi gärm ‘another thunder’ are homonymous. The phrase could either be interpreted as the direct object of a reflexive impersonal or as the subject of a reflexive passive. If the first interpretation is correct, the variant sa would be expected while the masculine form of the past participle would be unusual (cf., however, Example 15). According to the second interpretation, the masculine participle would be expected while se would not. The French original does not help with resolving this ambiguity because it significantly differs from the translation. Therefore, the example does not present a clear counterexample to the pattern in Table 3. Perhaps, the above-described ambiguity has contributed to the deviating pattern.

The fourth potential counterexample is 34b, repeated in 42 for the sake of convenience. Here, we find a da + imperative-construction with the reflexive se. The French original, in the respective position, contains an impersonal clause with the subject pronoun on ‘one’. Accordingly, we would expect to find an impersonal construction also in Resian. Although se lajej could also be interpreted as a passive construction (‘is read’) with möj lïbrin constituting the subject of the clause, I therefore believe that we are rather dealing with an impersonal reflexive with the clitic se. Thus, the example goes against the overall pattern observed in Table 3.

In conclusion, it can be said that it is possible to find alternative explanations for the four counterexamples, although, at least in one case (Example 42) the alternative account does not seem convincing. However, even if the alternative explanations are not accepted, the numbers in Table 3 clearly speak for a generalization of sa as a marker of impersonal clauses in this doculect, be it, as an ongoing process.

Now, the reason why the generalization of sa is so interesting consists in its implications for the analysis of impersonal reflexives. When an allomorph becomes associated with one specific meaning while the other allomorphs of the morpheme at hand become associated with the remaining meanings, the allomorph turns into a morpheme of its own. One would expect that this happens only – or is at least much more likely to happen – if the morpheme among its meanings and functions has one that is significantly different from the others. Otherwise, there would be no basis for developing an allomorph into a distinctive element. The difference would have to manifest itself at least in semantics, but differences in grammar would likely be favorable for the reanalysis.

Turning back to reflexive impersonals in Resian, this means that the difference between reflexive clitics in impersonal constructions, on the one hand, and reflexive clitics in all other formations, on the other hand, must have been significant enough for sa to be generalized to the former. And indeed, the syntactic and semantic properties of reflexive impersonals discussed in Sections 2, 3.2 and 3.3 set them off against other reflexive construction. As discussed further, these properties have been interpreted as indicative of the impersonal reflexive being a subject clitic. Accordingly, the restriction of sa to impersonals can be interpreted as a consequence of a reanalysis of the impersonal reflexive as a subject clitic.

Arguing that the reanalysis provides additional evidence for the subject-character of impersonal reflexives would, however, be circular. For this reason the evidence from the Resian clitic placement rules presented in Section 3.1 is so significant. The latter suggest that impersonal reflexives, with regard to their position in the clause, are associated with subject clitics. To my knowledge, this is the only evidence that directly associates impersonal reflexives with subject clitics from a Slavic language. This is important because the arguments presented in Section 2, as convincing as they may be, may to some extent be rooted in certain theoretical conceptions. For instance, the fact that the subject of reflexive impersonals can antecede reflexive and reciprocal pronouns (see Section 2) could also be explained by assuming that reflexives license null arbitrary subjects. The latter is often the case with 3rd-person plural forms (e.g., Siewierska and Papastathi 2011; Sifaki 2021; Mantovan 2022). Likewise, Rivero and Milojević Sheppard’s (2003: 109–110) argument regarding the control of passive complements presented above (see Section 2) depends on Jaeggli’s (1986) analysis of passive complements.^[26] The evidence from the clitic placement rules, on the other hand, is independent of this type of abstract analysis.

The fact that in Resian two reflexive clitics can appear in the same clause (cf. Section 3.3.1), at first glance, seems to further corroborate the assumption of a significant difference between impersonal reflexives and other reflexives. At least, according to Junghanns’s (2002: 79–80) constraint, Slavic clitics are only allowed in the same clause if they do not have a similar function. However, the possibility to use two adjacent reflexive clitics in Resian impersonal constructions can also be understood as a consequence of different placement rules applying to them. Since the data discussed in Section 3.4 suggest that impersonal reflexives are placed like subject clitics while other reflexives are placed like object clitics, it is reasonable to assume that they do not occupy the same slot in the clause. Accordingly, the adjacent occurrence of reflexive clitics in impersonal constructions could also be explained with reference to their different domains. However, regardless of which explanation is preferred, reflexive impersonals derived from reflexive verbs provide evidence for a significant difference between the two involved reflexive clitics, be it with regard to their meaning or function, or to their behavior as clitics.

The use of impersonal reflexives in optative construction discussed in Section 3.3.2 is less suggestive with regard to the status of the reflexive clitic. If it could be shown that the realization of an overt subject – clitic or not – is obligatory in these constructions, one could argue that the reflexive must be the subject in cases like those in Example 34. However, as mentioned in Section 3.3.2, at least, for the 1st person singular, the investigated resource contains evidence for pro-drop. At present, we therefore have to limit ourselves to noticing that the evidence provided by the optative construction fits well with the analysis of impersonal reflexives as subject clitics. However, they do not contribute independent evidence to it at this point.

To sum up, the data extracted from The little prince point to a reanalysis of sa as a subject clitic with an arbitrary human referent in the investigated doculect. Of the few examples contradicting this analysis, a majority can be accounted for by alternative explanations. Nevertheless, at this point, it seems appropriate to speak of a change in progress. Once completed, there would be no reason why sa should be associated with the reflexive meaning. It would have developed into a separate subject pronoun having the syntactic and semantic properties described throughout this paper.

3.6 Aspects of language contact

At the beginning of this article (Section 1), it has been said that reflexive impersonals represent a genuine Resian phenomenon. And indeed, the fact that we find largely identical constructions in standard Slovenian, Polish, and Croatian dialects speaks for an origin rooted in Slavic. However, this does not exclude that Romance languages of the area had an influence on some properties of Resian impersonal constructions. For instance, in Section 3.2 it has been suggested that certain unexpected agreement patterns in Resian reflexive impersonals could be due to contact with Italian. Another language which has had a significant influence on Resian is Friulian (Rhaeto-Romance) (Benacchio 2002; Šekli 2010; Skubic 1997; Steenwijk 1996).

In this section, I intend to point out two further aspects of the properties of Resian impersonal constructions that could be related to language contact: the association of impersonal reflexives with subject clitics and the possibility to derive reflexive impersonals from reflexive verbs without the deletion of reflexive clitics. Firm conclusions about whether contact was indeed involved here cannot be drawn at the moment. However, I believe that it is still important to identify aspects of language contact in the development of Resian impersonals, so as to provide a good basis for studying the diachronic development of this construction in the future.

The first topic concerns the association of impersonal reflexives with subject clitics in Friulian. The latter language, most probably, provided the example for introducing subject clitics into Resian and for remodelling the Resian clitic placement rules (Šekli 2010) as a whole. For this reason, it is interesting to note that, according to Haiman and Benincà (1992: 160–161), the Friulian reflexive si should be reckoned as a subject clitic in impersonal constructions, and that this is to some extent also reflected in its position in the clause. Like subject clitics, impersonal si either precedes object clitics preverbally (Example 43a) or “displaces” them into postverbal position (Example 43b).

While this is not the exact same pattern as in Resian, where object clitics follow the verb only in imperatives, the association of impersonal reflexives with subject clitics via their place in the clause could have had an impact on Resian.

The second topic that seems noteworthy concerns the possibility to form reflexive impersonals from reflexive verbs. Unlike in Polish, where this is impossible, and unlike in standard Slovenian, where it can occur, but requires the deletion of one of the two reflexives (see Section 3.3.1), we do find reflexive impersonal constructions with two reflexive clitics in Friulian and in standard Italian. Of these two languages, Italian, with regard to this option, shows greater resemblance to Resian. As can be seen in the examples in 44, Italian employs two phonologically different reflexive clitics in these constructions. However, unlike in Resian, ci is not simply a variant of si, but presents the accusative form of the 1st-person personal pronoun (Monachesi 1996: 134).^[27]

In Friulian, the reflexive clitics do not occur adjacent to each other in the respective constructions. Rather, one of the two reflexives si moves to the postverbal position (cf. Example 45). Examples such as 43b, where the object clitic is enclitic to the verb, suggest that it is the non-impersonal reflexive that is targeted by the displacement in Example 45.^[28]

Especially the similarities between the Italian and Resian reflexive impersonal constructions with reflexive verbs are so striking that one would be tempted to assume an influence from the former language onto the latter. For instance, the Resian combination sa se could be interpreted as a structural loan of Italian ci si. Resian would then have utilized the phonological variant sa as the impersonal reflexive with reflexive verbs. From here, sa could then have been generalized as the impersonal reflexive. This scenario could find support in the fact that Steenwijk (1992: 187) recognized the use of sa with reflexive verbs, but, apparently, did not yet observe a distribution similar to that in Table 3.

To verify whether this is a possible scenario a detailed investigation of the diachrony of Resian reflexive impersonals would be needed. Unfortunately, the limited attestation of Resian casts doubt on whether such an endeavor could be successful. In any case, a diachronic study is beyond the scope of the present paper which is primarily concerned with a synchronic analysis of impersonals. Nevertheless, it is worth noting that language contact may have played a role also in the development of Resian reflexive impersonals, even though the roots of this construction must be sought in Slavic itself (cf. Section 2).