Presentation of the data

The informants are subdivided in 18 female and 18 male speakers, equally stratified according to age (20 to 34, 35 to 54, and older than 54 years of age) and education (primary, secondary, and postsecondary) with 12 informants (6 females and 6 males) in each group. From these data, all noun phrases with possessive pronouns have been extracted.

The overall distribution of possessive variants collected from the corpus data is shown in Table 1. Table 1 only lists cases with an overt possessive marker. The middle column shows the 3340 instances of a prenominal weak or clitic possessive (e.g., mi hermano ‘my brother’), the next column shows the 31 cases with a postnominal strong possessive (e.g., el/un hermano mío ‘the/a brother of mine’), and the last column contains the six cases of possessive pronouns (e.g., los míos ‘mine’).Footnote ¹⁶ This is a proportion of 99.1% prenominal versus 0.9% postnominal possessors and shows the marginal status of postnominal possessive marking in this variety. Of 3340 nouns with a prenominal possessive, 3197 cases are of the type mi hermano ‘my brother’. In these 95.7% of the cases, the possessive is not preceded by any other article or adjective. Only 142 cases of the pleonastic possessive could be found (cf. the upper five lines in Table 1). In 10 of these examples, the possessive combines with a demonstrative (ese mi hermano ‘this my brother’) and in another 10 with an adjective, which can be preceded or not by an article (4 and 6 cases, respectively, e.g., (el) otro mi hermano ‘(the) other my brother’).

The majority of the tokens of the pleonastic possessive have an article in initial position, which is definite in only two cases (la mi madrastra ‘the my stepmother’), and indefinite in 120 cases (un mi hermano ‘a my brother’). This confirms observations from earlier studies that in contemporary Central American Spanish this construction is largely confined to the indefinite article and, to a minor degree, to demonstratives (Company Company, Reference Company Company2005:136, Reference Company Company and Company Company2009:843; Gómez Seibane, Reference Gómez Seibane2010; Martin, Reference Martin and England1978:111, Reference Martin1985:384; Pato Maldonado, Reference Pato Maldonado, Palacios and García2002:143n114). Medieval Spanish was very different in this regard, because the definite article still prevailed over the indefinite article in this construction, as Company Company (Reference Company Company2005:136, Reference Company Company and Company Company2009:843) reported. At some point during the 16th century, the co-occurrence of the definite article with the prenominal possessive virtually disappeared, which paved the way for the current situation.

Summing up, in the vast majority of cases, Guatemalan Spanish parallels Standard Spanish in that the prenominal possessive occurs as the sole determiner. Postnominal strong possessives only occur sporadically. The pleonastic possessive, if it occurs at all, is most of the times introduced by the indefinite article and only rarely by a demonstrative or an adjective. This is why, in the following, I will restrict my attention to only those cases of the pleonastic possessive with an indefinite article (Indef+Poss+N, e.g., un mi hermano ‘a my brother’). It will be shown that this construction is a sociolinguistic variable. The question of how to circumscribe its variable context is the topic of the next section.

Circumscribing the variable context of the pleonastic possessive (Indef+Poss+N)

If the pleonastic possessive is indeed one of several alternative forms, then the other forms of ‘saying the same thing’ need to be identified. Clearly, la mi hermana ‘the my sister’ and mi hermana ‘my sister’ are not the same thing, because they are definite expressions. The analysis must therefore be limited to indefinite contexts only. These are illustrated in (9) and (10).

1. (9) Indef+Poss+N

   Un hermano de mi esposa está casado con una mi hermana . (GUA080)

   a brother of my wife is married to a my sister

   ‘A brother of my wife is married to a sister of mine.’

2. (10) Indef+N+POSS

   Antes se había venido un hermano mío para acá. (GUA022)

   before refl had come a brother mine over here

   ‘Before, a brother of mine had come over here.’

Bare nouns with postnominal strong possessives (N+POSS) had to be excluded, because they only occur in predicatives, as in (7a) (siendo amigo mío ‘being a friend of mine’), or in the fixed expression Dios mío ‘my God’ and are hence not part of the variable context. According to the principle of accountability, “any variable form (a member of a set of alternative ways of ‘saying the same thing’) should be reported with the proportion of cases in which the form did occur in the relevant environment, compared to the total number of cases in which it might have occurred” (Labov, Reference Labov1972:94; italics removed). Bare nouns do not qualify as an alternative to the pleonastic possessive. There is, however, another form that can systematically be used instead of the pleonastic possessive and that has largely gone unnoticed in the foregoing discussion. This is the use of the indefinite article in possessive contexts without an overt possessive pronoun (Indef+N).Footnote ¹⁷ The following example shows that Indef+N can indeed be used as an alternative to Indef+Poss+N.

1. (11)

   1. a. Indef+Poss+N

      tenía una mi bicicleta de carrera (GU021)

      ‘I had a my racing bike.’

   2. b. Indef+N

      entonces mi mamá tenía una mesa para planchar (GUA073)

      ‘Then my mom had an ironing board.’

In conclusion, Indef+N+POSS and Indef+N are the pertinent variants that need to be considered alongside Indef+Poss+N. It is difficult to determine which of the Indef+N tokens in the corpus are truly eligible candidates for the variable context, because possessive marking is not restricted to cases of physical possession (Cifuentes Honrubia, Reference Cifuentes Honrubia2015:21). As Seiler (Reference Seiler1981:6) stated, “linguistic POSSESSION consists of the representation of a relationship between a substance and another substance. Substance A, called the POSSESSOR, is prototypically [+ animate], more specifically [+ human], and still more specifically [+ EGO] or close to the speaker.” I therefore took the property of a relationship between the possessor and the possessee as a criterion for the inclusion of the individual Indef+N tokens into the analysis. This relationship can be very diverse and includes ownership, authorship, as well as places of residence and origin. As the pleonastic possessive only occurs with human possessors, this restriction has also been imposed on the Indef+N tokens.

Table 2 shows that the pleonastic possessive is not only a minor variant as compared to the total number of possessive-marked noun phrases, but also within its own variable context. The overt realization of a prenominal possessive in indefinite contexts is clearly a marked variant as opposed to the unmarked Indef+N variant, which represents by far the majority of tokens. Postnominal strong possessives (Indef+N+POSS) are virtually absent from the data, occurring only in some isolated cases.

However, even though the pleonastic possessive is a rare phenomenon, it occurs frequently enough not to dismiss it as mere “noise” in the data. Considering its relatively small number of tokens, the question comes up which factors favor its choice over the majority variant, Indef+N. As already set out, these factors might also inform us about the morphosyntactic status of the indefinite article in this variety. According to the Overlap Model of grammaticalization put forth in Heine (Reference Heine1993:48–53), the source and target items (i.e., the less and more grammaticalized forms) of a grammaticalization chain may coexist during a certain time span in a language. This entails that more and less grammaticalized forms of the indefinite article might coexist synchronically. My prediction is that these coexisting forms are contextually distributed. Earlier, arguments from the research literature on the pleonastic possessive were presented that supported an analysis of it in terms of a structural remnant from Old Spanish. If this is indeed the case, it could be assumed that the use of the indefinite article in this construction shows more properties of an earlier stage of grammaticalization than in all other indefinite noun phrases (Indef+N). The following analysis of factor groups shall help to verify this hypothesis. In the present study, the dependent variable consists of three different variants: Indef+Poss+N (i.e., the pleonastic possessive), as in (11a) (una mi bicicleta de carrera ‘a my racing bike’); Indef+N, as in (11b) (una mesa para planchar ‘an ironing board’); and Indef+N+POSS, as in (10) (un hermano mío ‘a brother of mine’). As the latter occurs in only three cases (see Table 1), a binary approach has been chosen comparing the pleonastic possessive with its nonoccurrences, in other words, the sum of the other two variants, Indef+N and Indef+N+POSS.

Selecting the independent variables

The preferential conditions for the occurrence of the pleonastic possessive, discussed in the literature, have been operationalized into factor groups (independent variables) whose influence on the choice of the pleonastic possessive shall be tested. Some of these allow assessing the state of grammaticalization of the indefinite article in this very construction.

The first is specificity. According to the literature reviewed, the indefinite article passes through a stage of a specificity marker on its grammaticalization path (stage III in Heine [Reference Heine1997:72–73]). If it is true that the pleonastic possessive constitutes a structurally conservative context in which the indefinite article preferentially appears in its less grammaticalized form, we would expect to find a favoring effect of specificity on this variant as opposed to the Indef+N variant, compare (12) with (13).

1. (12) specific noun

   la iglesia había comprado un su terreno ahí y hicieron piscinas (GUA001)

   ‘The church has bought a[n its] plot of land there and they made swimming pools.’

2. (13) nonspecific noun

   le digo a ella que su papá y yo tenemos que luchar para que ella sea alguien y se reciba- y tenga un su buen trabajito (GU013)

   ‘I tell her that her dad and I must fight so that she is someone and receives- and has a [her] good job.’

The second factor group is thematic persistence. It has been claimed that one of the main functions of the pleonastic possessive is the introduction of a thematically persistent referent into the discourse (Company Company, Reference Company Company and Company Company2009). Thematic persistence has also been claimed to be indicative of an earlier grammaticalization stage of the indefinite article, when it turns from a numeral into a presentative marker (stage II in Heine [Reference Heine1997:72]). If the indefinite article is less grammaticalized within the pleonastic possessive than within Indef+N contexts, the presentative function is expected to favor it. This is illustrated in (14) and (15). A noun phrase has been coded as thematically persistent when its referent is taken up as a discourse topic within the following three sentences. This is the case in (14) where the referent is introduced by the indefinite noun phrase un mi primo ‘a my cousin’ and serves as a topic in the following utterances. In (15), the referent of unos sus ocho años ‘some his eight years’ is not taken up in the subsequent discourse.

1. (14) presentative marker (thematic persistence)

   De muy patojo de catorce años me vine, porque un mi- un mi primo tenía una su panadería y me ofreció trabajo y me dijo que- que me iba a pagar bien y que no sé qué y sí me pagaba bien porque yo era ayudante. (GUA007)

   ‘I came here as a very small child, at the age of fourteen, because a my- a my cousin had a his bakery and he offered me a job and he told me that- that he would pay me well and that I don't know what and indeed he paid me well because I was helpful.’

2. (15) no thematic persistence

   No, con ellos no. Mis hijos no conocían el puerto, o sea R ya tendría ¿qué? R tendría que como unos sus ocho años. Entonces, bueno, caminábamos y caminábamos y “espérate ya vamos a llegar” cuando llegamos verdad al puerto de San José y todo. (GU050)

   ‘No, with them no. My children did not know the harbor, that is to say R already must have been what? R must have been some [his] eight years old. Then, well, we walked and walked and ‘wait, we are just arriving’ when we actually arrived at the harbor of San José and everything.’

The third factor group is postnominal modification. The argument that a relative clause, an apposition or a prepositional phrase modifying the noun indicate an early stage of grammaticalization of the indefinite article has been put forward by Company Company (Reference Company Company2005:146) who stated that “the necessity to add information about the nucleus [i.e., the noun] shows, in my view, that these NPs had a strong quantificational force in Old Spanish, that they required additional referential information in order to identify the referent, and it proves that the construction [i.e., the pleonastic possessive] must have emerged when the grammaticalization of the numeral un(o)-una as an indefinite article was quite advanced but has not yet come to fully be a mere presenter of entities, i.e., an article” (my translation).Footnote ¹⁸ I therefore expect postnominal modification, as the relative clause in (16), to favor the pleonastic possessive.

1. (16) En mi niñez <eh> recuerdo que una vez con unos mis primos que ahorita se encuentran en Michigan <eh> <time lapse> J y R son hijos de una mi hermana. Con ellos nos gustaba irnos a lo que- a las famosas cataratas … (GU021)

   ‘In my childhood, eh, I remember that once with some my cousins, who now reside in Michigan- J and R are children of a my sister. We liked to go with them to the- to the famous waterfalls …’

The next three factor groups relate to the discourse-functional properties of the pleonastic possessive. The first is the reference of the possessee to human or nonhuman entities. This is to verify the claim that the pleonastic possessive has actually evolved from a marker of exclusively human referents in Old Spanish, as in (17), into a marker of predominantly nonhuman referents, as in (18).

1. (17) O non sea atal commo el ome que dezían que quería leer gramática, que se fue para un su amigo que era sabio (Anónimo, 1984:92, cited in Company Company [Reference Company Company2005:144, example (6a)])

   ‘Or he would not be like the man about whom they said that he liked to read grammars and that he went to a his friend who was wise.’ (My translation)

2. (18) … porque un negocio de- de ver- de verduras o de frutas que se va a traer en un mi picop y irme a vender a los- a los departamentos, incluso a aquí en Guate … (GUA001)

   ‘… because a business of vegetables or fruits which are carried on a my pickup and which I will sell in the departments, even here in Guate …’

The person and number specifications of the possessive are also analyzed. These are lexically encoded on the possessive, if present, see (19a), (20a), and (21), but they are covert when the possessive is not expressed, as in the Indef+N construction, see (19b) and (20b). Third person singular and plural possessives were not coded separately since the possessive shows number syncretism, see (21a) and (21b).

1. (19)

   1. a. First person singular

      Compré un mi traje, color melón … (GUA001)

      ‘I bought a my melon colored suit … ’

   2. b. First person plural

      Gracias a Dios que tenemos un trabajito . (GUA007)

      ‘Thank God that we have a little job .’

2. (20) Second person singular

   1. a. Tú tenés derecho quizás a tomarte un tu litro . (GUA002)

      ‘You have perhaps the right to take one your liter.’

   2. b. No te mereces una novia tan fea. (GUA097)

      ‘You don't deserve such an ugly girlfriend.’

3. (21)

   1. a. Third person singular

      … le compramos un su hueso … (GUA073)

      ‘… we bought him a his bone … ’

   2. b. Second and third person plural

      De vez en cuando les invitamos a una su cerveza o algo así para que se pongan así bonitos, verdad. (GUA012)

      ‘From time to time we invite them to a their beer or something like that so that they feel at ease, right.’

This factor group shall verify the claim by Company Company (Reference Company Company and Company Company2009:846–853) that the former preference of third person contexts has been replaced by first and second person contexts due to the preferential use of the pleonastic possessive in narratives.

Number has also been coded as a marker of agreement between the possessive and the possessee, cf. (22). According to Lipski (Reference Lipski1994:295), combinations of the type Indef+Poss+N “are invariably singular,” a claim that is already disproved by example (22a). I will test whether at least a favoring effect of the singular can be confirmed.

1. (22)

   1. a. El año nuevo casi siempre la pasábamos con- yo con unas mis amigas …   (GUA064)

      ‘We used to spend the New Year almost always with- me with some my friends   …’

   2. b. Voy a ir a visitar a una mi amiga . (GUA002)

      ‘I will go and visit a my friend.’

It has been claimed by a number of authors that the pleonastic possessive expresses some kind of affective meaning (e.g., Pato Maldonado, Reference Pato Maldonado, Palacios and García2002). In this respect, Nieuwenhuijsen (Reference Nieuwenhuijsen2007) mentioned the preferential use of diminutive endings on the possessee. This has been coded as a separate factor group:

1. (23) Sí se va a comer una su gallinita- gallinita de campo, que eso sí es verdad. (GUA079)

   ‘Indeed one will eat a his free-range hen, that is true.’

The difference between the indefinite article un-una ‘a’ and an indefinite adjective such as algún ‘some’ is taken into account, because the latter also occurs in the pleonastic possessive, see (24).

1. (24) Voy a ir a ver a algún mi familiar a un hospital. (GUA002)

   ‘I will go and see some my relative at a hospital.’

Algún has a quantificational meaning. According to Alonso-Ovalle and Menéndez-Benito (Reference Alonso-Ovalle and Menéndez-Benito2010:28), it “conveys that there are at least two individuals that can satisfy the existential claim.” If the pleonastic possessive has retained some of its former partitive function (Martin, Reference Martin and England1978, Reference Martin1985), this could be reflected by a preferred use in contexts of algún.

The last linguistic factor group tested is prenominal adjectival modification. Some authors investigating the pleonastic possessive have restricted their attention to those cases in which the article-possessive-noun sequence is not interrupted by any adjective, see Nieuwenhuijsen (Reference Nieuwenhuijsen2007:332n51) who did not justify this decision, however. Cardinaletti (Reference Cardinaletti, Alexiadou and Wilder1998:18) and Alexiadou (Reference Alexiadou, Fuß and Trips2004:36) stated that in languages with prenominal weak adjectival possessives, the possessive always precedes other modifiers of the noun, such as adjectives. Instead of excluding these cases from the analysis altogether, the presence or absence of prenominal adjectives has been included as a separate factor group:

1. (25) Entonces yo le digo a ella que su papá y yo tenemos que luchar para que ella sea alguien y se reciba- y tenga un su buen trabajito . (GU013)

   ‘Then I tell her that her dad and I have to struggle for her to be someone and to obtain- to have a her good little job.’

The logistic regression analysis: Results and interpretation

Table 3 shows the results of a variable rule analysis (Sankoff, Reference Sankoff, Ammon, Dittmar and Mattheier1988) using GoldVarb X (Sankoff, Tagliamonte, & Smith, Reference Sankoff, Tagliamonte and Smith2005) listing the linguistic independent variables according to their relative strength on the choice of the pleonastic possessive as measured by the range between the highest and the lowest factor weight.Footnote ¹⁹

Two factor groups had to be excluded from the variable rule analysis due to a too small number of tokens. The first is the type of determiner. Among all 2091 tokens, 155 occur with an indefinite adjective (e.g., algún ‘some’) in initial position, but only two of these occur in the Indef+Poss+N variant (cf. example (24), algún mi familiar ‘some my relative’). The reason for this might be an advanced state of lexicalization of the pleonastic possessive. It does not seem to be indefiniteness alone that triggers the use of the pleonastic possessive, since both un-una and algún-alguna are indefinite. Rather, I suppose that the pleonastic possessive is lexically limited to the word sequences un(a/s) mi(s), un(a/s) tu(s), and un(a/s) su(s). The second factor group that was excluded is the type of prenominal modification, be it by an adjective, a quantifier, or a numeral. Prenominal modification rarely occurs in the data as a whole (n = 173 of 2091, 8.3%) and only six instances of it can be found in the pleonastic possessive (3.5%).

Of the remaining seven factor groups, six turn out to have a significant effect on the choice of the pleonastic possessive while one, number agreement between possessor and possessee, is nonsignificant. Since the pleonastic possessive is represented by a total of only 120 tokens, a verification of seven factor groups might entail unreliable results. According to Harrell (Reference Harrell2015:72–73),Footnote ²⁰ “a fitted regression model is likely to be reliable when the number of predictors (or candidate predictors if using variable selection) p is less than m/10 or m/20, where m is the ‘limiting sample size.’” In the case of a binary response variable, m is min(n ₁,n ₂) where “n ₁ and n ₂ are the marginal frequencies of the two response levels.” Applied to our case, n ₁ is the number of applications of the pleonastic possessive, n = 120, and n ₂ is the number of its nonapplications, n = 1971. The number of independent variables tested should therefore not exceed six (m/20, with m = 120). After a first regression analysis including all seven independent variables, the second analysis shown in Table 3 is limited to those six variables that are significant predictors of the choice of the pleonastic possessive in the first run.

Human reference and a diminutive suffix on the possessee are the strongest predictors of variant choice. Together with the slight but significant favoring effect of first person singular possessors, they constitute the preferential discourse-pragmatic context of occurrence in contemporary Guatemalan Spanish, see example (26).

1. (26) Una mi abuelita me enseñó hacer pocas comidas. (GU013)

   ‘A my granny taught me to prepare few dishes.’

These results confirm that the pleonastic possessive is preferentially used to refer to a person with whom the speaker entertains an affectionate relationship.

With regard to human reference, Company Company (Reference Company Company2005:144; Reference Company Company and Company Company2009:815) reported that the historical precursor of the pleonastic possessive in Old Spanish categorically occurred with human possessees. Pato Maldonado (Reference Pato Maldonado, Palacios and García2002:148–149) and Nieuwenhuijsen (Reference Nieuwenhuijsen2007:334) stated that this distribution has dramatically changed and that in contemporary Guatemalan Spanish, nonhuman possessees outweigh human possessees. Is this a contradiction to the results displayed in Table 3? An examination of the total numbers reveals that it is not. Among 120 cases of the pleonastic possessive, a minority of 52 tokens have a human possessee (43%). This is lower than the rate of human possessees in the other possessive-marked contexts (poss+N, Def+N+POSS, Indef+N+POSS, N+POSS, poss+N+POSS, Def+POSS). Here, we observe 1869 human possessees out of a total of 3234 tokens (58%). The present data hence confirm the observations made by Pato Maldonado (Reference Pato Maldonado, Palacios and García2002) and Nieuwenhuijsen (Reference Nieuwenhuijsen2007). However, the authors have not analyzed the pleonastic possessive within its variable context. The favoring effect of human reference only becomes measurable when the principle of accountability is observed and the rate of human possessees is assessed in those cases where the pleonastic possessive actually did occur compared to those cases where it could have occurred but did not. It is only when the variable context is taken into account that we see more than a triplication of the pleonastic possessive within the domain of human reference (18%, n = 52 of 287) as compared to the entire variable context (6%, n = 120 of 2091). I interpret the fact that possessees with human reference still favor the use of the pleonastic possessive within its variable context as an indicator of its evolutionary descent from the Indef+Poss+N construction in Old Spanish.

As regards the grammatical person of the possessive, the pleonastic possessive is preferentially used in reference to the speaker himself or herself. This is shown by the favoring effect of first person singular as opposed to third person possessives, thus confirming the observation made by Company Company (Reference Company Company and Company Company2009:846–853). The small range of this factor group does however not necessarily entail that it is less important than the others in the choice of the pleonastic possessive. This is so because second person singular possessives (tu ‘your’) had to be excluded from the variable rule analysis due to their too small number of occurrences. Only 2 of 53 tokens with tu ‘your’ could be found with the pleonastic possessive (4%). This is a potentially disfavoring effect whose confirmation would require an overall larger number of tokens. For now, we can only note a frequency-based tendency of an avoidance of the pleonastic possessive in speech directed at the interlocutor.Footnote ²¹

Thematic persistence of the noun is the factor group with the third largest range in the analysis, following diminutive endings and human reference. It is directly related to the grammaticalization of the indefinite article. The observation that thematically persistent nouns favor the use of the pleonastic possessive (factor weight: .67) can be interpreted as a case of persistence that is, according to Hopper (Reference Hopper, Traugott and Heine1991:22), a basic principle of grammaticalization: “When a form undergoes grammaticization from a lexical to a grammatical function, so long as it is grammatically viable some traces of its original lexical meanings tend to adhere to it, and details of its lexical history may be reflected in constraints on its grammatical distribution.” In the present case, it seems that the indefinite article has in part retained its former function of a presentative marker (stage II on the grammaticalization pathway of the indefinite article according to Heine [Reference Heine1997:72]).

This assumption is supported by two other effects—specificity and postnominal modification—which play, however, a less decisive role as shown because of their small ranges. If it is true, as suggested by the favoring effect of thematic persistence, that some of the functions associated with earlier grammaticalization stages of the indefinite article persist into its contemporary usage within the pleonastic possessive, we would expect that its former function of a specific marker also has its repercussions in the modern days. The results of the variable rule analysis confirm this expectation. Nouns with a specific reference and those that are postnominally modified by a relative clause favor the pleonastic possessive. The latter effect, postnominal modification, however, does not provide homogeneous results. While postnominal adjectives and prepositional phrases are virtually never used with the pleonastic possessive, the rate of the pleonastic possessive doubles in the context of postnominal relative clauses (n = 27 of 227, 12%) as compared to the overall data. Company Company (Reference Company Company and Company Company2009:848) observed that the use of postnominal modification gradually declined between the 13th and 16th centuries. Whereas in that time span more than half of the Indef+Poss+N constructions were still postnominally modified, this rate dropped down to 27% in her data from 21st-century Central American Spanish (Company Company, Reference Company Company and Company Company2009:851, cuadro 23). This is very close to the rate of 25% of postnominal modification in the present data (n = 30 of 119).Footnote ²² According to the author, “the loss of specification in Modern Indigenous Spanish means that the form un-una introducing these NPs weakened its quantificational value and consolidated the value of a simple presenter of entities” (ibid.:851, my translation).Footnote ²³ In other words, un-una in the Indef+Poss+N construction has made some progress on its grammaticalization pathway toward a simple article though this process has probably not yet come to completion. The results provided by Company Company (Reference Company Company and Company Company2009:849, 851) also coincide with those reported here in that postnominal relative clauses, prepositional phrases, and adjectives are differently affected by this decline. However, the present study diverges as concerns the direction of this effect. While Company Company (Reference Company Company and Company Company2009) reported a decline of postnominal relative clauses and an increase of prepositional phrases and adjectives, in the present study, postnominal modification of Indef+Poss+N is virtually restricted to relative clauses. No postnominal adjectives and only 3 of 248 instances of postnominal prepositional phrases occur in this context (1.2%), as in (27) (and were therefore excluded from the variable rule analysis).

1. (27) Ahí saqué un mi curso de casi- de cuatro años de operador de computación. (GUA080)

   ‘There I took a my course of almost- of four years as a computer operator.’

For the time being, it remains an open question why the results of the present study and those of Company Company's (Reference Company Company and Company Company2009) investigation diverge in this way. Be that as it may, the combined favoring effects of thematic persistence, specificity of the possessee, and postnominal relative clauses show that the pleonastic possessive preferentially occurs in contexts that are indicative of an early step of grammaticalization of the indefinite article. It therefore seems justified to conclude that even though indefinite un-una has progressed along its grammaticalization pathway, it has not yet turned into a mere presenter of entities, but that it rather retains some of its former quantificational value, to use Company Company's words. Put differently, the pleonastic possessive constitutes a context in which un-una has not yet turned into a mere marker of indefiniteness in D°. It still occurs in some specifier position to the left of D° where it retains a discourse-pragmatic value.

The results of the variable rule analysis show, by its very nature, variable tendencies. In other words, the data also contain tokens of the pleonastic possessive in which the indefinite article behaves like a fully grammaticalized element in that it occurs with nonspecific and thematically nonpersistent nouns, as in (28).

1. (28) … entonces todavía se acostumbra eso unas sus tortillas con chicharrón con su limoncito y sus tortillas con moronga … (GUA061)

   ‘… then one even gets used to it, some his tortillas with greaves with its lime berry and its tortillas with blood sausage …’

The Overlap Model by Heine (Reference Heine1993) accounts for this variation in that different grammaticalization stages may coexist synchronically. The claim made here is that the synchronic distribution of these stages can be context-sensitive. Thus, the pleonastic possessive constitutes a context in which earlier stages of un-una are more likely to be represented than outside of this context. A possible reason for this special behavior might be the hypothesis that the pleonastic possessive has lexicalized to a certain extent. If the restriction of this construction to un-una and its rare occurrence not only with demonstratives but also with indefinite adjectives such as algún-alguna ‘some’ is indeed a reflex of its increasing lexicalization, it could be the case that the concurrent demotivation constitutes an obstacle for an ongoing grammaticalization of its subcomponents, in this case the indefinite article.Footnote ²⁴

The nonsignificance of number agreement between possessor and possessee shows that not all functional constraints indicative of an early step of grammaticalization are at work. The capacity of indefinite un-una to refer to more than one item emerges at a relatively late stage in the grammaticalization pathway (Heine, Reference Heine1997:73). Obviously, the pleonastic possessive is as likely in the context of singular possessees (n = 108 of 1935, 6%) as in that of plural ones (n = 12 of 156, 8%), contradicting Lipski's claim that the pleonastic possessive is restricted to possessives with singular marking (Reference Lipski1994:295). Again, this result could only be obtained by taking into account the variable context. If we had restricted our attention to the 120 tokens of the pleonastic possessive only, we would have noted a preponderance of 90% of singular (n = 108) as opposed to only 10% of plural possessives (n = 12). Indeed, many of the claims about the pleonastic possessive made in the previous literature suffer from the fact that the authors have not taken into account the variable context. As the present analysis shows, high rates of occurrence, as in the case of singular possessives and possessees with nonhuman reference, do not necessarily imply a favoring effect of these factors, but might even turn out to disfavor the variant under consideration. This shows, pace Company Company (Reference Company Company1995, Reference Company Company and Company Company2005:133–134), that relying on rates of occurrence alone can be misleading and guide the analyst to even counterfactual conclusions (see also Poplack & Torres Cacoullos, Reference Poplack, Torres Cacoullos, MacWhinney and O'Grady2015:288; Rinke & Elsig, Reference Rinke and Elsig2010).

To sum up, the results of the variable rule analysis show that the pleonastic possessive is favored by factors indicating an early stage of grammaticalization of the indefinite article, that is, by thematically persistent and specific nouns with postnominal relative clauses. This observation conforms with the assumption that the pleonastic possessive is a relic inherited from Old Spanish. On the functional side, the pleonastic possessive preferentially occurs with a first person singular possessor who entertains an affectionate relationship with a human possessee, as evidenced by the diminutive ending. The preference of the pleonastic possessive by human possessees has also been inherited from medieval Spanish.