Variable Constraints on "Anomalous" 'se los' in Mexican Spanish

Scott Schwenter The Ohio State University

LSRL 44-University of Western Ontario May 2, 2014

The Phenomenon “Anomalous” (Company 1998) pronominalization of plural dative referent by accusative pronoun  

Canonical:  Yo se LO di (DO = el libro, IO = ellos)  

Non-canonical:  Yo se LOS di (DO = el libro, IO = ellos)  

Dative 'se' is invariant and cannot be pluralized ! Plural number of the dative referent is compensatorily realized by ’-s' on the accusative pronoun.

Background Kany 1945: “...popular speech in many regions of Spanish America generally insists on indicating plurality of the indirect object se by adding an s to the immediately following direct object, lo or la, making them los and las, even though the object referred to is singular.”  

Company 1992/1998/2001/2006:  •  Standard explanation: invariability of 'se' compared to 'lo' and 'la’ !

need for plural marking on accusative pronouns  

•  Indirect objects are overwhelmingly more animate than direct objects, and this leads to a “reinstatement” of the number features of the IOs on the DO pronoun (“cannabilistic datives”)

•  “nowadays se los seems to be lexicalized, a single pronoun, selos, totally unanalyzable for most speakers”  

• Company 1992/1998/2001/2006

•  “Data from Rivarola (1985) and DeMello (1992: 171, 174) show that most Spanish speakers do not recognize two objects in this clitic sequence, but they interpret the sequence selos, selas, seles as having only one object, the Dat recipient.” (But see Moreno de Alba 2013)

•  Resolution of ambiguity explanation insufficient: 1.  “Many examples” where referent-NP of the DAT clitic occurs in the same

sentence or immediately preceding the one with the pronoun sequence: Es muy importante entonces que los jurados acepten la ley como se LAS da el juez

2.  In “a few examples” the ACC clitic may also take on the GENDER of the DAT referent:

Si ellas me quieren comprar el caballo, yo se LAS venderé

Background  

Bonet (1995) & Harris (1996) (summarized in Ordóñez 2011)

•  “Parasitic plurals” where the plurality of the IO surfaces on the accusative clitic (*ses)

•  There exists a complex clitic of the form [SE LO]-S

•  For Harris, there are “P-effect” dialects, which pluralize in such cases, and “normative” dialects, which do not

•  Binary distinction between two types ! no mention of variation between the “anomalous” pluralized cases and the non-pluralized “normative” ones in a given dialect

Generative Viewpoint  

Questions  • Which factors condition the variable assignment of the dative plurality onto the accusative clitic? This variability, although amply documented (DeMello 1992; Moreno de Alba 2013), has been essentially ignored in the literature.  

• Is the 'se los' combination really becoming lexicalized as a single, unanalyzable pronoun, as several scholars have proposed (DeMello 1992; Company 1998, 2001)?  

• What relationship is there, if any, between the variability of se lo vs. se los and the discourse accessibility of the indirect object referent?

Hypotheses

•  If the (plural) indirect object referent is further away in the prior discourse, then it is more likely that los will be substituted for lo.

•  If a clarifying indirect object prepositional phrase (e.g. a ustedes) is absent, then lo will be realized as los for disambiguation of the indirect object referent.

•  Pluralization of lo will be more likely in proclitic position (se los + V) than in enclitic (V + se los) position.

Factor Groups •  Dependent variable (se lo/la vs. se los/las)  •  Mode (oral vs. written vs. quote in text)  •  Animacy (human vs. animate vs. inanimate vs. propositional)  •  Tense (preterit vs. imperfect vs. present vs. future vs. imperative)  •  Temporal reference (past vs. present vs. future)  •  Verb person (1st vs. 2nd vs. 3rd)  •  Verb number (singular vs. plural)  •  DO gender (feminine vs. masculine vs. propositional)  •  IO gender (feminine vs. masculine)  •  Presence of Prep. Phrase (present vs. not present)  •  Referential Distance (close vs. far)  •  Following Segment (consonant vs. vowel vs. pause)  •  Corpus (CREA vs. Monterrey)  •  Enclitic/Proclitic (enclitic vs. proclitic)

Methods Variable Context (Envelope of variation):  

*Ditransitive verbs  *Dative and Accusative Clitic Cluster  *BOTH clitics must be third-person (i.e. 'se' and 'lo(s)/la(s), respectively)  *Dative referent MUST be plural  *Accusative referent MUST be singular  

Corpora:  *CREA Corpus (searching Mexico only)  *Corpus de Monterrey

Results: Examples

1)  Después les autorizaron una televisión chiquita. Primero la desarman toda, la revisan, la vuelven a armar y hasta entonces se las entregan. (México, CREA)

2)  Es muy importante entonces que los jurados acepten la ley como se las da el juez (México, CREA)

3)  Porque Dios no perdona, muchachos, eso sí se los prometo.(México, CREA)

4)  ...pero con ésa ni contar porque era más Govea que Zárate y además les dijo que aunque lo supiera no se los diría. (México, CREA)

Results • Original question: "Why does 'se los' appear for 'se lo'?”

• Revised question: "Why does 'se lo' appear for 'se los'?"  

• Quantitative results reveal 'se los’ as the dominant strategy, 'se lo' is much less common in the Mexican data:  

74 (28%) se lo/la  188 (72%) se los/las  

• Significant Factors:  " subject/verb number  " presence of prepositional phrase " referent distance  

Analy'cal  Methods  •  Fixed effects model using the independent

variables was fitted to the data, using the glm function in R

•  Selection of variables was done using random forests and also through step function

•  Nested models were compared using ANOVA to select best model; interaction terms/inference trees/xtabs to check interactions

Overall  Results-­‐Se  Los  Estimate SE zValue p-value

(Intercept) -1.44 0.39 -3.719 <0.01

Verb Number (Ref. level = Plural)

Singular 1.65 0.33 5.045 <0.01 Prep. Phrase (Ref. level = Present)

Absent 1.51 0.34 4.386 <0.01

Ref. Distance (Ref. level = 0-2)

3+ Clauses 1.02 0.53 1.923 0.05

NB: Positive estimate ! greater probability of the”anomalous” plural variant

Results

Results for prepositional phrase and referential distance go hand in hand:

•  Co-occurring prepositional phrase ! indirect object referent is immediately accessible in the same sentence as the pronoun cluster

•  Therefore, the plurality of the indirect object referent is manifested within the sentence by virtue of the NP itself, i.e. it is easily contextually accessible

•  When no PP present, pluralization of the accusative pronoun is significantly more likely

Results

•  Referential distance: when dative referent is only 0-2 clauses away in prior discourse, non-pluralization is significantly more likely

•  Counters Company’s view that se los is more frequent in such contexts

•  When dative referent is located further back in the prior discourse (3+ clauses), the addition of ’-s' can be seen as a resumptive anaphoric strategy, to aid the interlocutor in processing the plural IO referent

Results (RD of 0-2) With PP Without PP TOTAL

Se lo/la 57% (29/51) 24% (40/170) 31% (69/221) Se los/las 43% (22/51) 76% (130/170) 69% (152/221)

• When the last reference to the dative referent is found at 0-2 clauses back, there is a difference of 33% between tokens with and without an accompanying PP

• When a dative PP is present, “normative” se lo is actually more frequent overall (57% vs. 43%); without a PP, “anomalous” se los is the overwhelmingly preferred variant (76% vs. 24%)

•  This difference between tokens with and without an accompanying PP is statistically significant (Χ2=20.3, df=1, p<.001)

Results (RD of 3+) With PP Without PP TOTAL

Se lo/la 17% (1/6) 11% (4/35) 12% (5/41) Se los/las 83% (5/6) 89% (31/35) 88% (36/41)

• Data here are much scarcer, but the results suggest that tokens with and without a PP are less different from each other with greater RD

• The percentage of pluralization remains more or less the same with or without a PP, ! Pluralization occurs more frequently the further the distance to the last mention of the dative referent in the preceding discourse

Results •  The number of the subject (SG vs. PL) is the factor that

most affects the variation between lo/la and los/las.

•  Singular subjects are significantly more likely to lead to pluralized accusatives than plural subjects

•  Chi-square tests show that there is a significant difference between 3rd person singular and 3rd person overall rate (X2 = 7.73, df = 1, p = .005).

•  No significant difference between singular and overall rate when subject is 1st/2nd person (X2=0.173, df = 1, p = .678)

Results  • Therefore, singular subject number displays an independent effect on the variation, but only when the subject is 3rd person. When the subject is 1st or 2nd person, there is no independent effect of number.

• Additionally, a crosstab of subjects and presence/absence of a dative PP reveals an important distinction between singular and plural subjects:

Singular  Subject   Plural  Subject  Without  Dative  PP   89%  (137/154)   47%  (24/51)  

With  Dative  PP   49%  (20/41)   44%  (7/16)  

CIT-­‐Significant  Factors  

Plural Singular

PP Absent PP Present

Results  •  Notice that in the other three cases (SG subjects w/ a

dative PP, PL subjects both w/ and w/o a dative PP), the % of pluralization (se los for se lo) is nearly the same.

•  The discourse-pragmatic function of se los pluralization ! marker of anaphoric (in)accessibility; no reason to believe that the function of pluralization would vary across the other cases

•  Competing motivations: clause-level argument marking for distinctiveness and discourse-level accessibility marking for processing

Results  •  WHY is there such a huge difference between

singular subjects without an accompanying dative PP and the other types of subjects (with or without an accompanying dative PP)?

•  Because it is only in the case of these subjects that differentiation of number between subject and indirect object is being realized by an added s on lo/la; in the other cases there is either disambiguation via a PP, or via a plural subject

Conclusion •  Corpus research ! crucial for understanding constraints on

se los/se lo. •  Consideration of decontextualized tokens neglects discourse-

pragmatic accessibility constraints in naturally-occurring language use

•  Quantitative analysis reveals that speakers are sensitive to discourse-pragmatic factors when choosing between lo/la or los/las ! Disambiguation of anaphoric reference  

•  BOTH the indirect object AND the subject are implicated in the variability; given the similarities in their prototypical referential characteristics, this should not be surprising

References Bonet, Eulàlia. 1995. Feature structure of Romance clitics. Natural Language and

Linguistic Theory 13.607-47.

Company, Concepción. 1992. Un cambio en proceso: "El libro, ¿Quién se los prestó?" Scripta philologica in honorem Juan M. Lope Blanch, vol. 2, ed by Elizabeth Luna Traill, 349-62. México: UNAM.  

Company, Concepción. 1998. The interplay between form and meaning in language change. Grammaticalization of cannibalistic datives in Spanish. Studies in Language 22.529-66.  

Company, Concepción. 2001. Multiple dative-marking grammaticalization. Spanish as a special kind of primary object language. Studies in Language 25.1-47.

Company, Concepción. 2006. El objeto indirecto. Sintaxis histórica de la lengua española. Primera parte: La frase verbal, vol. 1, ed. by Concepción Company, 479-574. México: UNAM/Fondo de Cultura Económica.

DeMello, George. 1992. Se los for se lo in the spoken cultured Spanish of eleven cities. Hispanic Journal 13.165-79.

References Harris, James. 1996. The morphology of Spanish clitics. Evolution and revolution in

lingustic theory, ed. by Héctor Campos and Paula Kempchinsky, 168-97. Washington, DC: Georgetown University Press.

Hewson, John. 1981. More on Spanish selo. Linguistics 19.439-47.

Kany, Charles. 1945. American Spanish Syntax. Chicago: University of Chicago Press.

Lope Blanch, Juan M. 1953. Observaciones sobre la sintaxis del español hablado en México. México: IHMIC.

Moreno de Alba, José G. 2013. Notas de gramática dialectal (El Atlas Lingüístico de México). México: UNAM.

Ordóñez, Francisco 2011. Clitics in Spanish. The handbook of Hispanic linguistics, ed. by José Ignacio Hualde, Antxon Olarrea, and Erin O’Rourke, 423-51. Oxford: Blackwell.

References Rivarola, José Luis. 1985. Se los por se lo. Lexis 9.239-44.

Romero Morales, Juan. 2008. Los dativos en el español. Madrid: Arco/Libros.

Schwenter, Scott A. 2006. Null objects across South America. Selected Proceedings of the 8th Hispanic Linguistics Symposium, ed. By Timothy L. Face and Carol A. Klee, 23-36. Somerville, MA: Cascadilla Proceedings Project.