Intra-system variability and change in nominal and verbal morphology
Barbara E. Bullock and Almeida Jacqueline Toribio
The Pennsylvania State University
It is a relatively uncontroversial assumption in studies of language variation that contact between languages, particularly contact defined by extensive bilingualism, may both enhance linguistic variation and accelerate linguistic change at all levels of the grammar (see Dorian 1986, Silva-Corvalán 1986, 1991, 1994/2000, Thomason 2001). If it can reasonably be argued that individual bilingualism hastens inherent grammatical evolution, then it follows that the changes witnessed at the individual level should parallel internal diachronic developments. In this work, we draw on data from French-English and Spanish-English bilinguals to demonstrate how the fluctuations in the expression of nominal and verbal features in bilingual speech are indeed reminiscent of common processes of historical linguistic development in that change in the specification of grammatical features need not signal any fundamental loss of formal linguistic properties. In essence, in both historical and bilingual contexts, the grammar remains unchanged. However, since a ‘new grammar’ could conceivably arise from external influence, our analysis considers the extent and the limits of structural convergence, a vehicle of language change that is unique to bilingual contexts. It has been suggested that language loss via convergence is selective and does not affect all linguistic domains equally. Thus, phonology may be affected in convergence (Paradis 2000, Bullock & Gerfen forthcoming a) while syntax may remain unchanged, and within syntax proper, the computational core may be impervious to change while the periphery may be vulnerable to external influence.
Our inquiry is similar to that of others: “If bilinguals display patterns of language loss, of particular interest for linguistic theory is to identify potential vulnerable areas of grammatical knowledge, and to seek reasons to explain why systematic patterns of erosion or incompleteness, if they exist, look the way they do” (Montrul forthcoming). In this paper, we seek an explanation for why the patterns of loss in bilingual speech mirror those of diachronic change. Ultimately, we propose that language change over time is not the immediate result of incomplete acquisition or faulty transmission rather change is the result of the acquisition of a system that is not a complete replication (Bullock & Gerfen forthcoming a), a transmission pattern that is especially accelerated in a bilingual contexts. Our work adds to the literature in historical Romance linguistics in both concrete and theoretical ways. First, we provide new empirical evidence from unique sources that speak to the issue of change in progress in a naturalistic bilingual context. And by interpreting our evidence in light of theoretical advances in the study of bilingualism, we are able to probe directly the longstanding problem of the role of external convergence in language change.
The paper is organized as follows: in §1, we provide an overview of the structural variability and convergence that may be observed in language contact situations. It will be demonstrated that the components of the grammar of contact varieties of French and Spanish may be differentially affected by external influence; that is, phonology may be subject to inter-systemic convergence (Paradis 2000, Bullock & Gerfen, forthcoming a, b), whereas the formal computational syntax may be impervious to external influence (Schwartz & Sprouse 1996, Gavruseva & Lardière 1996, Lardière 1998a, 1988b, 2000, Sorace 2000, Haznedar 2001, Toribio 2001, Montrul 2003). In §2, we consider extant evidence from studies of heritage and second-language bilinguals whose non-target forms may be variously ascribed to deficiencies in morphological spell-out or to convergence at the syntax-pragmatic interface (Sorace 1990, 1993, Silva Corválan 1991, 1994/2000, Paradis & Navarro 2003, Montrul forthcoming, Toribio forthcoming). §3 introduces empirical evidence from two case studies of bilinguals—one French-English and one Spanish-English– in which the speakers present ample evidence of morphological reduction and of apparent syntactic convergence in the absence of standardized norms. In §4, the patterns observed in those data are likened to general diachronic changes that cannot be unequivocally accredited to language contact. In situations similar to the cases under study, eventual syntactic change may arise from intergenerational transmission of reanalyzed morph-syntactic mappings (or from morphological reanalysis (King 2000)).That is, external contact does not directly induce formal linguistic change; rather, novel structures are inferred from non-target, though licit, native language forms (Meisel 2001).
1.Variability and convergence in language contact situations
It is well-established by linguistic research that a decline in the pattern or the frequency of use of a language in a bilingual setting may lead to significant variability and linguistic change, sometimes compressed into a small amount of time. The rapidity of change is documented most fully by Schmidt (1985) and that of variation in a homogeneous bilingual community by Dorian (1994). As these authors have shown, variability may be particularly manifest in communities or contexts where normative linguistic pressures are lacking. Especially salient in such situations is the inconsistency of the expression of morphological features. The exponents of gender, number and person, or of tense-mood-aspect can be subject to both inter-speaker and intra-speaker variation. For example, in the French-English data to be discussed more fully in §3, the past participle of the verb mourir ‘to die’ is produced by one speaker as mort and by another as meuru. The former is the standard participle form although it has become syncretic with respect to gender, which can be phonologically signaled for this verb (mor(t), masc. versus mor[t]e , fem.). The latter is an analogical reformation, composed of the stem of the indicative singular, meur-, and an overgeneralization of the –u participle marker. These forms were produced in the same stretch of discourse when the study participants were conversing with one another. Especially revealing is the fact that these speakers are generally one another’s sole interlocutors yet they apparently have no consensus about which form is the correct one. Thus, in the absence of social pressure, variation persists.
Thomason (1997: 12) reminds us that “any and all contact-induced changes are possible,” yet much of the variation and change seen in the bilingual data under discussion in the present work and in the extant bilingual literature nonetheless follows a relatively unexceptional path. That is to say, the various kinds of neutralization revealed in bilingual studies —such as syncretism (shown above), analogical extensions, and paradigm leveling (also shown above)—occur regularly cross-dialectally and diachronically and are generally accompanied by competition, perhaps even free variation, between conservative and innovative forms. It is important to note that variable feature realization, including the absence of any exponent for a given feature, need not signal a concomitant loss of a grammatical category. Thus, the lack of gender realization on the past participle of mourir among our French-English bilinguals does not imply that the category of gender or its attendant agreement properties have been lost. Instead, variability of this sort may represent a deficiency in the mapping between morphology and syntax (i.e., morphological Spell-out) such as that which is frequently attested in L2 acquisition (Gavruseva & Lardière 1996, Schwartz and Sprouse 1996, Meisel 1997, Lardière 1998a, 1988b, 2000, Haznedar 2001, Prévost and White 1999, 2000, Herschensohn 2001).
Attendant to internally-induced changes, languages in contact situations often manifest transparent properties of inter-linguistic influence such as borrowings, semantic extensions, or syntactic calquing. It is, of course, possible to over-attribute change in bilingual speech to externally-induced interference. While contact may hasten change, it does not automatically follow that change is the result of external influence (Silva-Corvalán 1994/2000). Instead, the linguistic restructuring that occurs in bilingual contexts often parallels that which is attested in monolingual communities. However, structural similarity or overlap between languages may reveal areas of the grammatical system that are especially vulnerable to external interference. In such cases, language change may reasonably be imputed to external linguistic influence, or convergence. Much recent research on language variation and change in a bilinguals’ grammar suggests that the convergence of grammatical properties is either of a lexical nature (King 2000, Montrul 2003) or it occurs primarily at the interface of syntax and pragmatics/semantics (Silva Corválan 1991, 1994, Montrul forthcoming, Toribio forthcoming). In essence, while lexical and pragmatic aspects of a language may be borrowed, the purely formal aspects of a linguistic system appear to remain intact. This implies that within the syntax, at least, only the peripheral features are affected by convergence, not the core.
While syntax proper (i.e., the purely formal system) is arguably immune to convergence, there is evidence that the phonological system may be vulnerable to external influence. The impact of external influence on the phonology but not the syntax of bilingual children has been demonstrated by Paradis (2000) who suggests that her results may be merely an artifact of her methodology. However, Bullock & Gerfen (forthcoming a, b) similarly argue that the introduction of the American rhoticized schwa, as in sir, as a replacement for the mid front rounded vowels in the French vocalic system of Frenchville bilinguals is a change that can only be understood as externally-induced. That the phonetic system of bilinguals reflects convergence in measurable ways has been amply demonstrated (see Piske, Flege & MacKay 2001 for an overview). However, linguists generally assume that the phonological system is more abstract and formally organized than the phonetic system and thus, like syntax, probably resistant to systemically disruptive external influence (on phonology, see Cook 1989, 1991). In the case of Frenchville French, the “borrowed” rhoticized schwa has a number of direct formal reflexes; it has resulted in the loss of allophonic variation, a loss of possible functional contrasts, it has possibly increased phonological markedness, and perhaps most important, it has established a phonological contrast between the new rhoticized schwa and a French schwa that was not preexistent.
The differential impact of convergence over various domains of the grammar is not at issue in this analysis. However, we would like to tentatively suggest, following Bullock & Gerfen (forthcoming a), that the categorical separation of phonology and phonetics can probably not be maintained (Ohala 1981). As the phonetic system is open to convergence, so too is the phonology. What then distinguishes phonology from syntax with regard to convergence is that the surface realization of a formal syntactic feature is entirely incidental and dependent on the lexicon; in fact, features like case may have no concrete expression at all. Aspects of formal syntax have no necessary overt correspondents but even in its most formal aspects, phonology is dependent upon and, in very large part, cannot be divorced from phonetic features.
2.Target-deviant morpho-syntax
Numerous authors have examined non-target nominal and verbal forms on the part of second language bilinguals. While early research had sought to draw direct parallels between morphological inflections and their structural correlates (cf., Eubank 1993/1994; Vainikka and Young-Scholten 1998), more recent work converges on the dissociation between the exponents of inflection and the presence of higher functional categories (cf., Schwartz and Sprouse 1996, Lardière, 1998a, 1998b; Haznedar, 2001; Herschensohn 2001). Thus, for example, Herschensohn (2001) reports on a longitudinal study of two second-language learners of French who demonstrate knowledge of verb raising independently of the acquisition of paradigms associated with agreement; indeed, as shown in (1), the development of higher functional categories TP, AgrP, and CP precedes accuracy in TP-related morphology (cf., Haznedar 2001).
(1)A quelle heure est-ce que le musée ouvrer?
At what hour is it that the museum open-inf.
“At what time does the museum open?”
Cf., A quelle heure est-ce que le musée ouvre? (Herschensohn 2001, Emma III)
Analogous findings from Lardière’s (1998a, 1998b) examination of the naturalistic fossilized English-language productions of a native Chinese speaker lead her to conclude that areas of divergence from the target system reside in “precisely those PF areas increasingly seen as ‘external’ or ‘extraneous’ to the computational (CHL) component of the grammar (1998b: 370).”
The impoverished forms examined by researchers in first language attrition and convergence also fall outside the scope of the computational system, which is regulated by purely syntactic/formal properties such as Agreement and Case, and into the interpretive module, which is subject to properties of meaning such as Tense/Mood, Topic, and Focus (cf., Silva-Corvalán 1989, 1994/2000; Sánchez 1997; Toribio forthcoming; Zapata et al. 2002; Montrul 2003, forthcoming; Paradis and Navarro 2003). For instance, Montrul’s studies of heritage speakers of Spanish reveal variable behavior in the production and interpretation of tense, mood and aspect morphology (Montrul 2002), as in (2a), in lexical selection and expression of unaccusativity and unergativity (Montrul, in press), and in the distribution of subject and object arguments and semantically conditioned clitic-doubling (Montrul forthcoming), as in (2b).
(2)a.Cuando *caminó [pret.] por el bosque se encontró con el lobo. (Cf., caminaba [imp] ‘walked’)
‘When she walked through the forest she met up with the wolf.’ (Montrul 2002; early child L2)
El llegó a la casa de la abuela antes que ella llegó [pret.] (Cf., llegara [past subj] arrived’)
“He arrived at the grandmother’s house before she arrived.” (Montrul 2002; simultaneous bilingual)
- Y lo que hizo el lobo era también *comió la Caperucita. Lo que hizo el cazador para dar un lección al lobo fue cortar el estómago del lobo y *quitar la abuela y la Caperucita. (Cf., comer a la Caperucita… cortarle el estómago… y quirtar a la abuelita)
“And what the wolf did was also eat the Little Riding Hood. What the hunter did to teach a lesson to the wolf was to cut the wolf’s stomach and take out the grandmother and the Little Riding Hood.” (Montrul forthcoming, subject # 210, advanced)
Sánchez offers similar conclusions in her studies of bilinguals residing in a Quechua-Spanish contact situation: bilinguals evince convergent, non-target semantic features of definiteness and specificity in the use of Spanish null object pronouns (Sánchez 1997), in (3a), and convergent aspectual and discourse-oriented features in verbal selection and morphological specification (Sánchez forthcoming), in (3b).
(3)a.Nomás sacas [todas las yerbas]i y después que sacas [proi] echas abono.
“You just pull out all the herbs and after you pull them out you put fertilizer.” (Sánchez 1997)
b.Había una abuelita…no…una viejeta…vieja. Ya había…estaba sembrando y habé…ha encontrado un pájaro. De(s)pués el pajáro estaba enfermo... y después se llevó a su casa. Eso hizo,después, daba alimento.
“There was a granny, no an old woman. (She) was sowing and she found a bird. Then the bird was ill and then (she) took it to her house. She did that and she later gave (it) food.” Sánchez forthcoming).
The core linguistic system of all of these Spanish-English and Spanish-Quechua bilinguals approximates that of the full variety spoken by their Spanish monolingual counterparts, but those areas where the syntax interfaces with other cognitive or extra-grammatical areas are vulnerable to underspecification or respecification.
From the foregoing discussion, it should be evident that morphological errors and target-deviant semantic and pragmatic performance are to be attributed to deficiencies in mapping at Spell-out and the interface levels of lexical-semantics, syntax-semantics and discourse-pragmatics, rather than to discrepancies in formal (i.e., [-interpretable]) morpho-syntactic features. The ensuing paragraphs turn to the examination of data from two bilingual case studies that, despite their different natures, reveal strikingly similar morpho-syntactic properties, further confirming that variability and loss in the PF and interpretative interfaces are not necessarily accompanied by a similar fluctuation or degradation in core syntax.
3.Two Case Studies
Our study references two naturalistic language samples. One data set consists of the transcripts of audio-taped interviews with two brothers who reside in the formerly-French-speaking enclave community of Frenchville in Pennsylvania. The other set is a record of journal entries that depict the life of a Spanish-speaking agricultural worker as her family follows the ripening beet crops in south- and mid-western States. The French-speaking participants in the present analysis are two brothers, aged 69 and 72 years old at the time of the recordings. They are not literate in French. However, unlike other residents of the area, they have continued to speak French either occasionally with one another or with the first author of this article. Both have lived in Frenchville all of their lives, and both spoke French at home exclusively until they married and moved out of their parents’ home in their early twenties. Because their wives do not speak French, English is now the language of their homes. Both brothers terminated their formal educations by the age of 14. The data referred to in this paper were collected in structured but naturalistic field interviews with the first author in 2002. The interviews represent structured discourse in that they are monitored and recorded, however, the participants are comfortable and familiar with the interviewer with whom they have little trouble producing or comprehending French. Often, they do not hesitate to go beyond the questions asked, as in the following citation where one of the brothers offers, unsolicited, his view of language loss.
(4)Ça prend pas trop de temps pour le perdre si tu ne parler [sic] pas. Je parler [sic] avec lui de temps en temps un mot ou deux ou trios et c’est tout. C’est peut-êre quelque chose qu’on dit l’un l’autre qu’on ne peut pas dire à personne d’autre.
‘It doesn’t take much time to lose it if you don’t speak it. I speak with him from time to time a word, or two, or three but that’s all. It’s perhaps something that we say to one another that we can’t say to anyone else.’
As with the transcriptions of the Frenchville interviews, the language samples contained in the diary of the migrant worker are very revealing of the author. The diary entries in general demonstrate that the author has had insufficient training to instruct her in the normative conventions of Spanish orthography; it represents a simplification of the complex mapping between sound and graphemes to a few known values, and reproduces many of the phonological characteristics of her rural dialect. The lexicon of the text speaks to her occupational segregation in agricultural communities with other speakers of colloquial Mexican Spanish, as well as to contact with English. For example, there appear phonetically unincorporated forms such as stand by and workshop, and others that follow the structural organization of the Spanish language, e.g., fil ‘field’ and files ‘fields,’ which are phonologically and morphologically well-formed in Spanish, semantic extensions, e.g., atender for ‘to attend,’ rather than ‘attend to,’ registrar for ‘to register’ rather than ‘to check,’ and loan translations, such as pero ya mero no la hacía ‘I nearly didn’t make it (lit. trans.), which may be uninterpretable to the reader who has no knowledge of English.