Ouh que c'est laid! “Oh this is ugly!” is one of the comments among the 11,800 hits on Google for the sequence “la fille que je sors avec” [the girl I go out with]. Often the comments include the idea that the whole expression has been taken from English as a direct calque. The authors of the present keynote article, Poplack, Zentz and Dion (Poplack, Zentz & Dion, Reference Poplack, Zentz and Dion2011, this issue), argue convincingly that this type of preposition stranding in Canadian French cannot be ascribed to language contact with English. Using sound and accountable methodology, derived from the research paradigm of variationist sociolinguistics, they manage to disprove the hypothesis of a direct causal link between the expression in Canadian French and its supposed earlier use in English. Thus, an icon of language contact, both in popular perception and in many not-so-well-informed academic sources, has been shattered.

If this well-known phenomenon is not due to language contact, what other candidates for language contact will survive the test? On the whole, Poplack and her team have come out on the skeptical side as regards language contact phenomena. Poplack and Pousada's early (Reference Poplack, Pousada, Fishman and Keller1982) paper already had the title “No case for convergence”. Poplack's work has at least had the salutary effect of raising the level of argumentation in the language contact field.

In my commentary I will first briefly discuss earlier claims about contact-induced language change in relative clauses, in part based on the discussion in Appel and Muysken (Reference Appel and Muysken1987). Then I will turn to areal linguistics, and finally to cross-linguistic bilingual priming studies. I then conclude by briefly returning to the case discussed by Poplack, Zentz and Dion.

In Appel and Muysken (Reference Appel and Muysken1987) five cases of potential changes in relativization strategies are discussed, which I will very briefly summarize here.

1. 1. Nadkarni (Reference Nadkarni1975) has argued that speakers of Konkani (Indo-European, related to Marathi) have adopted relative clause structures modeled on the Dravidian language Kannada:

   1. (1)

Having a question word at the beginning and an interrogative marker at the end of the relative clause is a typical Kannada feature.

1. 2. Similarly, it has been claimed that in Turkish non-participial relative clauses the Persian-origin ki is used:

   1. (2)

Lewis (Reference Lewis1975), who also provides this example, notes several problems with it. It represents an older more Persian-influenced style, not the modern colloquial language (although it is found in European “heritage language” Turkish as well). There is also an older Turkish word resembling the element ki, which may have played a role.

1. 3. It has also been claimed that Spanish has influenced Nahuatl relative clause formation (Karttunen, Reference Karttunen, Steever, Walker and Mufwene1976).

   1. (3)

Using question words, such as tlen “which” in (3), is an innovation. Karttunen argues, however, that this possibility is an addition to the language, rather than a structural change. The original construction remains available as well.

1. 4. In southern Peruvian Quechua, it has also been argued (Schwartz, Reference Schwartz1971) that Spanish has led to the use of interrogative elements in relative clauses (here acc = accusative, dub = dubitative, loc = locative, prog = progressive):

   1. (4)

However, here the evidence is not completely convincing. There are early sources with question words, the construction has special features, etc.

1. 5. The final case cited by Appel and Muysken (Reference Appel and Muysken1987) is the Canadian French construction under consideration. Here, however, we argue, following Bouchard (Reference Bouchard and Lefebvre1982) and in line with the present target article, that English influence is not very likely.

Altogether, we can conclude from these five examples that the strength of the evidence for contact-induced language change in relative clauses varies. It has been accepted for cases from the Indian subcontinent like Konkani, but in other circumstances alternative explanations and other factors appear to play a role as well. There has been a vast literature on the topic of language contact in relative clause formation subsequent to our original study. It is impossible, however, to summarize it here. Two issues stand out in this literature: (i) bracketing devices: as in the studies cited, additional bracketing apparatus may come from a second language due to language contact; and (ii) reference tracking: strategies to mark who was involved in what action may be carried over from one language to the other. Relative clause formation may be identified as a vulnerable area for language contact, given the fact that (a) there is already much language-internal variation in many cases, involving competing strategies; (b) it is an area where languages tend to change fairly rapidly; and (c) relative clause formation involves complex information packaging, an area prone to cross-linguistic influence.

The second question I wanted to point to is the ever increasing role attributed to areal influence in linguistic typology (see Haspelmath, Reference Haspelmath, Haspelmath, König, Oesterreicher and Raible2001, Reference Haspelmath, Aikhenvald and Dixon2004, to cite just a few sources), inspired of course by Nichols (Reference Nichols1992), but gathering momentum in the wake of the publication of the WALS (Haspelmath, Dryer, Gil & Comrie, Reference Haspelmath, Dryer, Gil and Comrie2005). Although there are as yet no appropriate and widely accepted quantitative methodologies, there is a growing trend in typology to consider areal explanations for the distribution of linguistic phenomena, at the expense of typological universals and in some cases postulated genealogical units. It is worthwhile to consider this trend in the light of the reservations in the work of Poplack and her colleagues concerning the role of language contact.

Third, to complicate matters even more, there is a growing language processing literature on cross linguistic priming, which needs to be taken into account here. Following the seminal work of Bock (Reference Bock1986) and Pickering and Branigan (Reference Pickering and Branigan1999) on monolingual syntactic priming, a number of recent studies (Arai, van Gompel & Scheepers, Reference Arai, van Gompel and Scheepers2007; Hartsuiker, Pickering & Veltkamp, Reference Hartsuiker, Pickering and Veltkamp2004; Kootstra & Sahin, Reference Kootstra and Sahin2011; Schoonbaert, Hartsuiker & Pickering, Reference Schoonbaert, Hartsuiker and Pickering2007; Shin & Christianson, Reference Shin and Christianson2009; Weber & Indefrey Reference Weber and Indefrey2009) have shown that priming also occurs cross-linguistically: a pattern that has been processed in one language may then also occur in a second language, leading of course to interference. There are both production and comprehension results, and both behavioral and neuro-imaging studies, confirming this effect, and a wide range of syntactic constructions are being studied in this paradigm. Hopefully future research will clarify the conditions under which such priming takes place. Cross-linguistic priming may be the prime mechanism (more central than code-switching, as hinted at in the present keynote paper) to explain contact-induced language change, and given the ease under which we find priming, by itself the prediction would be that this type of change is quite commonplace and frequent, in contrast with the rigorous research of Poplack and colleagues.

Above I have briefly referred to three research areas relevant to the present discussion: first, many suggestions of contact-induced language change in the area of relative clause formation; second, increasing emphasis on areal, and hence contact, effects in explaining the distribution of typological features; and third, increasing processing evidence for cross-linguistic priming effects.

How can we reconcile this with the findings of the Poplack, Zentz and Dion keynote article? A first possibility is that the construction involved – preposition stranding – differs from the ones in priming, areal linguistics, and contact-induced language. I am not familiar with studies on preposition stranding from this area. It is certainly an emblematic type of language-contact construction, very much part of public discourse, and hence a construction where speakers will avoid using elements from one language in the other. A second possibility is that the bilingual community involved tries to avoid this type of cross-linguistic transfer, given the overall socio-political relation between the languages. Possibly, both explanations interact. More experimental work could be done in this area to ascertain this, and further disentangle the various factors involved.