Poplack, Zentz and Dion (PZD; Poplack, Zentz & Dion, Reference Poplack, Zentz and Dion2011, this issue) examine the often unquestioned assumption that the existence of preposition stranding (PS) in Canadian French is linked to the presence of a contact situation with English in the North American context. Although this issue has been the topic of previous research from a syntactic perspective (Bouchard, Reference Bouchard and Lefebvre1982; Vinet, Reference Vinet1979, Reference Vinet1984), to my knowledge, it has never been explored using variationist sociolinguistic methods applied to a large corpus of spontaneous speech, with emphasis on code-switchers as potential agents of contact-induced change.
The results of the study point to the conclusion that PS in Canadian French is not attributable to convergence with English. Code-switchers are thus not the agent of change in this context. The general conclusion is that while it makes sense, superficially, to see a link between PS in French and in English, the two constructions are actually quite different, which reinforces the futility of attributing the French construction to contact with English. Following PZD and others, I will use the terms PS for the English construction and orphan preposition (OP) for the French construction.
My initial reaction to the paper relates to its originality and tremendous usefulness, which lies in the variationist angle used to examine an issue which has never before been studied through quantitative methods. The conclusions reached are in complete agreement with qualitative (theoretical) analyses of the construction. PZD state (under section heading: “Stranding or orphaning?”):
Thus, remarkably, we arrive at the exact same major conclusions as the syntacticians, using an entirely different approach and actual usage data. We conclude, as did they, that the French phrase-final prepositions in relative clauses that look like English stranded prepositions can in fact be analyzed as French orphan prepositions extended to the relative clause context.
That this convergence between variationist and theoretical approaches needed to be qualified as “remarkable” points to the regrettable fact that these approaches still exist as two solitudes, without much interaction or productive dialogue. PZD must be commended for their serious consideration of available syntactic analyses and their willingness to incorporate them into their examination of the facts. This convergence also reveals the importance for syntacticians who study variation to be vigilant in the use and manipulation of their data, which are in most cases not obtained using accepted sociolinguistic methodology. It is a question of accountability, and this paper clearly shows that the accuracy, reality, and relevance of the examples used can be subjected to rigorous verification based on corpus analysis.
I will not comment here on a central theme of this paper, namely the debate surrounding the existence of contact-induced change; a recent issue of Bilingualism: Language and Cognition (14 (2), April 2011) provides an extensive discussion on this question. However, I will return to this in the second part of my commentary. In the first part, I examine the relevance of PZD's contribution by linking it to current research on null (or missing) arguments and explore new research questions.
Null arguments have played a central role in the development of generative approaches to syntactic descriptions. The idea that linguistic expressions are the result of a derivation, involving movement and structure-building, led to the postulation of the existence of empty categories in syntactic representations. Chomsky (Reference Chomsky1981, Reference Chomsky1982) proposed that empty categories can be either the result of movement or base-generated as such. This led to some important work on a particular type of null argument, namely null subjects found in tensed clauses in many languages. Pro-drop thus became a widely studied property, before other types of null arguments (and even other types of null subjects; Huang, Reference Huang1984) slowly emerged as potentially relevant to the general discussion. Null direct objects are part of this picture and have been extensively discussed in the theoretical and acquisition literature; see Pérez-Leroux, Pirvulescu and Roberge (Reference Pérez-Leroux, Pirvulescu and Roberge2008) and references cited there. More relevant to this commentary is the fact that the missing argument found in OP constructions has not yet been fully integrated into this general discussion on the nature of the null argument phenomena, beyond Zribi-Hertz's (Reference Zribi-Hertz1984) original proposal, which linked a pro complement to an operator. One reason for this was a lack of tools (or imagination) for such integration but recent developments in Minimalism (Chomsky, Reference Chomsky1995, Reference Chomsky2005, among others) open intriguing possibilities that are captured in the unified approach to argument drop presented in Sigurðsson (Reference Sigurðsson2011). He argues that all definite arguments independently of whether they are overt or null must be linked to a constituent in the edge of the clause (the local C-domain). This C/edge linking is computational (syntactic) in nature and interacts with pragmatics through context scanning. A variety of left-peripheral heads can act as C/edge linkers (CLn) and probe for a goal under Agree. The structure in (1) provides a general picture of how this works.
(1) 
In this conception, C/edge linking is not what allows an argument to be null, except in discourse drop languages. Null subjects in Italian, for instance, are pronouns (φ features) incorporated to T. Although he concentrates on null subjects and objects, Sigurðsson (Reference Sigurðsson2011, p. 290) uses “the term (direct) object to refer to objects of both verbs and prepositions (the facts discussed here do not suggest any relevant distinction between prepositional and direct verbal objects)”; however, he does not explicitly discuss argument drop in PPs.
Orphan preposition constructions, such as the one exemplified in (2a) (PZD's (26a)), fit naturally with the model in (1), as shown in the simplified structure in (2b).
- (2)
a. Lui avait trouvé ce charbon-là, puis il se chauffait avec [] (O).
“He had found that coal there, and he was warming
b.
As for PS, the most natural assumption is that extraction out of PP is excluded in French but not in English.
We are then left with how to treat OP in relative clauses in Canadian French as in (3), PZD's (10c).
(3) Puis il y a bien des affaires j'avais de la misère avec (S).
“And there are lots of things I had trouble with.”
Assuming that no extraction out of PP has taken place, this construction appears to involve an extension of the OP mechanism to a PS context, where the A΄ binder of the null complement of the preposition has been replaced by or reanalyzed as a C/edge linker. Two questions arise. First, pragmatic context scanning clearly plays a crucial role in OP as seen above, especially when the missing argument corresponds to a discourse context entity. However, for OP in relative clauses, we have to assume that the head of the relative counts as the immediate and the only possible controller for the C/edge linker. Second, since in Sigurðsson's approach the fact that an argument can be silent is mostly determined independently of C/edge linking, we still need to settle on what makes it possible for the complement of prepositions to be null. To do so, we can build on recent analyses of null objects of verbs. Cummins and Roberge (Reference Cummins and Roberge2005) propose that the null complement of a transitive verb used intransitively, which is normally interpreted as a non-referential and prototypical object of that verb, see (4a), can acquire a pragmatically inferred interpretation in certain contexts, as exemplified in (4b).
- (4)
a. I don't like to read ø.
b. Since I have to wait for you, I'll buy a magazine and I'll read ø.
In (4b), the direct object of read has exactly the same status as the one in (4a); however, the context allows for linking with magazine. Cummins and Roberge adopt the structure in (5) proposed by Hale and Keyser (Reference Hale and Keyser2002) for denominal unergative verbs (dance in their case) and extend it to all verbs.
(5)
For Hale and Keyser (Reference Hale and Keyser2002, p. 92) the relation between the verb and its complement in (5) is one of classificatory licensing whereby “the verb identifies the complement to some sufficient extent”.
Now, the fundamental grammaticality contrast in OP constructions between semantically strong and weak prepositions, in both the classic construction and its extension to relative clauses, suggests that the analysis of null objects above should be extended to prepositions. In this approach, a preposition used intransitively, as it is in OP constructions, would be represented as in (6). Weak prepositions (à “to”, “at” and de “of”) can be assumed not to identify their complements (or to identify them too weakly) resulting in their exclusion from OP constructions.
(6)
The null complement would then be open to C/edge linking, resulting in an OP construction or can remain non-linked and retain its “unergative”, i.e. non-referential prototypical, interpretation. While this may not be a widespread possibility, it is attested in Canadian French with the preposition, which, according to PZD, accounts for the bulk of the tokens of OP in their corpus, namely avec “with”. An example is given in (7), which, in the absence of a potential linker, is interpreted as per the translation provided.
(7) Moi avec ø.
“Me too.”
We have thus seen that the Canadian French data examined in PZD provide a new data set that can be used to test the applicability of Sigurðsson's (Reference Sigurðsson2011) approach to argument drop.
I now return to the question of the English influence on OP in Canadian French. As we have seen, and as confirmed by PZD, previous syntactic studies had convincingly demonstrated that OP in relative clauses could not be the result of direct borrowing of an English-type PS construction. As PZD point out (under section heading: “Discussion”), “[t]he apparently widespread belief that surface similarity can (or should) simply be equated with structural similarity makes contact-induced change a logical inference”. This is especially true “among laypeople” and in the prescriptive literature. PZD are quite categorical in their rejection of this unsubstantiated point of view, in the case of the variety represented in their corpus; yet I would like to argue that contact-induced change in this particular construction, i.e. the use of PS (as disctinct from OP), in certain varieties of French is not only hypothetically possible but also attested. Interestingly, PZD's article indirectly provides us with the two phenomena that are necessary (although not sufficient) to conclude that PS exists in a particular variety of French. First, semantically weak prepositions (essentially à and de) should be possible without a complement. Second, complementless prepositions should be found in “true” extraction contexts: wh-questions and pseudo-passives. The two phenomena are illustrated with the English examples in (8) and (9), respectively.
- (8)
a. This is the person that I was talking to.
b. That is all I can think of.
- (9)
a. Who did you vote for?
b. This bed was slept in.
In my works with Ruth King and Nicole Rosen, cited in PZD, we provide evidence that in certain varieties of Canadian French (Western Canada and the Maritimes) sentences equivalent to (8) and (9) are grammatical; see Roberge and Rosen's (Reference Roberge and Rosen1999) (12b, c) and (13b), cited here in (10).
- (10)
a. Qui tu as fait le gâteau pour?
“Who did you bake the cake for?”
b. Le ciment a été marché dedans.
“The concrete was stepped in.”
c. Où il vient de?/Quelle heure elle a arrivé à?
“Where does he come from?/What time did he arrive at?”
Although our work was not corpus-based, these varieties are undeniably in a more intense contact situation than the more central varieties (Quebec and Ontario). Given the linguistic facts and the contact situation, an influence from English appears more likely. The question of course is to determine the exact nature of the change and how it could have occurred (i.e. be implemented in the grammar). A detailed discussion of potential explanations goes beyond the scope of this commentary but two natural options present themselves. First, the contact situation may have led to lexical borrowing of some English prepositions into the French varieties and, if the possibility of PS is somehow linked to a property of prepositions, then the construction could have spread to those regional French grammatical systems; see King (Reference King2000) for an extensive discussion. Second, the contact situation could have acted as a catalyst to further extend PS beyond relative clauses (without movement out of PP) to movement constructions as in (10). Roberge (Reference Roberge and Brasseur1998) and Roberge and Rosen (Reference Roberge and Rosen1999) argue that this change could have arisen through a shift in the C-system in those varieties. Given this view, PS has not been “borrowed” directly as a construction; contact with English would have directed the change in a manner not attested across-the-board in other French varieties.
Poplack, Zentz and Dion (PZD; Poplack, Zentz & Dion, Reference Poplack, Zentz and Dion2011, this issue) examine the often unquestioned assumption that the existence of preposition stranding (PS) in Canadian French is linked to the presence of a contact situation with English in the North American context. Although this issue has been the topic of previous research from a syntactic perspective (Bouchard, Reference Bouchard and Lefebvre1982; Vinet, Reference Vinet1979, Reference Vinet1984), to my knowledge, it has never been explored using variationist sociolinguistic methods applied to a large corpus of spontaneous speech, with emphasis on code-switchers as potential agents of contact-induced change.
The results of the study point to the conclusion that PS in Canadian French is not attributable to convergence with English. Code-switchers are thus not the agent of change in this context. The general conclusion is that while it makes sense, superficially, to see a link between PS in French and in English, the two constructions are actually quite different, which reinforces the futility of attributing the French construction to contact with English. Following PZD and others, I will use the terms PS for the English construction and orphan preposition (OP) for the French construction.
My initial reaction to the paper relates to its originality and tremendous usefulness, which lies in the variationist angle used to examine an issue which has never before been studied through quantitative methods. The conclusions reached are in complete agreement with qualitative (theoretical) analyses of the construction. PZD state (under section heading: “Stranding or orphaning?”):
Thus, remarkably, we arrive at the exact same major conclusions as the syntacticians, using an entirely different approach and actual usage data. We conclude, as did they, that the French phrase-final prepositions in relative clauses that look like English stranded prepositions can in fact be analyzed as French orphan prepositions extended to the relative clause context.
That this convergence between variationist and theoretical approaches needed to be qualified as “remarkable” points to the regrettable fact that these approaches still exist as two solitudes, without much interaction or productive dialogue. PZD must be commended for their serious consideration of available syntactic analyses and their willingness to incorporate them into their examination of the facts. This convergence also reveals the importance for syntacticians who study variation to be vigilant in the use and manipulation of their data, which are in most cases not obtained using accepted sociolinguistic methodology. It is a question of accountability, and this paper clearly shows that the accuracy, reality, and relevance of the examples used can be subjected to rigorous verification based on corpus analysis.
I will not comment here on a central theme of this paper, namely the debate surrounding the existence of contact-induced change; a recent issue of Bilingualism: Language and Cognition (14 (2), April 2011) provides an extensive discussion on this question. However, I will return to this in the second part of my commentary. In the first part, I examine the relevance of PZD's contribution by linking it to current research on null (or missing) arguments and explore new research questions.
Null arguments have played a central role in the development of generative approaches to syntactic descriptions. The idea that linguistic expressions are the result of a derivation, involving movement and structure-building, led to the postulation of the existence of empty categories in syntactic representations. Chomsky (Reference Chomsky1981, Reference Chomsky1982) proposed that empty categories can be either the result of movement or base-generated as such. This led to some important work on a particular type of null argument, namely null subjects found in tensed clauses in many languages. Pro-drop thus became a widely studied property, before other types of null arguments (and even other types of null subjects; Huang, Reference Huang1984) slowly emerged as potentially relevant to the general discussion. Null direct objects are part of this picture and have been extensively discussed in the theoretical and acquisition literature; see Pérez-Leroux, Pirvulescu and Roberge (Reference Pérez-Leroux, Pirvulescu and Roberge2008) and references cited there. More relevant to this commentary is the fact that the missing argument found in OP constructions has not yet been fully integrated into this general discussion on the nature of the null argument phenomena, beyond Zribi-Hertz's (Reference Zribi-Hertz1984) original proposal, which linked a pro complement to an operator. One reason for this was a lack of tools (or imagination) for such integration but recent developments in Minimalism (Chomsky, Reference Chomsky1995, Reference Chomsky2005, among others) open intriguing possibilities that are captured in the unified approach to argument drop presented in Sigurðsson (Reference Sigurðsson2011). He argues that all definite arguments independently of whether they are overt or null must be linked to a constituent in the edge of the clause (the local C-domain). This C/edge linking is computational (syntactic) in nature and interacts with pragmatics through context scanning. A variety of left-peripheral heads can act as C/edge linkers (CLn) and probe for a goal under Agree. The structure in (1) provides a general picture of how this works.
(1)
In this conception, C/edge linking is not what allows an argument to be null, except in discourse drop languages. Null subjects in Italian, for instance, are pronouns (φ features) incorporated to T. Although he concentrates on null subjects and objects, Sigurðsson (Reference Sigurðsson2011, p. 290) uses “the term (direct) object to refer to objects of both verbs and prepositions (the facts discussed here do not suggest any relevant distinction between prepositional and direct verbal objects)”; however, he does not explicitly discuss argument drop in PPs.
Orphan preposition constructions, such as the one exemplified in (2a) (PZD's (26a)), fit naturally with the model in (1), as shown in the simplified structure in (2b).
a. Lui avait trouvé ce charbon-là, puis il se chauffait avec [] (O).
“He had found that coal there, and he was warming
b.
As for PS, the most natural assumption is that extraction out of PP is excluded in French but not in English.
We are then left with how to treat OP in relative clauses in Canadian French as in (3), PZD's (10c).
(3) Puis il y a bien des affaires j'avais de la misère avec (S).
“And there are lots of things I had trouble with.”
Assuming that no extraction out of PP has taken place, this construction appears to involve an extension of the OP mechanism to a PS context, where the A΄ binder of the null complement of the preposition has been replaced by or reanalyzed as a C/edge linker. Two questions arise. First, pragmatic context scanning clearly plays a crucial role in OP as seen above, especially when the missing argument corresponds to a discourse context entity. However, for OP in relative clauses, we have to assume that the head of the relative counts as the immediate and the only possible controller for the C/edge linker. Second, since in Sigurðsson's approach the fact that an argument can be silent is mostly determined independently of C/edge linking, we still need to settle on what makes it possible for the complement of prepositions to be null. To do so, we can build on recent analyses of null objects of verbs. Cummins and Roberge (Reference Cummins and Roberge2005) propose that the null complement of a transitive verb used intransitively, which is normally interpreted as a non-referential and prototypical object of that verb, see (4a), can acquire a pragmatically inferred interpretation in certain contexts, as exemplified in (4b).
a. I don't like to read ø.
b. Since I have to wait for you, I'll buy a magazine and I'll read ø.
In (4b), the direct object of read has exactly the same status as the one in (4a); however, the context allows for linking with magazine. Cummins and Roberge adopt the structure in (5) proposed by Hale and Keyser (Reference Hale and Keyser2002) for denominal unergative verbs (dance in their case) and extend it to all verbs.
(5)
For Hale and Keyser (Reference Hale and Keyser2002, p. 92) the relation between the verb and its complement in (5) is one of classificatory licensing whereby “the verb identifies the complement to some sufficient extent”.
Now, the fundamental grammaticality contrast in OP constructions between semantically strong and weak prepositions, in both the classic construction and its extension to relative clauses, suggests that the analysis of null objects above should be extended to prepositions. In this approach, a preposition used intransitively, as it is in OP constructions, would be represented as in (6). Weak prepositions (à “to”, “at” and de “of”) can be assumed not to identify their complements (or to identify them too weakly) resulting in their exclusion from OP constructions.
(6)
The null complement would then be open to C/edge linking, resulting in an OP construction or can remain non-linked and retain its “unergative”, i.e. non-referential prototypical, interpretation. While this may not be a widespread possibility, it is attested in Canadian French with the preposition, which, according to PZD, accounts for the bulk of the tokens of OP in their corpus, namely avec “with”. An example is given in (7), which, in the absence of a potential linker, is interpreted as per the translation provided.
(7) Moi avec ø.
“Me too.”
We have thus seen that the Canadian French data examined in PZD provide a new data set that can be used to test the applicability of Sigurðsson's (Reference Sigurðsson2011) approach to argument drop.
I now return to the question of the English influence on OP in Canadian French. As we have seen, and as confirmed by PZD, previous syntactic studies had convincingly demonstrated that OP in relative clauses could not be the result of direct borrowing of an English-type PS construction. As PZD point out (under section heading: “Discussion”), “[t]he apparently widespread belief that surface similarity can (or should) simply be equated with structural similarity makes contact-induced change a logical inference”. This is especially true “among laypeople” and in the prescriptive literature. PZD are quite categorical in their rejection of this unsubstantiated point of view, in the case of the variety represented in their corpus; yet I would like to argue that contact-induced change in this particular construction, i.e. the use of PS (as disctinct from OP), in certain varieties of French is not only hypothetically possible but also attested. Interestingly, PZD's article indirectly provides us with the two phenomena that are necessary (although not sufficient) to conclude that PS exists in a particular variety of French. First, semantically weak prepositions (essentially à and de) should be possible without a complement. Second, complementless prepositions should be found in “true” extraction contexts: wh-questions and pseudo-passives. The two phenomena are illustrated with the English examples in (8) and (9), respectively.
a. This is the person that I was talking to.
b. That is all I can think of.
a. Who did you vote for?
b. This bed was slept in.
In my works with Ruth King and Nicole Rosen, cited in PZD, we provide evidence that in certain varieties of Canadian French (Western Canada and the Maritimes) sentences equivalent to (8) and (9) are grammatical; see Roberge and Rosen's (Reference Roberge and Rosen1999) (12b, c) and (13b), cited here in (10).
a. Qui tu as fait le gâteau pour?
“Who did you bake the cake for?”
b. Le ciment a été marché dedans.
“The concrete was stepped in.”
c. Où il vient de?/Quelle heure elle a arrivé à?
“Where does he come from?/What time did he arrive at?”
Although our work was not corpus-based, these varieties are undeniably in a more intense contact situation than the more central varieties (Quebec and Ontario). Given the linguistic facts and the contact situation, an influence from English appears more likely. The question of course is to determine the exact nature of the change and how it could have occurred (i.e. be implemented in the grammar). A detailed discussion of potential explanations goes beyond the scope of this commentary but two natural options present themselves. First, the contact situation may have led to lexical borrowing of some English prepositions into the French varieties and, if the possibility of PS is somehow linked to a property of prepositions, then the construction could have spread to those regional French grammatical systems; see King (Reference King2000) for an extensive discussion. Second, the contact situation could have acted as a catalyst to further extend PS beyond relative clauses (without movement out of PP) to movement constructions as in (10). Roberge (Reference Roberge and Brasseur1998) and Roberge and Rosen (Reference Roberge and Rosen1999) argue that this change could have arisen through a shift in the C-system in those varieties. Given this view, PS has not been “borrowed” directly as a construction; contact with English would have directed the change in a manner not attested across-the-board in other French varieties.