Grammaticalization of the French and Bulgarian Causative Constructions . Some Diachronic and Developmental Aspects

The present article deals with the processes of grammaticalization from a diachronic and developmental point of view. To compare grammatical changes in history and acquisition, we focus on the French and Bulgarian causative constructions (e.g. Fr. faire travailler qn vs. Blg. karam njakogo da raboti – ‘to make someone work’). A total of 113 French speakers (71 children and 42 adults) and 96 Bulgarian speakers (56 children and 40 adults) took part in this cross-linguistic study. Children were aged 3 to 6 years at the time of the study. Results show that historically, the French causative construction evolves from analytic devices to synthetic forms. As a compact structure, the faire + Vinf complex predicate requires argument rearrangement and clitic raising. These specificities explain why its acquisition is difficult and occurs at a late stage in French-speaking children. The Bulgarian causative construction evolves in the opposite direction, from synthetic devices to less grammaticalized structures. As an analytic form including two predicates followed by their own arguments, the karam NP da + Vpres periphrastic causative seems easier to acquire and its full command by children is achieved earlier. Finally, we suggest that there are some similarities between language history and language acquisition with regard to the stages of competition between two causative mechanisms and the stabilization of the new construction.


INTRODUCTION
Researchers have been interested in grammaticalization from a diachronic and developmental perspective for many years (Slobin, 2002;Givón, 2009;Diessel, 2012 inter alia).However, the abundant linguistic literature does not reveal any agreement regarding the way these two processes could be related.According to Alain Peyraube (2002), syntactic change has to be considered either from a formal (generative) or a functional-discursive (functional-cognitive) perspective.In the generative approach, the acquisition of language by children influences grammatical change that is non-progressive and independent of all functional, semantic and pragmatic considerations (Lightfoot, 1979).The proponents of the functional-cognitive approach, by contrast, consider that language evolution can be explained by the primary function of language -the communicative function.In this framework, linguistic change is related to various biological, psychological, historical, sociocultural, environmental and developmental factors that play a crucial role in characterizing linguistic structure (Langacker, 1999).Functionalists also argue that language history and language acquisition often seem parallel because they involve similar mechanisms of change (Ziegeler, 1997) or similar adaptive behaviors (Givón, 2009).
Looking for some points of convergence between diachrony and acquisition, scholars have compared various linguistic devices, such as adpositions or English past tense forms (Clark and Carpenter, 1989;Ziegeler, 1997;Bybee andSlobin, 1982a, 1982b).These studies have provided striking results.They have revealed some similarities between the two processes, in particular at the semantic-pragmatic level and more rarely at the morphological level.Bybee andSlobin (1982a, 1982b), for example, investigated the convergence between diachrony and acquisition at the grammatical level.They noticed a tendency to regularize the past tense forms of less frequent irregular English verbs (e.g.break -*breaked [broke], go -*goed [went]) in both historical and child language data1 .However, despite this similarity, there are differences between children's and adults' erroneous past tense forms.The former can be observed only during initial stages of language development and are never maintained in adulthood.The latter, by contrast, persist and are passed on to subsequent generations.It should also be noted that in the sociolinguistic framework, the transmission of change is determined by power and prestige (Labov, 2001).Since children do not constitute an influential group, their linguistic innovations should not have a strong effect on adult language use.The particularities of children's and adults' new productions allow Bybee andSlobin (1982a, 1982b) to conclude that adults rather than children are the instigators of morphological change.
In this paper, we propose to shed some light on the possible parallel between processes of grammatical evolution.We will attempt to do this by studying the functioning of the French and Bulgarian causative constructions in history and child language, at the grammatical and semantic levels.Our hypothesis is that there are some points of convergence between the historical change and the acquisition process of the causative constructions.

DIACHRONIC EVOLUTION OF THE FRENCH FAIRE + VINF CONSTRUCTION
We will first present the diachronic evolution of the French faire + Vinf construction (Chamberlain, 1986;Simone and Cerbasi, 2001).The three main steps characterizing this process of language change are summarized in Table 1.In Latin, we have bi-predicative constructions including one causative verb, the complementizer 'ut' or a noun phrase (NP) and a second verb in the subjunctive or infinitive (see Table 1, row 1).In Old and Middle French, we can observe the competition between two kinds of faire + Vinf constructions: the old one that still functions like a bi-predicative construction including an NP (e.g.Carles ferat l'ost returner.)and the new one that works as a complex predicate (e.g.Ki tant me fist cunquere.)(Table 1, row 2).Finally, in Modern French, the use of the faire + Vinf complex predicate becomes the norm for expressing causativity.As a complex predicate, this construction is strongly grammaticalized and it requires a special rearrangement of the causer and the causee arguments as well as clitic raising (Gaatone, 1976).It is not possible to insert clitics or an NP between the causative verb faire and the infinitive.To sum up, the French causative construction evolves from analytic forms (bi-predicative constructions) to more compact, synthetic forms (complex predicate).

DIACHRONIC EVOLUTION OF THE BULGARIAN KARAM NP DA + VPRES CONSTRUCTION
In this section, we will present the historical evolution of the Bulgarian karam NP da + Vpres construction (Vaillant, 1966;Haralampiev, 2001).Table 2 illustrates the key steps in this process of linguistic change.Table 2 shows that in Old Slavic, causativity is encoded through morphological devices in 'iti' (see examples in row 1).In Old and Middle Bulgarian, we can observe the fluctuation between several constructions.The first one functions like a complex predicate (e.g.Nebogu trepetati sŭtvori.) or a bi-predicative construction (e.g.Sŭtvorite člověky vŭzlešti.)(see Table 2, row 2).The second one is a new kind of bi-predicative construction where the lexical verb is no longer used in the infinitive but in the present, and it is also part of the da-construction.In fact, between the 12 th and the 15 th centuries, the Bulgarian synthetic infinitive is gradually replaced by an analytic one -the da-construction, which includes the conjunction 'da' and the main verb in the present tense.Finally, in Modern Bulgarian, the V1 caus + NP + da conj + V2 pres periphrastic causative becomes the conventional usage.As a bi-predicative construction, the Bulgarian causative is less grammaticalized; each verb in this structure is autonomous and followed by its own arguments (Novakova, 2010) (Table 2, row 3).
In brief, the historical evolution of the Bulgarian causative construction follows the opposite pathway to that of French: from compact, morphological devices to analytic forms (bi-predicative constructions)3 .

ACQUISITION OF THE FRENCH FAIRE + VINF CONSTRUCTION
The following section presents the key steps of the acquisition of the faire + Vinf complex predicate, based on Sarkar's longitudinal study (2002) with eight 1.9 to 4-year-old French-speaking children from Quebec (Canada) (see Table 3).The gradual emergence of the French causative construction in child language is marked by three main stages.The first stage is characterized by the omission of the causative verb 'faire'.These kinds of productions are known as transitive causatives (see Table 3, row 1).During the second step, the causative verb faire is still occasionally omitted.In addition, Sarkar observed the coexistence of two structures.The former corresponds to the conventional use of the faire + Vinf complex predicate and the latter is agrammatical, with an inappropriate NP insertion between the causative verb faire and the main verb in the infinitive (Table 3, row 2).Finally, in stage 3, children become able to use causative constructions properly, but only with intransitive verbs.For this reason, Sarkar (2002) considers the existence of an additional step 4 (beyond the age of 4), where children achieve an adult-like competence in producing causatives with transitive verbs (see Table 3, rows 3 & 4).
To sum up, according to Sarkar, the developmental path of the faire + Vinf complex predicate in child language goes from synthetic forms (lexical causatives or transitive causatives) to more complex structures including several arguments (causer, causee, object).Regarding Bulgarian, the acquisition of causative constructions in child language has never been observed.

THE PRESENT STUDY
In this section, we present new results on the acquisition of the French and Bulgarian causative constructions, based on a large-scale study including both young children and adults who participate in a production task.

Sample
Two hundred and nine French and Bulgarian native speakers took part in this cross-linguistic study.All details concerning the sample are summarized in Table 4.As shown in Table 4, in each language, the children were divided into three age groups: 3 to 4, 4 to 5 and 5 to 6 years.Adult speakers were solicited in order to allow a better assessment of children's production skills with causatives.All participants were observed during individual videotaped sessions.Interviews with the children were held in the kindergarten and those with adults at their workplace.

Experimental task
The production task was constructed using extracts from six cartoons including causative situations.These causative situations were based on six target verbs: rire/smeja se ("to laugh"), pleurer/plača ("to cry"), tomber/padam ("to fall down"), danser/tancuvam ("to dance"), boire/pija ("to drink") and manger/jam ("to eat").Children and adults were asked to watch each video and then to answer three questions.Table 5 provides an overview of the organization of the experimental sessions.Kakata kara bebeto da se smee (causativity).
The girl makes the baby laugh.
As illustrated in Table 5, the production task was built around three graduated questions.The first one focused on the causer argument (e.g.What is X doing?).In this case, two main productions were expected: one encoding a causative situation or another one depicting only the causer event (see examples row 2).When the participant used a causative construction, we proceeded with the next video.And when he/she did not, we continued with questions 2 and 3. Question 2 focused on the causee argument (e.g.What is Y doing?) and the target form described the related event (see Table 5, row 5).The final question 3 concerned the entire causative scene (e.g.What is X doing to Y?); this time, only causative mechanisms were expected at the production level (Table 5, row 7).

Results
First of all, we will discuss conventional uses of the French and Bulgarian causatives.Then, we will focus on the unusual productions related to these analytic devices.

Conventional uses of the French and Bulgarian causative constructions
By conventional use of the French and Bulgarian causatives, we mean a correct production of the sequences faire + Vinf / karam NP da + Vpres and also proper use of the causer and causee arguments.We calculated an accuracy ratio as follows: Number of times the participant properly produces the target causative / Number of times the participant uses this linguistic construction.1 (a & b) indicates that between 3 and 6 years, French and Bulgarian children are able to correctly produce the causative constructions of their language; the mean scores exceed 50%.However, Bulgarian children's performances are better, in particular in 4-to 5-year-olds.At this developmental stage, French children show a mean accuracy of 67%, while Bulgarian children already demonstrate an adult-like competence (an average of 93%) in producing periphrastic causatives.These results suggest that the Bulgarian karam NP da + Vpres less grammaticalized causative is easier to acquire and its emergence in child language is probably earlier (before the age of six).

Unconventional uses of the French and Bulgarian causative constructions
The unconventional uses of the French causative construction are illustrated in Table 6 below.We noted three specific cases of unusual productions of the faire + Vinf complex predicate in our study.The first one is related to causer and causee arguments.For instance, we found errors on the causee's syntactic function, and some cases of argument omission (see Table 6, rows 1 & 2).Our data also revealed some cases of overgeneralization errors.These agrammatical productions are characterized by the improper addition of the causative verb faire to a lexical causative, which can in itself convey causativity (Table 6, row 3).The last specific case of incorrect use observed in our French data includes productions where the entire faire + Vinf structure is affected.In particular, we found examples with a conjugated lexical verb, and some factitive ellipses with missing causative verb and causer argument (see Table 6, rows 4 & 5).Finally, transitive causatives and NP insertions mentioned in Sarkar's (2002) experimental work also occurred in our study.However, it is worth noting that between 3 and 6 years, these errors are extremely unusual in child language and we did not obtain any significant statistical results (Table 6, rows 6 & 7).
Turning now to the unconventional uses of the Bulgarian karam NP da + Vpres construction, the cases observed in the study are listed in Table 7.The unusual productions found in Bulgarian data can also be divided into three categories.The first one is characterised by the inappropriate position of the causee argument (see Table 7, row 1).The second category includes the periphrastic ellipses.In this case, the periphrastic causative is reduced to the simple da-construction; the causative verb karam and the causer and causee arguments are omitted (Table 7, row 2).Finally, we observed some productions that are very unusual for this developmental stage, namely constructions with a less appropriate or an inappropriate causative verb and some transitive causatives (Table 7,rows 3 & 4).This kind of errors did not reach the statistical level of significance, however.

GENERAL DISCUSSION AND CONCLUSION
In this final section, we will point out the possible convergence between historical change and language evolution in children, taking into account only the French faire + Vinf construction.In diachrony, the French causative construction evolves from analytic forms (bi-predicative constructions in Latin) to more compact mechanisms (a complex predicate in Modern French).By contrast, in child language, the evolution of this construction follows the opposite pathway: from synthetic forms (lexical causatives or transitive causatives) to more analytic devices (a complex predicate).We can therefore notice two points of convergence between these grammaticalization processes.At steps 2 and 3, we can observe the competition between two alternative constructions, before the final standardization of the new usage.However, we note that this similarity between language change and language acquisition is not specific to causatives.It also can be associated with all kinds of linguistic devices, which determines its general nature.
In conclusion, we would like to highlight two points.The first one is related to the gradual emergence of the French and Bulgarian causatives in child language.As a strongly grammaticalized construction, faire + Vinf requires argument rearrangement and clitic raising.These particularities explain its late stabilization in child productions (at around 6 years and beyond).On the other hand, the Bulgarian karam NP da + Vpres is less grammaticalized by nature; it involves two autonomous predicates, each of them occurring with its own arguments.For that reason, we believe that the Bulgarian periphrastic causative is easier to acquire and its full command by children is achieved earlier (before the age of 6).
The second important point is related to the comparison between language changes in history and acquisition.As long as causatives are concerned, there is no clear evidence that the acquisition of grammatical devices repeats their diachronic evolution or that grammatical development in history originates from changes in child language.According to Slobin (2002), language changes in history and acquisition have rather to be considered as two independent processes, because children and adults are involved in different communicative tasks.Children have to discover meanings that are present in the ambient language.They create novel forms and novel meanings, but their innovations do not survive into adulthood and thus, have no effect on adult language.In contrast, adult speakers, who already have a well-established linguistic system, sometimes extend the meaning of existing expressions to novel meanings by pragmatic inference.Their innovations persist and are transmitted from generation to generation.francuskojęzycznych.Bułgarska konstrukcja kauzatywna ewoluuje w przeciwnym kierunku, od form syntetycznych do struktur mniej gramatycznych.Jako forma analityczna zawierająca dwa predykaty, którym towarzyszą ich własne argumenty, peryfrastyczna konstrukcja kauzatywna błg.karam NP da + Vpres wydaje się łatwiejsza do nauczenia, a jej pełne opanowanie przez dzieci osiągane jest wcześniej.Na koniec sugerujemy, że istnieją pewne podobieństwa między historią języka a nabywaniem języka w związku z okresem konkurowania między dwoma mechanizmami kauzatywnymi i stabilizacją nowej konstrukcji.

Figure
Figure 1(a & b)  indicates that between 3 and 6 years, French and Bulgarian children are able to correctly produce the causative constructions of their language; the mean scores exceed 50%.However, Bulgarian children's performances are better, in particular in 4-to 5-year-olds.At this developmental stage, French children show a mean accuracy of 67%, while Bulgarian children already demonstrate an adult-like competence (an average of 93%) in producing periphrastic causatives.These results suggest that the Bulgarian karam NP da + Vpres less grammaticalized causative is easier to acquire and its emergence in child language is probably earlier (before the age of six).

Table 1 .
Diachronic evolution of the French faire + Vinf construction 2

Table 2 .
Diachronic evolution of the Bulgarian karam NP da + Vpres construction piti (drink) → poiti (to make drink) mrěti (die) → umoriti (to make die) caus + NP + da conj + V2 pres Nebogu trepetati sŭtvori.Lit: He made tremble the poor man Sŭtvorite člověky vŭzlešti!'Make the men lie down' Văzljaze da vidit'/vidjat' Isusa.'He/She went out to see Jesus' caus + NP + da conj + V2 pres Učiteljat kara učenicite da četat uroka.Lit: The teacher makes that the students read the lesson Toj kara učenicite da go četat.Lit: He makes that the students read it Toj gi kara da četat uroka.Lit: He makes that they read the lesson Toj gi kara da go četat.Lit: He makes that they read it

Table 4 .
Sample of subjects 4

Table 5 .
Structure of the production task

Table 6 .
Unconventional uses of the French faire + Vinf construction

Table 7 .
Unconventional uses of the Bulgarian karam NP da + Vpres construction Table 8 briefly summarizes the characteristics of the two grammaticalization processes.

Table 8 .
Processes of grammaticalization of the French faire + Vinf construction