Acquisition of Putting Event Verbs in French

Acquisition of putting event verbs in French* 1. Introduction It has been argued that the frequent use of a small number of generic verbs favours the child’s abstraction of the verb’s meaning (Clark, 1978, 1996; Sandhofer, Smith & Luo 2000, Slobin, 1996). Goodman & Sethuraman (2006) propose that the consistent adult use of light verbs in a given syntactic frame facilitates the child’s acquisition of other verbs used for the same kind of activity in that frame. Goldberg, Casenhiser & Sethuraman (2004) attribute a learning advantage to high frequency generic verbs due to considerable regularity between the syntactic patterns in which the verbs are observed and their semantics. For instance, verbs found in the pattern “X causes Y to move Z path/location” (Subject Verb Object Oblique, John puts/places/pushes the glass on the table) were more often than would be expected by chance to be associated with caused-motion in the speech directed to 27 English-speaking children as well as in the early speech of the children. The light generic verb mettre (‘put’) is very frequent in spoken French, it can be used to denote a multitude of putting events (caused-motion actions) and it is acquired early on (Chenu & Jisa, 2005). In addition to the generic verb mettre, French has a host of other more semantically specific verbs that can be used for the same caused-motion situations as mettre (for example, accrocher, ‘to hook onto’, attacher, ‘to attach to’, enfiler ‘to slip on’). These more specific verbs are observed infrequently in child-directed speech (Chenu & Jisa, 2006). We predicted that the frequent use of the generic verb mettre (‘put’) in a consistent syntactic frame would lead to the acquisition of other verbs encoding caused-motion events. Our findings, however, do not confirm the prediction. We attribute our results to the fact that caused-motion verbs in French confront the child with two contrasting patterns: generic mettre follows a satellite-framed pattern while the other more semantically specific verbs follow a verb-framed pattern (Talmy, 2000). In the following section the contrasting types of putting event verbs will be described in the attempt to show how mettre differs semantically from the more specific verbs. Subsequently we will present a longitudinal study of putting event verbs observed in three caretaker-child dyads. In the conclusion we speculate on the role of frequency and typological shifts. 2. Two different patterns in French: distributed vs conflated action & relation Four meaning components, illustrated in Figure 1, are necessary for the description of putting events (Bowerman, Brown, Eisenbeiss, Narasimhan & Slobin, 2002). In this section we detail the two contrasting semantic patterns for encoding putting events in French. FIGURE ACTION GROUND RELATION the object that is caused to move the placement action (caused motion toward a GROUND) the location toward or from which the FIGURE is moved the resulting spatial relation between the FIGURE and the GROUND Figure 1 - Meaning components in caused motion activities (adapted from Bowerman, Brown, Eisenbeiss, Narasimhan & Slobin, 2002) The generic verb mettre distributes the semantics of ACTION and RELATION between the verb and the preposition, i.e., mettre dans (‘put in’), mettre sur (‘put on’), mettre entre (‘but between’), mettre à côté de (‘put next to’). In contrast, more specific verbs such as accrocher, ‘to hook onto’ or enfiler (‘slip into’) conflate ACTION and RELATION. Accrocher, for example specifies that the FIGURE be attached to the GROUND in a hooked RELATION. The verb accrocher also specifies that the GROUND should have a prominent part which can be conceived of as a hook. The verb enfiler specifies that the FIGURE be moved through the GROUND and that the resulting RELATION should be a tight fit, such as putting an arm through a sleeve, or socks on a foot or stringing pearls on a string. Thus, the lexicalisation patterns for the more specific verbs conflate two meaning components, ACTION and RELATION, whereas the generic verb mettre distributes the meaning components between the verb and an oblique locative expression. In addition to conflating ACTION and RELATION, the semantically more specific verbs constrain aspects of the FIGURE and the GROUND. The generic verb mettre places no restrictions on the FIGURE. In contrast, the more specific verbs construe the FIGURE. For example, the verb verser, ‘to pour in/onto’ requires that the FIGURE be liquid (water, juice) or composed of a number of small elements (flour, sand, rice). The verb accrocher, ‘to hook onto’, specifies that the FIGURE has a part which can be conceived of as a hook. Mettre places no restrictions on the GROUND whereas the more specific verbs construe the GROUND. With a verb such as insérer (‘to insert into’) the GROUND is construed of as being larger than the FIGURE, the relationship between the GROUND and the FIGURE must be in a tight fit and the GROUND must be conceived of as a container. From an historical perspective many of the semantically more specific caused motion verbs can be described as consisting of a prefix and either a verb or a noun (cf. Kopecka, 2006 for an excellent review). Kopecka shows how this earlier prefixing system leads to variable lexicalisation patterns. For example, the verb enfouir, ‘to bury in’, consists of the Latin prefix en(> Latin in-) and the verb fouir, ‘to dig’. The verb accourir, ‘to run into’ consists of the Latin prefix a(d)-, ‘to, toward’ and the verb courir ‘to run’. This lexicalisation pattern was highly productive in Old and Middle French and reveals an earlier satellite-framed character. The derived verbs can have either a verb or a noun base. When the derived form contains a verb as its base, as is the case with enfouir or accourir, the prefix generally gives information concerning the path of movement. When the derived form has a noun base, the GROUND element and the relation between the FIGURE and the GROUND are generally more specified. For example the verbs accrocher, ‘to hook onto’ (Latin a(d)-, ‘to, toward’ + croch-er), and décrocher, ‘to unhook’ (Latin dé(s)/dis-, ‘from, off, apart’ + croch-er) were derived by prefixing a- or dé- to the noun crochet, ‘hook’ and subsequently adding a verbalising suffix –er. In other verbs derived with noun bases it is clearly the GROUND element that is more saliently specified. For example the verb emboîter (em- (> Latin in-) boîte-er > boîte ‘box’) specifies that the GROUND be large enough to contain part of the FIGURE and the resulting relationship be a tight fit, for example, fitting together open blocks of different sizes, each one of the blocks slightly smaller than the others which serve as the GROUND. Another pattern of lexicalisation involves more salient specification of the FIGURE in verbs such as écrémer, ‘skim’, derived from crème, ‘cream’ or épépiner, ‘take seeds out’, derived from pépin, ‘seed’. The derivation of verbs through prefixing was extremely productive in Old and Middle French. A few of these prefixes are still productive in Modern French (i.e., dé- inversive and re- iterative) but the majority are no longer considered productive (Kopecka, 2006). And many of the derived verbs are no longer attested without the prefix. While the verb courir is frequent in Modern French, the verb fouir, ‘to dig’ is very rarely attested and the verb *sérer in insérer, ‘to insert’ no longer exits. In contrast to non-derived caused motion verbs such as mettre, ‘to put’, poser ‘to put down’, jeter ‘to throw’, placer, ‘to place’, the derived verbs place more restrictions on the preposition used to introduce the GROUND as illustrated in (1) with the derived verbs insérer, ‘to insert’ and enfouir ‘to bury’. (1) a. insérer la pièce de monnaie dans/*sur la machine. ‘insert the coin into/*onto the machine’ b. mettre la pièce de monnaie dans/sur la machine ‘put the coin into/on (top of) the machine’. c. enfouir le sac dans/*sur l’armoire ‘bury the bag in/*on the chest of drawers’ d. mettre le sac dans/sur l’armoire ‘put the bag in/on the chest of drawers’ Insérer is compatible with all spatial prepositions which construe the FIGURE and the GROUND as resulting in a tight fit relation and the FIGURE as being surrounded by the GROUND. Thus, it is not compatible with a preposition such as sur, ‘on’ which encodes contact between the FIGURE and the GROUND but not complete inclusion of the FIGURE by the GROUND. Enfouir, ‘to bury in’ is compatible with prepositions that construe the GROUND as a container. Thus, it is not compatible with sur, ‘on’ which encodes a support relation of the FIGURE by the GROUND, but not containment of the FIGURE by the GROUND. The non-derived caused motion verbs are compatible with all spatial prepositions. The more specific non-derived putting event verbs differ from mettre in the way the meaning components involved in motion events are combined. For example, verbs such as jeter, ‘to throw’ or rouler, ‘to roll’ specify the manner of movement of the FIGURE. Verbs such as coincer, ‘to stick into’ or planter, ‘to drive/stick into’ specify the resulting relation between the FIGURE and the GROUND. The semantically more specific verbs, both derived and non derived, also render somewhat unnecessary the expression of the GROUND semantic component and thus fit more consistently with the typical pattern in verb-framed languages (Bowerman, et al., 2002). The generic verb mettre calls for expression of the GROUND. The contrast between the semantically specific non derived verb poser ‘to put down’ (2a and 2c) and generic mettre (2b and 2d) is illustrated below. (2) a. b. c. d. Jean pose la tasse sur la table Jean met la tasse sur la table Jean pose la tasse. */?Jean met la tasse ‘John sets the cup down on the table.’ ‘John puts the cup down on the table.’ ‘Jean sets the cup down.’ ‘*Jean put the cup’ To conclude this section, the semantic specific verbs are of two types: derived verbs and non derived verbs. They are considered more specific than the generic verb mettre because they construe more specifically the FIGURE, the GROUND, the relation or the action. The generic verb mettre does not. In addition, for all the events which can be expressed using a specific verb, mettre can be used. The inverse, however, is obviously not possible. 3. Word order As mentioned in the introduction, Goldberg, et al. (2004) observed that English verbs found in the pattern “X causes Y to move Z path/location” (Subject Verb Object Oblique, John puts/places/pushes the glass on the table) were more often than would be expected by chance to be associated with caused-motion in the speech directed to 27 English-speaking children as well as in the early speech of the children. Word order in French constructions expressing caused motion shows much more variation than does English. When the object (= FIGURE) is a lexical noun or a pronoun in a declarative it follows the verb (3a and 3b). In imperatives the object (= FIGURE) also follows the verb (3c). However when the object (= FIGURE) is a clitic, as is often the case in conversation, it is placed before the verb in declaratives (3d). In addition, spoken French is characterised by a number of dislocation structures which alter the position of arguments (3e and 3f). (3) a. Jean met la tasse sur la table. 'Jean puts the cup on the table.' b. Jean met ça sur la table. 'Jean puts that on the table' c. Mets la tasse/la sur la table. 'Put the cup/it on the table' d. Jean la met sur la table. 'Jean puts it on the table' e. La tassei Jean lai met sur la table. 'The cupi Jean puts iti on the table.' f. Jean lai met sur la table la tassei. 'Jean puts iti on the table the cupi.' Thus, the pairing of one constructional pattern with the set of caused motion verbs is a less salient feature of French than of English (Chenu & Jisa, 2006). 4. Participants Three monolingual French-speaking children (two girls - Marie and Anaïs - and one boy Tim) and their caretakers were recorded every two weeks for approximately one hour in spontaneous interaction in the home. For this study twelve transcriptions per child and three transcriptions per caretaker were used. Table 1 gives the children’s ages in months and the total number of utterances produced by the children and the caretakers. In addition the children are classed into three MLU periods: MLU less than 2, MLU ranging from 2 to 3 and MLU at 3 or more. Anaïs Age in months 24-27 Total utterances 872 491 598 1106 MLU Anaïs’s caretaker Marie Marie’s caretaker 21-23 623 - 320 384 448 602 Tim 21-23 808 - 439 536 603 645 <2 <2 <2 Age in months 31-33 23-25 28-29 Total utterances 685 1283 507 985 MLU Age in months 821 - 441 653 483 449 677 - 711 587 668 605 2-3 2-3 2-3 35-37 29-31 30-32 Total utterances 681 636 746 804 MLU >3 1151 - 539 670 492 525 >3 Tim’s caretaker 777 - 465 524 573 501 931 - 734 - 800 - >3 Table 1 – Description of the corpus at each time point as defined by MLU: Age in months and total utterances produced by Anaïs, Marie and Tim and their caretakers. 5. The study of putting event verbs a. Distribution of mettre and the semantically more specific verbs All of the caused motion verbs were identified in the utterances produced by the children and the caretakers, following the guidelines given in Bowerman, et al. (2002). Given in Appendix 1 is the full list of putting event verbs observed in the corpus. Diagrams 1a, 1b and 1c show the distribution (in tokens) of the generic verb mettre, the non derived and the derived semantically more specific verbs that were observed in the three dyads at the three time points. At the first time point the children use very few putting event verbs. Mettre is more frequent than the more specific verbs for Marie at the first time point. In contrast, Anaïs and Tim do not show a predominance of mettre until the second time point. From the second time point on the children produce more putting event verbs and all three of them show a predominance of mettre, as do their caretakers. 160 mettre 140 non derived derived 120 100 80 60 40 20 0 Anaïs; 24-27 Months Anaïs; 31-33 Months Anaïs; 35-37 Months Anaïs caretaker Diagram 1a - Distribution (in tokens) of the generic verb mettre, the non derived and the derived semantically more specific verbs for Anaïs at the three time points 160 mettre 140 non derived derived 120 100 80 60 40 20 0 Marie; 21-23 Months Marie; 23-25 Months Marie; 29-31 Months Marie caretaker Diagram 1b - Distribution (in tokens) of the generic verb mettre, the non derived and the derived semantically more specific verbs for Marie at the three time points 160 mettre 140 non derived derived 120 100 80 60 40 20 0 Tim; 21-23 Months Tim; 28-29 Months Tim; 30-32 Months Tim caretaker Diagram 1c - Distribution (in tokens) of the generic verb mettre, the non derived and the derived semantically more specific verbs for Tim at the three time points b. Coding of the semantic components and the syntactic frame All of the constructions containing putting event verbs were coded for semantic components and syntactic frame. Figure 2, below, illustrates the semantic and syntactic coding of generic mettre, ‘put’ and more specific poser, ‘put down’. The categorization of the semantic components is based on Bowerman, et al. (2002). The coding of the syntactic frame takes into account 1) the varying word order in French of the verb and the object and 2) the expression or omission of the oblique locative expression. Semantic Components Syntactic Frame Jean met/pose la tasse sur la table ‘Jean puts/puts+down the cup on the table’ verb+figure+ground verb+object+locative Jean la met/pose sur la table ‘Jean puts/puts+down it on the table’ figure+verb+ground object+verb+locative Jean met/pose la tasse ‘Jean puts/puts+down the cup’ verb+figure verb+object Jean la met/pose ‘Jean puts/puts+down it’ figure+verb object+verb Figure 2 – Semantic components and syntactic frame coding For the current study we concentrate on the semantic component GROUND encoded or not by a locative expression. As mentioned earlier, the semantically specific verbs conform more consistently to a verb-framed pattern in that the GROUND can be omitted. Generic mettre requires that the GROUND be mentioned and conforms more consistently to a satellite-framed language. 6. Results The results show that the semantic components and the syntactic frames differ significantly for mettre as opposed to the other putting event verbs. In particular the GROUND semantic component encoded with a locative expression is mentioned more often with the generic verb mettre in putting event constructions for all three children and for their caretakers. In contrast, the GROUND semantic component (locative expression) is omitted with the more specific verbs which conflate action and relation. Thus, the different types of verbs elicit different semantic and syntactic patterns. Diagrams 2a and 2b, 3a and 3b, 4a and 4b, show the distribution of the expression of the GROUND with the semantically specific verbs, both derived and non derived, and the expression of the GROUND component with the generic verb mettre. Anaïs; 24-27 months Anaïs; 31-33 months mettre with ground Anaïs; 35-37 months Anaïs caretaker mettre without ground 0 10 20 30 40 50 60 70 80 Diagram 2a - Distribution of the semantic component GROUND with generic mettre (in tokens) for Anaïs Anaïs; 24-27 months Anaïs; 31-33 months specific verbs with ground Anaïs; 35-37 months Anaïs caretaker specific verbs without ground 0 10 20 30 Diagram 2b - Distribution of the semantic component tokens) for Anaïs 40 50 GROUND 60 70 80 with more specific verbs (in mettre with ground Marie; 21-23 months Marie; 23-25 months mettre without ground Marie; 29-31 months Marie caretaker 0 10 20 30 40 50 60 70 80 Diagram 3a - Distribution of the semantic component GROUND with generic mettre (in tokens) for Marie Marie; 21-23 months Marie; 23-25 months specific verbs with ground Marie; 29-31 months Marie caretaker specific verbs without ground 0 10 20 30 Diagram 3b - Distribution of the semantic component tokens) for Marie 40 50 GROUND 60 70 80 with more specific verbs (in Tim; 21-23 months Tim; 28-29 months mettre with ground Tim; 30-32 months Tim caretaker mettre without ground 0 10 20 30 40 50 60 70 80 Diagram 4a - Distribution of the semantic component GROUND with generic mettre (in tokens) for Tim Tim; 21-23 months Tim; 28-29 months specific verbs with ground Tim; 30-32 months Tim caretaker specific verbs without ground 0 10 20 30 Diagram 4b - Distribution of the semantic component tokens) for Tim 40 50 GROUND 60 70 80 with more specific verbs (in A series of individual ² analyses of the children’s productions reveal that for all three children the expression of the GROUND with mettre is much more frequent than the omission of the GROUND with mettre (Anaïs, p > 0.02; Marie, p > 0.0001; Tim, p > 0.005). The omission of the GROUND component is much more frequent with specific verbs than is the expression of the GROUND component with the specific verbs (Anaïs, p > 0.0001; Marie, p > 0.0001; Tim, p > 0.001). Finally, the expression of the GROUND with mettre is much more frequent than the expression of the GROUND with the more specific verbs (Anaïs, p > 0.001; Marie, p > 0.001; Tim, p > 0.001). 7. Discussion Our results show that the pattern in which the very frequent generic verb mettre occurs does not match that of the other putting event verbs. Thus, there is no evidence that the generic mettre syntactic frame provides a learning advantage for the other semantically more specific verbs. This lack of a learning advantage is due to the fact that French is a typologically mixed language. In French, verbs which follow the satellite-framed pattern encourage the child to express the GROUND in a locative phrase while the verbs following the verb-framed pattern do not. Mettre (satellite-framed pattern) requires mention of the GROUND to encode the RELATION: the semantic components ACTION and RELATION are distributed in the construction. More specific verbs (verbframed pattern) do not require mention of the GROUND to encode the RELATION: ACTION and RELATION are conflated in the verb. Thus, the two typologically contrasting verb types (distributed vs conflated verbs, ie. generic mettre vs specific caused-motion verbs in this study) confront the child with two different semantic and syntactic patterns. And the children’s production reflect these two contrasting patterns. 8. Conclusion The frequency of mettre can be attributed to the fact that it is a basic verb, it is easily accessible and it facilitates rapid mapping of a putting event situation and an appropriate means of expression. In conversational face-to-face caretaker-child interaction, more often than not the FIGURE and the GROUND are present in the visual context and the relation between them is highly predictable. Thus, the semantically more specific verbs are communicatively less motivated if one follows the Gricean maxims which dictate that a speaker give enough information – not too much and not too little – in the message. We can illustrate how a more specific verb can become communicatively motivated with an example from our corpus. Tim’s mother repeatedly asks him to hang up his jacket on a wooden peg by saying tu mets ta veste là?, ‘you put your jacket there?’. After a few unsuccessful attempts Tim lets his jacket fall on the floor. Tim’s mother, by this time very displeased with him says accroche-la comme il faut!, ‘hook it on properly!’. Without a specific communicative motivation, there is no reason to use a semantically specific verb. The verb mettre is perfectly appropriate given that the relation between the jacket and the peg on the wall are visible and the resulting relation between them is highly predictable. However, when Tim’s mother wants to draw specific attention to the resulting relationship between the jacket and the peg, she uses a semantically more specific verb. The results indicated that the frequent use of mettre in a stable syntactic frame does not provide children with a learning advantage for the more specific putting event verbs. We attributed this finding to the typologically hybrid status of Modern French. A question for which we can offer no answer is the following: How does frequency influence typological change? Mettre which follows a satellite-framed pattern is much more frequent than the other more specific verbs which follow a verb-framed pattern. To what extent will a frequent basic verb, such as mettre, move French to a less hybrid system? Slobin has suggested that semantically general basic or light verbs are at the beginning of grammaticization clines because: ‘….when they are used in a conversational context, they contrast with the more specific verbs that could be used in that context, thereby allowing the hearer to infer that a more specific meaning may not have been intended. This opens the way for the kinds of pragmatic inferencing and reanalysis that lie at the heart of grammaticization.’ (Slobin, 1997:24). Future research will be undertaken in two directions. First, we will attempt to show that the frequent use of mettre is not specific to caretaker-child interaction, but rather specific to spoken French in general. Second, we are currently studying the communicative motivations behind the children’s use of the semantically more specific verbs.