Questions make up approximately one-third of mothers’ speech to their young children (Cameron-Faulkner, Lieven, & Tomasello, Reference Cameron-Faulkner, Lieven and Tomasello2003). Children's answers to questions have been examined almost exclusively with respect to syntactic and semantic aspects of question–answer pairs (e.g., Ervin-Tripp, Reference Ervin-Tripp and Hayes1970; Peterson & Biggs, Reference Peterson and Biggs1997; Tyack & Ingram, Reference Tyack and Ingram1977). However, an answer is expected above all to be informative for the questioner, that is, to provide the questioner with information that is new, rather than with information that is already given (Kiefer, Reference Kiefer and Meyer1988). What is given information is partly determined by the wording of the question; that is, information that is already mentioned in the question is assumed to be known to the questioner (Lambrecht, Reference Lambrecht1994). However, in addition, the context in which a question is uttered often also provides given information in addition to that which is included in the question itself (Kiefer, Reference Kiefer and Meyer1988). This context can be either verbal or visual: information can have been mentioned already and/or can be available in the context and the visual focus of attention. In either case it can be considered as “given” (Chafe, Reference Chafe and Li1976). That is, when an element is given in the preceding context (i.e., prior to the question) there is no need to present it as new information (e.g., in a lexical noun) in the answer, even when it constitutes part of the requested information. Instead, it may be either omitted or marked as given, for example, by using a pronoun.

Giving an appropriate answer to a question while taking the preceding context into account requires a twofold effort: (a) one needs to pay attention to the question and determine what information is already given and what information is requested, and (b) one needs to judge with respect to the preceding context what part of the requested information is already given and tailor the answer accordingly. Because questions narrowly focus on the information required, they are an interesting topic for understanding how children coordinate given and new information within and between utterances in discourse. One might expect this to be a more difficult task than simply following the preceding utterance of an interlocutor in a discourse-coherent way. There is also the question of whether prior mention in discourse makes it easier or more difficult to identify given information by contrast with the information being presented only visually.

Previous research has shown that children as young as 2 years are very sensitive to the preceding discourse in general, that is, they express elements that have not been mentioned yet, and which are thus new information in discourse, in an informative way, and elements that have been mentioned before, noninformatively (e.g., Allen, Reference Allen2000; Clancy, Reference Clancy, Sohn and Haig1997; Narasimhan, Budwig, & Murty, Reference Narasimhan, Budwig and Murty2005; Serratrice, Reference Serratrice2005). With respect to question answering, it has been shown that 2-year-old children's choice of referring expressions in answers to questions is strongly influenced by whether a particular referent was already mentioned in the question or not (Campbell, Brooks, & Tomasello, Reference Campbell, Brooks and Tomasello2000; Wittek & Tomasello, Reference Wittek and Tomasello2005). However, to our knowledge, the issue of whether the preceding discourse context prior to a question influences children's answers has only been examined in one study so far. Matthews, Lieven, Theakston, and Tomaello (Reference Matthews, Lieven, Theakston and Tomasello2006) investigated children's referring expressions in answers to questions depending on whether or not a certain character has been mentioned in the preceding discourse context prior to the question. Children watched video clips (e.g., a clown jumping) and were asked a sentence-focus question (“What happened?”). This type of question requires the answer to encode the whole event (Salomo, Lieven, & Tomasello, Reference Salomo, Lieven and Tomasello2009; cf. Lambrecht, Reference Lambrecht1994), such as “The clown is dancing.” Matthews et al. (Reference Matthews, Lieven, Theakston and Tomasello2006) concentrated on nominal expressions, and varied in two conditions whether or not the questioner mentioned the performing character (i.e., the clown) prior to the question. More precisely, in one condition the experimenter asked the child “That sounds like fun! What happened?”, whereas in the other condition the experimenter overheard the name of the character involved and remarked: “Was that the clown? Oh! What happened?” When the character has been mentioned in the preceding discourse context, it is not necessary to express it informatively (i.e., in a lexical noun) in the answer, but rather it can be expressed as a pronoun.

Matthews et al. (Reference Matthews, Lieven, Theakston and Tomasello2006) found that 3- and 4-year-old children replied with a lexical noun and a verb (e.g., “The clown is dancing”) when the experimenter had not yet mentioned the name, but they gave a pronoun and verb response (e.g., “He is dancing”) when the character had been named before. Even children as young as 2 years of age adapted their responses according to whether or not there had been prior mention of a noun by the questioner, specifically by varying their use of nouns as opposed to verbs. Thus, the referring expression used often appeared to be the outcome of deciding whether to comment on the character or the action that was performed: when there was no prior mention of the character, 2-year-olds were more likely to simply name it (e.g., “clown”), but when the character had been previously mentioned, children omitted that argument more often and reference to the character's action became more likely (e.g., “dancing”). Thus, although 2-year-old children showed sensitivity to the interlocutor's question with respect to the actor of the event, they appeared to not consider the interlocutor's knowledge state with respect to the character's action.

In the current study we examined children's expression of verb phrases (i.e., verb and patient) in their answers to questions depending on the preceding discourse context; that is, we focused not only on expressions referring to nominals, but we also analyzed children's reference to actions. One possible reason for the young children's performance in Matthews et al.'s (Reference Matthews, Lieven, Theakston and Tomasello2006) study might be that What happened? questions are difficult for young children to answer, mainly because it is not clear to the child what exactly is being asked for. There is also the possibility that in the absence of a clear question focus, nominals are easier to access than verbs, given the suggestion that English-speaking children represent the category of nouns earlier than that of verbs (e.g., Gentner, Reference Gentner and Kuczaj1982). In our study, therefore, we used a question type that has been shown to be easier for 2-year-old children to answer (Salomo et al., Reference Salomo, Lieven and Tomasello2009): a predicate-focus question (What's X doing?). Predicate-focus questions ask directly for an action, and if this action is a transitive one (as in the current study), the patient is also required for the answer to be fully informative.

There were two conditions in which the question we asked was identical, whereas the givenness/newness of the action and the patient were manipulated prior to the question. In one condition, the action was discourse-new, because children watched a series of video clips prior to the question in which the same participants (agent and patient) were involved in different activities successively (action-new condition). In the other condition, the patient was discourse-new, because an agent did the same action successively to different patients (patient-new condition). During the last scene children were asked “What's AGENT doing now?”

We expected children to express the patient in their answers in the same way as in Matthews et al.'s (Reference Matthews, Lieven, Theakston and Tomasello2006) study, that is, by using lexical nouns when the patient was the new element (patient-new condition), and by using pronouns or null references when the patient was given (action-new condition). However, because our question focused on an action, we, in addition, expected the children to include the verb in their answers in most cases. However, the verb might occur less often when it corresponds to the given element (patient-new condition) than when it is the new element (action-new condition). In Study 1 the manipulation of the newness/givenness of the elements was done visually (using video clips) and orally (with accompanying utterances). In Study 2 only the visual information was provided to compare the difference that the provision of a verbal discourse context for given information made to the children's expression of given and new information in their answers.

STUDY 1

Results

Our main interest was the children's realization of the action word and patient in their answers to the predicate-focus question. Therefore, in our analysis we focused on these two elements.Footnote ² An overview of the distribution of the children's response types can be seen from Table 1. For each of these types we conducted paired-samples t tests. These revealed that children made more action-only responses in the action-new condition, t (15) = 6.619, p < .001, more patient-only responses in the patient-new condition, t (15) = −2.764, p < .05, and more action + patient responses in the patient-new condition, t (15) = −3.503, p < .01. Thus, children used more multiword utterances in answering questions in the patient-new condition than in the action-new condition.

Table 1.

Mean percentage of use of different response types for the two conditions in Study 1

Adding together columns 1 and 3 as well as columns 2 and 3 in Table 1, we found that children expressed overall the action more often in the action-new condition than in the patient-new condition, t (15) = 2.764, p < .05, paired-samples t test, and the patient more often in the patient-new condition than in the action-new condition, t (15) = −6.619, p < .001, paired-samples t test.

In a further analysis we looked at children's patient expressions in more detail (see Figure 1). Paired-samples t tests were performed for each of the patient types (except pronoun responses, which were too infrequent). As noted above, null references occurred very frequently in the action-new condition (75.0%), whereas they were rare in the patient-new condition (9.4%), t (15) = 6.619, p < .001. However, lexical nouns were used significantly more often in the patient-new condition (87.5%) than in the action-new condition (21.9%), t (15) = −6.012, p < .001. Thus, children realized the patient as a lexical noun more often in the patient-new condition and used more null references for the patient in the action-new condition.

Figure 1.

The mean percentage of use of different referring expressions for both conditions in Study 1.

To summarize, when the action was the new element in the series of succeeding scenes, children were very likely to respond to the question with a one-word utterance, which was a single verb. However, when the patient was the new element, children used many more multiword utterances when answering the question: they expressed the patient typically as a lexical noun in their answers and in addition, they included the verb very often, even though less often than in the action-new condition.

STUDY 2

Results

An overview of the distribution of the children's response types can be seen in Table 2. For each of these types we conducted paired-samples t tests. Those revealed that, as in Study 1, children made significantly more action-only responses in the action-new condition, t (14) = 3.055, p < .01. Furthermore, a tendency for more patient-only responses in the patient-new condition was found, t (14) = −1.784, p = .096, as well as a tendency for more action + patient responses in the patient-new condition, t (14) = −1.974, p = .068.

Table 2.

Mean percentage of use of different response types for the two conditions in Study 2

Adding together columns 1 and 3 as well as columns 2 and 3 in Table 2, we found that, as in Study 1, children expressed the patient more often in the patient-new condition than in the action-new condition, t (14) = −3.055, p < .01, paired-samples t test). However, contrary to what we have found in Study 1, children expressed the action equally often in both conditions (paired-samples t test, ns).

In a further analysis we looked at children's patient expressions in more detail (see Figure 2). Paired-samples t tests were performed for each of the patient types (except pronoun responses, which were too infrequent). Null references occurred very frequently in the action-new condition (78.1%), whereas they were less frequent in the patient-new condition (46.9%), t (15) = 2.611, p < .05. Lexical nouns were used 53.1% in the patient-new condition, but only 18.8% in the action-new condition, t (15) = −2.905, p < .05. Thus, as in Study 1, children dropped the patient more often when it was the given element (action-new condition), and they expressed it more often as a lexical noun phrase when it was new (patient-new condition).

Figure 2.

The mean percentage of use of different referring expressions for both conditions in Study 2.

Comparison of Study 1 and Study 2. To analyze the effect of the absence versus presence of the verbal description of the videos to children's answers, we compared children's answers in Study 1 and Study 2 with respect to (a) response types, (b) overall expression of action and of patient, and (c) patient type.

For each the response types, a 2 (Study)×2 (Condition) analysis of variance (ANOVA) was performed with the mean proportion of response type as the dependent variable. Neither a main effect for study nor an interaction for study and condition was found for any of the three response types (action-only, patient-only, action + patient).Footnote ³ With respect to the overall expression of action and patient, a 2 (Study)×2 (Condition) ANOVA was performed for each of them with the mean proportion of actions and patients as the dependent measure, respectively. Again, neither a main effect for study nor an interaction for study and condition was found for either of them. Finally, for the realization of the patient, a 2 (Study)×2 (Condition) ANOVA was performed for each of the patient types (except for pronouns, which were too infrequent) with the mean proportion of patient type as the dependent variable. The ANOVA on patients that were expressed as lexical noun phrases revealed a marginally significant main effect for study, F (1, 30) = 3.518; p = .070, as well as a marginally significant interaction for study and condition, F (1, 30) = 3.769; p = .062. The ANOVA on patients that were null referents revealed a marginally significant main effect for study, F (1, 30) = 4.017; p = .054, as well as a significant interaction for study and condition, F (1, 30) = 4.892; p < .05.

Therefore, when the experimenter verbally described the context videos (Study 1), children in the patient-new condition expressed the patient more often in a lexical noun phrase than when the experimenter did not verbally describe the scenes (Study 2). That is, when the preceding context was only perceptually available, children dropped the new patient more often as opposed to in an additional discourse context.

GENERAL DISCUSSION

This paper investigated 2-year-old children's answers to predicate-focus questions depending on the preceding context in which we manipulated the givenness/newness of either the action or the patient of that action. Although previous studies have exclusively focused on children's noun phrases in question answering, this is the first study that looked at children's verb phrases in that context.

We found that when the action was the new and contrasting element in a series of succeeding scenes, children were very likely to respond to the question with a one-word utterance, which was a single verb. However, when the patient was the new element, children used many more multiword utterances when answering the question: they expressed the patient typically as a lexical noun in their answers, and in addition, they included the verb very often, although less often than when the action was the new element. Thus, our findings suggest that children are very sensitive to both the predicate-focus question and newness as expressed in successively contrasting events.

Our results contribute to and extend previous findings on children's ability to encode new and given information. There is some early evidence that children show sensitivity to the flow of information in discourse. For instance, in a series of studies, Greenfield and colleagues (Baker & Greenfield, Reference Baker and Greenfield1988; Greenfield & Smith, Reference Greenfield and Smith1976; Greenfield & Zukow, Reference Greenfield, Zukow and Nelson1978) have shown that English-speaking children at the one-word stage verbalize the most informative element of the discourse situation. The verbalized element is usually the most variable aspect of the event and the one that reduces the uncertainty in the situation. Constant and redundant elements are those that are taken for granted and thus are left unexpressed. More recent studies with slightly older children have found that children pay attention to discourse pragmatics in choosing whether to represent an argument as a lexical noun or simply drop it (e.g., Allen, Reference Allen2000; Clancy, Reference Clancy, Sohn and Haig1997; Narasimhan et al., Reference Narasimhan, Budwig and Murty2005; Serratrice, Reference Serratrice2005). They showed that children produce lexical nouns more often when the argument they wish to represent is informative than when it is not. Indeed, this is our finding for the expression of the patient in the current study as well, that is to say, the patient was expressed as a full noun when it was new, but was simply omitted when it was given.

However, the case of question answering is in some sense different from that of other discourse contexts. A question sets up certain information expectations that must be realized in the information structure of the answer. That is to say, if the speaker is asking a predicate-focus question (“What is X doing?”), the addressee is expected to answer with a verb phrase, no matter whether the action is new or given information. Therefore, there are two factors that are relevant when answering questions in a given-new framework: (a) the focus of the question, and (b) newness in context. If children were only sensitive to the question, no difference between the two conditions would have been found. If they were sensitive to newness alone, the answers would have consisted of the new element only. However, because we found that children's answers included the new element as well as the verb in most cases, we can conclude that children at this early age are sensitive to both the question and newness, at least in a situation where the question is predicate focused and the contrasts in successive scenes make what is new and what given relatively accessible. The further finding, that even though children included a verb most often in their answers but did so more often in the action-new than in the patient-new condition, possibly results from the fact that the verb was the new element in the action-new condition while it was the given element in the patient-new condition.

Although previous studies have found that young children choose appropriate referring expressions for noun phrases when answering questions (Campbell et al., Reference Campbell, Brooks and Tomasello2000; Matthews et al., Reference Matthews, Lieven, Theakston and Tomasello2006; Wittek & Tomasello, Reference Wittek and Tomasello2005), we can now extend that finding to verb phrases as well. A further interesting finding was that children expressed new patients more often when there is a verbal discourse context prior to the questions (Study 1) than when there is not (Study 2). Because language directs our attention, it is possible that without any speech in the preceding context it is more difficult for young children to decide what is needed in an appropriate answer. Note that the verbal discourse provided the child with salient contrastive cues as to what is new.

One might further argue that children were primed by the verbal descriptions of the experimenter in Study 1 in which case they would be expected to produce more verb + patient answers. However, no difference was found in terms of multiword utterances (i.e., verb + patient answers) across studies. Furthermore, even with verbal descriptions (Study 1), children overwhelmingly produced one-word utterances in the action-new condition. This contradicts the assumption that the discourse context primed the children. Therefore, we can rule out the possibility of priming effects. An alternative explanation might be that in describing the context scenes, the experimenter naturally highlights the new information in the second context scene, which might make the new element more salient to the children.

However, there are limitations in our assumptions concerning children's production of verb phrases as answers to questions. We still do not know whether children would readily encode the action of an event without a question that focuses as strongly on a verb answer as does the predicate-focus question we used. It would be interesting to find how children reacted to a more neutral question, such as the one used by Matthews et al. (Reference Matthews, Lieven, Theakston and Tomasello2006): What happened? Furthermore, because we used a different design than that of previous studies, we do not know how children would answer predicate-focus questions if the interlocutor was truly naive about the event about which the child is supposed to give information; that is, the interlocutor did not watch the film with the child. Therefore, further studies might apply a design in which something that is given for the child is new to the interlocutor. Finally, in Study 1, all the cues for the child were going in the same direction: visual, discourse, and contrast between successive scenes. In real life, this is unlikely to be the case and, as we see from Study 2, removal of the discourse context, seems to make it somewhat less clear to the child what the answer requires. A further study would be to repeat the study with scenes that are not so narrowly contrasted and see whether children perform equally well or whether they become more successful with development.

One strong suggestion emerging from the current study is that children at the age of 28 months are able to express verb phrases as answers to predicate-focus questions at least in this tightly contrastive situation. They seem to know that an answer referring to the action is required to that kind of question, no matter whether the action that is encoded by the verb is new or given. Furthermore, our results suggest that language supports children in establishing what is given and new information.