Abstract
Children’s ability to ascribe beliefs to themselves and others has been shown to develop in the late preschool and early school years. This ability to represent, that is to think about, beliefs known to be false is described as metarepresentational development. This article extends these findings to the domain of linguistic representations by exploring how children come to judge statements as true or false of a given reality. It presents two models of judgment, one based on the pragmatics of the expression, the other on the basis of the semantics of the expression. Two studies show that when preschoolers object to what someone has said, they are disagreeing with the speaker, a pragmatic judgment, rather than judging the truth or falsity of what the speaker has said, a semantic judgment. True/false judgments are metalinguistic judgments, later acquired, suggesting a new consciousness of linguistic form.
Keywords
Introduction
Our interest in young children’s judgments of truth or falsity was triggered by the interesting but puzzling finding by Alison Dewsbury (unpublished) that appeared to indicate an inability to judge statements containing negatives as true or false. Kindergarten and first grade children (5- and 6-year-olds) were presented with affirmative and negative statements that were either true or false of an accompanying picture. For example, a picture of a man with a hat was accompanied by either the statement ‘The man has a hat’ or ‘The man has no hat.’ Children were to indicate their judgment by a correct mark (√) for true and an (X) for not true. Surprisingly they marked both with the sign indicating correct. Symmetrically, when presented with a picture of a man with no hat and asked the same questions, they marked both with an X, indicating not correct as follows: Picture: Man with a hat Man with no hat The man has a hat √ The man has a hat X The man has no hat √ The man has no hat X
From a logical point of view, they failed half the items; they had particular difficulty with true negatives. However, we noted that the pattern would be explicable if we assumed that the children were agreeing or disagreeing with the speaker’s affirmative and negative assertions rather than doing what the task had required, namely, judging truth or falsity. Thus, in the first case when the man had a hat, they agreed with the speaker’s affirmative claim that he had a hat by responding ‘Yes he does (have a hat)’ and after the claim that he had no hat by disagreeing with the negative statement and signaling their disagreement by insisting ‘Yes he does (have a hat),’ hence the check. (Note the difference in stress on the word ‘Yes’ in the two cases, the first soto voce, the second stressed). Bolinger (1977, p. 38) recorded just such a case when a listener ignored the embedding form (Do you suppose he’s coming?) and responded directly to the content (Yes, he is). Olson and Hildyard (1981) reported many similar ‘errors’ in young children’s responses to questions. Perhaps agreeing or disagreeing with the speaker’s assertions is a simpler matter than judging the truth or falsity of the content asserted and further implying that true/false judgments are in some way special.
Children come to understand these facts about false beliefs when they are 4 or 5 years of age (Gopnik & Astington, 1988; Perner, Leekam, & Wimmer, 1987), an understanding associated with advances in their understanding of the pragmatics of language (Fernandez, 2013). Furthermore, children’s understanding of mental representations appears to parallel their understanding of some other forms of representation such as pictures (Zaitchik, 1990). So, how are statements different from beliefs and pictures? Indeed, one well-known position is that metalinguistic and metacognitive representations are merely two names for the same thing (Doherty & Perner, 1998). To understand the peculiarities of linguistic representation we distinguish two uses of language, one pragmatic, concerned with communication, the other more narrowly semantic, concerned with truth of the expression.
Judging agreement
‘What is truth? said jesting Pilate, and would not stay for an answer,’ wrote Francis Bacon in Of Truth (1601). For centuries, philosophers have disagreed on the answer to Pilate’s question. Our common-sense understanding of truth is perhaps best captured by correspondence theories, which argue that truth is conformity to reality. However, Tarski (1944) defined truth in terms of conformity to language: ‘A true sentence is one which says that the state of affairs is so and so, and the state of affairs is indeed so and so.’ More recently, Goodman has written that ‘Truth pertains solely to what is said, and literal truth solely to what is said literally’ (Goodman, 1978, p. 18). That is, truth pertains to statements in relation to events not to the events themselves.
But what is a statement? Olson (1977) distinguished oral utterances from written textual statements on the basis of the presence or absence of indicators of how a text was to be taken. Written statements, he argued, tend to lose their tie to context and to the writer’s intention. Harris (2009) called such written statements ‘unsponsored’ that is detached from the speaker’s personal intention and Gellner (1988) took such impersonal statements as definitional of modern forms of thought. Frege (in Geach & Black, 1960) showed how this was possible by claiming that a simple declarative sentence is actually composed of two parts, a ‘force’ and a ‘predicate.’ Speech act theory (Austin, 1962; Searle, 1969) makes the same distinction, calling one the ‘illocutionary force’ the other ‘propositional content.’ The propositional content specifies the relation between symbol and referent, allowing judgment of truth or falsity, a semantic judgment, while the illocutionary force specifies how the expression is to be taken by the person to whom it is addressed – as a suggestion, a command, a claim, and so on, a pragmatic judgment. Although all expressions have both force and content, writing, like quoted speech, tends to preserve the content divorced from the force (Olson & Oatley, in press). The content in the form of a statement is a unique candidate for truth judgments, semantics uncluttered by pragmatics. But if, as Donaldson (1978), French and Nelson (1985), and others have shown, children’s language use is primarily contextual and action-oriented, pragmatic force and semantic content may not be so easily distinguished by young children.
We thus distinguish statements as described above from assertions. Ordinary assertions, as mentioned, combine force and content and children may reject inappropriate assertions by agreeing or disagreeing with the speaker. Written or quoted statements such as ‘The man has no hat’ leave the force unstated and are normally seen as calling for a judgment of truth or falsity of the statement per se. Certainly that is the theorist’s intention. But if children treat the expression as an assertion by some imagined speaker they may respond by agreeing or disagreeing with the assertion of an imagined speaker rather than by judging his or her statement as true or false.
Ramsey (1927/1990) argued that to say that ‘p’ and ‘It is true that p’ is to say the same thing (see also Misak, 2013, p. 206). Although it is understood that any assertion is believed to be true by a speaker, to explicitly say, ‘It is true that …’ changes the focus of the sentence. Rather than making a claim about the world (i.e., p), a claim is made about the statement expressing the proposition p (i.e., that p is true). Therefore, judging a statement as true or false is a metalinguistic judgment, a judgment about the language in which a proposition is expressed.
There has been a great deal of interest in children’s understanding of representations, and in particular of such mental representations as beliefs (see Astington & Baird, 2005). On the one hand, children are able to reject false assertions at about 2 years of age, that is, almost as soon as they begin to talk. At this age, children are reported to deny obvious falsehoods. When shown an apple and told, ‘This is a biscuit,’ such children reply, ‘No, apple’ (Pea, 1982). Clearly the assertion is false, and clearly the child has rejected it. We might infer that the child has rejected it because it is judged to be false, but we must remember that is our inference. They may have rejected it for pragmatic reasons, namely, they do not agree with what the speaker said.
When children are explicitly asked to judge statements as true or false the pattern is much more complex. Both rejection and denial of statements may be expressed by the particle ‘no’ and this 2-year-olds appear to have mastered. Truth and falsity are not equivalent to affirmation and negation in that negative statements can be true. Thus, ‘This is not an apple,’ is true of an orange. So, while young children can affirm and negate, this is not grounds for thinking that they have concepts of truth and falsity, even if they may be said to use the truth of an assertion in accepting or rejecting it.
This somewhat paradoxical claim (i.e., that young children are able to affirm or deny assertions but not able to judge the truth or falsity of statements) requires some additional comment. Statements express propositions whereas beliefs are attitudes to propositions. The standard psycholinguistic model places the truth and falsity of statements at the core of thought and knowledge (Fodor, 1975; Foss & Hakes, 1978). Pragmatic theory on the other hand takes the speech act as basic while subordinating the propositional content to the pragmatic function. Divorcing the pragmatic from the semantic yields linguistic objects referred to above as ‘statements.’ Statements may be judged as true or false. However, and this is the claim of this article, for young children language is first and foremost pragmatic, the attempt by a speaker to assure agreement, whether as assent or compliance, from a listener (Olson, 1980). Responses to speech acts consist of agreements or disagreements with the speaker. Only later and under special conditions would the proposition be extracted as a statement amenable to judgments of truth or falsity. This is not to suggest that children’s understanding involves scant attention to the language or to the way the world is. On hearing an utterance ‘The grass is green,’ the child may judge of the grass that it is green without judging that the proposition ‘The grass is green’ is true. The latter, we have said, is a metalinguistic judgment, a judgment about a statement rather than a judgment about the world. And as is well known primarily through research on children’s theory of mind, other metarepresentational judgments, perhaps because of their complex grammatical form, are difficult for young children (Astington & Baird, 2005).
Judging truth
Studies of adults’ judgments of the truth or falsity of affirmative and negative sentences were the focus of sentence verification paradigms that were popular some three decades ago. Wason (1959, 1980) showed that while it was a simple matter to judge affirmative sentences as true or false, negatives presented more difficulty; a regular ordering for difficulty was found: least difficult true affirmative (TA) ⇓ false affirmative (FA) ⇓ false negative (FN) most difficult true negative (TN)
Wason (1965) later showed that true negative sentences are readily judged as true if they are used to mark exceptionality. If every other animal in a picture is a cat, it is plausible to say of the only dog that it is not a cat and is easily processed by the listener. Indeed, even 2-year-olds can understand true negation in a context of plausible denial (De Villiers & Flusberg, 1975). Just how pragmatic factors, both factors of context of alternatives and factors of speech act type (statements vs. assertions as described above), affect judgment requires further examination. Syntactic models of truth/false judgments of sentences by adults (Carpenter & Just, 1975; Clark & Chase, 1972; Trabasso, Rollins, & Shaughnessy, 1971) and by children (Akiyama & Gillory, 1983; Kim, Shatz, & Akiyama, 1990) indicated that judging a negative statement as true is difficult because it requires the comparison of additional constituents. However, young children make more errors on such items than could be expected on the basis of chance, as seen in Study 1 reported below. Speech act theory brings into prominence the question of just what listeners assume the task to require, specifically what is involved in agreeing with a speaker as opposed to judging statements as true or false.
Our hypothesis is that true/false judgments are peculiar to the processing of disembodied statements, that is, as autonomous texts divorced from a speaker, typical of written texts and requiring relatively sophisticated metalinguistic skills (Olson, 1977, 2012). Young children, we suggest, will focus on the social pragmatic function of an expression. If so they will take it as an assertion by a speaker and their ‘yes’ and ‘no’ will indicate either agreement or disagreement with what the speaker said rather than a judgment of truth or falsity of the statement.
We report two studies. In the first children are given the task of judging the truth or falsity of statements that express propositions about pictorial displays. In an attempt to specify the interpretation required and yet to allow the children to make a simple Yes/No judgment, the statement was embedded in the construction ‘Does the picture show that … .’ To say ‘Yes’ to such a statement is taken as asserting the truth of the statement.
Study 1
Method
Participants were 48 children from a public school and a day-care in a mixed-income neighborhood of Toronto. Sixteen children were tested at each of three age levels: mean ages: 4;6, 6;7, and 8;7. There were approximately equal numbers of boys and girls in each age group. Children were tested individually.
Task 1 (judging)
There were eight cards with a simple picture on each one. The child was shown a card with, for example, a picture of a green apple on it, and was asked, ‘Does the picture show that … [for example] … The apple is red?’ The child had to answer ‘Yes’ or ‘No’. Four types of complement sentence were used: true affirmative (TA), false affirmative (FA), true negative (TN), and false negative (FN). A pretest determined that children knew the items and color terms used in the test sentences.
Task 2 (matching)
There were 16 cards with two simple pictures on each. Children were shown a card, they heard an affirmative or negative sentence, and had to point to the picture that matched the sentence. For example, the child was shown a card with pictures of a red and a green apple on it, and was asked, ‘Which picture shows … [for example] … The apple is not green?’
The order of tasks and of items within each task was balanced over participants at each age level.
Results
Task 1 (judging)
Children at all age levels responded almost completely correctly to true and false affirmative sentences, saying ‘Yes’ and ‘No’ appropriately. However, only the 8-year-olds responded at above chance levels to true negative sentences (z = 5.82, p < .05) and false negative sentences (z = 4.62, p < .05) (see Table 1) saying ‘Yes’ to true negatives and ‘No’ to false negatives.
Percentage of correct responses in Study 1.
Task 2 (matching)
Children of all ages were almost entirely correct in their picture choices (see Table 1).
Discussion
The younger children’s correct responses to the affirmative sentences in Task 1 show that they understood the task, and their performance in Task 2 shows that they comprehended the negative sentences. Why, then, do they make incorrect judgments for the negative sentences in Task 1? Why do they have such difficulty accepting true negatives as true? They have no difficulty in choosing the picture corresponding to the negative sentences. Hence, it is implausible to suggest that they just do not understand true negative sentences. The difficulty is in judging them as true, that is, in talking about the sentence rather than talking about the speaker’s assertions. In line with our hypothesis, the younger children in Task 1 may be treating the statements as assertions and agreeing or disagreeing with them. That is what their Yes/No responses marked. In English, agreement is marked by using the same affirmative/negative particle as the speaker, and disagreement by using the opposite one. Thus, to agree with a negative sentence we say ‘No’ and to disagree by insisting ‘Yes’: for example, A says, ‘It’s not raining.’ B, who agrees says, ‘No it’s not.’ C, who has just come in from outside where it has started to rain, disagrees and says, ‘Yes it is.’ This is how the younger children use Yes/No in Task 1 (see Table 2), whereas the older children were making logical, truth-functional judgments (see Table 2). Put another way, the younger children were making a pragmatic judgment, the older a semantic metalinguistic one.
Comparison of truth-functional and agreement/disagreement responses.
‘Yes it does show that the apple is red.’
‘Yes it is red.’
The fact that children had little difficulty with Task 2 reflects the fact that there was no opportunity to agree or disagree or to judge truth or falsity. Matching indicates only what the speaker would say when faced with any display.
Study 2
If our hypothesis is correct, then young children may be expected to respond correctly to negative statements if the criterion chosen is that of appropriate pragmatic force rather than the semantic judgment of truth/falsity. This was tested in Study 2. The task was embedded in a context that was meaningful to the child (Donaldson, 1978), and more importantly, the emphasis shifted from judging the truth of a sentence to judging the correct performance of a speaker. In this way, the child was asked to make judgments about what a speaker would say about the world, and not about the truth of the statement about the world. In addition, the task asked for a behavioral, rather than a verbal Yes/No response. The method is also appropriate for controlling for a possible affirmative response bias (Mehrani, 2011) as it is no longer the child’s assertions that are being judged.
Method
Participants were 48 children from day-cares and nursery schools in a variety of neighborhoods in Toronto. There were 24 4-year-olds (7 boys and 17 girls, mean age 4;6) and 24 5-year-olds (11 boys and 13 girls, mean age 5;5). Children were tested individually.
Task 1 (judging)
The children were introduced to a puppet, a monkey, and were told that he was learning to talk but he sometimes made mistakes. The children’s task was to help him. The monkey told them about some pictures, but because he was shy, he whispered to the experimenter who passed the statement on to the child. Task materials and sentences were similar to those used in Study 1: eight cards with a simple picture on each, and eight test sentences, two true affirmative (TA), two false affirmative (FA), two true negative (TN), and two false negative (FN). For each one the child was shown a picture (e.g., a red apple) and was told, ‘Monkey says … [for example] … The apple is not green’ (TN). If the children judged that the monkey had said it right they gave him a sticker, and if they judged that he had said it wrong, they told him, ‘No sticker, monkey.’ As in Study 1, a pretest determined that children knew the items and color terms used in the test sentences, and that they understood how they were to respond to the monkey.
Task 2 (matching)
As in Study 1, children were shown pairs of pictures and asked to point to the one that matched a given affirmative or negative sentence.
The order of tasks and of items within each task was balanced over participants at each age level.
Results
Children understood and enjoyed the task, and responded at above chance levels to all sentence types for both Task 1 (judging) and Task 2 (matching) (see Table 3). There is no indication of a bias towards sentences containing an affirmative.
Percentage of correct responses in Study 2.
Discussion
The children’s ability to judge the monkey’s statements and to match the sentences to the pictures clearly indicates that the children are not misunderstanding true negative sentences. The children seem able to determine if assertions accord with what a speaker should say to a listener, a pragmatic judgment. As Study 1 indicated, however, children have difficulty with reflecting on the semantics of the representations themselves, a metarepresentational task. Research on children’s theory of mind (e.g., Perner et al., 1987), however, indicates that children begin to develop some metarepresentational abilities around the age of 4 or 5 years. Hence, the difficulty seems not to be metarepresentational judgements in general but rather a failure to understand that even in a pragmatic, communicative context it is possible to make a metalinguistic semantic true/false judgment.
General discussion
Pragmatic theories of language hold that speech acts consist of an illocutionary force and a propositional content. Ordinarily, the content is embedded within the force of an expression to make up, for example, an assertion. Children, we have argued, make judgments by agreeing or disagreeing with the assertions of speakers. Truth/falsity judgments on the other hand require that the child abstract the semantic content of the assertion, setting aside its illocutionary force, treating it as what we have called a statement. Even when this semantic content is made explicit by expressions such as ‘Is it true that … ?’ younger children disregard that embedding and continue to treat the expression as an assertion by a speaker with whom they then agree or disagree. They lack, or at least fail to apply, the metalinguistic concept of truth.
Children in Study 1, Task 1 (judging) understand true negative assertions and accept or reject them appropriately. What appeared as errors on the true negatives are in fact correct ways of marking agreement or disagreement with the speaker’s assertion. This is seen in Study 2, where the younger children appropriately awarded a sticker to mark acceptance of a true assertion, and withheld a sticker to mark rejection of a false one (see Table 2). They lack only the ability to treat the assertions as statements, that is, as verbal objects that may be judged as true or false. Judging truth, as Goodman (1978) said, depends upon knowing something about literal meanings.
It could be argued that both rejecting assertions and denying truth are equally metalinguistic in that the sentences in fact are true or false. However, when children agree or disagree they are commenting on pictured world, whereas when they judge truth or falsity they are commenting on the statement. In the former case, they say ‘Yes, he does,’ in the latter ‘‘Yes, it is true.’ The difference is captured in the pronoun, ‘he’ referring to the man with the hat, ‘it’ referring to the sentence. Hence, only the latter is a comment on the expression and so is metalinguistic.
In both agreeing/disagreeing with a speaker’s utterance and in judging truth/falsity children are recognizing when the claims are appropriate or inappropriate. They differ only in how the claim is attributed, to the person or to the statement. The detection of the anomaly is more primitive than the linguistic expressions young children use to mark that recognition. It is only the latter that we capture in experiments requiring verbal judgments and verbal responses from children. This kind of argument has recently been raised in studies of children’s theory of mind (He, Bolz, & Baillargeon, 2012; San Juan & Astington, 2012). Apparently, it is easier to detect a match or mismatch with ongoing expectations about a person’s actions than to make an actual judgment of truth or falsity of the beliefs of persons or the truth of statements.
If these findings hold up, they have important implications for the study of language in general and children’s language in particular. The standard paradigm for language processing takes statements rather than assertions as the basic unit of language. This study suggests that statements amenable to judgments of truth and falsity are not basic units of ordinary language but rather peculiar structures, associated primarily with quoted speech and with writing, in which content has been arbitrarily cut off from the pragmatic illocutionary force of the utterances in which it is normally embedded. As Brandom (2011, p. 23) has argued, ‘semantics must answer to pragmatics.’
In sum, the present study suggests, first, that concepts of truth and falsity are metalinguistic concepts that children acquire in the late preschool and early school years, presumably as they learn to deal with written language and learn to detach the truth of an expression from its assertion by a speaker, an important step in linguistic and metalinguistic development. Second, it is clear that even the youngest children recognize the fact that the content expressed is true or false. That is the basis of their agreement or disagreement. Yet to judge true or false requires concepts of true and false and that is what appears to be missing. They are missing, we have argued, because true and false are higher order metalinguistic concepts that apply to linguistic objects, namely statements, abstracted from the pragmatic assertions of speakers. This is an important step in the developing consciousness of language.
Footnotes
Acknowledgements
We are grateful Elizabeth Lee for her help with data collection, and to the Natural Sciences and Engineering Research Council of Canada for financial support to JWA.