Children's use of referring expressions in spontaneous discourse: Implications for theory of mind development

Available online at www.sciencedirect.com

Journal of Pragmatics 56 (2013) 43--57 www.elsevier.com/locate/pragma

Children’s use of referring expressions in spontaneous discourse: Implications for theory of mind development Jeanette K. Gundel *, Kaitlin Johnson University of Minnesota, United States Received 2 February 2011; received in revised form 5 April 2013; accepted 13 April 2013

Abstract This paper refines and extends the empirical base of previous work on children’s use of referring expressions in spontaneous discourse within the Givenness Hierarchy framework, and further develops implications of this work for our understanding of the development of a theory of mind. The study supports earlier findings that children use definite and indefinite articles, demonstrative determiners, and demonstrative and personal pronouns appropriately by age 3 or earlier. It also provides further support for two stages in mind-reading ability. The first, implicit and non-representational, includes the ability to assess memory and attention states such as familiarity and attention; the second, representational and more conscious, includes the ability to assess propositional, epistemic states such as knowledge and belief. Distinguishing the two stages provides a possible explanation for why children learn to use forms in ways consistent with their encoded information about how to mentally access an intended referent before they fully exhibit the metarepresentational ability to calculate pragmatic inferences, such as that involved in assessing how much information is relevant for the addressee. © 2013 Elsevier B.V. All rights reserved. Keywords: Reference; Cognitive status; Theory of mind; Procedural meaning; Implicature

1. Introduction A characteristic feature of human linguistic communication is that while linguistic form constrains possible interpretations, it grossly underdetermines a speaker’s intended meaning in a given context of use. The same sentence can have different possible interpretations, but general pragmatic principles that guide language production and interpretation typically allow us to determine which of the possible interpretations is the one the speaker intended. This feature is rooted in the interactive/discursive (as distinct from purely representational) function of language. The absence of a simple one to one mapping between linguistic form and a speaker’s intended meaning is perhaps most evident in the case of nominal expressions. This is obvious for pronouns (e.g. it, that), which encode little if any conceptual content or descriptive information; but it is also true for full descriptions such as the speakers at this conference, the woman in the blue dress, and so on. The same (pro)nominal expression can be used to refer to many different things and different things can be referred to with the same expression. Thus, children acquiring their language learn, among many other things, that an entity, a particular cat, for example, can be referred to as a cat, that cat, the cat I saw chasing a bird, it, and so on. The child must also learn that the different possibilities change depending on what can be assumed about the knowledge, memory and attention state of the interlocutor at a given point in the discourse. The aim of the present work is to extend the empirical base of previous corpus analytic investigations of young children’s use of

* Corresponding author. Tel.: +1 612 624 7564. E-mail address: [email protected] (J.K. Gundel). 0378-2166/$ -- see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.pragma.2013.04.003

44

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

different types of nominal expressions conducted within the Givenness Hierarchy framework (Gundel et al., 2007; Gundel, 2009, 2011), and to further develop and refine a proposal about how children’s use of such expressions can help inform our understanding of their ability to assess the mental states of others. Section 2 introduces the theoretical framework assumed in this research, the Givenness Hierarchy (Gundel et al., 1993 and subsequent work). In section 3 we present results of our analysis of the use of referring expressions in spontaneous conversations of English-speaking children aged 1;9--2;8, and in section 4 we discuss implications of this research for our understanding of children’s development of a theory of mind. 2. The givenness hierarchy Gundel et al. (1993) propose that determiners (e.g. the, this, that) and pronouns (e.g. it, this, that) provide the addressee with information about how to mentally access the referent, by encoding, as part of their conventional meaning, the cognitive (memory and attention) status of the intended referent for the addressee at the point just before the form is encountered. The cognitive statuses comprise the following ‘Givenness Hierarchy’, where, for purpose of illustration, English forms are listed below the cognitive statuses they directly encode. (1) Givenness Hierarchy (GH) and Associated English Forms in focus > activated > familiar it1 this/that/this N that N

uniquely > identifiable > referential > the N indefinite this N

type identifiable aN

GH statuses are (assumed) mental states, not linguistic entities; the linguistic forms that encode these statuses provide procedural information about how to access the referent, as described in (2) below. (2) It this/that/thisN that N the N Indefinite this N aN

associate associate associate associate associate associate

representation in focus of attention representation in working memory representation in memory unique representation with determiner phrase2 unique representation type representation

This makes it possible to explain facts like those exemplified in (3) and (4). (3)

A1. B. A2.

(4)

A. B. B’

Do you have a pet, Randy? Yeah. Currently we have a poodle. Really? I read somewhere that the poodle is one of the most intelligent dogs around. (adapted from the Switchboard corpus, Godfrey et al., 1992) Sam snores. That’s rude It’s rude.

In (3), the phrase the poodle in A2 can be interpreted as referring to Randy’s poodle, introduced in B. However, it can also be interpreted generically to refer to the entire class of poodles. Both interpretations are licit here, since the article ‘the’ simply encodes the information that the addressee is to associate a unique representation, and either the recently mentioned poodle or the whole (unique) class of poodles would satisfy this requirement. In the former case, B could easily associate a unique representation by selecting the recently activated representation of his poodle. In the latter case, B could associate a familiar representation of the class of poodles or simply construct such a representation based on the information encoded by ‘poodle’. Given the context, A probably intended the generic interpretation, and this is how B (and

1

The pronoun it here stands for all personal pronouns. The term ‘determiner phrase’ (DP) is used here to refer to the whole referring phrase, either a pronoun (a determiner with no NP complement) or a determiner followed by a full noun phrase complement, here represented as ‘N’. 2

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

45

also the readers of this example) are most likely to interpret it; but both interpretations are possible, given the cognitive status information encoded by ‘the’. In (4), the pronoun that in B has two possible interpretations. It can refer to Sam’s snoring, which is directly inferable from, and thereby activated by, A’s statement; it can also refer to A’s speech act of saying that Sam snores, which would meet the condition of being activated by virtue of just having been uttered. The pronoun it in B’, by contrast, can only have the former interpretation, since ‘it’, unlike ‘that’, requires the referent to be not only activated, but in the addressee’s focus of attention. This is so given the independently motivated assumption that when someone says something, the utterance act is activated, but not brought into focus. The addressee’s focus of attention will be on a prominent part of the semantic content of the utterance. It is important to note that determiners and pronouns encode the cognitive (i.e. mental) status a referent has, not how it came to have that status. Thus, in (5), the pronoun it can be used to refer to the house, which is assumed to be in focus because it was mentioned in a syntactically prominent position in the previous sentence (5a) or because the addressee is looking at it (5b). (5)

a. b.

The house is huge. It has 3 bedrooms. (Speaker sees addressee looking at a house) It has 3 bedrooms.

Similarly, in (6), the pronoun that can be used to refer to the event of the addressee’s tooth being pulled, which can be assumed to be activated by virtue of being mentioned in the previous utterance (6a), or the speaker just having performed the act of pulling the tooth (6b). (6)

a. b.

A. I had my tooth pulled. B. Did that hurt? (dentist to patient after pulling his tooth) Did that hurt? (Jackendoff, 2002)

Statuses on the GH are in a unidirectional entailment (by their definition); they are not mutually exclusive, a feature which distinguishes the GH from other referential hierarchies, such as those proposed by Givo´n (1983), Prince (1981), and Ariel (1990).3 Thus, anything in focus is necessarily also activated (in awareness/working memory); anything activated is necessarily also familiar (in memory); anything familiar is also uniquely identifiable (the addressee can associate a unique representation); anything uniquely identifiable is also referential (the addressee can associate a referent by the time the sentence is processed); and anything referential is also type identifiable (there is a type to which the referent belongs). Forms are thus underspecified for cognitive statuses higher than the ones they explicitly encode; they do not exclude them. This one-way entailment has some important consequences. First, it results in a one to many mapping between cognitive statuses and different forms. For example, a form such as the English definite article ‘the’, which explicitly signals that the referent is uniquely identifiable, can be used to refer to something that is not only uniquely identifiable, but also activated or even in focus, as illustrated in the example in (3) above, where the whole class of dogs is activated by the phrase ‘a dog’ in the previous sentence. Second, unidirectional entailment of statuses on the GH gives rise to pragmatic inferences, specifically scalar implicatures (Horn, 1972), where use of a weaker (entailed) form implicates negation of the meaning associated with the stronger (entailing) form. For example, just as ‘some’ gives rise to the implicature ‘not all’, the indefinite article ‘a’ gives rise to the implicature ‘not uniquely identifiable (and therefore also not familiar and so on). Thus, the second occurrence of a student in (7) would normally be interpreted as referring to a different student than the one referred to by the first occurrence, which would be uniquely identifiable at that point. (7)

A student came by at 2; a student also came by at 5.

Similarly, the pronoun ‘that’, which explicitly signals only activation, often gives rise to the implicature ‘not in focus’, as in (8), where that is more easily interpreted as referring to the den than to the living room. Note that it on the other hand would be interpreted as referring to the living room, which is brought into focus in the first sentence, where it is mentioned first and sets the perspective for the rest of the sentence. (8)

Across from the living room is the den. And next to that is a small bedroom. (cf. and next to it.)

3 The Givenness Hierarchy is also distinct from the accessibility hierarchy of Ariel (1990) in that forms encode manner of accessibility of the referent, not degree of accessibility. For further discussion of this point, see Gundel (2010) and Gundel et al. (2012).

46

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

The claim that these interpretations are implicatures rather than necessary inferences that arise from conventional meanings of the forms in question is supported by the fact that, like other implicatures, they arise only when the extra information that would be conveyed by the strong (entailing) form is relevant. Thus, as Grice (1975) points out, a car in (9) is not necessarily interpreted as a car the addressee cannot uniquely identify. Since it is the property of being a car and not the identity of the car in question that is relevant, use of the indefinite article does not implicate that the referent is not uniquely identifiable to the addressee. (9)

I’m not going along; I’ve been sitting in a car all day.

Similarly, the pronoun that in (10) is interpreted as referring to the in focus kitchen. Since the distinction between being in focus or merely activated is irrelevant here, as there are no other competing activated entities in this context, use of that does not give rise to the implicature that the referent is not in focus. (10) The kitchen is to your left, and next to that/it is the den. In sum, within the GH framework, referring expressions explicitly encode two kinds of information: descriptive information about the conceptual content of the referent (e.g. that it is a cup or a blue hat), and procedural information about how the referent is to be accessed in the mind of the addressee (in focus, activated, familiar, uniquely identifiable, referential, type identifiable). The latter is encoded by various determiners/pronouns as part of their conventional meaning, thus allowing the addressee to further restrict possible interpretations consistent with the conceptual content encoded by the expression. In addition, pragmatic inferences (specifically scalar implicatures) that result from the unidirectional entailment of statuses on the GH further restrict interpretations, when relevant, in contexts when necessary conditions for use of more than one form are met. To use a form appropriately, a speaker must therefore assess what cognitive status the intended referent has (or can be expected to have) for the addressee and also how much information about cognitive status is relevant. 3. Development of reference production and understanding by children Given assumptions of the Givenness Hierarchy framework outlined in the previous section, children need to understand several things in order to appropriately produce and accurately understand referring expressions in their language. They must know the linguistic meanings of these forms, the conceptual content as well as the cognitive status information encoded by different pronouns and determiners. For example, children acquiring English must learn that the determiner ‘that’ in ‘that dog’ encodes the procedural meaning that the referent is to be accessed from among the entities already in memory (familiar) and that the pronoun ‘it’ encodes the information that the referent is already in the addressee’s focus of attention. In addition to this linguistic knowledge, children must be able to assess if something has a particular cognitive status, for example is the addressee’s attention focused on it, just as they must be able to assess whether something is a dog or not. They must also assess how much information about cognitive status (or conceptual content) is necessary and relevant, i.e. when to use the most informative form and when a weaker form is sufficient or would result in an implicature that the information encoded by the stronger form does not apply. These non-linguistic abilities involve being able to assess an interlocutor’s mental state in relation to the object referred to and understanding that this may be different from their own. That is, it involves, in some sense, what has been called ‘theory of mind’ (Premack and Woodruff, 1978; Baron-Cohen, 1995). Naturalistic studies of spontaneous conversational discourse by children acquiring various languages have shown that children use the full range of referring forms (definite and indefinite articles, demonstratives, personal pronouns), appropriately, by age 3 or earlier (e.g. Bennett-Kastor, 1983; Schnell de Acedo, 1994; Gundel and Page, 1998; Gundel et al., 1999; Bittner, 2002; Gundel et al., 2007; Salazar-Orvig et al., 2010, inter alia). Similar conclusions are suggested by recent experimental work (e.g. O’Neill, 1996; Wittek and Tomasello, 2005; Matthews et al., 2006;). The present study builds on this work, focusing not only on whether the full range of forms is used and whether this use appears appropriate, but also on whether the contexts in which the forms are used are sufficiently broad (e.g. including reference to objects that are not in the immediate environment, to have allowed the children to make an error in assessing the addressee’s mental state. 4. The study 4.1. Methodology The present study is a corpus analysis of the use of referring expressions in spontaneous conversation by children aged 3 and younger. It extends the empirical base of previous work on children’s use of nominal expressions within the

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

47

Table 1 Details of transcript files. Transcript

Gender

Age

MLU

Valian 02a Valian 04a Valian 15b Valian 19a Valian 21b Bloom,Peter14 Brown,Eve11 Brown,Adam03 Brown,Adam07

F ?a F M M M F M M

1;9.21 1;9.20 2;5.05 2;8.15 2;6.12 2;5.22 1;11 2;4.03 2;6.03

1.9 1.9 3.3 3.3 4.3 2.8 2.8 2.3 2.5

a

This child’s gender is not indicated in the transcript information.

Givenness Hierarchy framework, and also refines the analysis and interpretation of the data, by distinguishing more carefully data that reflect procedural, and relatively automatic ‘mind-reading’ involved in using forms consistent with the cognitive statuses they conventionally encode from meta-representational ability involved in assessing epistemic states, including how much information is relevant for the addressee. 4.1.1. Data The analysis was conducted in sum on nine corpus files from the CHILDES database (MacWhinney, 2000). We analyzed one transcript from each of five different children from the Valian (1991) corpus, three transcript files from the Brown (1973) corpus and the first 900 lines of one transcript file from the Bloom (1970) corpus. The files are transcriptions of audio recordings of sessions ranging from 30 to 60 minutes in length. The recording sessions took place at the child’s home, day care, or in a college playroom. At least one parent and investigator were present during each session. The age and MLU information for the nine files we analyzed are detailed in Table 1. 4.1.2. Coding The following established procedure used in the Givenness Hierarchy research was employed: (1) each coder independently identified personal and demonstrative pronouns and phrases headed by demonstrative determiners or by definite and indefinite articles in each of the files. (2) The intended interpretation of each of the phrases was identified and coded for the highest cognitive status the speaker could reasonably assume it had for the addressee at the point just before it was mentioned. Referring forms used by the children were coded using a set of coding guidelines that provide sufficient conditions for each cognitive status. The statuses are assessed from highest to lowest. If a referent does not meet any of the criteria for the highest status, in focus, the coder proceeds to the next highest, activated, and so on4. For example, in (11) below, the referent of this is coded as activated, according to the coding guideline that an entity is activated if it is present in the immediate perceptual environment. However, it cannot be coded as having a status higher than activated because the referent has not been mentioned in a syntactically prominent position in the preceding sentence, nor is there evidence that it meets any of the other criteria for in focus status listed in the coding guidelines.5 The referent of it, on the other hand, is coded as in focus because of the coding guideline criterion that anything mentioned earlier in the same sentence can be assumed to be in focus. (11)

CHI you touch this and you throw it to me (Valian 15b, 1487)

4 The coding guidelines were compiled during prior Givenness Hierarchy research and are available from Jeanette Gundel at gunde003@umn. edu. 5 As our analysis is based on audio recordings rather than video, it might be objected (as pointed out by a reviewer) that crucial information about objects present in the perceptual environment and non-verbal behaviors such as direction of eye-gaze, pointing, etc. would be lost. It should be noted, however, that with the exception of all but a few cases, the transcripts we analyzed contained sufficient information about such factors to determine the cognitive status of referents whose status could not be inferred from linguistic context alone. The transcripts included information about the setting (e.g. ‘child and adult are sitting at table which has a barn and various farm animals that the child is moving around’ or ‘(child) gives one of the guitars to Patsy.’ In those cases where the coders could not be reasonably confident about the status of the referent, they were instructed to assign the lower status (e.g. ‘activated’ when they could not be certain whether the referent could be expected to be at most activated, or actually in focus). The more conservative coding in such cases ensured that we did not miss any ‘errors’ because of lack of information about non-linguistic behaviors. For example, this in (11) was coded as at most activated, but the demonstrative would have been licit here even if the referent had been in focus, since anything in focus is also activated.

48

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

Graph 1. Children’s use of referring expressions.

Coding was performed separately by each author with spot checking for inter-coder reliability, which was high (94% agreement). Disagreements and unclear cases were resolved by consultation. 4.2. Results Overall, our findings were consistent with those of previous studies that found that children used the full range of forms, and used them appropriately, by age 3 or earlier.6 Graph 1 presents a summary of the children’s use of referring expressions. Each bar represents the different forms we analyzed: personal pronouns he/she/it7, demonstrative pronouns this/that, demonstrative determiners this N and that N, the definite article the N and the indefinite article a N8. Each bar also shows the proportion of times that the referent of these DPs could be assumed to have a given cognitive status for the addressee at the moment just before the form was encountered. For example, the bar for demonstrative pronouns this/that shows that roughly 18% were used for referents that could be assumed to be in focus, and the remainder were used for referents that could be assumed to be at most activated (i.e. activated, but not in focus). 4.2.1. Personal pronouns Personal pronouns are used almost exclusively for in focus referents (>95% of the time). It is especially noteworthy that many of the objects referred to with these pronouns were originally introduced, either by the child or the adult, with a demonstrative pronoun or full noun phrase, as they could not yet be assumed to be in the addressee’s focus of attention, as illustrated in the examples (12)--(14). (12)

Valian

(13)

Valian 21b, 2;6.12 a. 148 MOT 149--50 CHI

15b 160 161

2;5..05 MOT that’s not a shark CHI what is it?

and can you get the blocks out? uhhuh. I touched them.

6 It should be noted, however, that some experimental studies (e.g. Emslie and Stevenson, 1981; Warden, 1981) have suggested that children don’t master the full range of referring forms (pronouns, demonstratives, definite vs. indefinite article) until age 7 or even later. This is especially true for studies that involved elicited narrative tasks (e.g. Kail and Hickmann, 1992), which impose an additional cognitive burden on the subjects (see Hickmann, 2003 for an extensive review of the literature on this topic.) 7 The plural forms for each type of referring expression (e.g. they, those) were also analyzed and are included in the data presented in Graph 1. 8 There were no instances of indefinite ‘this’ in the data. This form is used very infrequently, if at all, by children aged 3 and younger.

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

b. c.

(14)

877 878 272 273

Brown

MOT CHI CHI CHI

49

well, what are we gonna do with these animals here? well, we’re going to give them a rest well, this is how the dog will behave he’ll walk along

Eve 1;11.8 63 FAT tie that shoe 64 EVE papa tie it

Children rarely used personal pronouns for referents that were not in focus, and when they did, they corrected after feedback from the adult, as in (15) and (16). (15)

(16)

Valian 21b 2;6.12 (barn mentioned 4 utterances back) a. 899 CHI I’m going to clean it with my popper 900 MOT you’re going to what? 901 CHI clean the barn with my popper b. (no recent mention of grass) 989 CHI he’s going to mow it MOT he’s going to mow it? 990 991 CHI he need to mow the grass 19a, 2;8.15 31 CHI it’s hot 32 MOT what’s that? 33 CHI that’s hot 34 CHI the beater’s hot 35--6 MOT oh, the beater’s hot. yes, it is.

4.2.2. Demonstrative pronouns Demonstrative pronouns are used almost exclusively for referents that are at least (and usually at most) activated, as in (17)--(22). (17)

Valian 02a a. 427 428 b. 729 730 731 732 733 734

1;9.21 CHI CHI MOT CHI MOT CHI MOT CHI

right there what is that? he’s playing with a car have that you don’t want Marco to have the car? no why not? CHILD have it

The use of that in line 730 of (7b) is especially interesting. Since the car is introduced in a non-prominent syntactic position (object of a preposition) the child would not be justified in assuming that it is in focus, and have it would have been questionable here. After the car has been mentioned a second time, however, he appropriately uses a personal pronoun (it) in line 734. (18)

Valian 19a 2;8.15 ( pause: sound of crashing dishes) 206 MOT how many cups do you think there are here 207 CHI that was the dishwasher making noise 208 MOT oh was it?

In line 207, the child is referring to the background noise of crashing dishes, indicated in the transcript in parentheses. While the mother can be assumed to be aware of the noise, there is no reason to assume that her attention is focused on it. Hence the child can appropriately use only that in referring to the noise for the first time.

50

(19)

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

Valian 15b 2;5.05 1089 MOT There once was a girl who had one little curl, right in the middle of her forehead 1090 CHI and when she was bad her was horrid 1091 CHI what does that mean? 1092 MOT it means terrible

In line 1091, the form that refers to the word ‘horrid’ or to the whole utterance. Either way, it is an utterance act that is being referred to. Since speech acts can be assumed to be activated (i.e. in awareness), but focus of attention is on some aspect of the content of the speech act, not on the act itself, that is again the only appropriate form here (cf. #’what does it mean’ in this context). But once the utterance act is referred to, it can be subsequently referred to with it, as in 1092. Sometimes demonstratives are used for in focus referents, which is licit, as anything in focus is also activated. (20)

Valian 557 558 559

(21)

Valian 2;5.05 25 INV do you know what those are? 26 CHI what are those? (cf. also ok ‘what are they’)

(22)

Brown, Adam03, 2;4.03 141 ADA truck, look 142 MOT oh, it’s a truck 143 ADA oh no, busy bulldozer 144 MOT oh no, it’s a busy bulldozer 145 ADA dat busy bulldozer truck

4a 1;9.20 CHI the rabbit MOT that’s not a rabbit, that’s a mouse CHI that’s a rabbit (cf. also ok - it’s a rabbit)

As with the few incorrect uses of personal pronouns, demonstrative pronouns used for entities that are not yet activated are corrected after feedback, as in (23) below. (23)

Brown, 1219 1220 1221

Adam07, 2;6.03 ADA monkey get dat MOT what? ADA monkey get de penny

As noted in Gundel et al. (1993:287--288), the proximal demonstrative this in English (both pronoun and determiner) signals, and thus requires, not only activation, but speaker-activation. That is, its referent must be an entity introduced by the speaker or within the speaker’s context space. Occurrences of this used by the children in the examined transcripts were consistent with this restriction. Following are some examples. (24)

Bloom, 1970, Peter14, 2;5.23 a. 3006 PET This is a guitar % has long wall in arms in ‘guitar’ position 3007 PET awoh % holding another wall 3008 LOI what’s this? 3009 LOI is that a guitar? 3010 PET hm 3011 LOI is that a guitar too? 3012 PET yeah 3013 PET two three guitars # here one# you % gives one of the ‘guitars’ to Patsy 3014 PAT oh thank you 3015 PET that’s a guitar

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

51

b. % looks at Lois’ big wall guitar, pointing to wall Lois is holding, then to smaller one he has 3059 PET No. that’s my guitar and this is your guitar 4.2.3. Demonstrative determiners The demonstrative determiner that was always used for referents that are familiar or higher, consistent with the restriction that this form signals familiarity. (25)

Valian 713 714 715 716 717

(26)

Valian 04a; 1:9.20 670 MOT I’m sorry I brought it up, Child, because I’m not sure where your keys are 671 MOT Oh you know what 672 MOT You can have these for now, you just have to remember to given them back. 673 CHI not that keys 674 MOT not those keys because those aren’t yours

(27)

Bloom, 1970, Peter 2;5.23 a. 570 PET see that light 571 LOI mmhmm 572 PET I turn that light on in the living room %act touches light in house9 b. %act takes Lois’ pencil, but can’t put it with his because microphone is on top of pad 3392 PET I want take that microphone off

04a; 1;9.20 CHI back door MOT back door CHI not that door MOT I know that door is more fun, but it’s too cold to open CHI open it. That door open

As with the demonstrative pronoun this, the demonstrative determiner this is used for entities that are not only activated, but speaker-activated, as in the following examples. (28)

Bloom, 1970, Peter 2;5.23 347 PET having trouble. I found another light 348 %act pulling out another lamp 349 LOI mm# yes you did 351 PET turn this light

(29)

Valian 106 107 108

21b; 2;6.12 CHI um, what’s this in this thing here CHI what goes in here MOT I think it’s called a cement mixer.

4.2.4. Definite article With only one or two possible exceptions, the definite article the is always used for entities that are at least uniquely identifiable, as in (30)--(33). (30)

Brown, 1973, Eve 1;11.8 %act looking at photographs 49 FAT Eve # please 50 EVE no # let me hold it 51 EVE Eve in the snow

9 This example, and other examples below, contain more instances of information about non-linguistic behaviors available in the transcripts, marked as ‘%act’ (see footnote 5).

52

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

In (30), Eve uses a definite article to refer to the snow, which is activated (since it is apparently visible in the picture), and is therefore also uniquely identifiable. A personal pronoun (it) would clearly not have been appropriate here, as there is no reason to think the snow is in her father’s focus of attention. Attention is focused on the photograph, not specific details of it. Although being activated would otherwise license the use of a demonstrative pronoun, this or that, these forms would not allow the referent to be uniquely identified, as they do not contain sufficient conceptual content to distinguish the snow from other activated entities in the picture. (31)

Brown, 1243 1244 1245 1246 1247

1973, Adam 2;6.3 ADA what dat fire engine doin MOT there isn’t a fire engine there MOT there’s just a fireman on a ladder ADA what the fireman doing? MOT he may be going to help fight fire

In (31) the fireman is at least activated, and probably in focus in 1246, as he has been introduced in a syntactically prominent position in the previous utterance by the mother. A personal pronoun (he) would therefore have been appropriate. But since anything in focus is also uniquely identifiable, the definite article phrase is appropriate here as well. (32)

Bloom, 1970, 357 PET 358 LOI 359 PET 360 LOI 361 PET

Peter 2;5.23 what’s over there? over there behind Jenny? yeah that’s a house who’s go in the house

In (32), the child again uses a definite article to refer to something that has just been introduced with an indefinite article by the adult. The house is at least activated as it was mentioned in the immediately preceding utterance, and is hence also uniquely identifiable. (33)

Valian 19, 2;8.15 79 MOT where’s her checkbook? 80 CHI right over here 81 CHI on the table

In (33) the table is at least uniquely identifiable by virtue of being in the visual context. Example (34) illustrates one of a few possible errors in definite article uses, where the referent may be at most type identifiable. (34)

Bloom, 4300 4301 4302 4303 4304 4305 4306 4307

1970, Peter 2:5.23 LOI that guitar’s almost as big as you are %act (puts her wall back in house) PET it’s the guitar!! LOI well I think’s it’s gonna be a wall right now PET it’s the guitar!! %act bringing Lois her guitar PET it’s the guitar!! LOI oh well # I don’t want to play the guitar any more

If Peter is merely identifying the wall as a guitar in 4302, then it is at most type identifiable, not uniquely identifiable, or even referential, and a definite article is inappropriate. It is likely, however, that he intends to identify it as the wall-guitar that he referred to earlier (see (24) above). In that case, it is familiar, and hence also uniquely identifiable and not an error. 4.2.5. Indefinite article Finally, since the indefinite article only requires the lowest cognitive status, type identifiable, there are technically no errors in the use of this form. The form itself does appear to be acquired later than the other forms, however, and is frequently omitted in the earlier transcripts from younger children, at times when children consistently use the other determiners and pronominal forms. Example (35) illustrates Eve’s resistance to using the indefinite article at an age when

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

53

she is consistently using pronouns and the definite article (see example (30) above). Note that she knows when to use it, as shown by her self-correction in line 252. (35)

Brown, 1973, 248 MOT 249 EVE 251 MOT 252 EVE 254 MOT 255 EVE 257 MOT 258 EVE

Eve 1:11.8 what do you want? I want sandwich you want what? a sandwich sandwich yeah well# what do you want to drink? I want I want sandwich

Example (36) shows a similar instance of indefinite article omission from the Valian corpus. (36)

Valian 04a, 1;9.20 75 MOT then he looked way down, what did he see? 76 CHI boat 77 MOT a boat!

However, when the indefinite article is used, it is usually used correctly, even in the same transcript, as in (35) above and in (37). (37)

Valian 04a, 1;9.20 (looking at a picture book) 623 CHI bird 624 MOT bird 625 CHI on a tree 626 MOT on the tree

Notice that even though the mother uses the definite article in 626, since the tree is uniquely identifiable from the picture they are looking at, the child’s use of the indefinite article in 625 is not incorrect because the tree is type identifiable, and marking it as uniquely identifiable is not relevant. 5. Discussion and possible implications for theory of mind development As shown in the previous sections, appropriate use of referring forms (specifically certain pronouns and determiners) involves the ability to make reasonable assumptions about the addressee’s memory and attention state in relation to the intended referent, i.e. it involves the attribution of mental states to others. Children begin using such forms (with very few errors) toward the end of their second year, and use the full range of such forms appropriately by age 3 or earlier. Thus, these children appear to be sensitive to the mental states of others when using such forms. The appropriate use of referring forms by children aged 3 and younger is consistent with findings that they can correctly infer some mental states (e.g. intention, attention, desire) before age 3 (cf. Tomasello and Haberl, 1993; Baron-Cohen, 1995; Repacholi and Gopnik, 1997). However, standard theory of mind tests with children all over the world find that they are typically unable to pass such tests before the age of 4. This naturally raises the question of whether the mind-reading ability involved in appropriate use of referring forms is more like the mind-reading abilities that have been shown to develop before age 3 (even before age 2) and less like abilities measured by standard theory of mind tests, which typically develop after age 4. A number of factors may be relevant here. First, the ability to attribute mental states of the type encoded by pronouns and determiners (e.g. awareness/activation, attention, familiarity) is more like attribution of mental states that is known to develop before age 3, in the sense that these states do not require representation of others’ propositional attitudes10. The one possible exception here is the status ‘uniquely identifiable’. However, it is important to note that the majority of definite article phrases in our data are at least familiar. At most uniquely identifiable entities, i.e. ones that are uniquely identifiable, but not familiar, and require the

10 It should be noted, however, that of the mental states encoded by determiners and pronouns only attention has been discussed extensively in the theory of mind literature.

54

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

addressee to be able to construct a new unique representation solely on the basis of conceptual content encoded in the phrase or by association with a bridging inference to an already activated entity are rare in the speech of children aged 3 and younger. This may be because assessing the ability to construct a new representation requires more conscious, representational knowledge. Second, the mind-reading ability involved in using determiners and pronouns in spontaneous verbal interactions is arguably implicit and automatic. This is in contrast to the mind-reading abilities involved in theory of mind tests, which are explicit and representational, and are measured by asking subjects to consciously reflect on and articulate beliefs and actions of others. In support of this distinction, there is evidence that 3 year-olds can assess belief implicitly, e.g. they can look to where they think people with false beliefs will look, and can decide who (not) to believe, even if they cannot provide a verbal report of the source of belief (Robinson and Whitcombe, 2003; Koenig et al., 2004). Finally, the GH account of the contribution of determiners/pronouns to the distribution and interpretation of referring forms, specifically that they encode procedural information about how to mentally access the referent, is consistent with the view that the information encoded by these forms manipulates conceptual representations, but is not representational itself (Wilson and Sperber, 1993). As such, it is inaccessible to consciousness. For example, Matsui et al. (2006) investigated 3--6 year-old Japanese-speaking children’s ability to use information about speaker’s degree of certainty and quality of evidence with respect to an expressed proposition when encoded by open-class items, e.g. epistemic verbs ‘know’, ‘think’, vs. highfrequency, closed-class, sentence-final particles, which convey procedural (implicit) rather than conceptual (explicit) information. Their results suggest that children can use information encoded in particles earlier than information encoded in verbs, and understanding of particles (unlike epistemic verbs) does not correlate with fully representational false-belief understanding. The assumption that procedural information about cognitive status encoded by pronouns and determiners, like information encoded by other closed-class items, is non-representational and inaccessible to consciousness would explain why adults, including many linguists, cannot articulate explicitly why they use it vs. that or the vs. a. Distinguishing different stages in mind-reading ability -- implicit, procedural before explicit, conceptual, and (meta-) representational, makes it possible to explain why children use referring forms appropriately, consistent with the cognitive status they require, before they can make choices in referring forms based on the quantity of relevant information, and specifically before they are fully aware of the pragmatic inferences that can result from providing too little or too much information. Unlike assessment of such states as attention and familiarity, determining how much information about these states should be conveyed requires conscious reflection on what the addressee has represented and how much information would be relevant in a given context. Our finding that children learn to use determiners/pronouns that encode cognitive status appropriately before they exhibit sensitivity to implicatures associated with these forms in different contexts is also consistent with results from experimental studies which suggest that sensitivity to scalar implicatures associated with use of modals and quantifiers develops relatively late in children (Noveck, 2001; Verbuk, 2007; Papafragou and Musolino, 2003; inter alia). For example, like the children in these studies, young children sometimes use ‘a’ when it would have been more appropriate to provide the stronger information about cognitive status encoded by ‘the’. Thus, in an experimental study with preschoolers (3--3.6 and 4--5.) Maratsos (1974, 1976) found that when told a story about a boy and a girl and then asked ‘who was making a noise’, the older group said ‘the girl’ 94% of time, but the younger group used ‘the girl’ only 54% of the time, and ‘a girl’ the rest of the time. Both groups were correct in that they used a form whose meaning was consistent with the cognitive status of the referent, as the girl was uniquely identifiable and anything uniquely identifiable is also type identifiable. However, ‘a girl’ is less adult-like in this context as it provides less information than would be relevant about the cognitive status of the referent, and ‘a’ thus typically implicates that the referent is not uniquely identifiable to the addressee. Similar examples of children’s use of the weaker ‘a’, when it would be relevant to use the strong ‘the’ are occasionally found in spontaneous conversations, as in example (38) (38) [Peter 2:0.26] a. 528 PET the baby’s taking a bath 529 LYN mmmhm. getting all clean 530 %act Lynn puts baby in bed 531 LYN now the baby can go to bed. there 532 PET a baby’s going to bed b. 730 %act Peter putting girl in chair 731 PET eat lunch; what’s that. 732 LYN that’s a chair. 733 PET can a, can a girl sit on a chair Finally, in an experimental study of children’s use of the definite and indefinite article, Schaeffer and Matthewson (2005) found that children used both definite and indefinite articles in situations where adults always used an indefinite

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

55

article. Subjects were shown a picture of Mickey Mouse who just finished drawing a house. When the investigator asked ‘‘What did Mickey Mouse do?’’, the adults always answered ‘‘He drew a house’’. However, only about half of the children responded using an indefinite article. The rest responded ‘‘He drew the house.’’ Schaeffer and Matthewson attempt to explain these findings by proposing that the children lack a pragmatic parameter (the Concept of Non-Shared Assumptions), requiring them to distinguish between their own belief state and that of their interlocutor. That is, if something is identifiable to them, then they treat it as identifiable to the addressee. But this explanation is inconsistent with the fact that the house actually is identifiable to the investigator in this context, as it is in the perceptual environment. It is also inconsistent with findings of children’s use of the definite article in spontaneous conversations, like those discussed here, which indicate that children rarely make mistakes using the definite article in environments where the referent is not uniquely identifiable to the addressee. An alternative explanation for these results, which would be expected within the GH framework, is that the children assume (correctly) that the investigator can uniquely identify the referent. However, their use of the definite article here is inconsistent with the pragmatic fact that this form provides more information than is necessary in the context of this experiment, as it is the property of being a house and not the particular identity of the house that is relevant. Adults thus use the indefinite article when conditions for using the definite article are met, if token/unique identifiability is irrelevant, as it is in this case (and in examples like (9) in section 2 above). Children, on the other hand, like adults, use both the indefinite and definite article for entities that are uniquely identifiable, but they are not yet sensitive to making the choice based on pragmatic inferences that could arise when irrelevant information is provided. They thus use both forms in this experiment. Evidence of relatively late acquisition of sensitivity to providing as much information as necessary, but not more than is relevant for the addressee, is also seen in children’s use of demonstratives. Gundel et al. (1993) invoke relevance of cognitive status information provided by demonstratives to explain the relatively low frequency of demonstrative determiners relative to the definite article and demonstrative pronouns relative to personal pronouns in adult usage across languages. They argue that for pronouns, which encode little, if any, conceptual content, information about cognitive status is crucial. Unstressed personal pronouns or zero, which explicitly signal that the referent is in focus, are thus typically used whenever conditions for their use are met, and use of demonstrative pronouns, which only signal activation, often implicates that the referent is not in focus. For full nominal phrases, on the other hand, the stronger cognitive status information provided by demonstrative determiners is usually unnecessary, given conceptual content encoded in the rest of the phrase; so the weaker definite article or bare nominal form is used more frequently. If children are less sensitive to relevance and scalar implicature, this would also explain the relatively more frequent use of demonstratives by children as compared to adults (see Gundel and Page, 1998, inter alia). 6. Conclusion This paper has broadened the empirical base of previous work on children’s acquisition of referring expressions within the Givenness Hierarchy framework, supporting and further refining the hypothesis that the appropriate use of referring expressions involves the ability to take into account the mental states of others in at least two ways: (1) the ability to appropriately assess what cognitive status the intended interpretation has for the addressee at a given point in the discourse and (2) the ability to assess how much information about conceptual content and cognitive status of the intended referent is sufficient and relevant for the addressee. We have suggested that the ability described in (1), which is necessary for assessing the appropriateness of using a particular determiner or pronoun, is implicit, non-representational and relatively automatic; it corresponds to a kind of mind-reading ability that develops at a relatively early age in children. This would explain why children by the age of 3 are able to use the full range of cognitive status signaling forms more or less correctly. The ability described in (2), on the other hand, appears to require more conscious reasoning about the epistemic states of others, and as such corresponds to a kind of mind-reading ability that is typically not fully developed until after the age of 4. This would explain the relatively late development of sensitivity to pragmatic inferences and of the role of relevance in children’s use of referring expressions, including the relatively high frequency of demonstratives in the speech of children aged 3 and younger. While analysis of naturalistic, spontaneous discourse clearly shows that children use the full range of forms, and use them appropriately by age three and earlier, it does not offer sufficient opportunity for children to make errors based on cognitive status alone, since most of the things they refer to are at least activated for the addressee. Moreover, except for the relatively high frequency of demonstratives, evidence of children’s ability to assess how much information about cognitive status is relevant for the addressee does not occur with sufficient frequency in spontaneous conversation to allow any firm conclusions about the development of this ability.11 Investigations of spontaneous conversation must

11 A similar observation is made by Wicklund, 2012, who investigated high functioning autistic children’s use of referring forms in conversational settings.

56

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

therefore be complemented by experimental studies which explicitly construct contexts where such abilities can be tested in a more controlled setting.12 References Ariel, Mira, 1990. Accessing Noun-Phrase Antecedents. Routledge, London. Baron-Cohen, Simon, 1995. Mindblindness: An Essay on Autism and Theory of Mind. MIT Press, Cambridge, MA. Bennett-Kastor, Tina, 1983. Noun phrases and coherence in child narratives. Journal of Child Language 10 (1), 35--49. Bittner, Dagmar, 2002. Emergence of grammatical complexity and markedness in the acquisition of verb and noun phrases in German. Future Challenges for Natural Linguistics 5--56. Bloom, Lois, 1970. Language Development: Form and Function in Emerging Grammars. MIT Press, Cambridge, MA. Brown, Roger, 1973. A First Language: The Early Stages. Harvard University Press, Cambridge, MA. Emslie, Hazel C., Stevenson, Rosemary J., 1981. Pre-school children’s use of the articles in definite and indefinite referring expressions. Journal of Child Language 8 (2), 313--328. Givo´n, Talmy, 1983. Topic-Continuity in Discourse: A Quantitative Cross-Language Study. John Benjamins, Amsterdam. Godfrey, J., Holliman, E., McDaniel, J., 1992. SWITCHBOARD: telephone speech corpus for research and development. In: IEEE International Conference on Acoustics, Speech, and Signal Processing 1992, ICASSP-92, vol. 1. Grice, Paul H., 1975. Logic and conversation. Syntax and Semantics 3, 41--58. Gundel, Jeanette, 2009. Children’s use of referring expressions. What can it tell us about theory of mind? Cognitive Critique 1--28. Gundel, Jeanette K., 2010. Reference and accessibility from a Givenness Hierarchy perspective. International Review of Pragmatics 148--168. Gundel, Jeanette, 2011. Child language, theory of mind, and the role of procedural markers in identifying referents of nominal expressions. In: Escandell-Vidal, E., Leonetti, M., Ahern, A. (Eds.), Procedural Meaning: Problems and Perspectives (Current Research in the Semantics/ Pragmatics Interface, vol. 25. Emerald Group Publishing Limited, United Kingdom, pp. 205--231. Gundel, Jeanette, Page, Sherri, 1998. The givenness hierarchy and children’s use of referring expressions. Paper presented at the Annual Meeting of the Linguistic Society of America, January 6, 1998. Gundel, Jeanette, Hedberg, Nancy, Zacharski, Ron, 1993. Cognitive status and the form of referring expressions in discourse. Language 69 (2), 274--307. Gundel, Jeanette, Sera, Maria, Page, Sherri, 1999. The acquisition of referring expressions in English and Spanish. Paper presented at the Child Language Research Forum, Stanford, April 10, 1999. Gundel, Jeanette, Ntelitheos, Dmitris, Kowalsky, Melinda, 2007. Children’s use of referring expressions: some implications for theory of mind. ZAS Working Papers in Linguistics 39, pp. 1--23. Gundel, Jeanette, Hedberg, Nancy, Zacharski, Ron, 2012. Underspecification of cognitive status in reference production: some empirical predictions. Topics in Cognitive Science 4 (2), 249--268. Hickmann, Maya, 2003. Children’s Discourse: Person, Space and Time Across Languages. Cambridge University Press, Cambridge, MA. Horn, Laurence, 1972. On the semantic properties of logical operators in English: UCLA dissertation. Jackendoff, Ray, 2002. Foundations of Language. Oxford University Press, London and New York. Johnson, Kaitlin R., 2012. Development of scalar implicatures and the indefinite article. PhD. Dissertation. University of Minnesota, Minneapolis. Kail, Michèle, Hickmann, Maya, 1992. French children’s ability to introduce referents in narratives as a function of mutual knowledge. First Language 12 (34), 73. Koenig, Melissa A., Clements, Fabrice, Harris, Paul L., 2004. Trust in testimony: children’s use of true and false statements. Psychological Science 10, 694--698. MacWhinney, Brian, 2000. The CHILDES Project: Tools for Analyzing Talk. Lawrence Erlbaum Associates Inc., US. Maratsos, Michael, 1974. Preschool children’s use of definite and indefinite articles. Child Development 45, 446--455. Maratsos, Michael, 1976. The Use of Definite and Indefinite Reference in Young Children. Cambridge University Press, Cambridge. Matsui, Tomoko, Yamamoto, Taeko, McCagg, Peter, 2006. On the role of language in children’s early understanding of others as epistemic beings. Cognitive Development 21 (2), 158--173. Matthews, Danielle, Lieven, Elena, Theakston, Anna, Tomasello, Michael, 2006. The effect of perceptual availability and prior discourse on young children’s use of referring expressions. Applied Psycholinguistics 27 (3), 403--422. Noveck, Ira A., 2001. When children are more logical than adults: experimental investigations of scalar implicature. Cognition 78 (2), 165--188. O’Neill, Daniela K., 1996. Two-year-old children’s sensitivity to a parent’s knowledge state when making requests. Child Development 67, 659-677. Papafragou, Anna, Musolino, Julien, 2003. Scalar implicatures: experiments at the semantic--pragmatic interface. Cognition 86 (3), 233--282. Premack, David, Woodruff, Guy, 1978. Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences 1 (4), 515--526. Prince, Ellen, 1981. Toward a taxonomy of given/new information. In: Cole, P. (Ed.), Radical Pragmatics. Academic Press, New York, pp. 223-254. Repacholi, Betty M., Gopnik, Alison, 1997. Early reasoning about desires. Evidence from 14- and 18-month-olds. Developmental Psychology 33 (1), 12--21. Robinson, Elizabeth J., Whitcombe, E.L., 2003. Children’s suggestibility in relation to their sources of knowledge. Child Development 74, 48--62. Salazar Orvig, Anne, Marcos, Haydée, Morgenstern, Aliyah, Hassan, Rouba, Leber-Marin, Jocelyne, Parès, Jacques, 2010. Dialogical beginnings of anaphora: the use of third person pronouns before the age of 3. Journal of Pragmatics 42 (7), 1842--1865.

12 See, for example, Johnson, 2012 for a recent experimental study comparing pragmatic inferences in the use of the definite and indefinite articles a and the in English by 5 and 7 year old children.

J.K. Gundel, K. Johnson / Journal of Pragmatics 56 (2013) 43--57

57

Schaeffer, Jeanette, Matthewson, Lisa, 2005. Grammar and pragmatics in the acquisition of article systems. Natural Language and/Linguistic Theory 23, 53--101. Schnell de Acedo, Beatrice, 1994. Early morphological development: the acquisition of articles in Spanish. In: Sokolov, J., Snow, C. (Eds.), Handbook of research in language development using CHILDES. LEA, Hillsdale, NJ, pp. 210--253. Tomasello, Michael, Haberl, Katharina, 1993. Understanding attention: 12- and 18-moth-olds know what’s new for other persons. Developmental Psychology 39, 906--912. Valian, Virginia, 1991. Syntactic subjects in the early speech of American and Italian children. Cognition 40 (1--2), 21--81. Verbuk, Anna, 2007. Why children do not compute irrelevant scalar implicatures. Paper presented at the Annual Meeting of the Linguistic Society of America. Anaheim, CA, Jaunuary 7, 2007. Warden, David, 1981. Learning to identify referents. British Journal of Psychology 78, 93--99. Wicklund, Mark, 2012. Use of referring expressions by autistic children in spontaneous conversations: does impaired metarepresentational ability affect reference production? Ph.D. Dissertation. University of Minnesota, Minneapolis. Wilson, Deirdre, Sperber, Dan, 1993. Linguistic form and relevance. Lingua 90 (2), 1--25. Wittek, Angelika, Tomasello, Michael, 2005. Young children’s sensitivity to listener knowledge and perceptual context in choosing referring expressions. Applied Psycholinguistics 26 (4), 541--558.